diff --git a/.circleci/config.yml b/.circleci/config.yml index 55fa9410845..63012ce3fc0 100644 --- a/.circleci/config.yml +++ b/.circleci/config.yml @@ -1483,7 +1483,7 @@ jobs: - run: name: Run tests command: | - uv run --no-sync python -m pytest -vv tests/local_testing/test_basic_python_version.py -k "not v2_resolver" + uv run --no-sync python -m pytest -vv tests/local_testing/test_basic_python_version.py -k "not legacy_resolver" installing_litellm_on_python_3_13: docker: @@ -1507,9 +1507,9 @@ jobs: - run: name: Run tests command: | - uv run --no-sync python -m pytest -v tests/local_testing/test_basic_python_version.py -k "not v2_resolver" + uv run --no-sync python -m pytest -v tests/local_testing/test_basic_python_version.py -k "not legacy_resolver" - installing_litellm_on_python_v2_migration_resolver: + installing_litellm_on_python_legacy_migration_resolver: docker: - *python312_image - image: cimg/postgres:16.0@sha256:b125148bc76e8e8eee5eb3ad6020a3a14110a14e8192f1c645128afebe2e2f84 @@ -1536,10 +1536,10 @@ jobs: url: tcp://localhost:5432 timeout: "60" - run: - name: Run v2 migration resolver proxy smoke test + name: Run legacy migration resolver proxy smoke test command: | uv run --no-sync python -m pytest -vv \ - tests/local_testing/test_basic_python_version.py::test_litellm_proxy_server_config_no_general_settings_v2_resolver + tests/local_testing/test_basic_python_version.py::test_litellm_proxy_server_config_no_general_settings_legacy_resolver helm_chart_testing: machine: @@ -2879,7 +2879,8 @@ jobs: command: | if grep -q "Error: P1001: Can't reach database server at" docker_output.log && \ (grep -q "Database setup failed after multiple retries" docker_output.log || \ - grep -q "ERROR: Application startup failed. Exiting." docker_output.log); then + grep -q "ERROR: Application startup failed. Exiting." docker_output.log || \ + grep -q "Database migration cannot proceed" docker_output.log); then echo "Expected error found. Test passed." else echo "Expected error not found. Test failed." @@ -3011,7 +3012,7 @@ workflows: filters: *main_branches - installing_litellm_on_python_3_13: filters: *main_branches - - installing_litellm_on_python_v2_migration_resolver: + - installing_litellm_on_python_legacy_migration_resolver: filters: *main_branches - helm_chart_testing: requires: diff --git a/CLAUDE.md b/CLAUDE.md index d9e9e8f1586..6af8390b1af 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -37,7 +37,7 @@ Same applies for filing bug reports and feature requests, with .github/ISSUE_TEM If you're resolving a linear ticket, in the "## Linear ticket" section of the PR, say "Resolves LIT-1234", replacing "LIT-1234" with the actual ticket id that you're resolving. If you don't have the ticket id, don't make one up or search for it. Just leave the section blank -Never use `pytest` commands or the like as "Screenshots / Proof of Fix". We prefer curl'ing a live proxy instance running on localhost:4000 (I like to run it with `python litellm/proxy/proxy_cli.py --config litellm/proxy/dev_config.yaml --detailed_debug --reload --use_v2_migration_resolver 2>&1 | tee litellm.log`; the Admin UI dev server is `npm run dev` in `ui/litellm-dashboard`, served on port 3000) and showing both the command run and the output. Also, it should hit real LLM provider APIs, not mocks, and cost real $$$ because that is the most realistic test. The proof of fix should be exactly what the end user / customer would see / do. The run logs in PR #27703 is a prime example of how to do it (not a huge fan of using a python test script that future me and the team will have no visibility into; I prefer just curl commands or a short list of bash commands (e.g., using `for`)). If it's a UI thing, just tell me which URLs to go to (e.g., http://localhost:4000/ui/?page=logs), where to click, what fields to fill out, etc. along with the other commands to run in an ordered list, and I'll do it myself and post the screenshots after you make the PR +Never use `pytest` commands or the like as "Screenshots / Proof of Fix". We prefer curl'ing a live proxy instance running on localhost:4000 (I like to run it with `python litellm/proxy/proxy_cli.py --config litellm/proxy/dev_config.yaml --detailed_debug --reload 2>&1 | tee litellm.log`; the Admin UI dev server is `npm run dev` in `ui/litellm-dashboard`, served on port 3000) and showing both the command run and the output. Also, it should hit real LLM provider APIs, not mocks, and cost real $$$ because that is the most realistic test. The proof of fix should be exactly what the end user / customer would see / do. The run logs in PR #27703 is a prime example of how to do it (not a huge fan of using a python test script that future me and the team will have no visibility into; I prefer just curl commands or a short list of bash commands (e.g., using `for`)). If it's a UI thing, just tell me which URLs to go to (e.g., http://localhost:4000/ui/?page=logs), where to click, what fields to fill out, etc. along with the other commands to run in an ordered list, and I'll do it myself and post the screenshots after you make the PR If you ever write any human-facing text (pull requests, issues, commit messages, discussion posts, github comments, release notes, docs, etc.), always follow these guidelines to sound less AI-y: - don't use emojis diff --git a/README.md b/README.md index 68aaa09ec98..92757fcbbc1 100644 --- a/README.md +++ b/README.md @@ -354,6 +354,8 @@ curl -X POST 'http://0.0.0.0:4000/v1/chat/completions' \ | [Petals (`petals`)](https://docs.litellm.ai/docs/providers/petals) | ✅ | ✅ | ✅ | | | | | | | | | [Pinstripes (`pinstripes`)](https://docs.litellm.ai/docs/providers/pinstripes) | ✅ | ✅ | ✅ | | | | | | | | | [Predibase (`predibase`)](https://docs.litellm.ai/docs/providers/predibase) | ✅ | ✅ | ✅ | | | | | | | | +| [Qwen AI Platform (`qwen_ai_platform`)](https://docs.litellm.ai/docs/providers/qwencloud) | ✅ | ✅ | ✅ | ✅ | ✅ | | | | | ✅ | +| [QwenCloud (`qwencloud`)](https://docs.litellm.ai/docs/providers/qwencloud) | ✅ | ✅ | ✅ | ✅ | ✅ | | | | | ✅ | | [Recraft (`recraft`)](https://docs.litellm.ai/docs/providers/recraft) | | | | | ✅ | | | | | | | [Replicate (`replicate`)](https://docs.litellm.ai/docs/providers/replicate) | ✅ | ✅ | ✅ | | | | | | | | | [Sagemaker Chat (`sagemaker_chat`)](https://docs.litellm.ai/docs/providers/aws_sagemaker) | ✅ | ✅ | ✅ | | | | | | | | diff --git a/basedpyright-code-budget.json b/basedpyright-code-budget.json index a07b9352659..df52069e71f 100644 --- a/basedpyright-code-budget.json +++ b/basedpyright-code-budget.json @@ -1,6 +1,6 @@ { "reportAny": { - "limit": 14765 + "limit": 14076 }, "reportArgumentType": { "limit": 2216 @@ -24,7 +24,7 @@ "limit": 19 }, "reportExplicitAny": { - "limit": 4493 + "limit": 4128 }, "reportFunctionMemberAccess": { "limit": 7 @@ -54,10 +54,10 @@ "limit": 0 }, "reportMissingParameterType": { - "limit": 5607 + "limit": 5601 }, "reportMissingTypeArgument": { - "limit": 15310 + "limit": 15306 }, "reportMissingTypeStubs": { "limit": 40 @@ -105,13 +105,13 @@ "limit": 109 }, "reportUnknownMemberType": { - "limit": 38368 + "limit": 38350 }, "reportUnknownParameterType": { - "limit": 19633 + "limit": 19626 }, "reportUnknownVariableType": { - "limit": 29908 + "limit": 29890 }, "reportUnnecessaryCast": { "limit": 111 @@ -123,7 +123,7 @@ "limit": 5 }, "reportUnnecessaryIsInstance": { - "limit": 828 + "limit": 826 }, "reportUntypedBaseClass": { "limit": 0 diff --git a/enterprise/pyproject.toml b/enterprise/pyproject.toml index cac98b69793..8360c0a077d 100644 --- a/enterprise/pyproject.toml +++ b/enterprise/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "litellm-enterprise" -version = "0.1.62" +version = "0.1.63" description = "Package for LiteLLM Enterprise features" readme = "README.md" requires-python = ">=3.9" @@ -26,7 +26,7 @@ required-version = ">=0.10.9" module-root = "" [tool.commitizen] -version = "0.1.62" +version = "0.1.63" version_files = [ "pyproject.toml:^version", "../pyproject.toml:litellm-enterprise==", diff --git a/helm/litellm/templates/backend/deployment.yaml b/helm/litellm/templates/backend/deployment.yaml index 5c0431fc0bd..0db2f0b3d43 100644 --- a/helm/litellm/templates/backend/deployment.yaml +++ b/helm/litellm/templates/backend/deployment.yaml @@ -7,6 +7,10 @@ metadata: {{- include "litellm.commonLabels" . | nindent 4 }} app.kubernetes.io/component: backend spec: + {{- with .Values.backend.strategy }} + strategy: + {{- toYaml . | nindent 4 }} + {{- end }} selector: matchLabels: {{- include "litellm.backend.selectorLabels" . | nindent 6 }} diff --git a/helm/litellm/templates/gateway/deployment.yaml b/helm/litellm/templates/gateway/deployment.yaml index d5363d0096e..5030ba2c9dc 100644 --- a/helm/litellm/templates/gateway/deployment.yaml +++ b/helm/litellm/templates/gateway/deployment.yaml @@ -7,6 +7,10 @@ metadata: {{- include "litellm.commonLabels" . | nindent 4 }} app.kubernetes.io/component: gateway spec: + {{- with .Values.gateway.strategy }} + strategy: + {{- toYaml . | nindent 4 }} + {{- end }} selector: matchLabels: {{- include "litellm.gateway.selectorLabels" . | nindent 6 }} diff --git a/helm/litellm/templates/migrations-job.yaml b/helm/litellm/templates/migrations-job.yaml index 9cd8397f794..8d33081e72f 100644 --- a/helm/litellm/templates/migrations-job.yaml +++ b/helm/litellm/templates/migrations-job.yaml @@ -7,6 +7,8 @@ # # Running this pre-upgrade closes the window where new application pods would # otherwise serve traffic against the previous release's unmigrated schema. +# Argo CD users can swap the Helm hook for a PreSync hook through +# `migrationJob.hooks`, which re-runs the Job on every sync. apiVersion: batch/v1 kind: Job metadata: @@ -14,10 +16,18 @@ metadata: labels: {{- include "litellm.commonLabels" . | nindent 4 }} app.kubernetes.io/component: migrations + {{- if or .Values.migrationJob.hooks.helm.enabled .Values.migrationJob.hooks.argocd.enabled }} annotations: + {{- if .Values.migrationJob.hooks.helm.enabled }} helm.sh/hook: pre-install,pre-upgrade helm.sh/hook-delete-policy: before-hook-creation - helm.sh/hook-weight: "0" + helm.sh/hook-weight: {{ .Values.migrationJob.hooks.helm.weight | default "0" | quote }} + {{- end }} + {{- if .Values.migrationJob.hooks.argocd.enabled }} + argocd.argoproj.io/hook: PreSync + argocd.argoproj.io/hook-delete-policy: BeforeHookCreation + {{- end }} + {{- end }} spec: backoffLimit: {{ .Values.migrationJob.backoffLimit }} ttlSecondsAfterFinished: {{ .Values.migrationJob.ttlSecondsAfterFinished }} diff --git a/helm/litellm/templates/ui/deployment.yaml b/helm/litellm/templates/ui/deployment.yaml index 91d6de39ea6..b992b347bad 100644 --- a/helm/litellm/templates/ui/deployment.yaml +++ b/helm/litellm/templates/ui/deployment.yaml @@ -7,6 +7,10 @@ metadata: {{- include "litellm.commonLabels" . | nindent 4 }} app.kubernetes.io/component: ui spec: + {{- with .Values.ui.strategy }} + strategy: + {{- toYaml . | nindent 4 }} + {{- end }} selector: matchLabels: {{- include "litellm.ui.selectorLabels" . | nindent 6 }} diff --git a/helm/litellm/tests/migration_job_hooks_tests.yaml b/helm/litellm/tests/migration_job_hooks_tests.yaml new file mode 100644 index 00000000000..650d2700429 --- /dev/null +++ b/helm/litellm/tests/migration_job_hooks_tests.yaml @@ -0,0 +1,63 @@ +suite: test migrations Job hook annotations +templates: + - migrations-job.yaml +values: + - ./values/required.yaml +tests: + - it: runs as a Helm pre-install / pre-upgrade hook by default + asserts: + - equal: + path: metadata.annotations["helm.sh/hook"] + value: pre-install,pre-upgrade + - equal: + path: metadata.annotations["helm.sh/hook-delete-policy"] + value: before-hook-creation + - equal: + path: metadata.annotations["helm.sh/hook-weight"] + value: "0" + - notExists: + path: metadata.annotations["argocd.argoproj.io/hook"] + + - it: adds the Argo CD PreSync hook when asked + set: + migrationJob.hooks.argocd.enabled: true + asserts: + - equal: + path: metadata.annotations["argocd.argoproj.io/hook"] + value: PreSync + - equal: + path: metadata.annotations["argocd.argoproj.io/hook-delete-policy"] + value: BeforeHookCreation + + - it: drops the Helm hook so Argo CD owns the Job + set: + migrationJob.hooks.argocd.enabled: true + migrationJob.hooks.helm.enabled: false + asserts: + - equal: + path: metadata.annotations["argocd.argoproj.io/hook"] + value: PreSync + - notExists: + path: metadata.annotations["helm.sh/hook"] + - notExists: + path: metadata.annotations["helm.sh/hook-delete-policy"] + - notExists: + path: metadata.annotations["helm.sh/hook-weight"] + + - it: renders an ordinary Job when both hooks are disabled + set: + migrationJob.hooks.helm.enabled: false + asserts: + - notExists: + path: metadata.annotations + - equal: + path: kind + value: Job + + - it: honours a custom Helm hook weight + set: + migrationJob.hooks.helm.weight: "-5" + asserts: + - equal: + path: metadata.annotations["helm.sh/hook-weight"] + value: "-5" diff --git a/helm/litellm/tests/rollout_strategy_tests.yaml b/helm/litellm/tests/rollout_strategy_tests.yaml new file mode 100644 index 00000000000..b12e2073c7c --- /dev/null +++ b/helm/litellm/tests/rollout_strategy_tests.yaml @@ -0,0 +1,66 @@ +suite: test rolling update strategy on the component deployments +templates: + - gateway/deployment.yaml + - gateway/configmap.yaml + - backend/deployment.yaml + - ui/deployment.yaml +values: + - ./values/required.yaml +tests: + - it: leaves the strategy to Kubernetes defaults when unset + asserts: + - notExists: + path: spec.strategy + + - it: renders the configured strategy on each deployment + set: + gateway.strategy: + type: RollingUpdate + rollingUpdate: + maxUnavailable: 0 + maxSurge: 1 + backend.strategy: + type: RollingUpdate + rollingUpdate: + maxUnavailable: "25%" + maxSurge: 2 + ui.strategy: + type: Recreate + asserts: + - equal: + path: spec.strategy + value: + type: RollingUpdate + rollingUpdate: + maxUnavailable: 0 + maxSurge: 1 + template: gateway/deployment.yaml + - equal: + path: spec.strategy + value: + type: RollingUpdate + rollingUpdate: + maxUnavailable: 25% + maxSurge: 2 + template: backend/deployment.yaml + - equal: + path: spec.strategy + value: + type: Recreate + template: ui/deployment.yaml + + - it: keeps a component on the cluster default when only another one sets a strategy + set: + gateway.strategy: + type: Recreate + asserts: + - equal: + path: spec.strategy.type + value: Recreate + template: gateway/deployment.yaml + - notExists: + path: spec.strategy + template: backend/deployment.yaml + - notExists: + path: spec.strategy + template: ui/deployment.yaml diff --git a/helm/litellm/values.yaml b/helm/litellm/values.yaml index d0c80fd6f6f..378c3b7a618 100644 --- a/helm/litellm/values.yaml +++ b/helm/litellm/values.yaml @@ -75,6 +75,22 @@ serviceAccounts: # generate` — the migration engine doesn't need the generated client. migrationJob: enabled: true + # Which controller is responsible for running the Job. + # + # `helm.enabled` renders the Helm pre-install / pre-upgrade hook, so the Job + # runs whenever `helm upgrade` sees a change to apply. `argocd.enabled` + # renders an Argo CD PreSync hook instead, which runs the Job on every sync + # even when the rendered manifests are unchanged: the way to re-run + # migrations on demand from a GitOps pipeline. Turning the Helm hook off + # while the Argo CD hook is on leaves the Job out of Helm's own upgrade + # path, which is what Argo CD users want since Argo, not Helm, applies the + # manifests. + hooks: + helm: + enabled: true + weight: "0" + argocd: + enabled: false backoffLimit: 4 ttlSecondsAfterFinished: 120 # Wall-clock budget for the whole Job, shared across every `backoffLimit` @@ -257,6 +273,15 @@ gateway: initialDelaySeconds: 5 periodSeconds: 10 timeoutSeconds: 10 + # Rolling update tuning for the gateway Deployment. Empty by default, so + # Kubernetes applies its own RollingUpdate defaults (25% maxSurge / + # 25% maxUnavailable). Example, for a surge-only rollout behind a load + # balancer that must never lose capacity: + # type: RollingUpdate + # rollingUpdate: + # maxUnavailable: 0 + # maxSurge: 1 + strategy: {} # Optional startupProbe. Empty by default, so existing installs are unchanged # and liveness/readiness apply from container start. Set it to gate # liveness/readiness until a slow cold start finishes — a high failureThreshold @@ -369,6 +394,8 @@ backend: initialDelaySeconds: 5 periodSeconds: 10 timeoutSeconds: 10 + # Same shape as gateway.strategy. + strategy: {} # Optional startupProbe; same shape as gateway.startupProbe. Empty by default. startupProbe: {} hpa: @@ -433,6 +460,8 @@ ui: httpGet: { path: /, port: http } initialDelaySeconds: 2 periodSeconds: 10 + # Same shape as gateway.strategy. + strategy: {} # Optional startupProbe; same shape as gateway.startupProbe. Empty by default. startupProbe: {} hpa: diff --git a/litellm-proxy-extras/litellm_proxy_extras/utils.py b/litellm-proxy-extras/litellm_proxy_extras/utils.py index b8032dd0d28..c088609dad7 100644 --- a/litellm-proxy-extras/litellm_proxy_extras/utils.py +++ b/litellm-proxy-extras/litellm_proxy_extras/utils.py @@ -7,7 +7,8 @@ import subprocess import tempfile import time from pathlib import Path -from typing import Optional +from types import MappingProxyType +from typing import Final, Optional from litellm_proxy_extras._logging import logger from litellm_proxy_extras.replica_identity import ( @@ -50,6 +51,17 @@ _SPEND_LOGS_PK_CLAUSE_RE = re.compile( re.IGNORECASE, ) +_PRISMA_ATTEMPTS: Final = 4 + +_TRANSIENT_PRISMA_FAILURES: Final = MappingProxyType( + { + "deadlock detected": "a deadlock on the migration advisory lock (a concurrent migrate deploy)", + "P1001": "an unreachable database server", + "P1002": "a database server that timed out", + } +) + + PARTITIONED_SPEND_LOGS_PUSH_ERROR = ( "LiteLLM_SpendLogs is a partitioned table (see db_scripts/partition_spend_logs.sql), " "so its primary key must include the partition key (\"startTime\"). `prisma db push` " @@ -274,6 +286,23 @@ class ProxyExtrasDBManager: env=prisma_env, ) + @staticmethod + def _transient_prisma_failure(stderr: str) -> str | None: + """Why a failed prisma command is worth retrying, or None. + + v1 retried every failure, so it absorbed a database that was not up yet + or another instance holding the migration lock. v2 fails fast, which is + right for a broken migration and wrong for these. + """ + return next( + ( + reason + for marker, reason in _TRANSIENT_PRISMA_FAILURES.items() + if marker in stderr + ), + None, + ) + @staticmethod def _is_permission_error(error_message: str) -> bool: """ @@ -655,7 +684,7 @@ class ProxyExtrasDBManager: @staticmethod def _setup_database_v2(use_migrate: bool) -> bool: """ - v2 migration resolver (opt-in via --use_v2_migration_resolver). + v2 migration resolver (what the proxy CLI selects by default). Runs `prisma migrate deploy` and handles standard recovery paths (P3005 baseline, P3009/P3018 idempotent errors). Critically, it does @@ -676,20 +705,46 @@ class ProxyExtrasDBManager: original_dir = os.getcwd() os.chdir(migrations_dir) try: - subprocess.run( - [_get_prisma_command(), "db", "push", "--accept-data-loss"], - timeout=prisma_command_timeout(), - check=True, - env=_get_prisma_env(), + for attempt in range(_PRISMA_ATTEMPTS): + try: + subprocess.run( + [_get_prisma_command(), "db", "push", "--accept-data-loss"], + timeout=prisma_command_timeout(), + check=True, + capture_output=True, + text=True, + env=_get_prisma_env(), + ) + return True + except subprocess.TimeoutExpired: + logger.info( + "prisma db push attempt %s timed out, retrying", + attempt + 1, + ) + time.sleep(random.randrange(5, 15)) + except subprocess.CalledProcessError as e: + stderr = e.stderr or "" + transient = ProxyExtrasDBManager._transient_prisma_failure( + stderr + ) + # Re-raise as RuntimeError so proxy_cli.py's + # `except RuntimeError` catches it and exits cleanly. + if transient is None or attempt == _PRISMA_ATTEMPTS - 1: + raise RuntimeError( + f"prisma db push failed.\n\nDetail: {e}" + f"\n\nPrisma error:\n{stderr}" + ) from e + logger.info( + "prisma db push attempt %s failed on %s, retrying. " + "Prisma error:\n%s", + attempt + 1, + transient, + stderr, + ) + time.sleep(random.randrange(5, 15)) + raise RuntimeError( + f"prisma db push failed after {_PRISMA_ATTEMPTS} attempts." ) - return True - except ( - subprocess.CalledProcessError, - subprocess.TimeoutExpired, - ) as e: - # Re-raise as RuntimeError so proxy_cli.py's - # `except RuntimeError` catches it and exits cleanly. - raise RuntimeError(f"prisma db push failed.\n\nDetail: {e}") from e finally: os.chdir(original_dir) @@ -699,7 +754,7 @@ class ProxyExtrasDBManager: original_dir = os.getcwd() os.chdir(migrations_dir) try: - for attempt in range(4): + for attempt in range(_PRISMA_ATTEMPTS): try: result = subprocess.run( [_get_prisma_command(), "migrate", "deploy"], @@ -814,16 +869,36 @@ class ProxyExtrasDBManager: f"Manual intervention required.\n\nPrisma error:\n{stderr}" ) from e - raise RuntimeError( - "Database migration failed and cannot be auto-recovered. " - f"Manual intervention required.\n\nPrisma error:\n{stderr}" - ) from e + transient = ProxyExtrasDBManager._transient_prisma_failure(stderr) + if transient is None: + raise RuntimeError( + "Database migration failed and cannot be auto-recovered. " + f"Manual intervention required.\n\nPrisma error:\n{stderr}" + ) from e + + if attempt == _PRISMA_ATTEMPTS - 1: + raise RuntimeError( + f"Database migration failed after " + f"{_PRISMA_ATTEMPTS} attempts on {transient}. " + "Check database connectivity and load." + f"\n\nPrisma error:\n{stderr}" + ) from e + + logger.info( + "prisma migrate deploy attempt %s failed on %s, retrying. " + "Prisma error:\n%s", + attempt + 1, + transient, + stderr, + ) + time.sleep(random.randrange(5, 15)) + continue raise RuntimeError( - "Database migration failed after 4 attempts (retry loop " - "exhausted by timeouts or repeated idempotent-recovery " - "continues). Check database connectivity, load, and " - "_prisma_migrations ledger state." + f"Database migration failed after {_PRISMA_ATTEMPTS} " + "attempts (retry loop exhausted by timeouts or repeated " + "idempotent-recovery continues). Check database connectivity, " + "load, and _prisma_migrations ledger state." ) finally: os.chdir(original_dir) @@ -871,10 +946,11 @@ class ProxyExtrasDBManager: Args: use_migrate: Whether to use prisma migrate instead of db push - use_v2_resolver: Opt into the v2 migration resolver (safer during + use_v2_resolver: Run the v2 migration resolver (safer during rolling deploys; does not run the diff-and-force recovery - that causes schema thrashing). Defaults to False for - backwards compatibility. + that causes schema thrashing). Defaults to False here so + direct callers keep the old behavior; the proxy CLI passes + True, so the proxy's runtime default is v2. Returns: bool: True if setup was successful, False otherwise @@ -892,7 +968,7 @@ class ProxyExtrasDBManager: @staticmethod def _run_migrations(use_migrate: bool, use_v2_resolver: bool) -> bool: if use_v2_resolver: - logger.info("Using v2 migration resolver (--use_v2_migration_resolver)") + logger.info("Using v2 migration resolver") return ProxyExtrasDBManager._setup_database_v2(use_migrate=use_migrate) schema_path = ProxyExtrasDBManager._get_prisma_dir() + "/schema.prisma" diff --git a/litellm-proxy-extras/pyproject.toml b/litellm-proxy-extras/pyproject.toml index d5741d479bf..0944f99ad54 100644 --- a/litellm-proxy-extras/pyproject.toml +++ b/litellm-proxy-extras/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "litellm-proxy-extras" -version = "0.4.91" +version = "0.4.92" description = "Additional files for the LiteLLM Proxy. Reduces the size of the main litellm package." readme = "README.md" requires-python = ">=3.9" @@ -26,7 +26,7 @@ required-version = ">=0.10.9" module-root = "" [tool.commitizen] -version = "0.4.91" +version = "0.4.92" version_files = [ "pyproject.toml:^version", "../pyproject.toml:litellm-proxy-extras==", diff --git a/litellm-proxy-extras/tests/__init__.py b/litellm-proxy-extras/tests/__init__.py deleted file mode 100644 index e69de29bb2d..00000000000 diff --git a/litellm-proxy-extras/tests/test_setup_database_fail_fast.py b/litellm-proxy-extras/tests/test_setup_database_fail_fast.py deleted file mode 100644 index 8d66bf872de..00000000000 --- a/litellm-proxy-extras/tests/test_setup_database_fail_fast.py +++ /dev/null @@ -1,242 +0,0 @@ -"""Regression tests for ProxyExtrasDBManager v2 migration resolver. - -The v2 resolver is opt-in via `--use_v2_migration_resolver` / the -`use_v2_resolver=True` kwarg. These tests exercise the v2 path; the v1 -(default) behavior is unchanged from pre-fix. -""" - -import subprocess -from unittest.mock import patch - -import pytest - -from litellm_proxy_extras.utils import ( - ProxyExtrasDBManager, - _max_migration_timestamp, - _migration_timestamp, -) - - -def _fake_migrate_deploy_failure(returncode: int, stderr: str): - def _run(*args, **kwargs): - raise subprocess.CalledProcessError( - returncode=returncode, - cmd=args[0], - stderr=stderr, - output="", - ) - - return _run - - -def test_v2_p3018_permission_error_raises_runtime_error(monkeypatch, tmp_path): - """v2: a permission failure during migrate deploy raises RuntimeError.""" - monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") - monkeypatch.setattr( - ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None - ) - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - - stderr = ( - "Error: P3018\nMigration name: 20250326162113_baseline\n" - "Database error code: 42501\npermission denied for schema public" - ) - with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): - with pytest.raises(RuntimeError, match="permission"): - ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) - - -def test_v2_non_idempotent_p3009_raises_runtime_error(monkeypatch, tmp_path): - """v2: a non-idempotent migration failure raises (no silent recovery).""" - monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") - monkeypatch.setattr( - ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None - ) - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - - stderr = ( - "Error: P3009\nMigration `20260101000000_genuinely_broken` failed\n" - 'Reason: syntax error at or near "BRKN" LINE 42' - ) - with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): - with pytest.raises(RuntimeError, match="cannot be auto-recovered"): - ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) - - -def test_strip_prisma_query_params_removes_connection_limit(): - """DATABASE_URLs with Prisma-specific params should be parseable by psycopg.""" - url = "postgresql://u:p@h:5432/db?connection_limit=100&pool_timeout=60&sslmode=require" - stripped = ProxyExtrasDBManager._strip_prisma_query_params(url) - assert "connection_limit" not in stripped - assert "pool_timeout" not in stripped - assert "sslmode=require" in stripped - - -def test_strip_prisma_query_params_passthrough_no_query(): - """URLs without query strings are returned unchanged.""" - url = "postgresql://u:p@h:5432/db" - assert ProxyExtrasDBManager._strip_prisma_query_params(url) == url - - -def test_migration_timestamp_extracts_leading_digits(): - assert _migration_timestamp("20260101000000_add_foo") == 20260101000000 - assert _migration_timestamp("20250326162113_baseline") == 20250326162113 - - -def test_migration_timestamp_returns_zero_on_malformed(): - assert _migration_timestamp("0_init") == 0 - assert _migration_timestamp("not_a_migration") == 0 - - -def test_max_migration_timestamp(): - names = {"20250326000000_a", "20260415000000_b", "20251115000000_c"} - assert _max_migration_timestamp(names) == 20260415000000 - - -def test_max_migration_timestamp_empty_set(): - assert _max_migration_timestamp(set()) == 0 - - -def test_v1_default_still_calls_resolve_all_migrations(monkeypatch, tmp_path): - """v1 (default) continues to call _resolve_all_migrations on the happy path. - - This is the existing buggy behavior — we're not fixing it in v1, only - offering v2 as opt-in. This test pins the default so that a future - inadvertent default flip is caught. - """ - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - - # Stub `prisma migrate deploy` to claim success with pending migrations - # applied, which is the code path that triggers the legacy post-migration - # sanity check (a call to _resolve_all_migrations). - class FakeResult: - stdout = "Applied migration.\n" - stderr = "" - - def fake_run(cmd, *args, **kwargs): - return FakeResult() - - resolve_called = {"n": 0} - - def fake_resolve(*args, **kwargs): - resolve_called["n"] += 1 - - monkeypatch.setattr("subprocess.run", fake_run) - monkeypatch.setattr(ProxyExtrasDBManager, "_resolve_all_migrations", fake_resolve) - - ok = ProxyExtrasDBManager.setup_database(use_migrate=True) # v2 flag NOT set - assert ok is True - assert resolve_called["n"] == 1, "v1 default should still invoke the legacy path" - - -def test_v2_db_push_wraps_subprocess_error_as_runtime_error(monkeypatch, tmp_path): - """v2: a failing `prisma db push` must raise RuntimeError, not leak - CalledProcessError past proxy_cli.py's `except RuntimeError`.""" - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - - stderr = "db push error" - with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): - with pytest.raises(RuntimeError, match="prisma db push failed"): - ProxyExtrasDBManager.setup_database(use_migrate=False, use_v2_resolver=True) - - -def test_v2_warn_ahead_of_head_swallows_db_errors(monkeypatch, tmp_path): - """_warn_if_db_ahead_of_head must never raise — it's informational. - - Non-connection DB errors (e.g. InsufficientPrivilege from a user - without SELECT on _prisma_migrations) must be caught, not propagated. - """ - import psycopg - - monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - - class _FakeConn: - def __enter__(self): - return self - - def __exit__(self, *a): - return False - - def execute(self, *a, **kw): - # Simulate an InsufficientPrivilege (subclass of DatabaseError). - raise psycopg.errors.InsufficientPrivilege("permission denied") - - def _fake_connect(*a, **kw): - return _FakeConn() - - monkeypatch.setattr("psycopg.connect", _fake_connect) - - # Must not raise. - ProxyExtrasDBManager._warn_if_db_ahead_of_head(str(tmp_path)) - - -def test_v2_resolve_specific_migration_failure_raises_runtime_error( - monkeypatch, tmp_path -): - """If marking a migration as applied fails inside P3009 idempotent - recovery, the subprocess error must be re-raised as RuntimeError so - proxy_cli.py catches it cleanly (instead of leaking CalledProcessError).""" - monkeypatch.setattr( - ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None - ) - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - monkeypatch.setattr( - ProxyExtrasDBManager, "_roll_back_migration", lambda *a, **kw: None - ) - - # First call: migrate deploy -> P3009 idempotent error. - # Recovery path tries _resolve_specific_migration; that also raises. - def _failing_resolve(*a, **kw): - raise subprocess.CalledProcessError( - returncode=1, - cmd="prisma migrate resolve --applied", - stderr="resolve failed", - output="", - ) - - monkeypatch.setattr( - ProxyExtrasDBManager, "_resolve_specific_migration", _failing_resolve - ) - - stderr = ( - "Error: P3009\nMigration `20260101000000_some_migration` failed\n" - "relation already exists" - ) - with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): - with pytest.raises( - RuntimeError, match="Failed to mark migration .* as applied" - ): - ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) - - -def test_v2_does_not_call_resolve_all_migrations(monkeypatch, tmp_path): - """v2 must never call _resolve_all_migrations — that's the bug it fixes.""" - monkeypatch.setattr( - ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None - ) - monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) - (tmp_path / "schema.prisma").write_text("// stub") - - class FakeResult: - stdout = "Applied migration.\n" - stderr = "" - - monkeypatch.setattr("subprocess.run", lambda *a, **kw: FakeResult()) - - resolve_called = {"n": 0} - monkeypatch.setattr( - ProxyExtrasDBManager, - "_resolve_all_migrations", - lambda *a, **kw: resolve_called.__setitem__("n", resolve_called["n"] + 1), - ) - - ok = ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) - assert ok is True - assert resolve_called["n"] == 0, "v2 must not invoke the diff-and-force recovery" diff --git a/litellm/__init__.py b/litellm/__init__.py index 1447e05fdf7..4eeececdb7e 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -659,6 +659,8 @@ aiml_models: Set = set() deepgram_models: Set = set() elevenlabs_models: Set = set() dashscope_models: Set = set() +qwencloud_models: Set = set() +qwen_ai_platform_models: Set = set() moonshot_models: Set = set() publicai_models: Set = set() darkbloom_models: Set = set() @@ -909,6 +911,10 @@ def _populate_provider_model_sets(model_cost_map: Dict) -> None: heroku_models.add(key) elif value.get("litellm_provider") == "dashscope": dashscope_models.add(key) + elif value.get("litellm_provider") == "qwencloud": + qwencloud_models.add(key) + elif value.get("litellm_provider") == "qwen_ai_platform": + qwen_ai_platform_models.add(key) elif value.get("litellm_provider") == "modelscope": modelscope_models.add(key) elif value.get("litellm_provider") == "moonshot": @@ -1072,6 +1078,8 @@ model_list = list( | deepgram_models | elevenlabs_models | dashscope_models + | qwencloud_models + | qwen_ai_platform_models | moonshot_models | publicai_models | darkbloom_models @@ -1178,6 +1186,8 @@ def _build_models_by_provider() -> dict: "elevenlabs": elevenlabs_models, "heroku": heroku_models, "dashscope": dashscope_models, + "qwencloud": qwencloud_models, + "qwen_ai_platform": qwen_ai_platform_models, "modelscope": modelscope_models, "moonshot": moonshot_models, "publicai": publicai_models, @@ -2014,6 +2024,24 @@ if TYPE_CHECKING: from .llms.dashscope.rerank.transformation import ( DashScopeRerankConfig as DashScopeRerankConfig, ) + from .llms.dashscope.qwencloud import ( + QwenCloudChatConfig as QwenCloudChatConfig, + ) + from .llms.dashscope.qwencloud import ( + QwenCloudEmbeddingConfig as QwenCloudEmbeddingConfig, + ) + from .llms.dashscope.qwencloud import ( + QwenCloudRerankConfig as QwenCloudRerankConfig, + ) + from .llms.dashscope.qwen_ai_platform import ( + QwenAIPlatformChatConfig as QwenAIPlatformChatConfig, + ) + from .llms.dashscope.qwen_ai_platform import ( + QwenAIPlatformEmbeddingConfig as QwenAIPlatformEmbeddingConfig, + ) + from .llms.dashscope.qwen_ai_platform import ( + QwenAIPlatformRerankConfig as QwenAIPlatformRerankConfig, + ) from .llms.modelscope.chat.transformation import ( ModelScopeChatConfig as ModelScopeChatConfig, ) diff --git a/litellm/_lazy_imports_registry.py b/litellm/_lazy_imports_registry.py index 1c833256598..e9199e1ec80 100644 --- a/litellm/_lazy_imports_registry.py +++ b/litellm/_lazy_imports_registry.py @@ -310,6 +310,8 @@ LLM_CONFIG_NAMES: Final = ( "GigaChatConfig", "GigaChatEmbeddingConfig", "DashScopeChatConfig", + "QwenCloudChatConfig", + "QwenAIPlatformChatConfig", "ModelScopeChatConfig", "MoonshotChatConfig", "DockerModelRunnerChatConfig", @@ -1172,6 +1174,14 @@ _LLM_CONFIGS_IMPORT_MAP: Final = { ".llms.dashscope.chat.transformation", "DashScopeChatConfig", ), + "QwenCloudChatConfig": ( + ".llms.dashscope.qwencloud", + "QwenCloudChatConfig", + ), + "QwenAIPlatformChatConfig": ( + ".llms.dashscope.qwen_ai_platform", + "QwenAIPlatformChatConfig", + ), "GDCGeminiConfig": ( ".llms.gdc.chat.transformation", "GDCGeminiConfig", diff --git a/litellm/caching/caching.py b/litellm/caching/caching.py index cefe6aae9ed..754815fce47 100644 --- a/litellm/caching/caching.py +++ b/litellm/caching/caching.py @@ -12,6 +12,7 @@ import hashlib import json import time import traceback +from collections.abc import Mapping from enum import Enum from typing import Any, Final @@ -506,7 +507,7 @@ class Cache: def _get_cache_logic( self, - cached_result: Any | None, + cached_result: object | None, max_age: float | None, ): """ @@ -538,8 +539,8 @@ class Cache: return cached_result @staticmethod - def _get_safe_cache_lookup_kwargs(kwargs: dict[str, Any]) -> dict[str, Any]: - cache_lookup_kwargs: Final[dict[str, Any]] = {} + def _get_safe_cache_lookup_kwargs(kwargs: Mapping[str, object]) -> dict[str, object]: + cache_lookup_kwargs: Final[dict[str, object]] = {} for prompt_kwarg in ("messages", "input"): if prompt_kwarg in kwargs: cache_lookup_kwargs[prompt_kwarg] = kwargs[prompt_kwarg] @@ -552,7 +553,7 @@ class Cache: @staticmethod def _update_metadata_from_cache_lookup_kwargs( - original_kwargs: dict[str, Any], cache_lookup_kwargs: dict[str, Any] + original_kwargs: Mapping[str, object], cache_lookup_kwargs: Mapping[str, object] ) -> None: original_metadata: Final = original_kwargs.get("metadata") cache_lookup_metadata: Final = cache_lookup_kwargs.get("metadata") diff --git a/litellm/caching/qdrant_semantic_cache.py b/litellm/caching/qdrant_semantic_cache.py index 4898700c403..c5876e993d3 100644 --- a/litellm/caching/qdrant_semantic_cache.py +++ b/litellm/caching/qdrant_semantic_cache.py @@ -12,7 +12,7 @@ import ast import asyncio import json import os -from typing import TYPE_CHECKING, Any, Final, cast +from typing import TYPE_CHECKING, Any, Final, Protocol, cast import litellm from litellm._logging import print_verbose @@ -39,6 +39,12 @@ if TYPE_CHECKING: from litellm.router import Router +class _QdrantCollectionDetailsResponse(Protocol): + """The qdrant `/collections/{name}` response, whose body is kept as an opaque JSON object.""" + + def json(self) -> dict[str, object]: ... + + class QdrantSemanticCache(BaseCache): CACHE_KEY_FIELD_NAME = "litellm_cache_key" embedding_max_input_tokens: int | None = None @@ -115,15 +121,15 @@ class QdrantSemanticCache(BaseCache): raise ValueError(f"Error from qdrant checking if /collections exist {collection_exists.text}") if collection_exists.json()["result"]["exists"]: - collection_details = self.sync_client.get( + collection_details: _QdrantCollectionDetailsResponse = self.sync_client.get( url=f"{self.qdrant_api_base}/collections/{self.collection_name}", headers=self.headers, ) - self.collection_info = collection_details.json() + self.collection_info: dict[str, object] = collection_details.json() print_verbose(f"Collection already exists.\nCollection details:{self.collection_info}") self._ensure_cache_key_payload_index() else: - quantization_params: dict[str, Any] + quantization_params: dict[str, dict[str, object]] if quantization_config is None or quantization_config == "binary": quantization_params = { "binary": { @@ -214,7 +220,7 @@ class QdrantSemanticCache(BaseCache): resolve_embedding_max_input_tokens(self.embedding_max_input_tokens, self.embedding_model, router), ) - def _get_embedding(self, prompt: str, metadata: dict[str, Any] | None = None) -> EmbeddingResponse: + def _get_embedding(self, prompt: str, metadata: dict[str, object] | None = None) -> EmbeddingResponse: """Embed via the proxy Router when it serves the model, else direct.""" try: from litellm.proxy.proxy_server import llm_model_list, llm_router @@ -241,7 +247,7 @@ class QdrantSemanticCache(BaseCache): num_retries=0, ) - async def _get_async_embedding(self, prompt: str, metadata: dict[str, Any] | None = None) -> EmbeddingResponse: + async def _get_async_embedding(self, prompt: str, metadata: dict[str, object] | None = None) -> EmbeddingResponse: try: from litellm.proxy.proxy_server import llm_model_list, llm_router except ImportError: diff --git a/litellm/completion_extras/litellm_responses_transformation/handler.py b/litellm/completion_extras/litellm_responses_transformation/handler.py index 727c39c16ec..f494d6610a1 100644 --- a/litellm/completion_extras/litellm_responses_transformation/handler.py +++ b/litellm/completion_extras/litellm_responses_transformation/handler.py @@ -45,14 +45,14 @@ class ResponsesToCompletionBridgeHandler: return bool(stream) @staticmethod - def _is_preformatted_cached_chat_stream(result: Any) -> bool: + def _is_preformatted_cached_chat_stream(result: object) -> bool: from litellm.litellm_core_utils.streaming_handler import CustomStreamWrapper return isinstance(result, CustomStreamWrapper) and result.custom_llm_provider == "cached_response" @staticmethod def _coerce_response_object( - response_obj: Any, + response_obj: object, hidden_params: dict | None, ) -> "ResponsesAPIResponse": if isinstance(response_obj, ResponsesAPIResponse): @@ -78,8 +78,8 @@ class ResponsesToCompletionBridgeHandler: for _ in stream_iter: pass - completed: Final = getattr(stream_iter, "completed_response", None) - response_obj: Final = getattr(completed, "response", None) if completed else None + completed: Final[object] = getattr(stream_iter, "completed_response", None) + response_obj: Final[object] = getattr(completed, "response", None) if completed else None if response_obj is None: raise ValueError("Stream ended without a completed response") @@ -93,8 +93,8 @@ class ResponsesToCompletionBridgeHandler: async for _ in stream_iter: pass - completed: Final = getattr(stream_iter, "completed_response", None) - response_obj: Final = getattr(completed, "response", None) if completed else None + completed: Final[object] = getattr(stream_iter, "completed_response", None) + response_obj: Final[object] = getattr(completed, "response", None) if completed else None if response_obj is None: raise ValueError("Stream ended without a completed response") @@ -157,7 +157,7 @@ class ResponsesToCompletionBridgeHandler: def completion( self, *args, **kwargs ) -> Union[ - Coroutine[Any, Any, Union["ModelResponse", "CustomStreamWrapper"]], + Coroutine[None, None, Union["ModelResponse", "CustomStreamWrapper"]], "ModelResponse", "CustomStreamWrapper", ]: diff --git a/litellm/constants.py b/litellm/constants.py index c482ab0e39a..9a50797f517 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -13,6 +13,12 @@ DEFAULT_BATCH_SIZE: Final = int(os.getenv("DEFAULT_BATCH_SIZE", 512)) DEFAULT_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_FLUSH_INTERVAL_SECONDS", 5)) DEFAULT_S3_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_S3_FLUSH_INTERVAL_SECONDS", 10)) DEFAULT_S3_BATCH_SIZE: Final = int(os.getenv("DEFAULT_S3_BATCH_SIZE", 512)) +# https://docs.aws.amazon.com/AmazonS3/latest/userguide/object-keys.html +MAX_S3_OBJECT_KEY_BYTES: Final = 1024 +S3_BOUNDED_OBJECT_KEY_HEAD_BYTES: Final = 64 +S3_PREFIX_DIGEST_CHARS: Final = 16 +# s3 allows 2048 bytes of combined metadata headers, which Content-Disposition counts against +MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES: Final = 1024 DEFAULT_SQS_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_SQS_FLUSH_INTERVAL_SECONDS", 10)) DEFAULT_NUM_WORKERS_LITELLM_PROXY: Final = int(os.getenv("DEFAULT_NUM_WORKERS_LITELLM_PROXY", 1)) DYNAMIC_RATE_LIMIT_ERROR_THRESHOLD_PER_MINUTE = int(os.getenv("DYNAMIC_RATE_LIMIT_ERROR_THRESHOLD_PER_MINUTE", 1)) @@ -630,6 +636,8 @@ LITELLM_CHAT_PROVIDERS: Final = [ "nscale", "nebius", "dashscope", + "qwencloud", + "qwen_ai_platform", "modelscope", "moonshot", "publicai", @@ -799,6 +807,7 @@ openai_compatible_endpoints: Final[list] = [ "inference.api.nscale.com/v1", "api.studio.nebius.ai/v1", "https://dashscope-intl.aliyuncs.com/compatible-mode/v1", + "https://dashscope.aliyuncs.com/compatible-mode/v1", "https://api-inference.modelscope.cn/v1", "https://api.moonshot.ai/v1", "https://api.publicai.co/v1", @@ -872,6 +881,8 @@ openai_compatible_providers: Final[list] = [ "nscale", "nebius", "dashscope", + "qwencloud", + "qwen_ai_platform", "modelscope", "moonshot", "v0", @@ -902,6 +913,8 @@ openai_text_completion_compatible_providers: Final[list] = [ # providers that s "featherless_ai", "nebius", "dashscope", + "qwencloud", + "qwen_ai_platform", "modelscope", "moonshot", "publicai", @@ -1109,7 +1122,7 @@ nebius_models: Final[set] = set( ] ) -dashscope_models: Final[set] = set( +dashscope_models: Final[frozenset] = frozenset( [ "qwen-turbo", "qwen-plus", @@ -1124,6 +1137,10 @@ dashscope_models: Final[set] = set( ] ) +qwencloud_models: Final[frozenset] = frozenset(dashscope_models) + +qwen_ai_platform_models: Final[frozenset] = frozenset(dashscope_models) + nebius_embedding_models: Final[set] = set( [ "BAAI/bge-en-icl", @@ -1240,6 +1257,7 @@ BEDROCK_CONVERSE_MODELS: Final = [ "openai.gpt-oss-120b-1:0", "anthropic.claude-haiku-4-5-20251001-v1:0", "anthropic.claude-sonnet-4-5-20250929-v1:0", + "anthropic.claude-fable-5-1", "anthropic.claude-fable-5", "anthropic.claude-sonnet-5", "anthropic.claude-opus-5", diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py index 78f34abf766..b83e9b395a8 100644 --- a/litellm/cost_calculator.py +++ b/litellm/cost_calculator.py @@ -641,12 +641,12 @@ def cost_per_token( return xai_cost_per_token(model=model, usage=usage_block) elif custom_llm_provider == "lemonade": return lemonade_cost_per_token(model=model, usage=usage_block) - elif custom_llm_provider == "dashscope": + elif custom_llm_provider in ("dashscope", "qwencloud", "qwen_ai_platform"): from litellm.llms.dashscope.cost_calculator import ( cost_per_token as dashscope_cost_per_token, ) - return dashscope_cost_per_token(model=model, usage=usage_block) + return dashscope_cost_per_token(model=model, usage=usage_block, custom_llm_provider=custom_llm_provider) elif custom_llm_provider == "azure_ai": return azure_ai_cost_per_token( model=model, @@ -1910,12 +1910,15 @@ def ocr_cost( if credits is not None and cost_per_credit is not None: return cost_per_credit * credits, 0.0 - ocr_cost_per_page: float | None = None - if model_info is not None: - ocr_cost_per_page = model_info.get("ocr_cost_per_page") + ocr_cost_per_page: Final = model_info.get("ocr_cost_per_page") if model_info is not None else None + annotation_cost_per_page: Final = model_info.get("annotation_cost_per_page") if model_info is not None else None + annotation_rate: Final = annotation_cost_per_page if annotation_cost_per_page is not None else ocr_cost_per_page pages_processed: Final = response.usage_info.pages_processed - if pages_processed is None: + annotation_pages: Final = response.usage_info.pages_processed_annotation or 0 + has_billable_annotation_pages: Final = annotation_rate is not None and annotation_pages > 0 + + if pages_processed is None and not has_billable_annotation_pages: if cost_per_credit is not None or ocr_cost_per_page is None: # Surface missing usage data instead of silently under-reporting # cost. The previous behavior raised ValueError; we now return 0.0 @@ -1931,7 +1934,7 @@ def ocr_cost( return 0.0, 0.0 raise ValueError("OCR response pages_processed is None") - if ocr_cost_per_page is None: + if ocr_cost_per_page is None and not has_billable_annotation_pages: # No per-page pricing configured. Either the model is on credit-based # pricing (and credits weren't returned, so the credit branch above did # not match) or the model has no OCR pricing entry at all. Surface a @@ -1947,8 +1950,9 @@ def ocr_cost( ) return 0.0, 0.0 - total_ocr_processing_cost: Final[float] = ocr_cost_per_page * pages_processed - return total_ocr_processing_cost, 0.0 + ocr_pages_cost: Final = (ocr_cost_per_page or 0.0) * (pages_processed or 0) + annotation_pages_cost: Final = (annotation_rate or 0.0) * annotation_pages + return ocr_pages_cost + annotation_pages_cost, 0.0 def vector_store_search_cost( diff --git a/litellm/google_genai/main.py b/litellm/google_genai/main.py index b5815bd3f7c..c1822e4720d 100644 --- a/litellm/google_genai/main.py +++ b/litellm/google_genai/main.py @@ -52,10 +52,10 @@ class GenerateContentSetupResult(BaseModel): model_config: ClassVar[ConfigDict] = ConfigDict(arbitrary_types_allowed=True) model: str - request_body: dict[str, Any] + request_body: dict[str, object] custom_llm_provider: str generate_content_provider_config: BaseGoogleGenAIGenerateContentConfig | None - generate_content_config_dict: dict[str, Any] + generate_content_config_dict: dict[str, object] native_request_fields: dict[str, object] litellm_params: GenericLiteLLMParams litellm_logging_obj: LiteLLMLoggingObj @@ -68,7 +68,7 @@ class GenerateContentHelper: @staticmethod def mock_generate_content_response( mock_response: str = "This is a mock response from Google GenAI generate_content.", - ) -> dict[str, Any]: + ) -> dict[str, object]: """Mock response for generate_content for testing purposes""" return { "text": mock_response, @@ -239,9 +239,9 @@ async def agenerate_content( tools: ToolConfigDict | None = None, # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs. # The extra values given here take precedence over values defined on the client or passed to this method. - extra_headers: dict[str, Any] | None = None, - extra_query: dict[str, Any] | None = None, - extra_body: dict[str, Any] | None = None, + extra_headers: dict[str, object] | None = None, + extra_query: dict[str, object] | None = None, + extra_body: dict[str, object] | None = None, timeout: float | httpx.Timeout | None = None, # LiteLLM specific params, custom_llm_provider: str | None = None, @@ -307,9 +307,9 @@ def generate_content( tools: ToolConfigDict | None = None, # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs. # The extra values given here take precedence over values defined on the client or passed to this method. - extra_headers: dict[str, Any] | None = None, - extra_query: dict[str, Any] | None = None, - extra_body: dict[str, Any] | None = None, + extra_headers: dict[str, object] | None = None, + extra_query: dict[str, object] | None = None, + extra_body: dict[str, object] | None = None, timeout: float | httpx.Timeout | None = None, # LiteLLM specific params, custom_llm_provider: str | None = None, @@ -397,9 +397,9 @@ async def agenerate_content_stream( tools: ToolConfigDict | None = None, # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs. # The extra values given here take precedence over values defined on the client or passed to this method. - extra_headers: dict[str, Any] | None = None, - extra_query: dict[str, Any] | None = None, - extra_body: dict[str, Any] | None = None, + extra_headers: dict[str, object] | None = None, + extra_query: dict[str, object] | None = None, + extra_body: dict[str, object] | None = None, timeout: float | httpx.Timeout | None = None, # LiteLLM specific params, custom_llm_provider: str | None = None, @@ -492,9 +492,9 @@ def generate_content_stream( tools: ToolConfigDict | None = None, # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs. # The extra values given here take precedence over values defined on the client or passed to this method. - extra_headers: dict[str, Any] | None = None, - extra_query: dict[str, Any] | None = None, - extra_body: dict[str, Any] | None = None, + extra_headers: dict[str, object] | None = None, + extra_query: dict[str, object] | None = None, + extra_body: dict[str, object] | None = None, timeout: float | httpx.Timeout | None = None, # LiteLLM specific params, custom_llm_provider: str | None = None, diff --git a/litellm/images/main.py b/litellm/images/main.py index 1688087c2da..6a94e7c8df2 100644 --- a/litellm/images/main.py +++ b/litellm/images/main.py @@ -3,7 +3,7 @@ import contextvars import importlib from collections.abc import Coroutine from functools import partial -from typing import TYPE_CHECKING, Any, Final, Literal, Optional, cast, overload +from typing import TYPE_CHECKING, Final, Literal, Optional, cast, overload if TYPE_CHECKING: from litellm.images.utils import ImageEditRequestUtils @@ -151,7 +151,7 @@ def image_generation( *, aimg_generation: Literal[True], **kwargs, -) -> Coroutine[Any, Any, ImageResponse]: +) -> Coroutine[object, object, ImageResponse]: ... @@ -197,7 +197,7 @@ def image_generation( api_version: str | None = None, custom_llm_provider=None, **kwargs, -) -> ImageResponse | Coroutine[Any, Any, ImageResponse]: +) -> ImageResponse | Coroutine[object, object, ImageResponse]: """ Maps the https://api.openai.com/v1/images/generations endpoint. @@ -386,6 +386,8 @@ def image_generation( litellm.LlmProviders.VERTEX_AI, litellm.LlmProviders.OPENROUTER, litellm.LlmProviders.DASHSCOPE, + litellm.LlmProviders.QWENCLOUD, + litellm.LlmProviders.QWEN_AI_PLATFORM, ): if image_generation_config is None: raise ValueError(f"image generation config is not supported for {custom_llm_provider}") @@ -723,14 +725,14 @@ def image_edit( user: str | None = None, # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs. # The extra values given here take precedence over values defined on the client or passed to this method. - extra_headers: dict[str, Any] | None = None, - extra_query: dict[str, Any] | None = None, - extra_body: dict[str, Any] | None = None, + extra_headers: dict[str, object] | None = None, + extra_query: dict[str, object] | None = None, + extra_body: dict[str, object] | None = None, timeout: float | httpx.Timeout | None = None, # LiteLLM specific params, custom_llm_provider: str | None = None, **kwargs, -) -> ImageResponse | Coroutine[Any, Any, ImageResponse]: +) -> ImageResponse | Coroutine[object, object, ImageResponse]: """ Maps the image edit functionality, similar to OpenAI's images/edits endpoint. """ @@ -769,7 +771,7 @@ def image_edit( images: Final = image if isinstance(image, list) else ([image] if image is not None else []) headers_from_kwargs: Final = kwargs.get("headers") - merged_extra_headers: Final[dict[str, Any]] = {} + merged_extra_headers: Final[dict[str, object]] = {} if isinstance(headers_from_kwargs, dict): merged_extra_headers.update(headers_from_kwargs) if isinstance(extra_headers, dict): @@ -974,9 +976,9 @@ async def aimage_edit( user: str | None = None, # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs. # The extra values given here take precedence over values defined on the client or passed to this method. - extra_headers: dict[str, Any] | None = None, - extra_query: dict[str, Any] | None = None, - extra_body: dict[str, Any] | None = None, + extra_headers: dict[str, object] | None = None, + extra_query: dict[str, object] | None = None, + extra_body: dict[str, object] | None = None, timeout: float | httpx.Timeout | None = None, # LiteLLM specific params, custom_llm_provider: str | None = None, @@ -1044,7 +1046,7 @@ async def aimage_edit( ) -def __getattr__(name: str) -> Any: +def __getattr__(name: str) -> type["ImageEditRequestUtils"]: """Lazy import handler for images.main module""" if name == "ImageEditRequestUtils": # Lazy load ImageEditRequestUtils to avoid heavy import from images.utils at module load time diff --git a/litellm/integrations/SlackAlerting/slack_alerting.py b/litellm/integrations/SlackAlerting/slack_alerting.py index 94d734546be..c137164ecdb 100644 --- a/litellm/integrations/SlackAlerting/slack_alerting.py +++ b/litellm/integrations/SlackAlerting/slack_alerting.py @@ -545,7 +545,6 @@ class SlackAlerting(CustomBatchLogger): # Get the appropriate budget alert type handler budget_alert_class: Final = get_budget_alert_type(type) _id: Final = budget_alert_class.get_id(user_info) - user_info_json: Final = user_info.model_dump(exclude_none=True) user_info_str: Final = self._get_user_info_str(user_info) event_message = budget_alert_class.get_event_message() @@ -575,7 +574,22 @@ class SlackAlerting(CustomBatchLogger): webhook_event = WebhookEvent( event=event, event_message=event_message, - **user_info_json, + spend=user_info.spend, + max_budget=user_info.max_budget, + soft_budget=user_info.soft_budget, + token=user_info.token, + customer_id=user_info.customer_id, + user_id=user_info.user_id, + team_id=user_info.team_id, + team_alias=user_info.team_alias, + organization_id=user_info.organization_id, + user_email=user_info.user_email, + key_alias=user_info.key_alias, + projected_exceeded_date=user_info.projected_exceeded_date, + projected_spend=user_info.projected_spend, + event_group=user_info.event_group, + alert_emails=user_info.alert_emails, + max_budget_alert_emails=user_info.max_budget_alert_emails, ) await self.send_alert( message=event_message + "\n\n" + user_info_str, @@ -657,7 +671,7 @@ class SlackAlerting(CustomBatchLogger): """ Create a standard message for a budget alert """ - _all_fields_as_dict: Final = user_info.model_dump(exclude_none=True) + _all_fields_as_dict: Final[dict[str, object]] = user_info.model_dump(exclude_none=True) _all_fields_as_dict.pop("token") msg = "" for k, v in _all_fields_as_dict.items(): @@ -1006,7 +1020,7 @@ class SlackAlerting(CustomBatchLogger): except Exception: pass - async def model_added_alert(self, model_name: str, litellm_model_name: str, passed_model_info: Any): + async def model_added_alert(self, model_name: str, litellm_model_name: str, passed_model_info: object): base_model_from_user: Final = getattr(passed_model_info, "base_model", None) model_info = {} base_model = "" @@ -1973,7 +1987,7 @@ Model Info: try: message = f"`{event_name}`\n" - key_event_dict: Final = key_event.model_dump() + key_event_dict: Final[dict[str, object]] = key_event.model_dump() # Add Created by information first message += "*Action Done by:*\n" diff --git a/litellm/integrations/bitbucket/bitbucket_prompt_manager.py b/litellm/integrations/bitbucket/bitbucket_prompt_manager.py index 6a03e3ee93c..ff34bd91e31 100644 --- a/litellm/integrations/bitbucket/bitbucket_prompt_manager.py +++ b/litellm/integrations/bitbucket/bitbucket_prompt_manager.py @@ -3,6 +3,7 @@ BitBucket prompt manager that integrates with LiteLLM's prompt management system Fetches .prompt files from BitBucket repositories and provides team-based access control. """ +from collections.abc import Mapping from typing import TYPE_CHECKING, Any, Final from jinja2 import DictLoader, select_autoescape @@ -65,7 +66,7 @@ class BitBucketTemplateManager: def __init__( self, - bitbucket_config: dict[str, Any], + bitbucket_config: Mapping[str, object], prompt_id: str | None = None, ): self.bitbucket_config = bitbucket_config @@ -123,7 +124,7 @@ class BitBucketTemplateManager: template_content = content # Parse YAML frontmatter - metadata: dict[str, Any] = {} + metadata: dict[str, object] = {} if frontmatter_str: try: import yaml @@ -141,9 +142,9 @@ class BitBucketTemplateManager: metadata=metadata, ) - def _parse_yaml_basic(self, yaml_str: str) -> dict[str, Any]: + def _parse_yaml_basic(self, yaml_str: str) -> dict[str, object]: """Basic YAML parser for simple cases when PyYAML is not available.""" - result: Final[dict[str, Any]] = {} + result: Final[dict[str, object]] = {} for line in yaml_str.split("\n"): line = line.strip() if ":" in line and not line.startswith("#"): @@ -162,7 +163,7 @@ class BitBucketTemplateManager: result[key] = value.strip("\"'") return result - def render_template(self, template_id: str, variables: dict[str, Any] | None = None) -> str: + def render_template(self, template_id: str, variables: Mapping[str, object] | None = None) -> str: """Render a template with the given variables.""" if template_id not in self.prompts: raise ValueError(f"Template '{template_id}' not found") @@ -209,7 +210,7 @@ class BitBucketPromptManager(CustomPromptManagement): def __init__( self, - bitbucket_config: dict[str, Any], + bitbucket_config: Mapping[str, object], prompt_id: str | None = None, ): self.bitbucket_config = bitbucket_config @@ -234,7 +235,7 @@ class BitBucketPromptManager(CustomPromptManagement): def get_prompt_template( self, prompt_id: str, - prompt_variables: dict[str, Any] | None = None, + prompt_variables: Mapping[str, object] | None = None, ) -> tuple[str, dict[str, Any]]: """ Get a prompt template and render it with variables. @@ -267,12 +268,12 @@ class BitBucketPromptManager(CustomPromptManagement): self, user_id: str | None, messages: list[AllMessageValues], - function_call: dict[str, Any] | str | None = None, - litellm_params: dict[str, Any] | None = None, + function_call: Mapping[str, object] | str | None = None, + litellm_params: dict[str, object] | None = None, prompt_id: str | None = None, - prompt_variables: dict[str, Any] | None = None, + prompt_variables: Mapping[str, object] | None = None, **kwargs, - ) -> tuple[list[AllMessageValues], dict[str, Any] | None]: + ) -> tuple[list[AllMessageValues], dict[str, object] | None]: """ Pre-call hook that processes the prompt template before making the LLM call. """ @@ -316,9 +317,9 @@ class BitBucketPromptManager(CustomPromptManagement): except Exception as e: # Log error but don't fail the call - import litellm + from litellm._logging import verbose_proxy_logger - litellm._logging.verbose_proxy_logger.error("Error in BitBucket prompt pre_call_hook: %s", e) + verbose_proxy_logger.error("Error in BitBucket prompt pre_call_hook: %s", e) return messages, litellm_params def _parse_prompt_to_messages(self, prompt_content: str) -> list[AllMessageValues]: @@ -384,14 +385,14 @@ class BitBucketPromptManager(CustomPromptManagement): def post_call_hook( self, user_id: str | None, - response: Any, + response: object, input_messages: list[AllMessageValues], - function_call: dict[str, Any] | str | None = None, - litellm_params: dict[str, Any] | None = None, + function_call: Mapping[str, object] | str | None = None, + litellm_params: Mapping[str, object] | None = None, prompt_id: str | None = None, - prompt_variables: dict[str, Any] | None = None, + prompt_variables: Mapping[str, object] | None = None, **kwargs, - ) -> Any: + ) -> object: """ Post-call hook for any post-processing after the LLM call. """ diff --git a/litellm/integrations/cloudzero/transform.py b/litellm/integrations/cloudzero/transform.py index f0d4d67fc22..ffc8fe1c1f5 100644 --- a/litellm/integrations/cloudzero/transform.py +++ b/litellm/integrations/cloudzero/transform.py @@ -19,14 +19,29 @@ """Transform LiteLLM data to CloudZero AnyCost CBF format.""" from datetime import datetime -from typing import Any, Final +from typing import Final, SupportsFloat, SupportsIndex, SupportsInt import polars as pl +from typing_extensions import Buffer from ...types.integrations.cloudzero import CBFRecord from .cz_resource_names import CZEntityType, CZRNGenerator +def _as_int(value: object) -> int: + """The integer form of a spend table cell, computed the way :func:`int` computes it.""" + if isinstance(value, (str, Buffer, SupportsInt, SupportsIndex)): + return int(value) + raise TypeError(f"int() argument must be a string or a number, not {type(value).__name__!r}") + + +def _as_float(value: object) -> float: + """The floating point form of a spend table cell, computed the way :func:`float` computes it.""" + if isinstance(value, (str, Buffer, SupportsFloat, SupportsIndex)): + return float(value) + raise TypeError(f"float() argument must be a string or a number, not {type(value).__name__!r}") + + class CBFTransformer: """Transform LiteLLM usage data to CloudZero Billing Format (CBF).""" @@ -82,15 +97,15 @@ class CBFTransformer: return pl.DataFrame(cbf_data) - def _create_cbf_record(self, row: dict[str, Any]) -> CBFRecord: + def _create_cbf_record(self, row: dict[str, object]) -> CBFRecord: """Create a single CBF record from LiteLLM daily spend row.""" # Parse date (daily spend tables use date strings like '2025-04-19') usage_date: Final = self._parse_date(row.get("date")) # Calculate total tokens - prompt_tokens: Final = int(row.get("prompt_tokens", 0)) - completion_tokens: Final = int(row.get("completion_tokens", 0)) + prompt_tokens: Final = _as_int(row.get("prompt_tokens", 0)) + completion_tokens: Final = _as_int(row.get("completion_tokens", 0)) total_tokens: Final = prompt_tokens + completion_tokens # Create CloudZero Resource Name (CZRN) as resource_id @@ -154,7 +169,7 @@ class CBFTransformer: "time/usage_start": ( usage_date.isoformat() if usage_date else None ), # Required: ISO-formatted UTC datetime - "cost/cost": float(row.get("spend", 0.0)), # Required: billed cost + "cost/cost": _as_float(row.get("spend", 0.0)), # Required: billed cost "resource/id": resource_id, # CZRN (CloudZero Resource Name) # Usage metrics for token consumption "usage/amount": total_tokens, # Numeric value of tokens consumed @@ -187,7 +202,7 @@ class CBFTransformer: return CBFRecord(cbf_record) - def _parse_date(self, date_str) -> datetime | None: + def _parse_date(self, date_str: object) -> datetime | None: """Parse date string from daily spend tables (e.g., '2025-04-19').""" if date_str is None: return None diff --git a/litellm/integrations/custom_guardrail.py b/litellm/integrations/custom_guardrail.py index 8dc6881d23e..e87ac9521ae 100644 --- a/litellm/integrations/custom_guardrail.py +++ b/litellm/integrations/custom_guardrail.py @@ -2,6 +2,7 @@ import contextvars import hashlib import os import secrets +from collections.abc import Mapping from datetime import datetime from typing import TYPE_CHECKING, Any, ClassVar, Final, Literal, Optional, get_args @@ -227,13 +228,13 @@ class CustomGuardrail(CustomLogger): ) super().__init__(**kwargs) - def render_violation_message(self, default: str, context: dict[str, Any] | None = None) -> str: + def render_violation_message(self, default: str, context: Mapping[str, object] | None = None) -> str: """Return a custom violation message if template is configured.""" if not self.violation_message_template: return default - format_context: Final[dict[str, Any]] = {"default_message": default} + format_context: Final[dict[str, object]] = {"default_message": default} if context: format_context.update(context) try: @@ -661,7 +662,7 @@ class CustomGuardrail(CustomLogger): value: Final = self._get_admin_metadata(data).get("opted_out_global_guardrails") return value if isinstance(value, list) else [] - def _is_valid_response_type(self, result: Any) -> bool: + def _is_valid_response_type(self, result: object) -> bool: """ Check if result is a valid LLMResponseTypes instance. @@ -722,7 +723,7 @@ class CustomGuardrail(CustomLogger): return None return f"{_PRE_CALL_EXECUTED_TOKEN}:{name}" - def mark_pre_call_hook_ran(self, data: dict[str, Any]) -> None: + def mark_pre_call_hook_ran(self, data: dict[str, object]) -> None: """ Record that this guardrail's ``async_pre_call_hook`` already ran for this request, so the deployment-level hook does not run it a second time. @@ -747,7 +748,7 @@ class CustomGuardrail(CustomLogger): return data["metadata"] = {PRE_CALL_EXECUTED_GUARDRAILS_KEY: [marker]} - def _pre_call_hook_already_ran(self, data: dict[str, Any]) -> bool: + def _pre_call_hook_already_ran(self, data: dict[str, object]) -> bool: marker: Final = self._pre_call_marker() if marker is None: return False @@ -1170,7 +1171,7 @@ class CustomGuardrail(CustomLogger): This gets logged on downsteam Langfuse, DataDog, etc. """ # Convert None to empty dict to satisfy type requirements - guardrail_response: dict[str, Any] | str = {} if response is None else response + guardrail_response: dict[str, object] | str = {} if response is None else response # For apply_guardrail functions in custom_code_guardrail scenario, # simplify the logged response to "allow", "deny", or "mask" diff --git a/litellm/integrations/datadog/datadog.py b/litellm/integrations/datadog/datadog.py index 04f1c6dff15..866076a3c49 100644 --- a/litellm/integrations/datadog/datadog.py +++ b/litellm/integrations/datadog/datadog.py @@ -20,10 +20,11 @@ import time import traceback from collections.abc import Sequence from datetime import datetime as datetimeObj -from typing import Any, Final +from typing import TYPE_CHECKING, Any, Final import httpx from httpx import Response +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -62,6 +63,18 @@ from litellm.types.utils import StandardLoggingPayload from ..additional_logging_utils import AdditionalLoggingUtils +if TYPE_CHECKING: + from fastapi import HTTPException + + from litellm.proxy._types import UserAPIKeyAuth + + +class _DatadogLoggingKwargs(TypedDict, total=False): + """The subset of logging ``kwargs`` that the Datadog payload builder reads.""" + + standard_logging_object: ReadOnly[StandardLoggingPayload | None] + + # max number of logs DD API can accept @@ -87,6 +100,11 @@ def _resolve_dd_batch_size() -> int: return max(1, min(value, DD_MAX_BATCH_SIZE)) +def _span_attribute(span: object, name: str) -> object: + """Read an optional attribute off whatever span object the active tracer hands back.""" + return getattr(span, name, None) + + class DataDogLogger( CustomBatchLogger, AdditionalLoggingUtils, @@ -271,9 +289,9 @@ class DataDogLogger( self, request_data: dict, original_exception: Exception, - user_api_key_dict: Any, + user_api_key_dict: "UserAPIKeyAuth", traceback_str: str | None = None, - ) -> Any | None: + ) -> "HTTPException | None": """ Log proxy-level failures (e.g. 401 auth, DB connection errors) to Datadog. @@ -297,7 +315,7 @@ class DataDogLogger( status_code = int(_code) # Use project-standard sanitized user context when running in proxy - user_context: dict[str, Any] = {} + user_context: dict[str, object] = {} try: from litellm.proxy.litellm_pre_call_utils import ( LiteLLMProxyRequestSetup, @@ -553,8 +571,8 @@ class DataDogLogger( def create_datadog_logging_payload( self, - kwargs: dict | Any, - response_obj: Any, + kwargs: _DatadogLoggingKwargs, + response_obj: object, start_time: datetime.datetime, end_time: datetime.datetime, ) -> DatadogPayload: @@ -562,8 +580,8 @@ class DataDogLogger( Helper function to create a datadog payload for logging Args: - kwargs (Union[dict, Any]): request kwargs - response_obj (Any): llm api response + kwargs: request kwargs, read for its standard logging object + response_obj: llm api response start_time (datetime.datetime): start time of request end_time (datetime.datetime): end time of request @@ -625,7 +643,7 @@ class DataDogLogger( self, payload: ServiceLoggerPayload, error: str | None = "", - parent_otel_span: Any | None = None, + parent_otel_span: object = None, start_time: datetimeObj | float | None = None, end_time: float | datetimeObj | None = None, event_metadata: dict | None = None, @@ -659,7 +677,7 @@ class DataDogLogger( self, payload: ServiceLoggerPayload, error: str | None = "", - parent_otel_span: Any | None = None, + parent_otel_span: object = None, start_time: datetimeObj | float | None = None, end_time: float | datetimeObj | None = None, event_metadata: dict | None = None, @@ -696,7 +714,7 @@ class DataDogLogger( def _create_v0_logging_payload( self, - kwargs: dict | Any, + kwargs: dict, response_obj: Any, start_time: datetime.datetime, end_time: datetime.datetime, @@ -810,11 +828,11 @@ class DataDogLogger( if current_span is None: return None - trace_id: Final = getattr(current_span, "trace_id", None) + trace_id: Final = _span_attribute(current_span, "trace_id") if trace_id is None: return None - span_id: Final = getattr(current_span, "span_id", None) + span_id: Final = _span_attribute(current_span, "span_id") trace_context: Final[dict[str, str]] = {"trace_id": str(trace_id)} if span_id is not None: trace_context["span_id"] = str(span_id) diff --git a/litellm/integrations/datadog/datadog_llm_obs.py b/litellm/integrations/datadog/datadog_llm_obs.py index 704f0323e95..e5789965c6e 100644 --- a/litellm/integrations/datadog/datadog_llm_obs.py +++ b/litellm/integrations/datadog/datadog_llm_obs.py @@ -9,6 +9,7 @@ API Reference: https://docs.datadoghq.com/llm_observability/setup/api/?tab=examp import asyncio import json import os +from collections.abc import Mapping, Sequence from datetime import datetime from typing import Any, Final, Literal @@ -334,7 +335,7 @@ class DataDogLLMObsLogger(CustomBatchLogger): def _get_response_messages( self, standard_logging_payload: StandardLoggingPayload, call_type: str | None - ) -> list[Any]: + ) -> list[object]: """ Get the messages from the response object @@ -484,7 +485,7 @@ class DataDogLLMObsLogger(CustomBatchLogger): # Default fallback for unknown or passthrough operations return "llm" - def _ensure_string_content(self, messages: str | list[Any] | dict[Any, Any] | None) -> list[Any]: + def _ensure_string_content(self, messages: str | Sequence[object] | Mapping[object, object] | None) -> list[object]: if messages is None: return [] if isinstance(messages, str): @@ -495,11 +496,11 @@ class DataDogLLMObsLogger(CustomBatchLogger): return [str(messages.get("content", ""))] return [] - def _get_dd_llm_obs_payload_metadata(self, standard_logging_payload: StandardLoggingPayload) -> dict[str, Any]: + def _get_dd_llm_obs_payload_metadata(self, standard_logging_payload: StandardLoggingPayload) -> dict[str, object]: """ Fields to track in DD LLM Observability metadata from litellm standard logging payload """ - _metadata: Final[dict[str, Any]] = { + _metadata: Final[dict[str, object]] = { "model_name": standard_logging_payload.get("model", "unknown"), "model_provider": standard_logging_payload.get("custom_llm_provider", "unknown"), "id": standard_logging_payload.get("id", "unknown"), @@ -647,7 +648,7 @@ class DataDogLLMObsLogger(CustomBatchLogger): return spend_metrics - def _process_input_messages_preserving_tool_calls(self, messages: list[Any]) -> list[dict[str, Any]]: + def _process_input_messages_preserving_tool_calls(self, messages: Sequence[object]) -> list[dict[str, object]]: """ Process input messages while preserving tool_calls and tool message types. @@ -671,13 +672,13 @@ class DataDogLLMObsLogger(CustomBatchLogger): return processed @staticmethod - def _tool_calls_kv_pair(tool_calls: list[dict[str, Any]]) -> dict[str, Any]: + def _tool_calls_kv_pair(tool_calls: list[dict[str, Any]]) -> dict[str, object]: """ Extract tool call information into key-value pairs for Datadog metadata. Similar to OpenTelemetry's implementation but adapted for Datadog's format. """ - kv_pairs: Final[dict[str, Any]] = {} + kv_pairs: Final[dict[str, object]] = {} for idx, tool_call in enumerate(tool_calls): try: # Extract tool call ID @@ -712,11 +713,11 @@ class DataDogLLMObsLogger(CustomBatchLogger): return kv_pairs - def _extract_tool_call_metadata(self, standard_logging_payload: StandardLoggingPayload) -> dict[str, Any]: + def _extract_tool_call_metadata(self, standard_logging_payload: StandardLoggingPayload) -> dict[str, object]: """ Extract tool call information from both input messages and response for Datadog metadata. """ - tool_call_metadata: Final[dict[str, Any]] = {} + tool_call_metadata: Final[dict[str, object]] = {} try: # Extract tool calls from input messages diff --git a/litellm/integrations/dotprompt/prompt_manager.py b/litellm/integrations/dotprompt/prompt_manager.py index fd0b17ba746..9c82ff7c5ba 100644 --- a/litellm/integrations/dotprompt/prompt_manager.py +++ b/litellm/integrations/dotprompt/prompt_manager.py @@ -3,12 +3,21 @@ Based on Google's GenAI Kit dotprompt implementation: https://google.github.io/d """ import re +from collections.abc import Mapping from pathlib import Path from typing import Any, Final import yaml from jinja2 import DictLoader, select_autoescape from jinja2.sandbox import ImmutableSandboxedEnvironment +from typing_extensions import NotRequired, ReadOnly, TypedDict + + +class _PromptFileJson(TypedDict): + """JSON form of a .prompt file: rendered template text plus its frontmatter.""" + + content: ReadOnly[NotRequired[str]] + metadata: ReadOnly[NotRequired[dict[str, object]]] def strip_version_suffix(prompt_id: str) -> str | None: @@ -167,7 +176,7 @@ class PromptManager: template_id=prompt_id, ) - def _parse_frontmatter(self, content: str) -> tuple[dict[str, Any], str]: + def _parse_frontmatter(self, content: str) -> tuple[dict[str, object], str]: """Parse YAML frontmatter from prompt content.""" # Match YAML frontmatter between --- delimiters frontmatter_pattern: Final = r"^---\s*\n(.*?)\n---\s*\n(.*)$" @@ -178,7 +187,7 @@ class PromptManager: template_content = match.group(2) try: - frontmatter = yaml.safe_load(frontmatter_yaml) or {} + frontmatter: dict[str, object] = yaml.safe_load(frontmatter_yaml) or {} except yaml.YAMLError as e: raise ValueError(f"Invalid YAML frontmatter: {e}") else: @@ -191,7 +200,7 @@ class PromptManager: def render( self, prompt_id: str, - prompt_variables: dict[str, Any] | None = None, + prompt_variables: Mapping[str, object] | None = None, version: int | None = None, ) -> str: """ @@ -231,7 +240,7 @@ class PromptManager: except Exception as e: raise ValueError(f"Error rendering template '{prompt_id}': {e}") - def _validate_input(self, variables: dict[str, Any], schema: dict[str, Any]) -> None: + def _validate_input(self, variables: Mapping[str, object], schema: Mapping[str, str]) -> None: """Basic validation of input variables against schema.""" for field_name, field_type in schema.items(): if field_name in variables: @@ -291,7 +300,7 @@ class PromptManager: """Get a list of all available prompt IDs.""" return list(self.prompts.keys()) - def get_prompt_metadata(self, prompt_id: str) -> dict[str, Any] | None: + def get_prompt_metadata(self, prompt_id: str) -> dict[str, object] | None: """Get metadata for a specific prompt.""" template: Final = self.prompts.get(prompt_id) return template.metadata if template else None @@ -302,12 +311,12 @@ class PromptManager: if self.prompt_directory: self._load_prompts() - def add_prompt(self, prompt_id: str, content: str, metadata: dict[str, Any] | None = None) -> None: + def add_prompt(self, prompt_id: str, content: str, metadata: dict[str, object] | None = None) -> None: """Add a prompt template programmatically.""" template: Final = PromptTemplate(content=content, metadata=metadata or {}, template_id=prompt_id) self.prompts[prompt_id] = template - def prompt_file_to_json(self, file_path: str | Path) -> dict[str, Any]: + def prompt_file_to_json(self, file_path: str | Path) -> _PromptFileJson: """Convert a .prompt file to JSON format. Args: @@ -324,7 +333,7 @@ class PromptManager: return {"content": template_content.strip(), "metadata": frontmatter} - def json_to_prompt_file(self, prompt_data: dict[str, Any]) -> str: + def json_to_prompt_file(self, prompt_data: _PromptFileJson) -> str: """Convert JSON prompt data to .prompt file format. Args: diff --git a/litellm/integrations/galileo.py b/litellm/integrations/galileo.py index 23727801a6f..b27618993a3 100644 --- a/litellm/integrations/galileo.py +++ b/litellm/integrations/galileo.py @@ -6,10 +6,11 @@ import re import uuid from collections.abc import Mapping, Sequence from datetime import datetime, timezone, tzinfo -from typing import Any, Final, TypedDict, cast +from typing import Any, Final, Protocol, cast import httpx from pydantic import BaseModel, Field +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -35,6 +36,34 @@ GALILEO_CLOUD_API_BASE_URL: Final = "https://api.galileo.ai" GALILEO_MAX_IN_MEMORY_RECORDS: Final = 1000 +class _GalileoLoginBody(TypedDict): + """Decoded body of the Galileo login response.""" + + access_token: ReadOnly[str] + + +class _GalileoLoginResponse(Protocol): + """The login call's HTTP response, read for the access token it carries.""" + + def json(self) -> _GalileoLoginBody: ... + + +class _JsonResponse(Protocol): + """An HTTP response read only for whatever JSON body it decodes to.""" + + def json(self) -> object: ... + + +def _login_access_token(response: _GalileoLoginResponse) -> str: + """Read the bearer token out of a Galileo login response body.""" + return response.json()["access_token"] + + +def _decoded_body(response: _JsonResponse) -> object: + """Decode a response body without asserting anything about its shape.""" + return response.json() + + class GalileoStandardLoggingFields(TypedDict, total=False): call_type: str model: str @@ -156,7 +185,7 @@ class GalileoObserve(CustomLogger): }, ) galileo_login_response.raise_for_status() - access_token: Final = galileo_login_response.json()["access_token"] + access_token: Final = _login_access_token(galileo_login_response) self.headers = { "accept": "application/json", "Content-Type": "application/json", @@ -421,7 +450,7 @@ class GalileoObserve(CustomLogger): try: verbose_logger.debug( "Galileo Logger HTTP error response json: %s", - response.json(), + _decoded_body(response), ) except Exception: pass diff --git a/litellm/integrations/gitlab/gitlab_client.py b/litellm/integrations/gitlab/gitlab_client.py index 0690ccc8c15..813a2ef2821 100644 --- a/litellm/integrations/gitlab/gitlab_client.py +++ b/litellm/integrations/gitlab/gitlab_client.py @@ -4,12 +4,80 @@ Now supports selecting a tag via `config["tag"]`; falls back to branch ("main"). """ import base64 -from typing import Any, Final +from collections.abc import Mapping, Sequence +from typing import Any, Final, Protocol, TypedDict from urllib.parse import quote +from typing_extensions import ReadOnly + from litellm.llms.custom_httpx.http_handler import HTTPHandler +class GitLabFilePayload(TypedDict, total=False): + """A repository-files API entry.""" + + content: ReadOnly[str] + encoding: ReadOnly[str] + + +class GitLabTreeEntry(TypedDict, total=False): + """A repository-tree API entry.""" + + path: ReadOnly[str] + type: ReadOnly[str] + + +class GitLabBranch(TypedDict, total=False): + """A repository-branches API entry.""" + + name: ReadOnly[str] + type: ReadOnly[str] + + +class GitLabFileMetadata(TypedDict): + """The response headers a raw file request exposes as metadata.""" + + content_type: ReadOnly[str | None] + content_length: ReadOnly[str | None] + last_modified: ReadOnly[str | None] + + +class _FileJsonResponse(Protocol): + def json(self) -> GitLabFilePayload: ... + + +class _TreeJsonResponse(Protocol): + def json(self) -> Sequence[GitLabTreeEntry] | None: ... + + +class _ProjectJsonResponse(Protocol): + def json(self) -> Mapping[str, object]: ... + + +class _BranchesJsonResponse(Protocol): + def json(self) -> Sequence[GitLabBranch] | None: ... + + +def _file_payload(resp: _FileJsonResponse) -> GitLabFilePayload: + """The JSON body of a repository-files response.""" + return resp.json() + + +def _tree_entries(resp: _TreeJsonResponse) -> Sequence[GitLabTreeEntry]: + """The entries of a repository-tree response.""" + return resp.json() or [] + + +def _project_info(resp: _ProjectJsonResponse) -> Mapping[str, object]: + """The JSON body of a project response.""" + return resp.json() + + +def _branch_entries(resp: _BranchesJsonResponse) -> Sequence[GitLabBranch] | None: + """The JSON body of a repository-branches response.""" + return resp.json() + + class GitLabClient: """ Client for interacting with the GitLab API to fetch files. @@ -42,12 +110,12 @@ class GitLabClient: self.project: str | int = project self.access_token: str = str(access_token) - self.auth_method = config.get("auth_method", "token") # 'token' or 'oauth' + self.auth_method: str = config.get("auth_method", "token") # 'token' or 'oauth' self.branch = config.get("branch", None) if not self.branch: self.branch = "main" self.tag = config.get("tag") - self.base_url = config.get("base_url", "https://gitlab.com/api/v4") + self.base_url: str = config.get("base_url", "https://gitlab.com/api/v4") if not all([self.project, self.access_token]): raise ValueError("project and access_token are required") @@ -159,7 +227,7 @@ class GitLabClient: if resp.status_code == 404: return None resp.raise_for_status() - data: Final = resp.json() + data: Final = _file_payload(resp) content: Final = data.get("content") encoding: Final = data.get("encoding", "") if content and encoding == "base64": @@ -208,7 +276,7 @@ class GitLabClient: return [] resp.raise_for_status() - data: Final = resp.json() or [] + data: Final = _tree_entries(resp) files: Final[list[str]] = [] for item in data: if item.get("type") == "blob": @@ -229,13 +297,13 @@ class GitLabClient: raise Exception("Authentication failed. Check your GitLab token and auth_method.") raise Exception(f"Failed to list files in '{directory_path}': {e}") - def get_repository_info(self) -> dict[str, Any]: + def get_repository_info(self) -> Mapping[str, object]: """Get information about the project/repository.""" url: Final = f"{self.base_url}/projects/{self._project_enc}" try: resp: Final = self.http_handler.get(url, headers=self.headers) resp.raise_for_status() - return resp.json() + return _project_info(resp) except Exception as e: raise Exception(f"Failed to get repository info: {e}") @@ -247,18 +315,18 @@ class GitLabClient: except Exception: return False - def get_branches(self) -> list[dict[str, Any]]: + def get_branches(self) -> list[GitLabBranch]: """Get list of branches in the repository.""" url: Final = f"{self.base_url}/projects/{self._project_enc}/repository/branches" try: resp: Final = self.http_handler.get(url, headers=self.headers) resp.raise_for_status() - data: Final = resp.json() + data: Final = _branch_entries(resp) return data if isinstance(data, list) else [] except Exception as e: raise Exception(f"Failed to get branches: {e}") - def get_file_metadata(self, file_path: str, *, ref: str | None = None) -> dict[str, Any] | None: + def get_file_metadata(self, file_path: str, *, ref: str | None = None) -> GitLabFileMetadata | None: """ Get minimal metadata about a file via RAW endpoint headers at a given ref. diff --git a/litellm/integrations/langfuse/langfuse.py b/litellm/integrations/langfuse/langfuse.py index 296c2b5714e..9576eabaa34 100644 --- a/litellm/integrations/langfuse/langfuse.py +++ b/litellm/integrations/langfuse/langfuse.py @@ -89,7 +89,7 @@ def _extract_cache_read_input_tokens(usage_obj) -> int: # Check prompt_tokens_details.cached_tokens (used by Gemini and other providers) if hasattr(usage_obj, "prompt_tokens_details"): - prompt_tokens_details: Final = getattr(usage_obj, "prompt_tokens_details", None) + prompt_tokens_details: Final[object] = getattr(usage_obj, "prompt_tokens_details", None) if prompt_tokens_details is not None and hasattr(prompt_tokens_details, "cached_tokens"): cached_tokens: Final = getattr(prompt_tokens_details, "cached_tokens", None) if cached_tokens is not None and isinstance(cached_tokens, (int, float)) and cached_tokens > 0: @@ -623,9 +623,16 @@ class LangFuseLogger: ) # Apply custom masking function if provided - if masking_function is not None and callable(masking_function): - input = self._apply_masking_function(input, masking_function) - output = self._apply_masking_function(output, masking_function) + masked_input: Final[object] = ( + self._apply_masking_function(input, masking_function) + if masking_function is not None and callable(masking_function) + else input + ) + masked_output: Final[object] = ( + self._apply_masking_function(output, masking_function) + if masking_function is not None and callable(masking_function) + else output + ) clean_metadata = redact_user_api_key_info(metadata=clean_metadata) @@ -651,15 +658,15 @@ class LangFuseLogger: # Special keys that are found in the function arguments and not the metadata if "input" in update_trace_keys: - trace_params["input"] = input if not mask_input else "redacted-by-litellm" + trace_params["input"] = masked_input if not mask_input else "redacted-by-litellm" if "output" in update_trace_keys: - trace_params["output"] = output if not mask_output else "redacted-by-litellm" + trace_params["output"] = masked_output if not mask_output else "redacted-by-litellm" else: # don't overwrite an existing trace trace_params = { "id": trace_id, "name": trace_name, "session_id": session_id, - "input": input if not mask_input else "redacted-by-litellm", + "input": masked_input if not mask_input else "redacted-by-litellm", "version": clean_metadata.pop( "trace_version", clean_metadata.get("version", None) ), # If provided just version, it will applied to the trace as well, if applied a trace version it will take precedence @@ -669,9 +676,9 @@ class LangFuseLogger: trace_params[key.replace("trace_", "")] = clean_metadata.pop(key, None) if level == "ERROR": - trace_params["status_message"] = output + trace_params["status_message"] = masked_output else: - trace_params["output"] = output if not mask_output else "redacted-by-litellm" + trace_params["output"] = masked_output if not mask_output else "redacted-by-litellm" if debug is True or (isinstance(debug, str) and debug.lower() == "true"): debug_metadata: Final = { @@ -708,7 +715,7 @@ class LangFuseLogger: ("aws_region_name", aws_region_name, bool(aws_region_name)), ("cache_hit", kwargs.get("cache_hit") or False, self._supports_tags() and "cache_hit" in kwargs), ) - enrichments: Final[Mapping[str, Any]] = { + enrichments: Final[Mapping[str, object]] = { key: value for key, value, include in candidate_enrichments if include } @@ -802,8 +809,8 @@ class LangFuseLogger: "end_time": end_time, "model": model_name, "model_parameters": optional_params, - "input": input if not mask_input else "redacted-by-litellm", - "output": output if not mask_output else "redacted-by-litellm", + "input": masked_input if not mask_input else "redacted-by-litellm", + "output": masked_output if not mask_output else "redacted-by-litellm", "usage": usage, "usage_details": usage_details, "metadata": { @@ -825,8 +832,8 @@ class LangFuseLogger: prompt_management_metadata=prompt_management_metadata, langfuse_client=self.Langfuse, ) - if output is not None and isinstance(output, str) and level == "ERROR": - generation_params["status_message"] = output + if masked_output is not None and isinstance(masked_output, str) and level == "ERROR": + generation_params["status_message"] = masked_output if self._supports_completion_start_time(): generation_params["completion_start_time"] = kwargs.get("completion_start_time", None) @@ -935,7 +942,7 @@ class LangFuseLogger: return Version(self.langfuse_sdk_version) >= Version("2.7.3") @staticmethod - def _apply_masking_function(data: Any, masking_function: Callable[[Any], Any]) -> Any: + def _apply_masking_function(data: object, masking_function: Callable[[object], object]) -> object: """ Apply a masking function to data, handling different data types. @@ -1049,7 +1056,7 @@ def _add_prompt_to_generation_params( generation_params: dict, clean_metadata: dict, prompt_management_metadata: StandardLoggingPromptManagementMetadata | None, - langfuse_client: Any, + langfuse_client: object, ) -> dict: from langfuse import Langfuse from langfuse.model import ( diff --git a/litellm/integrations/opik/opik.py b/litellm/integrations/opik/opik.py index fae93f03d1e..ce47d7fe27a 100644 --- a/litellm/integrations/opik/opik.py +++ b/litellm/integrations/opik/opik.py @@ -4,9 +4,12 @@ Opik Logger that logs LLM events to an Opik server import asyncio import traceback +from collections.abc import Mapping from datetime import datetime from typing import Any, Final +from typing_extensions import ReadOnly, TypedDict, Unpack + from litellm._logging import verbose_logger from litellm.integrations.custom_batch_logger import CustomBatchLogger from litellm.llms.custom_httpx.http_handler import ( @@ -23,7 +26,7 @@ except Exception: opik_client = None -def _should_skip_event(kwargs: dict[str, Any]) -> bool: +def _should_skip_event(kwargs: Mapping[str, object]) -> bool: """Check if event should be skipped due to missing standard_logging_object.""" if kwargs.get("standard_logging_object") is None: verbose_logger.debug("OpikLogger skipping event; no standard_logging_object found") @@ -31,12 +34,24 @@ def _should_skip_event(kwargs: dict[str, Any]) -> bool: return False +class _OpikLoggerKwargs(TypedDict, total=False): + """Constructor options accepted by ``OpikLogger``.""" + + project_name: ReadOnly[str | None] + url: ReadOnly[str | None] + api_key: ReadOnly[str | None] + workspace: ReadOnly[str | None] + batch_size: ReadOnly[int | None] + flush_interval: ReadOnly[int | None] + max_queue_size: ReadOnly[int | None] + + class OpikLogger(CustomBatchLogger): """ Opik Logger for logging events to an Opik Server """ - def __init__(self, **kwargs: Any) -> None: + def __init__(self, **kwargs: Unpack[_OpikLoggerKwargs]) -> None: self.async_httpx_client = get_async_httpx_client(llm_provider=httpxSpecialProvider.LoggingCallback) self.sync_httpx_client = _get_httpx_client() @@ -95,7 +110,7 @@ class OpikLogger(CustomBatchLogger): async def async_log_success_event( self, - kwargs: dict[str, Any], + kwargs: dict[str, object], response_obj: Any, start_time: datetime, end_time: datetime, @@ -163,7 +178,7 @@ class OpikLogger(CustomBatchLogger): except Exception as e: verbose_logger.exception("OpikLogger failed to log success event - %s\n%s", e, traceback.format_exc()) - def _sync_send(self, url: str, headers: dict[str, str], batch: dict[str, Any]) -> None: + def _sync_send(self, url: str, headers: dict[str, str], batch: dict[str, object]) -> None: try: response: Final = self.sync_httpx_client.post( url=url, @@ -178,7 +193,7 @@ class OpikLogger(CustomBatchLogger): def log_success_event( self, - kwargs: dict[str, Any], + kwargs: dict[str, object], response_obj: Any, start_time: datetime, end_time: datetime, @@ -247,7 +262,7 @@ class OpikLogger(CustomBatchLogger): except Exception as e: verbose_logger.exception("OpikLogger failed to log success event - %s\n%s", e, traceback.format_exc()) - async def _submit_batch(self, url: str, headers: dict[str, str], batch: dict[str, Any]) -> None: + async def _submit_batch(self, url: str, headers: dict[str, str], batch: dict[str, object]) -> None: try: response: Final = await self.async_httpx_client.post( url=url, diff --git a/litellm/integrations/opik/opik_payload_builder/extractors.py b/litellm/integrations/opik/opik_payload_builder/extractors.py index 92a7eca7f3e..4dd3d40fae3 100644 --- a/litellm/integrations/opik/opik_payload_builder/extractors.py +++ b/litellm/integrations/opik/opik_payload_builder/extractors.py @@ -1,6 +1,7 @@ """Data extraction functions for Opik payload building.""" import json +from collections.abc import Mapping from typing import Any, Final from litellm import _logging @@ -35,8 +36,8 @@ def normalize_provider_name(provider: str | None) -> str | None: def extract_opik_metadata( - litellm_metadata: dict[str, Any], - standard_logging_metadata: dict[str, Any], + litellm_metadata: Mapping[str, Any], + standard_logging_metadata: Mapping[str, Any], ) -> dict[str, Any]: """ Merge Opik metadata from three sources in increasing priority order: @@ -97,7 +98,7 @@ def extract_span_identifiers( def extract_tags( - opik_metadata: dict[str, Any], + opik_metadata: Mapping[str, Any], custom_llm_provider: str | None, ) -> list[str]: """ @@ -122,7 +123,7 @@ def apply_proxy_header_overrides( project_name: str, tags: list[str], thread_id: str | None, - proxy_headers: dict[str, Any], + proxy_headers: Mapping[str, str], ) -> tuple[str, list[str], str | None]: """ Apply overrides from proxy request headers (opik_* prefix). @@ -148,7 +149,7 @@ def apply_proxy_header_overrides( thread_id = value elif param_key == "tags": try: - parsed_tags = json.loads(value) + parsed_tags: object = json.loads(value) if isinstance(parsed_tags, list): tags.extend(parsed_tags) except (json.JSONDecodeError, TypeError): @@ -158,11 +159,11 @@ def apply_proxy_header_overrides( def extract_and_build_metadata( - opik_metadata: dict[str, Any], - standard_logging_metadata: dict[str, Any], - standard_logging_object: dict[str, Any], - litellm_kwargs: dict[str, Any], -) -> dict[str, Any]: + opik_metadata: Mapping[str, object], + standard_logging_metadata: Mapping[str, object], + standard_logging_object: Mapping[str, object], + litellm_kwargs: Mapping[str, object], +) -> dict[str, object]: """ Build the complete metadata dictionary from all available sources. diff --git a/litellm/integrations/otel/plumbing/metrics.py b/litellm/integrations/otel/plumbing/metrics.py index c7e491c002a..e1623f4697f 100644 --- a/litellm/integrations/otel/plumbing/metrics.py +++ b/litellm/integrations/otel/plumbing/metrics.py @@ -11,9 +11,10 @@ identical metrics. The attribute cardinality filter is reused from v1 by import from collections.abc import Mapping from dataclasses import dataclass from datetime import datetime -from typing import Any, Final, TypeAlias +from typing import Any, Final, Literal, Protocol, TypeAlias from opentelemetry.metrics import Histogram, Meter +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -151,6 +152,29 @@ METRIC_ATTRIBUTE_CEILING: Final[frozenset[str]] = frozenset( BOUNDED_HIDDEN_PARAM_KEYS: Final[tuple[str, ...]] = ("model_id",) +class _TokenUsage(TypedDict, total=False): + """The token counts a response's ``usage`` carries, as the recorder reads them.""" + + prompt_tokens: ReadOnly[int] + completion_tokens: ReadOnly[int] + + +class _ResponseView(Protocol): + """The one read the recorder makes on a litellm response object.""" + + def get(self, key: Literal["usage"], /) -> _TokenUsage | None: ... + + +class _MetricKwargs(TypedDict, total=False): + """The logging kwargs the recorder reads directly.""" + + call_type: ReadOnly[str | None] + litellm_params: ReadOnly[Mapping[str, object] | None] + response_cost: ReadOnly[float | None] + completion_start_time: ReadOnly[datetime | float | str | None] + api_call_start_time: ReadOnly[datetime | float | str | None] + + def resolve_error_type(kwargs: Mapping[str, Any]) -> str: """The ``error.type`` value for a failed request. @@ -192,8 +216,8 @@ class GenAIMetricRecorder: def record( self, - kwargs: Mapping[str, Any], - response_obj: Any, + kwargs: _MetricKwargs, + response_obj: _ResponseView | None, start_time: datetime, end_time: datetime, ) -> None: @@ -218,7 +242,7 @@ class GenAIMetricRecorder: def record_failure( self, - kwargs: Mapping[str, Any], + kwargs: _MetricKwargs, start_time: datetime, end_time: datetime, ) -> None: @@ -342,7 +366,7 @@ class GenAIMetricRecorder: # Per-metric recording # ------------------------------------------------------------------ # - def _record_token_usage(self, response_obj: Any, common_attrs: dict) -> None: + def _record_token_usage(self, response_obj: _ResponseView | None, common_attrs: dict) -> None: if not response_obj: return usage: Final = response_obj.get("usage") @@ -353,7 +377,7 @@ class GenAIMetricRecorder: self._metrics.token_usage.record(usage.get("prompt_tokens", 0), attributes=in_attrs) self._metrics.token_usage.record(usage.get("completion_tokens", 0), attributes=out_attrs) - def _record_time_to_first_token(self, kwargs: Mapping[str, Any], common_attrs: dict) -> None: + def _record_time_to_first_token(self, kwargs: _MetricKwargs, common_attrs: dict) -> None: time_to_first_chunk: Final = time_to_first_chunk_seconds(kwargs) if time_to_first_chunk is None: return @@ -361,15 +385,14 @@ class GenAIMetricRecorder: def _record_time_per_output_token( self, - kwargs: Mapping[str, Any], - response_obj: Any, + kwargs: _MetricKwargs, + response_obj: _ResponseView | None, end_time: datetime, duration_s: float, common_attrs: dict, ) -> None: - completion_tokens = None - if response_obj and (usage := response_obj.get("usage")): - completion_tokens = usage.get("completion_tokens") + usage: Final = response_obj.get("usage") if response_obj else None + completion_tokens: Final = usage.get("completion_tokens") if usage else None if completion_tokens is None or completion_tokens <= 0: return diff --git a/litellm/integrations/s3.py b/litellm/integrations/s3.py index ddeb410c54a..8ce461eea5b 100644 --- a/litellm/integrations/s3.py +++ b/litellm/integrations/s3.py @@ -1,11 +1,18 @@ #### What this does #### # On success + failure, log events to Supabase +import hashlib from datetime import datetime from typing import Final, cast import litellm from litellm._logging import print_verbose, verbose_logger +from litellm.constants import ( + MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES, + MAX_S3_OBJECT_KEY_BYTES, + S3_BOUNDED_OBJECT_KEY_HEAD_BYTES, + S3_PREFIX_DIGEST_CHARS, +) from litellm.types.utils import StandardLoggingPayload @@ -133,9 +140,7 @@ class S3Logger: s3_file_name, ) - s3_object_download_filename: Final = ( - "time-" + start_time.strftime("%Y-%m-%dT%H-%M-%S-%f") + "_" + payload["id"] + ".json" - ) + s3_object_download_filename: Final = get_s3_object_download_filename(start_time, payload["id"]) from litellm.litellm_core_utils.safe_json_dumps import safe_dumps @@ -198,6 +203,47 @@ def resolve_sse_params( return algorithm, valid_key_id +S3_MIN_BOUNDED_FILE_NAME_BYTES: Final = 64 + + +def _truncate_to_utf8_bytes(value: str, max_bytes: int) -> str: + """Trim `value` so its UTF-8 encoding fits `max_bytes`, never splitting a character.""" + if max_bytes <= 0: + return "" + encoded: Final = value.encode("utf-8") + if len(encoded) <= max_bytes: + return value + return encoded[:max_bytes].decode("utf-8", errors="ignore") + + +def get_s3_object_download_filename(start_time: datetime, response_id: str) -> str: + """Content-Disposition filename for the uploaded object, bounded to the metadata header cap.""" + sanitized_response_id: Final = response_id.replace("/", "_").replace('"', "_") + file_name: Final = f"time-{start_time.strftime('%Y-%m-%dT%H-%M-%S-%f')}_{response_id}" + sanitized_file_name: Final = f"time-{start_time.strftime('%Y-%m-%dT%H-%M-%S-%f')}_{sanitized_response_id}" + budget: Final = MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES - len(b".json") + if len(sanitized_file_name.encode("utf-8")) <= budget: + return sanitized_file_name + ".json" + return _bounded_s3_file_name(file_name, sanitized_file_name, budget) + ".json" + + +def _bounded_s3_file_name(s3_file_name: str, sanitized_s3_file_name: str, max_bytes: int) -> str: + """As much of the file name as `max_bytes` allows, then the sha256 of the whole name.""" + digest: Final = hashlib.sha256(s3_file_name.encode("utf-8")).hexdigest() + head_budget: Final = min(S3_BOUNDED_OBJECT_KEY_HEAD_BYTES, max_bytes - len(digest) - 1) + head: Final = _truncate_to_utf8_bytes(sanitized_s3_file_name, head_budget) + return f"{head}_{digest}" if head else digest + + +def _bounded_s3_prefix(configured_prefix: str, max_bytes: int) -> str: + """As much of the configured prefix as fits, then a digest segment naming the full prefix.""" + digest_segment: Final = hashlib.sha256(configured_prefix.encode("utf-8")).hexdigest()[:S3_PREFIX_DIGEST_CHARS] + "/" + if max_bytes < len(digest_segment): + return "" + head: Final = _truncate_to_utf8_bytes(configured_prefix, max_bytes - len(digest_segment) - 1).rstrip("/") + return f"{head}/{digest_segment}" if head else digest_segment + + def get_s3_object_key( s3_path: str, prefix: str, @@ -205,12 +251,23 @@ def get_s3_object_key( s3_file_name: str, ) -> str: sanitized_s3_file_name: Final = s3_file_name.replace("/", "_") - s3_object_key = ( - (s3_path.rstrip("/") + "/" if s3_path else "") - + prefix - + start_time.strftime("%Y-%m-%d") - + "/" - + sanitized_s3_file_name - ) # we need the s3 key to include the time, so we log cache hits too - s3_object_key += ".json" - return s3_object_key + configured_prefix: Final = (s3_path.rstrip("/") + "/" if s3_path else "") + prefix + date_segment: Final = start_time.strftime("%Y-%m-%d") + "/" + # we need the s3 key to include the time, so we log cache hits too + s3_object_key: Final = configured_prefix + date_segment + sanitized_s3_file_name + ".json" + if len(s3_object_key.encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES: + return s3_object_key + + # shorten the response id first and only trim the configured prefix if that is what does not + # fit, so prefix scoped IAM policies and lifecycle rules keep matching + budget: Final = MAX_S3_OBJECT_KEY_BYTES - len(date_segment.encode("utf-8")) - len(b".json") + prefix_bytes: Final = len(configured_prefix.encode("utf-8")) + if prefix_bytes + S3_MIN_BOUNDED_FILE_NAME_BYTES <= budget: + bounded_file_name: Final = _bounded_s3_file_name(s3_file_name, sanitized_s3_file_name, budget - prefix_bytes) + return configured_prefix + date_segment + bounded_file_name + ".json" + + shortest_file_name: Final = _bounded_s3_file_name( + s3_file_name, sanitized_s3_file_name, S3_MIN_BOUNDED_FILE_NAME_BYTES + ) + bounded_prefix: Final = _bounded_s3_prefix(configured_prefix, budget - len(shortest_file_name.encode("utf-8"))) + return bounded_prefix + date_segment + shortest_file_name + ".json" diff --git a/litellm/integrations/s3_v2.py b/litellm/integrations/s3_v2.py index 9f6ae72fb3a..712ce41d09e 100644 --- a/litellm/integrations/s3_v2.py +++ b/litellm/integrations/s3_v2.py @@ -16,7 +16,11 @@ from urllib.parse import quote import litellm from litellm._logging import print_verbose, verbose_logger from litellm.constants import DEFAULT_S3_BATCH_SIZE, DEFAULT_S3_FLUSH_INTERVAL_SECONDS -from litellm.integrations.s3 import get_s3_object_key, resolve_sse_params +from litellm.integrations.s3 import ( + get_s3_object_download_filename, + get_s3_object_key, + resolve_sse_params, +) from litellm.litellm_core_utils.aws_partition import get_aws_dns_suffix from litellm.litellm_core_utils.safe_json_dumps import safe_dumps from litellm.litellm_core_utils.sensitive_data_masker import SensitiveDataMasker @@ -259,11 +263,11 @@ class S3Logger(CustomBatchLogger, BaseAWSLLM): now: Final = datetime.now(timezone.utc) audit_log_id: Final = audit_log.get("id", "unknown") - s3_path = cast(str | None, self.s3_path) or "" - s3_path = s3_path.rstrip("/") + "/" if s3_path else "" - - s3_object_key: Final = ( - f"{s3_path}audit_logs/{now.strftime('%Y-%m-%d')}/{now.strftime('%H-%M-%S')}_{audit_log_id}.json" + s3_object_key: Final = get_s3_object_key( + cast(str | None, self.s3_path) or "", + "audit_logs/", + now, + f"{now.strftime('%H-%M-%S')}_{audit_log_id}", ) element: Final = s3BatchLoggingElement( @@ -463,9 +467,7 @@ class S3Logger(CustomBatchLogger, BaseAWSLLM): ) verbose_logger.debug("s3_object_key=%s", s3_object_key) - s3_object_download_filename: Final = ( - f"time-{start_time.strftime('%Y-%m-%dT%H-%M-%S-%f')}_{standard_logging_payload['id']}.json" - ) + s3_object_download_filename: Final = get_s3_object_download_filename(start_time, standard_logging_payload["id"]) return s3BatchLoggingElement( payload=dict(standard_logging_payload), diff --git a/litellm/integrations/vector_store_integrations/vector_store_pre_call_hook.py b/litellm/integrations/vector_store_integrations/vector_store_pre_call_hook.py index aa29162ba1f..07d4f959489 100644 --- a/litellm/integrations/vector_store_integrations/vector_store_pre_call_hook.py +++ b/litellm/integrations/vector_store_integrations/vector_store_pre_call_hook.py @@ -13,7 +13,7 @@ from litellm._logging import verbose_logger from litellm.integrations.custom_logger import CustomLogger from litellm.types.llms.openai import AllMessageValues, ChatCompletionUserMessage from litellm.types.prompts.init_prompts import PromptSpec -from litellm.types.utils import StandardCallbackDynamicParams +from litellm.types.utils import CallTypes, StandardCallbackDynamicParams from litellm.types.vector_stores import ( LiteLLM_ManagedVectorStore, VectorStoreResultContent, @@ -226,7 +226,7 @@ class VectorStorePreCallHook(CustomLogger): self, request_data: dict, response: Any, - call_type: Any | None, + call_type: CallTypes | None, ) -> Any | None: """ Add search results to the response after successful LLM call. @@ -283,7 +283,7 @@ class VectorStorePreCallHook(CustomLogger): self, request_data: dict, response_chunk: Any, - call_type: Any | None, + call_type: CallTypes | None, ) -> Any | None: """ Add search results to the final streaming chunk. diff --git a/litellm/integrations/websearch_interception/handler.py b/litellm/integrations/websearch_interception/handler.py index aefd4fa3b47..dc61ee38a8c 100644 --- a/litellm/integrations/websearch_interception/handler.py +++ b/litellm/integrations/websearch_interception/handler.py @@ -416,25 +416,15 @@ class WebSearchInterceptionLogger(CustomLogger): if not tools: return None - is_responses_call: Final = call_type in (CallTypes.responses, CallTypes.aresponses) - has_websearch: Final = ( - any(is_web_search_tool_responses(tool) for tool in tools) - if is_responses_call - else any(is_web_search_tool(tool) for tool in tools) - ) + if call_type in (CallTypes.responses, CallTypes.aresponses): + return self._convert_responses_tools(kwargs=kwargs, tools=tools) + + # Check if any tool is a web search tool (native or already LiteLLM standard) + has_websearch: Final = any(is_web_search_tool(t) for t in tools) + if not has_websearch: return None - if self.search_tool_name: - try: - from litellm.proxy.proxy_server import llm_router - except ImportError: - llm_router = None - self._select_search_tool_from_router(llm_router=llm_router) - - if is_responses_call: - return self._convert_responses_tools(kwargs=kwargs, tools=tools) - verbose_logger.debug("WebSearchInterception: Converting native web_search tools to LiteLLM standard") # If the client sent an Anthropic-native web_search_* tool, mark the @@ -1641,36 +1631,34 @@ class WebSearchInterceptionLogger(CustomLogger): return None def _select_search_tool_from_router(self, llm_router: object) -> "_SearchToolConfig | None": - search_tools: Final = list(getattr(llm_router, "search_tools", []) or []) + if llm_router is None or not hasattr(llm_router, "search_tools"): + return None + search_tools: Final = tuple(getattr(llm_router, "search_tools", None) or ()) return self._select_search_tool_from_list(search_tools=search_tools, source="router") def _select_search_tool_from_list( self, - search_tools: list[_SearchToolConfig], + search_tools: Sequence[_SearchToolConfig], source: str, ) -> "_SearchToolConfig | None": if self.search_tool_name: - matching_tools = [tool for tool in search_tools if tool.get("search_tool_name") == self.search_tool_name] - if not matching_tools: - raise ValueError(f"Configured search tool '{self.search_tool_name}' was not found") - - selected_tool: Final = matching_tools[0] - litellm_params: Final = selected_tool.get("litellm_params") - selected_search_provider: Final = ( - litellm_params.get("search_provider") if isinstance(litellm_params, Mapping) else None + matching_tools: Final = tuple( + tool for tool in search_tools if tool.get("search_tool_name") == self.search_tool_name ) - if not isinstance(selected_search_provider, str) or not selected_search_provider.strip(): - raise ValueError( - f"Configured search tool '{self.search_tool_name}' does not define a valid search provider" + if matching_tools: + search_provider = (matching_tools[0].get("litellm_params", {}) or {}).get("search_provider") + verbose_logger.debug( + "WebSearchInterception: Found search tool '%s' from %s with provider '%s'", + self.search_tool_name, + source, + search_provider, ) - + return matching_tools[0] verbose_logger.debug( - "WebSearchInterception: Found search tool '%s' from %s with provider '%s'", + "WebSearchInterception: Search tool '%s' not found in %s, falling back to first available or perplexity", self.search_tool_name, source, - selected_search_provider, ) - return selected_tool if search_tools: first_tool: Final = search_tools[0] diff --git a/litellm/litellm_core_utils/get_llm_provider_logic.py b/litellm/litellm_core_utils/get_llm_provider_logic.py index 9b53b79bbe6..207e024ce0b 100644 --- a/litellm/litellm_core_utils/get_llm_provider_logic.py +++ b/litellm/litellm_core_utils/get_llm_provider_logic.py @@ -536,6 +536,14 @@ def get_llm_provider( ) +def _dashscope_family_chat_config(custom_llm_provider: str) -> "litellm.DashScopeChatConfig": + if custom_llm_provider == "qwencloud": + return litellm.QwenCloudChatConfig() + if custom_llm_provider == "qwen_ai_platform": + return litellm.QwenAIPlatformChatConfig() + return litellm.DashScopeChatConfig() + + def _get_openai_compatible_provider_info( model: str, api_base: str | None, @@ -785,11 +793,11 @@ def _get_openai_compatible_provider_info( api_base, dynamic_api_key, ) = litellm.HerokuChatConfig()._get_openai_compatible_provider_info(api_base, api_key) - elif custom_llm_provider == "dashscope": + elif custom_llm_provider in ("dashscope", "qwencloud", "qwen_ai_platform"): ( api_base, dynamic_api_key, - ) = litellm.DashScopeChatConfig()._get_openai_compatible_provider_info(api_base, api_key) + ) = _dashscope_family_chat_config(custom_llm_provider)._get_openai_compatible_provider_info(api_base, api_key) elif custom_llm_provider == "modelscope": ( api_base, diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py index 350e5403e4c..eb223c2f988 100644 --- a/litellm/litellm_core_utils/litellm_logging.py +++ b/litellm/litellm_core_utils/litellm_logging.py @@ -111,6 +111,7 @@ from litellm.types.mcp import MCPPostCallResponseObject from litellm.types.prompts.init_prompts import PromptSpec from litellm.types.rerank import RerankResponse from litellm.types.utils import ( + DEPLOYMENT_SCOPED_PRICING_FIELDS, CachingDetails, CallTypes, CostBreakdown, @@ -255,6 +256,7 @@ _STANDARD_LOGGING_METADATA_KEYS: Final[frozenset[str]] = frozenset(StandardLoggi # Cache custom pricing keys as frozenset for O(1) lookups instead of looping through 49 keys _CUSTOM_PRICING_KEYS: Final[frozenset[str]] = frozenset(CustomPricingLiteLLMParams.model_fields.keys()) +_MODEL_INFO_CUSTOM_PRICING_KEYS: Final[frozenset[str]] = _CUSTOM_PRICING_KEYS | DEPLOYMENT_SCOPED_PRICING_FIELDS sentry_sdk_instance = None capture_exception = None @@ -5033,7 +5035,9 @@ def use_custom_pricing_for_model(litellm_params: dict | None) -> bool: """ Check if the model uses custom pricing - Returns True if any of `SPECIAL_MODEL_INFO_PARAMS` are present in `litellm_params` or `model_info` + Returns True if any custom pricing field is present in `litellm_params`, or if + any custom pricing or deployment-scoped pricing field (such as + ``off_peak_pricing``) is present in the metadata ``model_info`` """ if litellm_params is None: return False @@ -5051,7 +5055,7 @@ def use_custom_pricing_for_model(litellm_params: dict | None) -> bool: model_info: dict = metadata.get("model_info", {}) or {} if model_info: - matching_keys = _CUSTOM_PRICING_KEYS & model_info.keys() + matching_keys = _MODEL_INFO_CUSTOM_PRICING_KEYS & model_info.keys() for key in matching_keys: if model_info.get(key) is not None: return True diff --git a/litellm/litellm_core_utils/llm_cost_calc/utils.py b/litellm/litellm_core_utils/llm_cost_calc/utils.py index 19e3f624268..b34c416cd40 100644 --- a/litellm/litellm_core_utils/llm_cost_calc/utils.py +++ b/litellm/litellm_core_utils/llm_cost_calc/utils.py @@ -2,10 +2,12 @@ ## Helper utilities for cost_per_token() import re -from collections.abc import Mapping +from collections.abc import Mapping, Sequence from dataclasses import dataclass +from datetime import datetime, timezone, tzinfo from types import MappingProxyType from typing import Any, Final, Literal, TypedDict, cast +from zoneinfo import ZoneInfo, ZoneInfoNotFoundError import litellm from litellm._logging import verbose_logger @@ -290,10 +292,187 @@ def _get_tiered_base_costs(model_info: ModelInfo, usage: Usage) -> tuple[float, ) +def _is_within_off_peak_window(off_peak_hours_utc: str | Sequence[str], current_time: datetime | None = None) -> bool: + """Return True if current_time (UTC, defaulting to now) falls inside any off-peak window. + + off_peak_hours_utc is a "HH:MM-HH:MM" string in UTC, or a list of such strings for providers + with multiple daily windows (e.g. ["16:30-00:30", "04:00-06:00"]). A window may wrap past + midnight, and a window whose start equals its end covers the whole day. The start is + inclusive and the end is exclusive; malformed windows are ignored. + + An aware current_time is converted to UTC. A naive one is taken to already be UTC rather + than being localised, so callers must pass datetime.now(timezone.utc), never datetime.now(), + or every window shifts by the host's offset. + """ + reference: Final = current_time if current_time is not None else datetime.now(timezone.utc) + now: Final = (reference.astimezone(timezone.utc) if reference.tzinfo is not None else reference).time() + windows: Final = (off_peak_hours_utc,) if isinstance(off_peak_hours_utc, str) else off_peak_hours_utc + for window in windows: + try: + start_str, end_str = window.split("-") + start = datetime.strptime(start_str.strip(), "%H:%M").replace(tzinfo=timezone.utc).time() + end = datetime.strptime(end_str.strip(), "%H:%M").replace(tzinfo=timezone.utc).time() + except (ValueError, AttributeError): + continue + if start < end: + if start <= now < end: + return True + elif now >= start or now < end: + return True + return False + + +_WEEKDAY_NUMBERS: Final = MappingProxyType( + { + "mon": 1, + "monday": 1, + "tue": 2, + "tues": 2, + "tuesday": 2, + "wed": 3, + "wednesday": 3, + "thu": 4, + "thur": 4, + "thurs": 4, + "thursday": 4, + "fri": 5, + "friday": 5, + "sat": 6, + "saturday": 6, + "sun": 7, + "sunday": 7, + } +) + + +def _normalize_weekday(value: object) -> int | None: + if isinstance(value, bool): + return None + if isinstance(value, int): + return value if 1 <= value <= 7 else None + if isinstance(value, str): + return _WEEKDAY_NUMBERS.get(value.strip().lower()) + return None + + +def _weekday_calendar(weekday_timezone: object) -> tzinfo: + if isinstance(weekday_timezone, str) and weekday_timezone.strip(): + try: + return ZoneInfo(weekday_timezone.strip()) + except (ValueError, ZoneInfoNotFoundError): + return timezone.utc + return timezone.utc + + +def _matches_weekdays(reference_utc: datetime, weekdays: object, weekday_timezone: object) -> bool: + """Return True when reference_utc falls on one of the rule's weekdays, read on the calendar + named by weekday_timezone (default UTC). An absent weekdays means every day. The calendar + matters even when UTC and vendor-local weekdays agree at every currently priced hour: a + window past 16:00 UTC is where an Asia/Shanghai weekday diverges from the UTC one. + """ + if weekdays is None: + return True + if isinstance(weekdays, str) or not isinstance(weekdays, Sequence): + return False + allowed: Final = frozenset(day for day in map(_normalize_weekday, weekdays) if day is not None) + return reference_utc.astimezone(_weekday_calendar(weekday_timezone)).isoweekday() in allowed + + +def _as_window_strings(value: object) -> tuple[str, ...]: + if isinstance(value, str): + return (value,) + if isinstance(value, Sequence): + return tuple(entry for entry in value if isinstance(entry, str)) + return () + + +def _is_off_peak(off_peak: Mapping[str, object], current_time: datetime | None = None) -> bool: + """Return True when current_time (UTC, defaulting to now) is off-peak under the block's + rules: the flat hours_utc windows, which apply every day, or any entry in windows, whose + hours apply only on its weekdays. + """ + reference: Final = current_time if current_time is not None else datetime.now(timezone.utc) + reference_utc: Final = ( + reference.astimezone(timezone.utc) if reference.tzinfo is not None else reference.replace(tzinfo=timezone.utc) + ) + flat_windows: Final = _as_window_strings(off_peak.get("hours_utc")) + if flat_windows and _is_within_off_peak_window(flat_windows, reference_utc): + return True + windows: Final = off_peak.get("windows") + if isinstance(windows, str) or not isinstance(windows, Sequence): + return False + weekday_timezone: Final = off_peak.get("weekday_timezone") + for rule in windows: + if not isinstance(rule, Mapping): + continue + rule_windows = _as_window_strings(rule.get("hours_utc")) + if not rule_windows: + continue + if not _matches_weekdays(reference_utc, rule.get("weekdays"), weekday_timezone): + continue + if _is_within_off_peak_window(rule_windows, reference_utc): + return True + return False + + +def _coerce_off_peak_rate(value: object, default: float) -> float: + if isinstance(value, bool): + return default + if isinstance(value, (int, float)): + return float(value) + if isinstance(value, str): + try: + return float(value) + except ValueError: + return default + return default + + +def _apply_off_peak_pricing( + model_info: ModelInfo, + current_time: datetime | None, + prompt_base_cost: float, + completion_base_cost: float, + cache_read_cost: float, +) -> tuple[float, float, float]: + """Swap in off-peak per-token rates when the current UTC time is inside one of the model's + off_peak_pricing rules, the every-day hours_utc windows or a day-of-week-qualified entry in + windows. An off-peak rate replaces the rate that would otherwise apply rather than + discounting it, so a model that also has tiered or above-threshold pricing bills the flat + off-peak rate for the whole request while the window is open. Any rate left unset in + off_peak_pricing falls back to the standard rate. + """ + off_peak: Final = model_info.get("off_peak_pricing") + if not isinstance(off_peak, Mapping) or not _is_off_peak(off_peak, current_time): + return prompt_base_cost, completion_base_cost, cache_read_cost + return ( + _coerce_off_peak_rate(off_peak.get("input_cost_per_token"), prompt_base_cost), + _coerce_off_peak_rate(off_peak.get("output_cost_per_token"), completion_base_cost), + _coerce_off_peak_rate(off_peak.get("cache_read_input_token_cost"), cache_read_cost), + ) + + +def _apply_off_peak_to_base_costs( + model_info: ModelInfo, + current_time: datetime | None, + base_costs: tuple[float, float, float, float, float], +) -> tuple[float, float, float, float, float]: + """Apply off-peak rates to an already-resolved set of base costs, whichever pricing path + produced them. Cache-creation rates are passed through untouched, since off_peak_pricing + has no field for them. + """ + prompt, completion, cache_creation, cache_creation_above_1hr, cache_read = base_costs + off_peak_prompt, off_peak_completion, off_peak_cache_read = _apply_off_peak_pricing( + model_info, current_time, prompt, completion, cache_read + ) + return (off_peak_prompt, off_peak_completion, cache_creation, cache_creation_above_1hr, off_peak_cache_read) + + def _get_token_base_cost( model_info: ModelInfo, usage: Usage, service_tier: str | None = None, + current_time: datetime | None = None, *, threshold_is_inclusive: bool = False, ) -> tuple[float, float, float, float, float]: @@ -311,7 +490,7 @@ def _get_token_base_cost( """ tiered_base_costs: Final = _get_tiered_base_costs(model_info=model_info, usage=usage) if tiered_base_costs is not None: - return tiered_base_costs + return _apply_off_peak_to_base_costs(model_info, current_time, tiered_base_costs) # Get service tier aware cost keys input_cost_key: Final = _get_service_tier_cost_key("input_cost_per_token", service_tier) @@ -345,12 +524,16 @@ def _get_token_base_cost( k for k in model_info if k.startswith("input_cost_per_token_above_") and not k.endswith(_SERVICE_TIER_SUFFIXES) ] if not threshold_keys: - return ( - prompt_base_cost, - completion_base_cost, - cache_creation_cost, - cache_creation_cost_above_1hr, - cache_read_cost, + return _apply_off_peak_to_base_costs( + model_info, + current_time, + ( + prompt_base_cost, + completion_base_cost, + cache_creation_cost, + cache_creation_cost_above_1hr, + cache_read_cost, + ), ) # Only sort the threshold keys (typically 1-2 keys instead of 66+) @@ -451,12 +634,16 @@ def _get_token_base_cost( except Exception: continue - return ( - prompt_base_cost, - completion_base_cost, - cache_creation_cost, - cache_creation_cost_above_1hr, - cache_read_cost, + return _apply_off_peak_to_base_costs( + model_info, + current_time, + ( + prompt_base_cost, + completion_base_cost, + cache_creation_cost, + cache_creation_cost_above_1hr, + cache_read_cost, + ), ) diff --git a/litellm/litellm_core_utils/prompt_templates/factory.py b/litellm/litellm_core_utils/prompt_templates/factory.py index 795fb36961e..e6402e8c1bd 100644 --- a/litellm/litellm_core_utils/prompt_templates/factory.py +++ b/litellm/litellm_core_utils/prompt_templates/factory.py @@ -1694,6 +1694,18 @@ def convert_function_to_anthropic_tool_invoke( raise e +def _find_server_tool_result( + tool_id: str, + web_search_results: Sequence[object] | None, + tool_results: Sequence[object] | None, +) -> dict[str, object] | None: + candidates: Final = (*(web_search_results or ()), *(tool_results or ())) + return next( + (result for result in candidates if isinstance(result, dict) and result.get("tool_use_id") == tool_id), + None, + ) + + def convert_to_anthropic_tool_invoke( tool_calls: list[ChatCompletionAssistantToolCall], web_search_results: list[Any] | None = None, @@ -1758,32 +1770,22 @@ def convert_to_anthropic_tool_invoke( context="Anthropic tool invoke", ) - # Check if this is a server-side tool (web_search, tool_search, etc.) - # Server tool IDs start with "srvtoolu_" - if tool_id.startswith("srvtoolu_"): - # Create server_tool_use block instead of tool_use - _anthropic_server_tool_use: dict[str, object] = { - "type": "server_tool_use", - "id": tool_id, - "name": tool_name, - "input": tool_input, - } - anthropic_tool_invoke.append(_anthropic_server_tool_use) - - # Add corresponding tool result if available. - # Check both web_search_results (web_search_tool_result / web_fetch_tool_result) - # and tool_results (bash_code_execution_tool_result, etc.) - _all_tool_results: list[Any] = [] - if web_search_results: - _all_tool_results.extend(web_search_results) - if tool_results: - _all_tool_results.extend(tool_results) - for result in _all_tool_results: - if result.get("tool_use_id") == tool_id: - anthropic_tool_invoke.append(result) - break + server_tool_result = ( + _find_server_tool_result(tool_id, web_search_results, tool_results) + if tool_id.startswith("srvtoolu_") + else None + ) + if server_tool_result is not None: + anthropic_tool_invoke.append( + { + "type": "server_tool_use", + "id": tool_id, + "name": tool_name, + "input": tool_input, + } + ) + anthropic_tool_invoke.append(server_tool_result) else: - # Regular tool_use sanitized_tool_id = _sanitize_anthropic_tool_use_id(tool_id) _anthropic_tool_use_param = AnthropicMessagesToolUseParam( type="tool_use", diff --git a/litellm/litellm_core_utils/realtime_streaming.py b/litellm/litellm_core_utils/realtime_streaming.py index 9125ed6e70a..8479e108d17 100644 --- a/litellm/litellm_core_utils/realtime_streaming.py +++ b/litellm/litellm_core_utils/realtime_streaming.py @@ -1500,6 +1500,6 @@ class RealTimeStreaming: pass -def client_sent_openai_beta_realtime_header(websocket: Any) -> bool: +def client_sent_openai_beta_realtime_header(websocket: _ScopedWebSocket) -> bool: """True when the client WebSocket includes ``OpenAI-Beta: realtime=v1``.""" return RealTimeStreaming._detect_beta_header(websocket) diff --git a/litellm/litellm_core_utils/streaming_chunk_builder_utils.py b/litellm/litellm_core_utils/streaming_chunk_builder_utils.py index 0e2139d688b..3978a01a5db 100644 --- a/litellm/litellm_core_utils/streaming_chunk_builder_utils.py +++ b/litellm/litellm_core_utils/streaming_chunk_builder_utils.py @@ -73,6 +73,18 @@ class _ContentChunk(TypedDict): choices: Sequence[_ContentChoice] +class _FunctionCallDelta(TypedDict): + function_call: ReadOnly[FunctionCall] + + +class _FunctionCallChoice(TypedDict): + delta: ReadOnly[_FunctionCallDelta] + + +class _FunctionCallChunk(TypedDict): + choices: ReadOnly[Sequence[_FunctionCallChoice]] + + class _AudioDelta(TypedDict, total=False): audio: ChatCompletionAudioDelta | None @@ -588,7 +600,7 @@ class ChunkProcessor: return tool_calls_list - def get_combined_function_call_content(self, function_call_chunks: list[dict[str, Any]]) -> FunctionCall: + def get_combined_function_call_content(self, function_call_chunks: Sequence["_FunctionCallChunk"]) -> FunctionCall: argument_list: Final = [] delta = function_call_chunks[0]["choices"][0]["delta"] function_call = delta.get("function_call", "") diff --git a/litellm/litellm_core_utils/streaming_handler.py b/litellm/litellm_core_utils/streaming_handler.py index f5f671e7585..480b1921c18 100644 --- a/litellm/litellm_core_utils/streaming_handler.py +++ b/litellm/litellm_core_utils/streaming_handler.py @@ -862,6 +862,8 @@ class CustomStreamWrapper: model_response: Final = ModelResponseStream(**args) if self.response_id is not None: model_response.id = self.response_id + elif model_response.id: + self.response_id = model_response.id if self.system_fingerprint is not None: model_response.system_fingerprint = self.system_fingerprint diff --git a/litellm/llms/a2a/chat/guardrail_translation/handler.py b/litellm/llms/a2a/chat/guardrail_translation/handler.py index 1c5ba951942..f1c7451796d 100644 --- a/litellm/llms/a2a/chat/guardrail_translation/handler.py +++ b/litellm/llms/a2a/chat/guardrail_translation/handler.py @@ -11,8 +11,11 @@ A2A Protocol Format: """ import json +from collections.abc import Sequence from typing import TYPE_CHECKING, Any, Final, Optional +from typing_extensions import ReadOnly, TypedDict + from litellm._logging import verbose_proxy_logger from litellm.llms.base_llm.guardrail_translation.base_translation import BaseTranslation from litellm.types.utils import GenericGuardrailAPIInputs @@ -23,6 +26,13 @@ if TYPE_CHECKING: from litellm.proxy._types import UserAPIKeyAuth +class _A2ATextPart(TypedDict, total=False): + """The subset of an A2A message part this handler reads text from.""" + + kind: ReadOnly[str] + text: ReadOnly[str] + + class A2AGuardrailHandler(BaseTranslation): """ Handler for processing A2A Protocol messages with guardrails. @@ -41,7 +51,7 @@ class A2AGuardrailHandler(BaseTranslation): data: dict, guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: Optional["LiteLLMLoggingObj"] = None, - ) -> Any: + ) -> dict: """ Process A2A input messages by applying guardrails to text content. @@ -214,12 +224,12 @@ class A2AGuardrailHandler(BaseTranslation): async def process_output_streaming_response( self, - responses_so_far: list[Any], + responses_so_far: list[object], guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: Optional["LiteLLMLoggingObj"] = None, user_api_key_dict: Optional["UserAPIKeyAuth"] = None, request_data: dict | None = None, - ) -> list[Any]: + ) -> list[object]: """ Process A2A streaming output by applying guardrails to accumulated text. @@ -305,11 +315,12 @@ class A2AGuardrailHandler(BaseTranslation): def _parse_streaming_responses( self, - responses_so_far: list[Any], - ) -> tuple[list[dict[str, Any] | None], list[tuple[int, dict[str, Any]]]]: + responses_so_far: list[object], + ) -> tuple[list[dict[str, object] | None], list[tuple[int, dict[str, object]]]]: """Parse JSON-RPC items, returning aligned parsed list and valid entries.""" - parsed: Final[list[dict[str, Any] | None]] = [None] * len(responses_so_far) + parsed: Final[list[dict[str, object] | None]] = [None] * len(responses_so_far) for i, item in enumerate(responses_so_far): + obj: dict[str, object] if isinstance(item, dict): obj = item elif isinstance(item, str): @@ -326,7 +337,7 @@ class A2AGuardrailHandler(BaseTranslation): def _collect_text_from_parsed_chunks( self, - valid_parsed: list[tuple[int, dict[str, Any]]], + valid_parsed: list[tuple[int, dict[str, object]]], ) -> tuple[str, list[int]]: """Collect text from parsed chunks, returning combined text and indices.""" from litellm.llms.a2a.common_utils import extract_text_from_a2a_response @@ -411,7 +422,7 @@ class A2AGuardrailHandler(BaseTranslation): def _extract_texts_from_parts( self, - parts: list[dict[str, Any]], + parts: Sequence[_A2ATextPart], path: tuple[str, ...], texts_to_check: list[str], task_mappings: list[tuple[tuple[str, ...], int]], diff --git a/litellm/llms/anthropic/chat/transformation.py b/litellm/llms/anthropic/chat/transformation.py index e1387a9068c..0bfe7dddd7f 100644 --- a/litellm/llms/anthropic/chat/transformation.py +++ b/litellm/llms/anthropic/chat/transformation.py @@ -7,6 +7,7 @@ from typing import TYPE_CHECKING, Any, Final, NoReturn, cast import httpx from pydantic import ValidationError +from typing_extensions import ReadOnly, TypedDict import litellm from litellm.constants import ( @@ -125,7 +126,25 @@ else: _ANTHROPIC_TOOL_NAME_INVALID_CHARS: Final = re.compile(r"[^a-zA-Z0-9_-]") _ANTHROPIC_TOOL_NAME_MAX_LEN: Final = 128 -_ENUM_TYPE_CHECKS: Final[Mapping[str, Callable[[Any], bool]]] = MappingProxyType( + +class _AnthropicUsageIteration(TypedDict, total=False): + """One entry of the ``usage.iterations`` array on an Anthropic response.""" + + input_tokens: ReadOnly[int | None] + output_tokens: ReadOnly[int | None] + cache_creation_input_tokens: ReadOnly[int | None] + cache_read_input_tokens: ReadOnly[int | None] + + +class _AnthropicToolResultBlock(TypedDict, total=False): + """A ``*_tool_result`` content block on an Anthropic response.""" + + type: ReadOnly[str] + tool_use_id: ReadOnly[str] + content: ReadOnly[object] + + +_ENUM_TYPE_CHECKS: Final[Mapping[str, Callable[[object], bool]]] = MappingProxyType( { "null": lambda v: v is None, "boolean": lambda v: isinstance(v, bool), @@ -440,7 +459,7 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): optional_params.pop("speed", None) @staticmethod - def _raise_invalid_reasoning_effort(model: str, value: Any, llm_provider: str) -> NoReturn: + def _raise_invalid_reasoning_effort(model: str, value: object, llm_provider: str) -> NoReturn: """Raise a ``BadRequestError`` for an unrecognised ``reasoning_effort``. Args: @@ -1466,7 +1485,9 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): ) if _tool_choice is not None: - optional_params["tool_choice"] = _tool_choice + optional_params["tool_choice"] = AnthropicConfig._apply_forced_tool_choice( + model=model, tool_choice=_tool_choice, drop_params=drop_params + ) elif param == "stream" and value is True: optional_params["stream"] = value elif param == "stop" and (isinstance(value, str) or isinstance(value, list)): @@ -1992,19 +2013,35 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): return data def _apply_output_config(self, data: dict, model: str, optional_params: dict) -> None: - """Validate and apply output_config to the request data.""" + """Validate and apply output_config to the request data. + + The ``drop_params`` gate here is an effort gate: ``format`` is a + structured-output field, not an effort field, so it survives the drop + and is vetted where it is consumed (the map's + ``supports_native_structured_output`` flag on emission paths). + """ if "output_config" not in optional_params: return output_config: Final = optional_params.get("output_config") if not output_config or not isinstance(output_config, dict): return - if litellm.drop_params is True and not self._model_supports_effort_param(model, self._resolved_provider): + if ( + litellm.drop_params is True + and any(key != "format" for key in output_config) + and not self._model_supports_effort_param(model, self._resolved_provider) + ): litellm.verbose_logger.warning( DROP_UNSUPPORTED_OUTPUT_CONFIG_WARNING, model, ) - optional_params.pop("output_config", None) - data.pop("output_config", None) + preserved_format: Final = output_config.get("format") + if preserved_format is None: + optional_params.pop("output_config", None) + data.pop("output_config", None) + return + format_only: Final = {"format": preserved_format} # mutable-ok: json body + optional_params["output_config"] = format_only # rebind-ok: out-param store + data["output_config"] = format_only # rebind-ok: out-param store return effort: Final = output_config.get("effort") valid_efforts: Final = ["high", "medium", "low", "xhigh", "max"] @@ -2059,22 +2096,22 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): self, completion_response: dict ) -> tuple[ str, - list[Any] | None, + list[object] | None, list[ChatCompletionThinkingBlock | ChatCompletionRedactedThinkingBlock] | None, str | None, list[ChatCompletionToolCallChunk], - list[Any] | None, - list[Any] | None, - list[Any] | None, + list[object] | None, + list[_AnthropicToolResultBlock] | None, + list[object] | None, ]: text_content = "" - citations: list[Any] | None = None + citations: list[object] | None = None thinking_blocks: list[ChatCompletionThinkingBlock | ChatCompletionRedactedThinkingBlock] | None = None reasoning_content: str | None = None tool_calls: Final[list[ChatCompletionToolCallChunk]] = [] - web_search_results: list[Any] | None = None - tool_results: list[Any] | None = None - compaction_blocks: list[Any] | None = None + web_search_results: list[object] | None = None + tool_results: list[_AnthropicToolResultBlock] | None = None + compaction_blocks: list[object] | None = None for idx, content in enumerate(completion_response["content"]): if content["type"] == "text": text_content += content["text"] @@ -2284,7 +2321,7 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): raw_speed: Final = _usage.get("speed") resolved_speed: Final = raw_speed if isinstance(raw_speed, str) else speed - iterations: Final[list[Any] | None] = _usage.get("iterations") + iterations: Final[Sequence[_AnthropicUsageIteration] | None] = _usage.get("iterations") if iterations: prompt_tokens = sum(it.get("input_tokens", 0) or 0 for it in iterations) completion_tokens = sum(it.get("output_tokens", 0) or 0 for it in iterations) @@ -2377,7 +2414,7 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): def _build_code_interpreter_results( self, - tool_results: list[Any], + tool_results: Sequence[_AnthropicToolResultBlock], code_by_id: dict[str, str], container_id: str | None, ) -> list[OutputCodeInterpreterCall]: @@ -2403,11 +2440,11 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): def _build_provider_specific_fields( self, completion_response: dict, - citations: list[Any] | None, + citations: Sequence[object] | None, thinking_blocks: list[ChatCompletionThinkingBlock | ChatCompletionRedactedThinkingBlock] | None, - web_search_results: list[Any] | None, - tool_results: list[Any] | None, - compaction_blocks: list[Any] | None, + web_search_results: Sequence[object] | None, + tool_results: Sequence[_AnthropicToolResultBlock] | None, + compaction_blocks: Sequence[object] | None, tool_calls: list[ChatCompletionToolCallChunk], ) -> dict[str, Any]: provider_specific_fields: Final[dict[str, Any]] = { diff --git a/litellm/llms/anthropic/common_utils.py b/litellm/llms/anthropic/common_utils.py index 9871001bf66..b1f927fd8f9 100644 --- a/litellm/llms/anthropic/common_utils.py +++ b/litellm/llms/anthropic/common_utils.py @@ -28,10 +28,15 @@ from litellm.types.llms.anthropic import ( ANTHROPIC_OAUTH_TOKEN_PREFIX, AllAnthropicToolsValues, AnthropicMcpServerTool, + AnthropicMessagesToolChoice, ) from litellm.types.llms.openai import AllMessageValues from litellm.types.proxy.model_listing import ModelInfoResponse +DROP_FORCED_TOOL_CHOICE_WARNING: Final = ( + "Downgrading forced tool_choice to 'auto' for model=%s (drop_params=True): this model rejects tool_choice type " + "'any'/'tool' with a 400 because thinking is always on and a forced call would skip it." +) DROP_DISABLED_THINKING_WARNING: Final = ( "Dropping `thinking={'type': 'disabled'}` for model=%s: thinking is always on for this model and cannot be " "disabled (the alternative is a provider 400). The model will still think adaptively, its response can contain " @@ -320,6 +325,41 @@ class AnthropicModelInfo(BaseLLMModelInfo): status_code=400, ) + @staticmethod + def forced_tool_use_downgraded(model: str, drop_params: bool) -> bool: + """True when the model map flags the model with + ``supports_forced_tool_use: false`` (Fable 5.1 / Mythos 5.1 400 on + ``any``/``tool``) and ``drop_params`` asks for the ``auto`` downgrade; + raises a clean client-side 400 for such models without ``drop_params``.""" + if AnthropicModelInfo._get_model_capability(model, "supports_forced_tool_use") is not False: + return False + if not (litellm.drop_params or drop_params): + raise litellm.utils.UnsupportedParamsError( + message=( + f"{model} does not support forced tool use (tool_choice='required' or a named tool). " + "Use tool_choice='auto' and tell the model in the prompt when to call the tool, or set " + "`litellm.drop_params = True` to downgrade to 'auto' automatically." + ), + status_code=400, + ) + litellm.verbose_logger.warning(DROP_FORCED_TOOL_CHOICE_WARNING, model) + return True + + @staticmethod + def _apply_forced_tool_choice( + model: str, + tool_choice: AnthropicMessagesToolChoice, + drop_params: bool, + ) -> AnthropicMessagesToolChoice: + if tool_choice["type"] not in ("any", "tool"): + return tool_choice + if not AnthropicModelInfo.forced_tool_use_downgraded(model, drop_params): + return tool_choice + disable_parallel: Final = tool_choice.get("disable_parallel_tool_use") + if disable_parallel is None: + return AnthropicMessagesToolChoice(type="auto") + return AnthropicMessagesToolChoice(type="auto", disable_parallel_tool_use=disable_parallel) + @staticmethod def _strip_version_suffix(model: str) -> str: at: Final = model.rfind("@") diff --git a/litellm/llms/anthropic/files/handler.py b/litellm/llms/anthropic/files/handler.py index 5fdf2ceff7f..dfd62ca575b 100644 --- a/litellm/llms/anthropic/files/handler.py +++ b/litellm/llms/anthropic/files/handler.py @@ -2,7 +2,7 @@ import asyncio import json import time from collections.abc import Coroutine -from typing import Any, Final +from typing import Final import httpx @@ -116,7 +116,7 @@ class AnthropicFilesHandler: api_key: str | None = None, timeout: float | httpx.Timeout = 600.0, max_retries: int | None = None, - ) -> HttpxBinaryResponseContent | Coroutine[Any, Any, HttpxBinaryResponseContent]: + ) -> HttpxBinaryResponseContent | Coroutine[object, object, HttpxBinaryResponseContent]: """ Retrieve file content from Anthropic. diff --git a/litellm/llms/azure/azure.py b/litellm/llms/azure/azure.py index 2bcc830851a..46a9dd1a531 100644 --- a/litellm/llms/azure/azure.py +++ b/litellm/llms/azure/azure.py @@ -2,7 +2,7 @@ import asyncio import json import time from collections.abc import Callable, Coroutine -from typing import Any, Final +from typing import Final import httpx from openai import ( @@ -374,7 +374,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): except Exception as e: status_code: Final = getattr(e, "status_code", 500) error_headers = getattr(e, "headers", None) - error_response: Final = getattr(e, "response", None) + error_response: Final[object] = getattr(e, "response", None) error_body: Final = getattr(e, "body", None) if error_headers is None and error_response: error_headers = getattr(error_response, "headers", None) @@ -392,7 +392,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): model: str, api_base: str, data: dict, - timeout: Any, + timeout: float | httpx.Timeout, dynamic_params: bool, model_response: ModelResponse, logging_obj: LiteLLMLoggingObj, @@ -502,7 +502,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): dynamic_params: bool, data: dict[str, object], model: str, - timeout: Any, + timeout: float | httpx.Timeout, max_retries: int, azure_ad_token: str | None = None, azure_ad_token_provider: Callable | None = None, @@ -578,7 +578,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): dynamic_params: bool, data: dict, model: str, - timeout: Any, + timeout: float | httpx.Timeout, max_retries: int, azure_ad_token: str | None = None, azure_ad_token_provider: Callable | None = None, @@ -634,7 +634,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): except Exception as e: status_code: Final = getattr(e, "status_code", 500) error_headers = getattr(e, "headers", None) - error_response: Final = getattr(e, "response", None) + error_response: Final[object] = getattr(e, "response", None) message: Final = getattr(e, "message", str(e)) error_body: Final = getattr(e, "body", None) if error_headers is None and error_response: @@ -754,7 +754,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): aembedding=None, headers: dict | None = None, litellm_params: dict | None = None, - ) -> EmbeddingResponse | Coroutine[Any, Any, EmbeddingResponse]: + ) -> EmbeddingResponse | Coroutine[object, object, EmbeddingResponse]: if headers: optional_params["extra_headers"] = headers if self._client_session is None: @@ -1268,7 +1268,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): headers["Authorization"] = f"Bearer {azure_ad_token}" # init AzureOpenAI Client - azure_client_params: Final[dict[str, Any]] = self.initialize_azure_sdk_client( + azure_client_params: Final[dict[str, object]] = self.initialize_azure_sdk_client( litellm_params=litellm_params or {}, api_key=api_key, model_name=model or "", diff --git a/litellm/llms/azure_ai/agents/handler.py b/litellm/llms/azure_ai/agents/handler.py index a13b1300e55..f7382190fca 100644 --- a/litellm/llms/azure_ai/agents/handler.py +++ b/litellm/llms/azure_ai/agents/handler.py @@ -51,15 +51,13 @@ else: AsyncHTTPHandler = Any -class _AzureRawAnnotation(TypedDict, total=False): - type: ReadOnly[str] +class _AzureRawAnnotation(ChatCompletionAnnotation, total=False): text: ReadOnly[str] start_index: ReadOnly[int] end_index: ReadOnly[int] - url_citation: ReadOnly[ChatCompletionAnnotationURLCitation] -_TransformedAnnotation: TypeAlias = ChatCompletionAnnotation | _AzureRawAnnotation +_TransformedAnnotation: TypeAlias = ChatCompletionAnnotation class _AzureText(TypedDict, total=False): @@ -223,18 +221,11 @@ class AzureAIAgentsHandler: """Build the ModelResponse from agent output.""" from litellm.types.utils import Choices, Message, Usage - message_kwargs: Final[dict[str, Any]] = { - "content": content, - "role": "assistant", - } - if annotations: - message_kwargs["annotations"] = annotations - model_response.choices = [ Choices( finish_reason="stop", index=0, - message=Message(**message_kwargs), + message=Message(content=content, role="assistant", annotations=annotations or None), ) ] model_response.model = model @@ -655,9 +646,6 @@ class AzureAIAgentsHandler: if data_str == "[DONE]": # Send final chunk with finish_reason - final_delta_kwargs: dict[str, Any] = {"content": None} - if collected_annotations: - final_delta_kwargs["annotations"] = collected_annotations final_chunk = ModelResponseStream( id=response_id, created=created, @@ -667,7 +655,7 @@ class AzureAIAgentsHandler: StreamingChoices( finish_reason="stop", index=0, - delta=Delta(**final_delta_kwargs), + delta=Delta(content=None, annotations=collected_annotations or None), ) ], ) diff --git a/litellm/llms/base_llm/ocr/transformation.py b/litellm/llms/base_llm/ocr/transformation.py index d1c77186ea8..3b302837032 100644 --- a/litellm/llms/base_llm/ocr/transformation.py +++ b/litellm/llms/base_llm/ocr/transformation.py @@ -75,6 +75,7 @@ class OCRUsageInfo(LiteLLMPydanticObjectBase): """Usage information from OCR response.""" pages_processed: int | None = None + pages_processed_annotation: int | None = None credits: float | None = None doc_size_bytes: int | None = None diff --git a/litellm/llms/bedrock/chat/converse_transformation.py b/litellm/llms/bedrock/chat/converse_transformation.py index 395d99a4caa..6f99f572686 100644 --- a/litellm/llms/bedrock/chat/converse_transformation.py +++ b/litellm/llms/bedrock/chat/converse_transformation.py @@ -588,6 +588,10 @@ class AmazonConverseConfig(BaseConfig): supported_params.append("context_management") return supported_params + @staticmethod + def _auto_tool_choice() -> ToolChoiceValuesBlock: + return ToolChoiceValuesBlock(auto={}) + def map_tool_choice_values( self, model: str, tool_choice: str | dict, drop_params: bool ) -> ToolChoiceValuesBlock | None: @@ -600,10 +604,14 @@ class AmazonConverseConfig(BaseConfig): status_code=400, ) elif tool_choice == "required": + if AnthropicModelInfo.forced_tool_use_downgraded(model, drop_params): + return self._auto_tool_choice() return ToolChoiceValuesBlock(any={}) elif tool_choice == "auto": - return ToolChoiceValuesBlock(auto={}) + return self._auto_tool_choice() elif isinstance(tool_choice, dict): + if AnthropicModelInfo.forced_tool_use_downgraded(model, drop_params): + return self._auto_tool_choice() # only supported for anthropic + mistral models - https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_ToolChoice.html specific_tool: Final = SpecificToolChoiceBlock( name=make_valid_bedrock_tool_name(tool_choice.get("function", {}).get("name", "")) diff --git a/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py b/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py index 40b90014f3b..8e709349400 100644 --- a/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py +++ b/litellm/llms/bedrock/chat/invoke_transformations/anthropic_claude3_transformation.py @@ -3,7 +3,6 @@ from typing import TYPE_CHECKING, Any, Final import httpx from litellm.anthropic_beta_headers_manager import filter_and_transform_beta_headers -from litellm.litellm_core_utils.litellm_logging import verbose_logger from litellm.litellm_core_utils.prompt_templates.factory import ( convert_to_anthropic_image_obj, ) @@ -16,17 +15,16 @@ from litellm.llms.bedrock.chat.invoke_transformations.base_invoke_transformation AmazonInvokeConfig, ) from litellm.llms.bedrock.common_utils import ( - convert_bedrock_invoke_output_format_to_inline_schema, + apply_bedrock_invoke_structured_output, get_anthropic_beta_from_headers, normalize_bedrock_opus_output_config_effort, normalize_custom_field_on_tools, normalize_tool_input_schema_types_for_bedrock_invoke, - pop_bedrock_invoke_output_config_format, + strip_unsupported_bedrock_invoke_output_config_keys, ) from litellm.types.llms.anthropic import ANTHROPIC_TOOL_SEARCH_BETA_HEADER from litellm.types.llms.openai import AllMessageValues from litellm.types.utils import ModelResponse -from litellm.utils import _supports_factory if TYPE_CHECKING: import tiktoken @@ -212,36 +210,14 @@ class AmazonAnthropicClaudeConfig(AmazonInvokeConfig, AnthropicConfig): anthropic_request.pop("model", None) anthropic_request.pop("stream", None) anthropic_request.pop("stream_chunk_size", None) - output_format: Final = anthropic_request.pop("output_format", None) - output_config_format: Final = pop_bedrock_invoke_output_config_format(anthropic_request) - if output_format: - convert_bedrock_invoke_output_format_to_inline_schema( - output_format=output_format, - request_body=anthropic_request, - ) - elif output_config_format: - convert_bedrock_invoke_output_format_to_inline_schema( - output_format=output_config_format, - request_body=anthropic_request, - ) - if not ( - _supports_factory( - model=model, - custom_llm_provider="bedrock", - key="supports_output_config", - ) - or AnthropicConfig._model_supports_effort_param(model, "bedrock") - ): - if anthropic_request.pop("output_config", None) is not None: - verbose_logger.warning( - "Bedrock Invoke: stripping unsupported `output_config` for " - "model=%s — neither `supports_output_config` nor any " - "`supports_*_reasoning_effort` flag is set in " - "model_prices_and_context_window.json. Add the capability " - "flag to the model JSON entry if this model accepts " - "`output_config`.", - model, - ) + apply_bedrock_invoke_structured_output( + model=model, + request_body=anthropic_request, + ) + strip_unsupported_bedrock_invoke_output_config_keys( + model=model, + request_body=anthropic_request, + ) if "anthropic_version" not in anthropic_request: anthropic_request["anthropic_version"] = self.anthropic_version diff --git a/litellm/llms/bedrock/common_utils.py b/litellm/llms/bedrock/common_utils.py index 9cbceb4880c..30a77d57f24 100644 --- a/litellm/llms/bedrock/common_utils.py +++ b/litellm/llms/bedrock/common_utils.py @@ -177,6 +177,95 @@ def convert_bedrock_invoke_output_format_to_inline_schema( request_body["messages"] = new_messages +def _bedrock_model_supports(model: str, key: str) -> bool: + from litellm.utils import _supports_factory + + return _supports_factory(model=model, custom_llm_provider="bedrock", key=key) + + +def apply_bedrock_invoke_structured_output( + model: str, + request_body: dict[str, object], # mutable-ok: edited in place like siblings +) -> None: + """ + Route Anthropic structured-output params to what the Bedrock model supports. + + Consumes the legacy top-level ``output_format`` and the newer + ``output_config.format``, keeping the pre-existing precedence of the legacy + field when a request carries both. Models flagged + ``supports_native_structured_output`` in the model map get the schema + forwarded as ``output_config.format``, which Bedrock relays to the model for + enforced structured output. For every other model the schema is inlined into + the last user message as best-effort text, with a warning because nothing + enforces it. + """ + legacy_output_format: Final = request_body.pop("output_format", None) + output_config_format: Final = pop_bedrock_invoke_output_config_format(request_body) + schema_format: Final = legacy_output_format if isinstance(legacy_output_format, dict) else output_config_format + if schema_format is None: + return + + if _bedrock_model_supports(model, "supports_native_structured_output"): + existing_output_config: Final = request_body.get("output_config") + if isinstance(existing_output_config, dict): + existing_output_config["format"] = schema_format + else: + request_body["output_config"] = {"format": schema_format} # rebind-ok: out-param # mutable-ok: json + return + + verbose_logger.warning( + "Bedrock Invoke: model=%s does not advertise `supports_native_structured_output` " + "in model_prices_and_context_window.json, so the JSON schema was inlined into " + "the last user message and is NOT enforced by the model.", + model, + ) + convert_bedrock_invoke_output_format_to_inline_schema( + output_format=schema_format, + request_body=request_body, + ) + + +def strip_unsupported_bedrock_invoke_output_config_keys( + model: str, + request_body: dict[str, object], # mutable-ok: edited in place like siblings +) -> None: + """ + Drop ``output_config`` keys the Bedrock model does not accept. + + ``format`` survives unconditionally: it is only attached for models whose map + entry advertises ``supports_native_structured_output``. Effort-bearing keys + survive only when the map flags ``supports_output_config`` or a + ``supports_*_reasoning_effort`` tier; otherwise they are dropped with a + warning so Bedrock does not reject the request. + """ + from litellm.llms.anthropic.chat.transformation import AnthropicConfig + + output_config: Final = request_body.get("output_config") + if not isinstance(output_config, dict): + return + if all(key == "format" for key in output_config): + return + if _bedrock_model_supports(model, "supports_output_config") or AnthropicConfig._model_supports_effort_param( + model, "bedrock" + ): + return + + verbose_logger.warning( + "Bedrock Invoke: stripping unsupported `output_config` keys for " + "model=%s: neither `supports_output_config` nor any " + "`supports_*_reasoning_effort` flag is set in " + "model_prices_and_context_window.json. Add the capability " + "flag to the model JSON entry if this model accepts " + "`output_config`.", + model, + ) + preserved_format: Final = output_config.get("format") + if preserved_format is None: + request_body.pop("output_config", None) + else: + request_body["output_config"] = {"format": preserved_format} # rebind-ok: out-param # mutable-ok: json + + def normalize_custom_field_on_tools(request_body: dict) -> None: """ Drop the ``custom`` field from each tool, first hoisting a boolean diff --git a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py index f74a290d773..6ff9f0155f9 100644 --- a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py +++ b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py @@ -29,14 +29,14 @@ from litellm.llms.bedrock.chat.invoke_transformations.base_invoke_transformation AmazonInvokeConfig, ) from litellm.llms.bedrock.common_utils import ( - convert_bedrock_invoke_output_format_to_inline_schema, + apply_bedrock_invoke_structured_output, ensure_bedrock_anthropic_messages_tool_names, get_anthropic_beta_from_headers, is_claude_4_5_on_bedrock, normalize_bedrock_opus_output_config_effort, normalize_custom_field_on_tools, normalize_tool_input_schema_types_for_bedrock_invoke, - pop_bedrock_invoke_output_config_format, + strip_unsupported_bedrock_invoke_output_config_keys, ) from litellm.llms.bedrock.request_metadata import ( bedrock_request_metadata_headers, @@ -51,7 +51,6 @@ from litellm.types.llms.openai import AllMessageValues from litellm.types.router import GenericLiteLLMParams from litellm.types.utils import GenericStreamingChunk, ModelResponseStream from litellm.types.utils import GenericStreamingChunk as GChunk -from litellm.utils import _supports_factory if TYPE_CHECKING: from litellm.litellm_core_utils.litellm_logging import Logging as _LiteLLMLoggingObj @@ -708,52 +707,25 @@ class AmazonAnthropicClaudeMessagesConfig( # 4. Remove `ttl` field from cache_control in messages (Bedrock doesn't support it for older models) self._remove_ttl_from_cache_control(anthropic_messages_request=anthropic_messages_request, model=model) - # 5. Convert structured-output params to inline schema. - # Bedrock Invoke doesn't support top-level `output_format`; its - # accepted `output_config` subset is also narrower than Anthropic's, so - # consume the newer `output_config.format` shape here instead of - # forwarding it as an unknown nested key. + # 5. Route structured-output params (`output_format` / + # `output_config.format`) to native enforcement or the inline-schema + # fallback, then strip `output_config` keys the model does not accept. + # Ref: https://github.com/BerriAI/litellm/issues/22797 existing_output_config: Final = anthropic_messages_request.get("output_config") if isinstance(existing_output_config, dict): anthropic_messages_request["output_config"] = dict(existing_output_config) - output_format: Final = anthropic_messages_request.pop("output_format", None) - output_config_format: Final = pop_bedrock_invoke_output_config_format(anthropic_messages_request) - if output_format: - convert_bedrock_invoke_output_format_to_inline_schema( - output_format=output_format, - request_body=anthropic_messages_request, - ) - elif output_config_format: - convert_bedrock_invoke_output_format_to_inline_schema( - output_format=output_config_format, - request_body=anthropic_messages_request, - ) + apply_bedrock_invoke_structured_output( + model=model, + request_body=anthropic_messages_request, + ) normalize_bedrock_opus_output_config_effort( model=model, output_config=anthropic_messages_request.get("output_config"), ) - - # 5a. Bedrock Invoke supports output_config (effort) for Claude 4.6+ models, - # but older models do not — strip it to avoid request rejection. - # Ref: https://github.com/BerriAI/litellm/issues/22797 - if not ( - _supports_factory( - model=model, - custom_llm_provider="bedrock", - key="supports_output_config", - ) - or AnthropicConfig._model_supports_effort_param(model, "bedrock") - ): - if anthropic_messages_request.pop("output_config", None) is not None: - verbose_logger.warning( - "Bedrock Invoke: stripping unsupported `output_config` for " - "model=%s — neither `supports_output_config` nor any " - "`supports_*_reasoning_effort` flag is set in " - "model_prices_and_context_window.json. Add the capability " - "flag to the model JSON entry if this model accepts " - "`output_config`.", - model, - ) + strip_unsupported_bedrock_invoke_output_config_keys( + model=model, + request_body=anthropic_messages_request, + ) # 5b. Hoist `custom.defer_loading` then drop `custom` (Bedrock doesn't support it) # Ref: https://github.com/BerriAI/litellm/issues/22847 @@ -774,9 +746,11 @@ class AmazonAnthropicClaudeMessagesConfig( if filtered_betas: anthropic_messages_request["anthropic_beta"] = filtered_betas + remaining_output_config: Final = anthropic_messages_request.get("output_config") if ( litellm.drop_params is True - and "output_config" in anthropic_messages_request + and isinstance(remaining_output_config, dict) + and any(key != "format" for key in remaining_output_config) and not AnthropicConfig._model_supports_effort_param(model, "bedrock") ): verbose_logger.warning( diff --git a/litellm/llms/bedrock/realtime/transformation.py b/litellm/llms/bedrock/realtime/transformation.py index 28c2e446d10..1f4c81d6491 100644 --- a/litellm/llms/bedrock/realtime/transformation.py +++ b/litellm/llms/bedrock/realtime/transformation.py @@ -7,7 +7,7 @@ Transforms between OpenAI Realtime API format and Bedrock Nova Sonic format. import base64 import json import uuid as uuid_lib -from typing import Any, Final, cast +from typing import Final, cast from pydantic import BaseModel @@ -633,7 +633,7 @@ class BedrockRealtimeConfig(BaseRealtimeConfig): List of Bedrock format messages (JSON strings) """ try: - json_message: Final = json.loads(message) + json_message: Final[dict[str, object]] = json.loads(message) except json.JSONDecodeError: verbose_logger.warning("Invalid JSON message: %s", message[:200]) return [] @@ -1182,7 +1182,7 @@ class BedrockRealtimeConfig(BaseRealtimeConfig): # Create a function call arguments done event # This is a custom event format that matches what clients expect - function_call_event: Final[dict[str, Any]] = { + function_call_event: Final[dict[str, object]] = { "type": "response.function_call_arguments.done", "event_id": f"event_{uuid.uuid4()}", "response_id": current_response_id, diff --git a/litellm/llms/black_forest_labs/image_edit/handler.py b/litellm/llms/black_forest_labs/image_edit/handler.py index 1ff02a6f8d9..178acb0de0d 100644 --- a/litellm/llms/black_forest_labs/image_edit/handler.py +++ b/litellm/llms/black_forest_labs/image_edit/handler.py @@ -8,9 +8,11 @@ then we poll until the result is ready. import asyncio import time -from typing import Any, Final +from collections.abc import Coroutine, Mapping +from typing import Final, Protocol import httpx +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -33,6 +35,42 @@ from ..common_utils import ( from .transformation import BlackForestLabsImageEditConfig +class _BFLSubmitBody(TypedDict, total=False): + """Decoded body of the BFL submit response, which hands back a polling URL.""" + + errors: ReadOnly[object] + polling_url: ReadOnly[str] + + +class _BFLPollBody(TypedDict, total=False): + """Decoded body of a BFL polling response.""" + + status: ReadOnly[str] + + +class _BFLSubmitResponse(Protocol): + """The submit call's HTTP response, read for its status, body text and decoded body.""" + + @property + def status_code(self) -> int: ... + + @property + def text(self) -> str: ... + + def json(self) -> _BFLSubmitBody: ... + + +class _BFLPollResponse(Protocol): + """A polling call's HTTP response, read only for the task status it carries.""" + + def json(self) -> _BFLPollBody: ... + + +def _poll_status(response: _BFLPollResponse) -> str | None: + """Read the task status out of a BFL polling response body.""" + return response.json().get("status") + + class BlackForestLabsImageEdit: """ Black Forest Labs Image Edit handler. @@ -53,10 +91,10 @@ class BlackForestLabsImageEdit: litellm_params: GenericLiteLLMParams | dict, logging_obj: LiteLLMLoggingObj, timeout: float | httpx.Timeout | None, - extra_headers: dict[str, Any] | None = None, + extra_headers: Mapping[str, object] | None = None, client: HTTPHandler | AsyncHTTPHandler | None = None, aimage_edit: bool = False, - ) -> ImageResponse | Any: + ) -> ImageResponse | Coroutine[object, object, ImageResponse]: """ Main entry point for image edit requests. @@ -185,7 +223,7 @@ class BlackForestLabsImageEdit: litellm_params: GenericLiteLLMParams | dict, logging_obj: LiteLLMLoggingObj, timeout: float | httpx.Timeout | None, - extra_headers: dict[str, Any] | None = None, + extra_headers: Mapping[str, object] | None = None, client: AsyncHTTPHandler | None = None, ) -> ImageResponse: """ @@ -281,7 +319,7 @@ class BlackForestLabsImageEdit: def _poll_for_result_sync( self, - initial_response: httpx.Response, + initial_response: _BFLSubmitResponse, headers: dict, sync_client: HTTPHandler, max_wait: float = DEFAULT_MAX_POLLING_TIME, @@ -356,8 +394,7 @@ class BlackForestLabsImageEdit: message=f"Polling failed: {response.text}", ) - data = response.json() - status = data.get("status") + status = _poll_status(response) verbose_logger.debug("BFL poll status: %s", status) @@ -383,7 +420,7 @@ class BlackForestLabsImageEdit: async def _poll_for_result_async( self, - initial_response: httpx.Response, + initial_response: _BFLSubmitResponse, headers: dict, async_client: AsyncHTTPHandler, max_wait: float = DEFAULT_MAX_POLLING_TIME, @@ -447,8 +484,7 @@ class BlackForestLabsImageEdit: message=f"Polling failed: {response.text}", ) - data = response.json() - status = data.get("status") + status = _poll_status(response) verbose_logger.debug("BFL poll status: %s", status) diff --git a/litellm/llms/black_forest_labs/image_generation/handler.py b/litellm/llms/black_forest_labs/image_generation/handler.py index 03e4999c5aa..879bef37b58 100644 --- a/litellm/llms/black_forest_labs/image_generation/handler.py +++ b/litellm/llms/black_forest_labs/image_generation/handler.py @@ -8,9 +8,11 @@ then we poll until the result is ready. import asyncio import time -from typing import Any, Final +from collections.abc import Coroutine, Mapping +from typing import Final, Protocol, TypedDict import httpx +from typing_extensions import ReadOnly import litellm from litellm._logging import verbose_logger @@ -33,6 +35,23 @@ from ..common_utils import ( from .transformation import BlackForestLabsImageGenerationConfig +class _BFLTaskPayload(TypedDict, total=False): + """The body BFL returns for a submitted or polled generation task.""" + + errors: ReadOnly[object] + polling_url: ReadOnly[str] + status: ReadOnly[str] + + +class _TaskJsonResponse(Protocol): + def json(self) -> _BFLTaskPayload: ... + + +def _task_payload(response: _TaskJsonResponse) -> _BFLTaskPayload: + """The JSON body of a BFL task submission or poll response.""" + return response.json() + + class BlackForestLabsImageGeneration: """ Black Forest Labs Image Generation handler. @@ -53,10 +72,10 @@ class BlackForestLabsImageGeneration: litellm_params: GenericLiteLLMParams | dict, logging_obj: LiteLLMLoggingObj, timeout: float | httpx.Timeout | None, - extra_headers: dict[str, Any] | None = None, + extra_headers: Mapping[str, str] | None = None, client: HTTPHandler | AsyncHTTPHandler | None = None, aimg_generation: bool = False, - ) -> ImageResponse | Any: + ) -> ImageResponse | Coroutine[object, object, ImageResponse]: """ Main entry point for image generation requests. @@ -187,7 +206,7 @@ class BlackForestLabsImageGeneration: litellm_params: GenericLiteLLMParams | dict, logging_obj: LiteLLMLoggingObj, timeout: float | httpx.Timeout | None, - extra_headers: dict[str, Any] | None = None, + extra_headers: Mapping[str, str] | None = None, client: AsyncHTTPHandler | None = None, ) -> ImageResponse: """ @@ -305,7 +324,7 @@ class BlackForestLabsImageGeneration: # Parse initial response to get polling URL try: - response_data: Final = initial_response.json() + response_data: Final = _task_payload(initial_response) except Exception as e: raise BlackForestLabsError( status_code=initial_response.status_code, @@ -350,7 +369,7 @@ class BlackForestLabsImageGeneration: message=f"Polling failed: {response.text}", ) - data = response.json() + data = _task_payload(response) status = data.get("status") verbose_logger.debug("BFL poll status: %s", status) @@ -396,7 +415,7 @@ class BlackForestLabsImageGeneration: # Parse initial response to get polling URL try: - response_data: Final = initial_response.json() + response_data: Final = _task_payload(initial_response) except Exception as e: raise BlackForestLabsError( status_code=initial_response.status_code, @@ -441,7 +460,7 @@ class BlackForestLabsImageGeneration: message=f"Polling failed: {response.text}", ) - data = response.json() + data = _task_payload(response) status = data.get("status") verbose_logger.debug("BFL poll status: %s", status) diff --git a/litellm/llms/codestral/completion/handler.py b/litellm/llms/codestral/completion/handler.py index 8c08b2bc33c..f8486d3b274 100644 --- a/litellm/llms/codestral/completion/handler.py +++ b/litellm/llms/codestral/completion/handler.py @@ -4,9 +4,10 @@ import json from collections.abc import Callable from functools import partial -from typing import Final +from typing import Final, Protocol import httpx +from typing_extensions import NotRequired, ReadOnly, TypedDict import litellm from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLogging @@ -23,6 +24,53 @@ from litellm.types.utils import TextChoices from litellm.utils import CustomStreamWrapper, TextCompletionResponse +class _CodestralChoiceMessage(TypedDict): + """`choices[].message` of a Codestral FIM completion.""" + + role: ReadOnly[NotRequired[str]] + content: ReadOnly[NotRequired[str | None]] + + +class _CodestralChoice(TypedDict): + """One entry of `choices` in a Codestral FIM completion.""" + + index: ReadOnly[int] + message: ReadOnly[NotRequired[_CodestralChoiceMessage]] + finish_reason: ReadOnly[NotRequired[str | None]] + logprobs: ReadOnly[NotRequired[dict[str, object] | None]] + + +class _CodestralUsage(TypedDict): + """Token accounting returned alongside a Codestral FIM completion.""" + + prompt_tokens: ReadOnly[NotRequired[int]] + completion_tokens: ReadOnly[NotRequired[int]] + total_tokens: ReadOnly[NotRequired[int]] + + +class _CodestralCompletionResponse(TypedDict): + """Body returned by the Codestral `/v1/fim/completions` endpoint.""" + + id: ReadOnly[NotRequired[str]] + created: ReadOnly[NotRequired[int]] + model: ReadOnly[NotRequired[str]] + object: ReadOnly[NotRequired[str]] + usage: ReadOnly[NotRequired[_CodestralUsage]] + choices: ReadOnly[NotRequired[list[_CodestralChoice]]] + + +class _CodestralHTTPResponse(Protocol): + """The Codestral completion response as this handler reads it.""" + + @property + def status_code(self) -> int: ... + + @property + def text(self) -> str: ... + + def json(self) -> _CodestralCompletionResponse: ... + + class TextCompletionCodestralError(Exception): def __init__( self, @@ -115,7 +163,7 @@ class CodestralTextCompletion: def process_text_completion_response( self, model: str, - response: httpx.Response, + response: _CodestralHTTPResponse, model_response: TextCompletionResponse, stream: bool, logging_obj: LiteLLMLogging, diff --git a/litellm/llms/dashscope/chat/transformation.py b/litellm/llms/dashscope/chat/transformation.py index 5ab7fbf3658..26e60fa959d 100644 --- a/litellm/llms/dashscope/chat/transformation.py +++ b/litellm/llms/dashscope/chat/transformation.py @@ -54,6 +54,9 @@ class DashScopeChatConfig(OpenAIGPTConfig): dynamic_api_key: Final = api_key or get_secret_str("DASHSCOPE_API_KEY") return api_base, dynamic_api_key + def _resolve_chat_api_base(self, api_base: str | None) -> str: + return api_base or "https://dashscope.aliyuncs.com/compatible-mode/v1" + def get_complete_url( self, api_base: str | None, @@ -66,10 +69,7 @@ class DashScopeChatConfig(OpenAIGPTConfig): """ If api_base is not provided, use the default DashScope /chat/completions endpoint. """ - if not api_base: - api_base = "https://dashscope.aliyuncs.com/compatible-mode/v1" - - if not api_base.endswith("/chat/completions"): - api_base = f"{api_base}/chat/completions" - - return api_base + resolved_api_base: Final = self._resolve_chat_api_base(api_base) + if resolved_api_base.endswith("/chat/completions"): + return resolved_api_base + return f"{resolved_api_base}/chat/completions" diff --git a/litellm/llms/dashscope/common_utils.py b/litellm/llms/dashscope/common_utils.py index 9a7dd4da8d3..b7c97893a15 100644 --- a/litellm/llms/dashscope/common_utils.py +++ b/litellm/llms/dashscope/common_utils.py @@ -2,9 +2,89 @@ Common utilities for the DashScope LLM provider. """ +from typing import TYPE_CHECKING + import httpx from litellm.llms.base_llm.chat.transformation import BaseLLMException +from litellm.secret_managers.main import get_secret_str + +if TYPE_CHECKING: + from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig + from litellm.llms.base_llm.image_generation.transformation import ( + BaseImageGenerationConfig, + ) + from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig + + +def get_dashscope_family_embedding_config(custom_llm_provider: str) -> "BaseEmbeddingConfig": + if custom_llm_provider == "qwencloud": + from litellm.llms.dashscope.qwencloud import QwenCloudEmbeddingConfig + + return QwenCloudEmbeddingConfig() + if custom_llm_provider == "qwen_ai_platform": + from litellm.llms.dashscope.qwen_ai_platform import ( + QwenAIPlatformEmbeddingConfig, + ) + + return QwenAIPlatformEmbeddingConfig() + from litellm.llms.dashscope.embed.transformation import DashScopeEmbeddingConfig + + return DashScopeEmbeddingConfig() + + +def get_dashscope_family_rerank_config(custom_llm_provider: str) -> "BaseRerankConfig": + if custom_llm_provider == "qwencloud": + from litellm.llms.dashscope.qwencloud import QwenCloudRerankConfig + + return QwenCloudRerankConfig() + if custom_llm_provider == "qwen_ai_platform": + from litellm.llms.dashscope.qwen_ai_platform import QwenAIPlatformRerankConfig + + return QwenAIPlatformRerankConfig() + from litellm.llms.dashscope.rerank.transformation import DashScopeRerankConfig + + return DashScopeRerankConfig() + + +def get_dashscope_family_image_generation_config( + custom_llm_provider: str, +) -> "BaseImageGenerationConfig": + if custom_llm_provider == "qwencloud": + from litellm.llms.dashscope.qwencloud import QwenCloudImageGenerationConfig + + return QwenCloudImageGenerationConfig() + if custom_llm_provider == "qwen_ai_platform": + from litellm.llms.dashscope.qwen_ai_platform import ( + QwenAIPlatformImageGenerationConfig, + ) + + return QwenAIPlatformImageGenerationConfig() + from litellm.llms.dashscope.image_generation.transformation import ( + DashScopeImageGenerationConfig, + ) + + return DashScopeImageGenerationConfig() + + +def resolve_dashscope_family_api_key(custom_llm_provider: str, api_key: str | None) -> str | None: + if custom_llm_provider == "dashscope": + return api_key or get_secret_str("DASHSCOPE_API_KEY") + return api_key or get_secret_str(f"{custom_llm_provider.upper()}_API_KEY") or get_secret_str("DASHSCOPE_API_KEY") + + +def missing_dashscope_family_key_message(custom_llm_provider: str) -> str: + if custom_llm_provider == "qwencloud": + return ( + "Missing API key for QwenCloud. Set QWENCLOUD_API_KEY or " + "DASHSCOPE_API_KEY environment variable or pass api_key parameter." + ) + if custom_llm_provider == "qwen_ai_platform": + return ( + "Missing API key for Qwen AI Platform. Set QWEN_AI_PLATFORM_API_KEY or " + "DASHSCOPE_API_KEY environment variable or pass api_key parameter." + ) + return "Missing API key for DashScope. Set DASHSCOPE_API_KEY environment variable or pass api_key parameter." class DashScopeError(BaseLLMException): diff --git a/litellm/llms/dashscope/cost_calculator.py b/litellm/llms/dashscope/cost_calculator.py index 771ce140f66..dd5bee1fe8b 100644 --- a/litellm/llms/dashscope/cost_calculator.py +++ b/litellm/llms/dashscope/cost_calculator.py @@ -110,7 +110,7 @@ def _calculate_completion_cost( return (breakdown.completion_tokens * output_cost) + (breakdown.reasoning_tokens * reasoning_cost) -def cost_per_token(model: str, usage: Usage) -> tuple[float, float]: +def cost_per_token(model: str, usage: Usage, custom_llm_provider: str = "dashscope") -> tuple[float, float]: """ Calculate cost per token for Dashscope models. @@ -119,11 +119,12 @@ def cost_per_token(model: str, usage: Usage) -> tuple[float, float]: Args: model: Model name without provider prefix usage: LiteLLM Usage block + custom_llm_provider: The provider id the request resolved to; dashscope or one of its brand aliases Returns: Tuple[float, float] - (prompt_cost_in_usd, completion_cost_in_usd) """ - model_info: Final = get_model_info(model=model, custom_llm_provider="dashscope") + model_info: Final = get_model_info(model=model, custom_llm_provider=custom_llm_provider) breakdown: Final = _extract_token_breakdown(usage) raw_tiers: Final = model_info.get("tiered_pricing") tiered_pricing: Final = raw_tiers if isinstance(raw_tiers, list) else None diff --git a/litellm/llms/dashscope/embed/transformation.py b/litellm/llms/dashscope/embed/transformation.py index 6d13f1e53f7..63ee984a65c 100644 --- a/litellm/llms/dashscope/embed/transformation.py +++ b/litellm/llms/dashscope/embed/transformation.py @@ -62,6 +62,17 @@ class DashScopeEmbeddingConfig(BaseEmbeddingConfig): # for drop_params=False before this method is called. return optional_params + def _resolve_api_key(self, api_key: str | None) -> str: + resolved_api_key: Final = api_key if api_key is not None else get_secret_str("DASHSCOPE_API_KEY") + if resolved_api_key is None: + raise ValueError( + "DashScope API key is required. Set 'DASHSCOPE_API_KEY' env var or pass api_key explicitly." + ) + return resolved_api_key + + def _resolve_embedding_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("DASHSCOPE_API_BASE") or DEFAULT_API_BASE + def validate_environment( self, headers: dict, @@ -72,17 +83,11 @@ class DashScopeEmbeddingConfig(BaseEmbeddingConfig): api_key: str | None = None, api_base: str | None = None, ) -> dict: - if api_key is None: - api_key = get_secret_str("DASHSCOPE_API_KEY") - if api_key is None: - raise ValueError( - "DashScope API key is required. Set 'DASHSCOPE_API_KEY' env var or pass api_key explicitly." - ) - default_headers: Final = { + return { "Content-Type": "application/json", - "Authorization": f"Bearer {api_key}", + "Authorization": f"Bearer {self._resolve_api_key(api_key)}", + **headers, } - return {**default_headers, **headers} def get_complete_url( self, @@ -93,8 +98,7 @@ class DashScopeEmbeddingConfig(BaseEmbeddingConfig): litellm_params: dict, stream: bool | None = None, ) -> str: - base = api_base or get_secret_str("DASHSCOPE_API_BASE") or DEFAULT_API_BASE - base = base.rstrip("/") + base: Final = self._resolve_embedding_api_base(api_base).rstrip("/") if base.endswith("/embeddings"): return base return f"{base}/embeddings" diff --git a/litellm/llms/dashscope/image_generation/transformation.py b/litellm/llms/dashscope/image_generation/transformation.py index a7f0e98865f..c0e278a96ef 100644 --- a/litellm/llms/dashscope/image_generation/transformation.py +++ b/litellm/llms/dashscope/image_generation/transformation.py @@ -91,6 +91,15 @@ class DashScopeImageGenerationConfig(BaseImageGenerationConfig): mapped[k] = v return mapped + def _resolve_api_key(self, api_key: str | None) -> str: + resolved_api_key: Final = api_key or get_secret_str("DASHSCOPE_API_KEY") + if not resolved_api_key: + raise ValueError("DASHSCOPE_API_KEY is not set") + return resolved_api_key + + def _resolve_image_api_base(self, image_api_base: str | None) -> str: + return image_api_base or get_secret_str("DASHSCOPE_API_BASE_IMAGE") or DEFAULT_API_BASE + def get_complete_url( self, api_base: str | None, @@ -103,7 +112,7 @@ class DashScopeImageGenerationConfig(BaseImageGenerationConfig): image_api_base: Final = ( api_base if api_base and not api_base.rstrip("/").endswith(CHAT_COMPATIBLE_MODE_PATH) else None ) - return image_api_base or get_secret_str("DASHSCOPE_API_BASE_IMAGE") or DEFAULT_API_BASE + return self._resolve_image_api_base(image_api_base) def validate_environment( self, @@ -115,10 +124,7 @@ class DashScopeImageGenerationConfig(BaseImageGenerationConfig): api_key: str | None = None, api_base: str | None = None, ) -> dict: - final_api_key: Final = api_key or get_secret_str("DASHSCOPE_API_KEY") - if not final_api_key: - raise ValueError("DASHSCOPE_API_KEY is not set") - headers["Authorization"] = f"Bearer {final_api_key}" + headers["Authorization"] = f"Bearer {self._resolve_api_key(api_key)}" headers["Content-Type"] = "application/json" return headers diff --git a/litellm/llms/dashscope/qwen_ai_platform.py b/litellm/llms/dashscope/qwen_ai_platform.py new file mode 100644 index 00000000000..9a44eaf574a --- /dev/null +++ b/litellm/llms/dashscope/qwen_ai_platform.py @@ -0,0 +1,62 @@ +from typing import Final + +from litellm.secret_managers.main import get_secret_str + +from .chat.transformation import DashScopeChatConfig +from .embed.transformation import DashScopeEmbeddingConfig +from .image_generation.transformation import DashScopeImageGenerationConfig +from .rerank.transformation import DashScopeRerankConfig + +QWEN_AI_PLATFORM_API_BASE: Final = "https://dashscope.aliyuncs.com/compatible-mode/v1" +QWEN_AI_PLATFORM_RERANK_API_BASE: Final = "https://dashscope.aliyuncs.com/compatible-api/v1/reranks" +QWEN_AI_PLATFORM_IMAGE_API_BASE: Final = ( + "https://dashscope.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation" +) + + +def _resolve_qwen_ai_platform_api_key(api_key: str | None) -> str | None: + return api_key or get_secret_str("QWEN_AI_PLATFORM_API_KEY") or get_secret_str("DASHSCOPE_API_KEY") + + +def _require_qwen_ai_platform_api_key(api_key: str | None) -> str: + resolved: Final = _resolve_qwen_ai_platform_api_key(api_key) + if resolved is None: + raise ValueError( + "Qwen AI Platform API key is required. Set 'QWEN_AI_PLATFORM_API_KEY' or 'DASHSCOPE_API_KEY' env var " + "or pass api_key explicitly." + ) + return resolved + + +class QwenAIPlatformChatConfig(DashScopeChatConfig): + def _get_openai_compatible_provider_info( + self, api_base: str | None, api_key: str | None + ) -> tuple[str | None, str | None]: + return self._resolve_chat_api_base(api_base), _resolve_qwen_ai_platform_api_key(api_key) + + def _resolve_chat_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("QWEN_AI_PLATFORM_API_BASE") or QWEN_AI_PLATFORM_API_BASE + + +class QwenAIPlatformEmbeddingConfig(DashScopeEmbeddingConfig): + def _resolve_api_key(self, api_key: str | None) -> str: + return _require_qwen_ai_platform_api_key(api_key) + + def _resolve_embedding_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("QWEN_AI_PLATFORM_API_BASE") or QWEN_AI_PLATFORM_API_BASE + + +class QwenAIPlatformRerankConfig(DashScopeRerankConfig): + def _resolve_api_key(self, api_key: str | None) -> str: + return _require_qwen_ai_platform_api_key(api_key) + + def _resolve_rerank_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("QWEN_AI_PLATFORM_API_BASE_RERANK") or QWEN_AI_PLATFORM_RERANK_API_BASE + + +class QwenAIPlatformImageGenerationConfig(DashScopeImageGenerationConfig): + def _resolve_api_key(self, api_key: str | None) -> str: + return _require_qwen_ai_platform_api_key(api_key) + + def _resolve_image_api_base(self, image_api_base: str | None) -> str: + return image_api_base or get_secret_str("QWEN_AI_PLATFORM_API_BASE_IMAGE") or QWEN_AI_PLATFORM_IMAGE_API_BASE diff --git a/litellm/llms/dashscope/qwencloud.py b/litellm/llms/dashscope/qwencloud.py new file mode 100644 index 00000000000..d8d53e340ef --- /dev/null +++ b/litellm/llms/dashscope/qwencloud.py @@ -0,0 +1,62 @@ +from typing import Final + +from litellm.secret_managers.main import get_secret_str + +from .chat.transformation import DashScopeChatConfig +from .embed.transformation import DashScopeEmbeddingConfig +from .image_generation.transformation import DashScopeImageGenerationConfig +from .rerank.transformation import DashScopeRerankConfig + +QWENCLOUD_API_BASE: Final = "https://dashscope-intl.aliyuncs.com/compatible-mode/v1" +QWENCLOUD_RERANK_API_BASE: Final = "https://dashscope-intl.aliyuncs.com/compatible-api/v1/reranks" +QWENCLOUD_IMAGE_API_BASE: Final = ( + "https://dashscope-intl.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation" +) + + +def _resolve_qwencloud_api_key(api_key: str | None) -> str | None: + return api_key or get_secret_str("QWENCLOUD_API_KEY") or get_secret_str("DASHSCOPE_API_KEY") + + +def _require_qwencloud_api_key(api_key: str | None) -> str: + resolved: Final = _resolve_qwencloud_api_key(api_key) + if resolved is None: + raise ValueError( + "QwenCloud API key is required. Set 'QWENCLOUD_API_KEY' or 'DASHSCOPE_API_KEY' env var " + "or pass api_key explicitly." + ) + return resolved + + +class QwenCloudChatConfig(DashScopeChatConfig): + def _get_openai_compatible_provider_info( + self, api_base: str | None, api_key: str | None + ) -> tuple[str | None, str | None]: + return self._resolve_chat_api_base(api_base), _resolve_qwencloud_api_key(api_key) + + def _resolve_chat_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("QWENCLOUD_API_BASE") or QWENCLOUD_API_BASE + + +class QwenCloudEmbeddingConfig(DashScopeEmbeddingConfig): + def _resolve_api_key(self, api_key: str | None) -> str: + return _require_qwencloud_api_key(api_key) + + def _resolve_embedding_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("QWENCLOUD_API_BASE") or QWENCLOUD_API_BASE + + +class QwenCloudRerankConfig(DashScopeRerankConfig): + def _resolve_api_key(self, api_key: str | None) -> str: + return _require_qwencloud_api_key(api_key) + + def _resolve_rerank_api_base(self, api_base: str | None) -> str: + return api_base or get_secret_str("QWENCLOUD_API_BASE_RERANK") or QWENCLOUD_RERANK_API_BASE + + +class QwenCloudImageGenerationConfig(DashScopeImageGenerationConfig): + def _resolve_api_key(self, api_key: str | None) -> str: + return _require_qwencloud_api_key(api_key) + + def _resolve_image_api_base(self, image_api_base: str | None) -> str: + return image_api_base or get_secret_str("QWENCLOUD_API_BASE_IMAGE") or QWENCLOUD_IMAGE_API_BASE diff --git a/litellm/llms/dashscope/rerank/transformation.py b/litellm/llms/dashscope/rerank/transformation.py index 98be4e4f2e7..3dd3996b2ee 100644 --- a/litellm/llms/dashscope/rerank/transformation.py +++ b/litellm/llms/dashscope/rerank/transformation.py @@ -58,19 +58,30 @@ class DashScopeRerankConfig(BaseRerankConfig): def __init__(self) -> None: pass + def _resolve_api_key(self, api_key: str | None) -> str: + resolved_api_key: Final = api_key if api_key is not None else get_secret_str("DASHSCOPE_API_KEY") + if resolved_api_key is None: + raise ValueError( + "DashScope API key is required. Set 'DASHSCOPE_API_KEY' env var or pass api_key explicitly." + ) + return resolved_api_key + + def _resolve_rerank_api_base(self, api_base: str | None) -> str: + if api_base is not None: + return api_base + return get_secret_str("DASHSCOPE_API_BASE_RERANK") or DEFAULT_RERANK_URL + def get_complete_url( self, api_base: str | None, model: str, optional_params: dict | None = None, ) -> str: - if api_base is None: - api_base = get_secret_str("DASHSCOPE_API_BASE_RERANK") or DEFAULT_RERANK_URL + resolved_api_base: Final = self._resolve_rerank_api_base(api_base) + if resolved_api_base == DEFAULT_RERANK_URL: + return resolved_api_base - if api_base == DEFAULT_RERANK_URL: - return DEFAULT_RERANK_URL - - cleaned: Final = api_base.rstrip("/") + cleaned: Final = resolved_api_base.rstrip("/") if cleaned.endswith("/reranks") or cleaned.endswith("/rerank"): return cleaned @@ -88,19 +99,12 @@ class DashScopeRerankConfig(BaseRerankConfig): optional_params: dict | None = None, litellm_params: Mapping[str, object] | None = None, ) -> dict: - if api_key is None: - api_key = get_secret_str("DASHSCOPE_API_KEY") - if api_key is None: - raise ValueError( - "DashScope API key is required. Set 'DASHSCOPE_API_KEY' env var or pass api_key explicitly." - ) - - default_headers: Final = { - "Authorization": f"Bearer {api_key}", + return { + "Authorization": f"Bearer {self._resolve_api_key(api_key)}", "accept": "application/json", "content-type": "application/json", + **headers, } - return {**default_headers, **headers} def get_supported_cohere_rerank_params(self, model: str) -> list: return ["query", "documents", "top_n", "return_documents"] diff --git a/litellm/llms/deepinfra/rerank/transformation.py b/litellm/llms/deepinfra/rerank/transformation.py index e52c56af82b..a3d0482af0a 100644 --- a/litellm/llms/deepinfra/rerank/transformation.py +++ b/litellm/llms/deepinfra/rerank/transformation.py @@ -2,10 +2,11 @@ Translate between Cohere's `/rerank` format and Deepinfra's `/rerank` format. """ -from collections.abc import Mapping -from typing import Any, Final +from collections.abc import Mapping, Sequence +from typing import Final, Protocol import httpx +from typing_extensions import ReadOnly, TypedDict from litellm._uuid import uuid from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj @@ -24,6 +25,36 @@ from litellm.types.rerank import ( ) +class _DeepinfraInferenceStatus(TypedDict, total=False): + """The ``inference_status`` block of a DeepInfra rerank response.""" + + status: ReadOnly[str] + runtime_ms: ReadOnly[float] + cost: ReadOnly[float] + tokens_generated: ReadOnly[int] + tokens_input: ReadOnly[int] + + +class _DeepinfraRerankResponse(TypedDict, total=False): + """Body of a DeepInfra ``/rerank`` response.""" + + scores: ReadOnly[Sequence[float]] + input_tokens: ReadOnly[int] + request_id: ReadOnly[str | None] + inference_status: ReadOnly[_DeepinfraInferenceStatus] + + +class _DeepinfraRerankResponseSource(Protocol): + """The DeepInfra ``/rerank`` HTTP response, read for the body it decodes to.""" + + def json(self) -> _DeepinfraRerankResponse: ... + + +def _deepinfra_rerank_body(response: _DeepinfraRerankResponseSource) -> _DeepinfraRerankResponse: + """Decode the body of a DeepInfra ``/rerank`` response.""" + return response.json() + + class DeepinfraRerankConfig(BaseRerankConfig): """ Deepinfra Rerank - Follows the same Spec as Cohere Rerank @@ -95,7 +126,7 @@ class DeepinfraRerankConfig(BaseRerankConfig): model: str, drop_params: bool, query: str, - documents: list[str | dict[str, Any]], + documents: list[str | dict[str, object]], custom_llm_provider: str | None = None, top_n: int | None = None, rank_fields: list[str] | None = None, @@ -150,7 +181,7 @@ class DeepinfraRerankConfig(BaseRerankConfig): litellm_params: dict = {}, ) -> RerankResponse: try: - response_json: Final = raw_response.json() + response_json: Final = _deepinfra_rerank_body(raw_response) logging_obj.post_call(original_response=raw_response.text) # Extract the scores from the response diff --git a/litellm/llms/gemini/interactions/transformation.py b/litellm/llms/gemini/interactions/transformation.py index dcd2e4e3471..6d0f211ed7b 100644 --- a/litellm/llms/gemini/interactions/transformation.py +++ b/litellm/llms/gemini/interactions/transformation.py @@ -12,9 +12,10 @@ Schema versioning: litellm.use_legacy_interactions_schema = True. Remove flag after June 8, 2026. """ -from typing import TYPE_CHECKING, Any, Final +from typing import TYPE_CHECKING, Any, Final, Protocol, TypeAlias import httpx +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -41,6 +42,53 @@ else: LiteLLMLoggingObj = Any +_JsonObject: TypeAlias = dict[str, object] + + +class _InteractionPayload(TypedDict, total=False): + """JSON body of an Interactions API interaction, keyed as ``InteractionsAPIResponse`` fields.""" + + id: ReadOnly[str | None] + object: ReadOnly[str | None] + model: ReadOnly[str | None] + agent: ReadOnly[str | None] + status: ReadOnly[str | None] + created: ReadOnly[str | None] + updated: ReadOnly[str | None] + outputs: ReadOnly[list[_JsonObject] | None] + steps: ReadOnly[list[_JsonObject] | None] + usage: ReadOnly[_JsonObject | None] + + +class _CancelPayload(TypedDict, total=False): + """JSON body of an Interactions API cancel response.""" + + id: ReadOnly[str | None] + status: ReadOnly[str | None] + + +class _InteractionPayloadSource(Protocol): + """An Interactions API HTTP response, read for the interaction body it decodes to.""" + + def json(self) -> _InteractionPayload: ... + + +class _CancelPayloadSource(Protocol): + """An Interactions API cancel HTTP response, read for the body it decodes to.""" + + def json(self) -> _CancelPayload: ... + + +def _interaction_body(response: _InteractionPayloadSource) -> _InteractionPayload: + """Decode the body of an Interactions API interaction response.""" + return response.json() + + +def _cancel_body(response: _CancelPayloadSource) -> _CancelPayload: + """Decode the body of an Interactions API cancel response.""" + return response.json() + + class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): """ Configuration for Google AI Studio Interactions API. @@ -143,7 +191,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): """ use_legacy: Final[bool] = litellm.use_legacy_interactions_schema - request_body: Final[dict[str, Any]] = {} + request_body: Final[dict[str, object]] = {} # Model or Agent (one required) if model: @@ -189,7 +237,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): and (not isinstance(response_format, dict) or "mime_type" not in response_format) ): # Wrap the legacy schema into the new polymorphic format. - new_rf: Final[dict[str, Any]] = { + new_rf: Final[dict[str, object]] = { "type": "text", "mime_type": response_mime_type, } @@ -215,7 +263,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): if image_config is not None: # Move image_config to response_format with type=image. - image_rf: Final[dict[str, Any]] = {"type": "image", **image_config} + image_rf: Final[_JsonObject] = {"type": "image", **image_config} existing_rf: Final = request_body.get("response_format") if existing_rf is None: request_body["response_format"] = image_rf @@ -239,7 +287,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): original_response=raw_response.text, additional_args={"complete_input_dict": {}}, ) - raw_json: Final = raw_response.json() + raw_json: Final = _interaction_body(raw_response) except Exception: raise GeminiError( message=raw_response.text, @@ -290,7 +338,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): logging_obj: LiteLLMLoggingObj, ) -> InteractionsAPIResponse: try: - raw_json: Final = raw_response.json() + raw_json: Final = _interaction_body(raw_response) except Exception: raise GeminiError( message=raw_response.text, @@ -355,7 +403,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig): logging_obj: LiteLLMLoggingObj, ) -> CancelInteractionResult: try: - raw_json: Final = raw_response.json() + raw_json: Final = _cancel_body(raw_response) except Exception: raise GeminiError( message=raw_response.text, diff --git a/litellm/llms/gemini/videos/transformation.py b/litellm/llms/gemini/videos/transformation.py index 6a1fc144c42..ff4c675b02f 100644 --- a/litellm/llms/gemini/videos/transformation.py +++ b/litellm/llms/gemini/videos/transformation.py @@ -1,4 +1,5 @@ import base64 +from collections.abc import Mapping from typing import TYPE_CHECKING, Any, Final import httpx @@ -54,8 +55,13 @@ def _convert_image_to_gemini_format(image_file) -> dict[str, str]: return {"bytesBase64Encoded": base64_encoded, "mimeType": mime_type} +def _json_payload(raw_response: httpx.Response) -> object: + """Read an HTTP response body as an opaque JSON payload.""" + return raw_response.json() + + def _usage_video_resolution_from_parameters( - parameters: dict[str, Any], + parameters: Mapping[str, object], ) -> str | None: """Normalize Veo ``parameters.resolution`` for usage and cost tracking.""" res: Final = parameters.get("resolution") @@ -97,7 +103,7 @@ class GeminiVideoConfig(BaseVideoConfig): video_create_optional_params: VideoCreateOptionalRequestParams, model: str, drop_params: bool, - ) -> dict[str, Any]: + ) -> dict[str, object]: """ Map OpenAI-style parameters to Veo format. @@ -111,7 +117,7 @@ class GeminiVideoConfig(BaseVideoConfig): All other params are passed through as-is to support Gemini-specific parameters. """ - mapped_params: Final[dict[str, Any]] = {} + mapped_params: Final[dict[str, object]] = {} # Get supported OpenAI params (exclude "model" and "prompt" which are handled separately) supported_openai_params: Final = self.get_supported_openai_params(model) @@ -312,11 +318,11 @@ class GeminiVideoConfig(BaseVideoConfig): - status: "processing" - usage: includes duration_seconds and optional video_resolution for cost calculation """ - response_data: Final = raw_response.json() + response_data: Final = _json_payload(raw_response) # Parse response using Pydantic model for type safety try: - operation_response: Final = GeminiLongRunningOperationResponse(**response_data) + operation_response: Final = GeminiLongRunningOperationResponse.model_validate(response_data) except Exception as e: raise ValueError(f"Failed to parse operation response: {e}") @@ -336,7 +342,7 @@ class GeminiVideoConfig(BaseVideoConfig): model=model, ) - usage_data: Final[dict[str, Any]] = {} + usage_data: Final[dict[str, float | str]] = {} if request_data: parameters: Final = request_data.get("parameters", {}) duration: Final = parameters.get("durationSeconds") or DEFAULT_GOOGLE_VIDEO_DURATION_SECONDS @@ -367,7 +373,7 @@ class GeminiVideoConfig(BaseVideoConfig): """ operation_name: Final = extract_original_video_id(video_id) url: Final = f"{api_base.rstrip('/')}/v1beta/{operation_name}" - params: Final[dict[str, Any]] = {} + params: Final[dict[str, object]] = {} return url, params @@ -403,9 +409,9 @@ class GeminiVideoConfig(BaseVideoConfig): } } """ - response_data: Final = raw_response.json() + response_data: Final = _json_payload(raw_response) # Parse response using Pydantic model for type safety - operation_response: Final = GeminiLongRunningOperationResponse(**response_data) + operation_response: Final = GeminiLongRunningOperationResponse.model_validate(response_data) operation_name: Final = operation_response.name is_done: Final = operation_response.done @@ -443,9 +449,9 @@ class GeminiVideoConfig(BaseVideoConfig): client: Final = litellm.module_level_client status_response: Final = client.get(url=status_url, headers=headers) status_response.raise_for_status() - response_data: Final = status_response.json() + response_data: Final = _json_payload(status_response) - operation_response: Final = GeminiLongRunningOperationResponse(**response_data) + operation_response: Final = GeminiLongRunningOperationResponse.model_validate(response_data) if not operation_response.done: raise ValueError( @@ -458,7 +464,7 @@ class GeminiVideoConfig(BaseVideoConfig): generated_samples: Final = operation_response.response.generateVideoResponse.generatedSamples download_url: Final = generated_samples[0].video.uri - params: Final[dict[str, Any]] = {} + params: Final[dict[str, object]] = {} return download_url, params @@ -480,7 +486,7 @@ class GeminiVideoConfig(BaseVideoConfig): api_base: str, litellm_params: GenericLiteLLMParams, headers: dict, - extra_body: dict[str, Any] | None = None, + extra_body: Mapping[str, object] | None = None, ) -> tuple[str, dict]: """ Video remix is not supported by Veo API. @@ -506,7 +512,7 @@ class GeminiVideoConfig(BaseVideoConfig): after: str | None = None, limit: int | None = None, order: str | None = None, - extra_query: dict[str, Any] | None = None, + extra_query: Mapping[str, object] | None = None, ) -> tuple[str, dict]: """ Video list is not supported by Veo API. @@ -547,7 +553,7 @@ class GeminiVideoConfig(BaseVideoConfig): """Video delete is not supported.""" raise NotImplementedError("Video delete is not supported by Google Veo.") - def transform_video_create_character_request(self, name, video, api_base, litellm_params, headers): + def transform_video_create_character_request(self, name, video: object, api_base, litellm_params, headers): raise NotImplementedError("video create character is not supported for Gemini") def transform_video_create_character_response(self, raw_response, logging_obj): diff --git a/litellm/llms/hosted_vllm/embedding/README.md b/litellm/llms/hosted_vllm/embedding/README.md index 2c58e16fc23..50474aabdeb 100644 --- a/litellm/llms/hosted_vllm/embedding/README.md +++ b/litellm/llms/hosted_vllm/embedding/README.md @@ -4,13 +4,12 @@ VLLM is a superset of OpenAI's `embedding` endpoint. ## `encoding_format` -For OpenAI-compatible embedding calls (including `openai/...` with a custom `api_base` pointing at vLLM), LiteLLM resolves `encoding_format` when it is not set on the request: +For OpenAI-compatible embedding calls (including `openai/...` with a custom `api_base` pointing at vLLM), LiteLLM resolves `encoding_format` when it is not set on the request. `hosted_vllm/...` models use a separate handler that never adds the field on its own, so this resolution applies to the `openai/...`-style routes only: 1. Explicit value on the embedding call (`encoding_format=...`). 2. Model config (`litellm_params.encoding_format` on the proxy `model_list` entry). 3. Environment variable `LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT` (e.g. in `.env` or container env). -4. Default **`float`**. -That avoids forwarding `encoding_format=None` to the provider/SDK where some servers behave poorly. +If none of those is set, or the winning value is the literal string `none`, the field is omitted from the upstream request entirely (LiteLLM also bypasses the OpenAI SDK's own base64 default), so OpenAI-compatible servers that reject `encoding_format` keep working. -To pass provider-specific parameters, see [provider-specific params](https://docs.litellm.ai/docs/completion/provider_specific_params). \ No newline at end of file +To pass provider-specific parameters, see [provider-specific params](https://docs.litellm.ai/docs/completion/provider_specific_params). diff --git a/litellm/llms/huggingface/embedding/transformation.py b/litellm/llms/huggingface/embedding/transformation.py index d3db3530109..f6fe7f2fa10 100644 --- a/litellm/llms/huggingface/embedding/transformation.py +++ b/litellm/llms/huggingface/embedding/transformation.py @@ -1,8 +1,9 @@ import json import os import time +from collections.abc import Sequence from copy import deepcopy -from typing import TYPE_CHECKING, Any, Final +from typing import TYPE_CHECKING, Any, Final, Protocol import httpx @@ -24,6 +25,8 @@ from litellm.utils import token_counter from ..common_utils import HuggingFaceError, hf_task_list, hf_tasks, output_parser if TYPE_CHECKING: + import tiktoken + from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj LoggingClass = LiteLLMLoggingObj @@ -31,6 +34,12 @@ else: LoggingClass = Any +class _TokenEncoding(Protocol): + """Tokenizer handle the caller passes in; only `encode` is used, to count completion tokens.""" + + def encode(self, text: str, /) -> Sequence[object]: ... + + tgi_models_cache = None conv_models_cache = None @@ -369,7 +378,7 @@ class HuggingFaceEmbeddingConfig(BaseConfig): model_response: ModelResponse, task: hf_tasks | None, optional_params: dict, - encoding: Any, + encoding: "_TokenEncoding | None", messages: list[AllMessageValues], model: str, ): @@ -439,9 +448,10 @@ class HuggingFaceEmbeddingConfig(BaseConfig): if output_text is not None and len(output_text) > 0: completion_tokens = 0 try: - completion_tokens = len( - encoding.encode(model_response["choices"][0]["message"].get("content", "")) - ) ##[TODO] use the llama2 tokenizer here + if encoding is not None: + completion_tokens = len( + encoding.encode(model_response["choices"][0]["message"].get("content", "")) + ) ##[TODO] use the llama2 tokenizer here except Exception: # this should remain non blocking we should not block a response returning if calculating usage fails pass @@ -469,7 +479,7 @@ class HuggingFaceEmbeddingConfig(BaseConfig): messages: list[AllMessageValues], optional_params: dict, litellm_params: dict, - encoding: Any, + encoding: "tiktoken.Encoding | None", api_key: str | None = None, json_mode: bool | None = None, ) -> ModelResponse: diff --git a/litellm/llms/openai/chat/gpt_transformation.py b/litellm/llms/openai/chat/gpt_transformation.py index d4747b2fb06..f3e41c054ef 100644 --- a/litellm/llms/openai/chat/gpt_transformation.py +++ b/litellm/llms/openai/chat/gpt_transformation.py @@ -4,7 +4,8 @@ Support for gpt model family import json import os -from collections.abc import AsyncIterator, Coroutine, Iterator +from collections.abc import AsyncIterator, Coroutine, Iterator, Mapping +from types import MappingProxyType from typing import TYPE_CHECKING, Any, Final, Literal, Optional, cast, overload from urllib.parse import urlparse @@ -19,6 +20,7 @@ from litellm.litellm_core_utils.llm_response_utils.convert_dict_to_response impo ) from litellm.litellm_core_utils.prompt_templates.common_utils import ( drop_tool_reference_parts_from_tool_messages, + flatten_top_level_schema_combinators, get_tool_call_names, hoist_images_from_tool_messages, ) @@ -65,6 +67,22 @@ else: LiteLLMLoggingObj = Any +_NO_TOOLS_UPDATE: Final[Mapping[str, object]] = MappingProxyType({}) + + +def _tool_with_flattened_parameters(tool: Mapping[str, object]) -> Mapping[str, object]: + function: Final = tool.get("function") + if not isinstance(function, dict): + return tool + parameters: Final = function.get("parameters") + if not isinstance(parameters, dict): + return tool + flattened: Final = flatten_top_level_schema_combinators(parameters) + if flattened is parameters: + return tool + return {**tool, "function": {**function, "parameters": flattened}} # mutable-ok: request tools are JSON dicts + + class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): """ Reference: https://platform.openai.com/docs/api-reference/chat/create @@ -325,7 +343,7 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): @overload def _transform_messages( self, messages: list[AllMessageValues], model: str, is_async: Literal[True] - ) -> Coroutine[Any, Any, list[AllMessageValues]]: + ) -> Coroutine[object, object, list[AllMessageValues]]: ... @overload @@ -341,7 +359,7 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): def _transform_messages( self, messages: list[AllMessageValues], model: str, is_async: bool = False - ) -> list[AllMessageValues] | Coroutine[Any, Any, list[AllMessageValues]]: + ) -> list[AllMessageValues] | Coroutine[object, object, list[AllMessageValues]]: """OpenAI no longer supports image_url as a string, so we need to convert it to a dict""" stripped_messages: Final = drop_tool_reference_parts_from_tool_messages(messages) hoisted_messages: Final = hoist_images_from_tool_messages(stripped_messages) @@ -397,6 +415,21 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): ) return messages, tools + def _targets_openai_hosted_endpoint( + self, + custom_llm_provider: str | None, + api_base: str | None, + ) -> bool: + if custom_llm_provider != "openai": + return False + resolved_api_base = api_base or litellm.api_base or os.getenv("OPENAI_BASE_URL") or os.getenv("OPENAI_API_BASE") + if not resolved_api_base: + return True + hostname: Final = urlparse(resolved_api_base).hostname + if hostname is None: + return True + return hostname == "openai.com" or hostname.endswith(".openai.com") + def _should_preserve_cache_control_for_endpoint( self, custom_llm_provider: str | None, @@ -408,15 +441,34 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): api_base. Those can understand cache_control, so it must survive there. Real OpenAI cannot, so it is still stripped for an openai.com host. """ - if custom_llm_provider != "openai": - return False - resolved_api_base = api_base or litellm.api_base or os.getenv("OPENAI_BASE_URL") or os.getenv("OPENAI_API_BASE") - if not resolved_api_base: - return False - hostname: Final = urlparse(resolved_api_base).hostname - if hostname is None: - return False - return hostname != "openai.com" and not hostname.endswith(".openai.com") + return custom_llm_provider == "openai" and not self._targets_openai_hosted_endpoint( + custom_llm_provider, api_base + ) + + def _flattened_tools_update_for_openai( + self, + optional_params: Mapping[str, object], + litellm_params: Mapping[str, object], + ) -> Mapping[str, object]: + """ + OpenAI's chat completions validator rejects tool `parameters` carrying + 'oneOf'/'anyOf'/'allOf'/'enum'/'const'/'not' at the top level for every + model family, unlike the Responses API, where GPT-5+ accepts them. + """ + tools: Final = optional_params.get("tools") + if not isinstance(tools, list): + return _NO_TOOLS_UPDATE + provider: Final = litellm_params.get("custom_llm_provider") + raw_api_base: Final = litellm_params.get("api_base") + if not self._targets_openai_hosted_endpoint( + provider if isinstance(provider, str) else None, + raw_api_base if isinstance(raw_api_base, str) else None, + ): + return _NO_TOOLS_UPDATE + flattened: Final = [ # mutable-ok: request tools are a JSON list + _tool_with_flattened_parameters(tool) if isinstance(tool, dict) else tool for tool in tools + ] + return MappingProxyType({"tools": flattened}) def transform_request( self, @@ -443,11 +495,14 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): optional_params["tools"] = tools optional_params.pop("max_retries", None) + if not optional_params.get("tools") and not optional_params.get("functions"): + optional_params.pop("tool_choice", None) return { "model": model, "messages": messages, **optional_params, + **self._flattened_tools_update_for_openai(optional_params, litellm_params), } async def async_transform_request( @@ -473,10 +528,13 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): if tools is not None and len(tools) > 0: optional_params["tools"] = tools if self.__class__._is_base_class: + if not optional_params.get("tools") and not optional_params.get("functions"): + optional_params.pop("tool_choice", None) return { "model": model, "messages": transformed_messages, **optional_params, + **self._flattened_tools_update_for_openai(optional_params, litellm_params), } else: ## allow for any object specific behaviour to be handled @@ -497,8 +555,12 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): return None tool_call_names: Final = get_tool_call_names(optional_params.get("tools", [])) try: - json_content: Final = json.loads(content) - if json_content.get("type") == "function" and json_content.get("name") in tool_call_names: + json_content: Final[object] = json.loads(content) + if ( + isinstance(json_content, dict) + and json_content.get("type") == "function" + and json_content.get("name") in tool_call_names + ): return ChatCompletionMessageToolCall( function=Function( name=json_content.get("name"), @@ -622,7 +684,7 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig): ## RESPONSE OBJECT try: - completion_response: Final = raw_response.json() + completion_response: Final[dict[str, object]] = raw_response.json() except Exception as e: response_headers: Final = getattr(raw_response, "headers", None) raise OpenAIError( diff --git a/litellm/llms/openai/chat/guardrail_translation/handler.py b/litellm/llms/openai/chat/guardrail_translation/handler.py index de15fefe943..ed628f55350 100644 --- a/litellm/llms/openai/chat/guardrail_translation/handler.py +++ b/litellm/llms/openai/chat/guardrail_translation/handler.py @@ -51,6 +51,7 @@ if TYPE_CHECKING: from litellm.integrations.custom_guardrail import CustomGuardrail from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj + from litellm.proxy._types import UserAPIKeyAuth class OpenAIChatCompletionsHandler(BaseTranslation): @@ -80,7 +81,7 @@ class OpenAIChatCompletionsHandler(BaseTranslation): data: dict, guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: "LiteLLMLoggingObj | None" = None, - ) -> Any: + ) -> dict: """ Process input messages by applying guardrails to text content. """ @@ -329,9 +330,9 @@ class OpenAIChatCompletionsHandler(BaseTranslation): response: "ModelResponse", guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: "LiteLLMLoggingObj | None" = None, - user_api_key_dict: Any | None = None, + user_api_key_dict: "UserAPIKeyAuth | None" = None, request_data: dict | None = None, - ) -> Any: + ) -> ModelResponse: """ Process output response by applying guardrails to text content. @@ -436,7 +437,7 @@ class OpenAIChatCompletionsHandler(BaseTranslation): responses_so_far: list["ModelResponseStream"], guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: "LiteLLMLoggingObj | None" = None, - user_api_key_dict: Any | None = None, + user_api_key_dict: "UserAPIKeyAuth | None" = None, request_data: dict | None = None, stream_transform_sink: StreamTransformSink | None = None, ) -> list["ModelResponseStream"]: @@ -486,7 +487,7 @@ class OpenAIChatCompletionsHandler(BaseTranslation): responses_so_far: list["ModelResponseStream"], guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: "LiteLLMLoggingObj | None", - user_api_key_dict: Any | None, + user_api_key_dict: "UserAPIKeyAuth | None", request_data: dict | None, ) -> list["ModelResponseStream"]: """Block-only streaming path: run the guardrail so an in-flight BLOCK can @@ -589,8 +590,8 @@ class OpenAIChatCompletionsHandler(BaseTranslation): def build_stream_error_items( self, exc: "HTTPException", - responses_so_far: Sequence[Any] | None = None, - ) -> Sequence[Any] | None: + responses_so_far: Sequence[object] | None = None, + ) -> Sequence[bytes] | None: import json from litellm.proxy.common_request_processing import sse_error_payload @@ -630,7 +631,7 @@ class OpenAIChatCompletionsHandler(BaseTranslation): responses_so_far: list["ModelResponseStream"], guardrail_to_apply: "CustomGuardrail", litellm_logging_obj: "LiteLLMLoggingObj | None", - user_api_key_dict: Any | None, + user_api_key_dict: "UserAPIKeyAuth | None", request_data: dict | None, sink: StreamTransformSink, ) -> None: @@ -794,7 +795,7 @@ class OpenAIChatCompletionsHandler(BaseTranslation): # Determine content source and tool calls based on choice type content = None - tool_calls: list[Any] | None = None + tool_calls: Sequence[object] | None = None if isinstance(choice, litellm.Choices): content = choice.message.content tool_calls = choice.message.tool_calls diff --git a/litellm/llms/openai/openai.py b/litellm/llms/openai/openai.py index 56495f0097d..1cfc6e06ee9 100644 --- a/litellm/llms/openai/openai.py +++ b/litellm/llms/openai/openai.py @@ -12,9 +12,14 @@ if TYPE_CHECKING: import openai from openai import AsyncOpenAI, OpenAI +from openai._base_client import make_request_options +from openai._constants import RAW_RESPONSE_HEADER +from openai._legacy_response import LegacyAPIResponse +from openai._types import RequestOptions +from openai.types import CreateEmbeddingResponse from openai.types.beta.assistant_deleted import AssistantDeleted from openai.types.file_deleted import FileDeleted -from pydantic import BaseModel +from pydantic import BaseModel, TypeAdapter from typing_extensions import overload import litellm @@ -329,6 +334,28 @@ class OpenAIChatCompletionResponseIterator(BaseModelResponseIterator): raise e +_EXTRA_HEADERS_ADAPTER: Final = TypeAdapter(dict[str, str] | None) +_EXTRA_QUERY_ADAPTER: Final = TypeAdapter(dict[str, object] | None) +_NO_EXTRA_HEADERS: Final[Mapping[str, str]] = types.MappingProxyType({}) +_SDK_OPTION_KEYS: Final = frozenset(("extra_headers", "extra_query", "extra_body")) + + +def _embedding_request_without_sdk_defaults( + data: Mapping[str, object], timeout: float | httpx.Timeout +) -> tuple[Mapping[str, object], RequestOptions]: + body: Final = { # mutable-ok: the SDK json-encodes the body and needs a plain dict + k: v for k, v in data.items() if k not in _SDK_OPTION_KEYS + } + extra_headers: Final = _EXTRA_HEADERS_ADAPTER.validate_python(data.get("extra_headers")) or _NO_EXTRA_HEADERS + options: Final = make_request_options( + extra_headers=types.MappingProxyType({**extra_headers, RAW_RESPONSE_HEADER: "true"}), + extra_query=_EXTRA_QUERY_ADAPTER.validate_python(data.get("extra_query")), + extra_body=data.get("extra_body"), + timeout=timeout, + ) + return body, options + + class OpenAIChatCompletion(BaseLLM, BaseOpenAILLM): def __init__(self) -> None: super().__init__() @@ -1177,19 +1204,15 @@ class OpenAIChatCompletion(BaseLLM, BaseOpenAILLM): data: dict, timeout: float | httpx.Timeout, logging_obj: LiteLLMLoggingObj, - ): - """ - Helper to: - - call embeddings.create.with_raw_response when litellm.return_response_headers is True - - call embeddings.create by default - """ - try: - raw_response = await openai_aclient.embeddings.with_raw_response.create(**data, timeout=timeout) - headers: Final = dict(raw_response.headers) - response: Final = raw_response.parse() - return headers, response - except Exception as e: - raise e + ) -> LegacyAPIResponse[CreateEmbeddingResponse]: + if "encoding_format" not in data: + body, options = _embedding_request_without_sdk_defaults(data, timeout) + bypass_response: Final = await openai_aclient.post( + "/embeddings", body=body, options=options, cast_to=CreateEmbeddingResponse + ) + assert isinstance(bypass_response, LegacyAPIResponse) + return bypass_response + return await openai_aclient.embeddings.with_raw_response.create(**data, timeout=timeout) @track_llm_api_timing() def make_sync_openai_embedding_request( @@ -1198,20 +1221,15 @@ class OpenAIChatCompletion(BaseLLM, BaseOpenAILLM): data: dict, timeout: float | httpx.Timeout, logging_obj: LiteLLMLoggingObj, - ): - """ - Helper to: - - call embeddings.create.with_raw_response when litellm.return_response_headers is True - - call embeddings.create by default - """ - try: - raw_response = openai_client.embeddings.with_raw_response.create(**data, timeout=timeout) - - headers: Final = dict(raw_response.headers) - response: Final = raw_response.parse() - return headers, response - except Exception as e: - raise e + ) -> LegacyAPIResponse[CreateEmbeddingResponse]: + if "encoding_format" not in data: + body, options = _embedding_request_without_sdk_defaults(data, timeout) + bypass_response: Final = openai_client.post( + "/embeddings", body=body, options=options, cast_to=CreateEmbeddingResponse + ) + assert isinstance(bypass_response, LegacyAPIResponse) + return bypass_response + return openai_client.embeddings.with_raw_response.create(**data, timeout=timeout) async def aembedding( self, @@ -1236,14 +1254,15 @@ class OpenAIChatCompletion(BaseLLM, BaseOpenAILLM): client=client, shared_session=shared_session, ) - headers, response = await self.make_openai_embedding_request( + raw_response: Final = await self.make_openai_embedding_request( openai_aclient=openai_aclient, data=data, timeout=timeout, logging_obj=logging_obj, ) + headers: Final = dict(raw_response.headers) logging_obj.model_call_details["response_headers"] = headers - stringified_response: Final = response.model_dump() + stringified_response: Final = raw_response.parse().model_dump() ## LOGGING logging_obj.post_call( input=input, @@ -1335,13 +1354,14 @@ class OpenAIChatCompletion(BaseLLM, BaseOpenAILLM): ) ## embedding CALL - headers: dict | None = None - headers, sync_embedding_response = self.make_sync_openai_embedding_request( + raw_response: Final = self.make_sync_openai_embedding_request( openai_client=openai_client, data=data, timeout=timeout, logging_obj=logging_obj, ) + headers: Final = dict(raw_response.headers) + sync_embedding_response: Final = raw_response.parse() ## LOGGING logging_obj.model_call_details["response_headers"] = headers diff --git a/litellm/llms/openai/responses/transformation.py b/litellm/llms/openai/responses/transformation.py index eadc087383a..01313e95878 100644 --- a/litellm/llms/openai/responses/transformation.py +++ b/litellm/llms/openai/responses/transformation.py @@ -1,10 +1,11 @@ from collections.abc import Mapping, Sequence from types import MappingProxyType -from typing import TYPE_CHECKING, Any, Final, cast, get_type_hints +from typing import TYPE_CHECKING, Any, Final, Protocol, cast, get_type_hints import httpx from openai.types.responses import ResponseReasoningItem from pydantic import BaseModel, ValidationError +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -14,6 +15,7 @@ from litellm.litellm_core_utils.llm_response_utils.convert_dict_to_response impo ) from litellm.litellm_core_utils.url_utils import encode_url_path_segment from litellm.llms.base_llm.responses.transformation import BaseResponsesAPIConfig +from litellm.responses.litellm_completion_transformation.custom_tools import TOOL_CALL_ITEM_ID_PREFIX_BY_TYPE from litellm.secret_managers.main import get_secret_str from litellm.types.llms.openai import * from litellm.types.responses.main import * @@ -35,6 +37,37 @@ else: _NO_TOOL_UPDATE: Final[Mapping[str, object]] = MappingProxyType({}) _MODEL_FAMILIES_REJECTING_TOP_LEVEL_SCHEMA_COMBINATORS: Final = ("gpt-4", "gpt-3.5", "chatgpt-4o", "o1", "o3", "o4") _PROVIDERS_WITH_COMBINATOR_REJECTING_VALIDATOR: Final = frozenset({LlmProviders.AZURE, LlmProviders.OPENAI}) +_PROVIDERS_VALIDATING_TOOL_CALL_ITEM_IDS: Final = frozenset({LlmProviders.AZURE, LlmProviders.OPENAI}) + + +class _DeleteResponseBody(TypedDict): + """Decoded body of the Responses API delete call.""" + + id: ReadOnly[str | None] + object: ReadOnly[str | None] + deleted: ReadOnly[bool | None] + + +class _DeleteResponse(Protocol): + """The delete call's HTTP response, read for the decoded body it carries.""" + + def json(self) -> _DeleteResponseBody: ... + + +class _JsonObjectResponse(Protocol): + """A Responses API HTTP response, read for the JSON object it decodes to.""" + + def json(self) -> dict[str, object]: ... + + +def _delete_response_body(response: _DeleteResponse) -> _DeleteResponseBody: + """Decode a delete response body into the id, object and deleted fields it carries.""" + return response.json() + + +def _json_object_body(response: _JsonObjectResponse) -> dict[str, object]: + """Decode a Responses API response body into its JSON object form.""" + return response.json() class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): @@ -179,8 +212,9 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): ) if sanitized_tools is not None: response_api_optional_request_params["tools"] = sanitized_tools + replay_safe_input: Final = self._drop_foreign_tool_call_item_ids(input) final_request_params: Final = dict( - ResponsesAPIRequestParams(model=model, input=input, **response_api_optional_request_params) + ResponsesAPIRequestParams(model=model, input=replay_safe_input, **response_api_optional_request_params) ) return final_request_params @@ -217,6 +251,23 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): return input, tools + def _drop_foreign_tool_call_item_ids(self, input: str | ResponseInputParam) -> str | ResponseInputParam: + if self.custom_llm_provider not in _PROVIDERS_VALIDATING_TOOL_CALL_ITEM_IDS or not isinstance(input, list): + return input + sanitized_items: Final = [self._without_foreign_tool_call_item_id(item) for item in input] + return cast("ResponseInputParam", sanitized_items) # cast-ok: items keep their shape, minus a rejected id + + @staticmethod + def _without_foreign_tool_call_item_id(item: object) -> object: + if not isinstance(item, dict): + return item + item_type: Final = item.get("type") + item_id: Final = item.get("id") + genuine_prefix: Final = TOOL_CALL_ITEM_ID_PREFIX_BY_TYPE.get(item_type) if isinstance(item_type, str) else None + if genuine_prefix is None or not isinstance(item_id, str) or item_id.startswith(genuine_prefix): + return item + return {key: value for key, value in item.items() if key != "id"} # mutable-ok: outgoing JSON request item + def _flatten_tool_schema_combinators_for_openai( self, model: str, @@ -469,7 +520,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): return None @staticmethod - def get_event_model_class(event_type: str) -> Any: + def get_event_model_class(event_type: str) -> type[BaseLiteLLMOpenAIResponseObject]: """ Returns the appropriate event model class based on the event type. @@ -583,7 +634,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): Transform the delete response API response into a DeleteResponseResult """ try: - raw_response_json: Final = raw_response.json() + raw_response_json: Final = _delete_response_body(raw_response) except Exception: raise OpenAIError(message=raw_response.text, status_code=raw_response.status_code) return DeleteResponseResult(**raw_response_json) @@ -618,7 +669,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): Transform the get response API response into a ResponsesAPIResponse """ try: - raw_response_json: Final = raw_response.json() + raw_response_json: Final = _json_object_body(raw_response) except Exception: raise OpenAIError(message=raw_response.text, status_code=raw_response.status_code) raw_response_headers: Final = dict(raw_response.headers) @@ -646,7 +697,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): ) -> tuple[str, dict]: encoded_response_id: Final = encode_url_path_segment(response_id, field_name="response_id") url: Final = f"{api_base}/{encoded_response_id}/input_items" - params: Final[dict[str, Any]] = {} + params: Final[dict[str, object]] = {} if after is not None: params["after"] = after if before is not None: @@ -665,7 +716,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): logging_obj: LiteLLMLoggingObj, ) -> dict: try: - return raw_response.json() + return _json_object_body(raw_response) except Exception: raise OpenAIError(message=raw_response.text, status_code=raw_response.status_code) @@ -699,7 +750,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): Transform the cancel response API response into a ResponsesAPIResponse """ try: - raw_response_json: Final = raw_response.json() + raw_response_json: Final = _json_object_body(raw_response) except Exception: raise OpenAIError(message=raw_response.text, status_code=raw_response.status_code) raw_response_headers: Final = dict(raw_response.headers) @@ -742,7 +793,10 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig): ) if sanitized_tools is not None: response_api_optional_request_params["tools"] = sanitized_tools - data: Final = dict(ResponsesAPIRequestParams(model=model, input=input, **response_api_optional_request_params)) + replay_safe_input: Final = self._drop_foreign_tool_call_item_ids(input) + data: Final = dict( + ResponsesAPIRequestParams(model=model, input=replay_safe_input, **response_api_optional_request_params) + ) return url, data diff --git a/litellm/llms/openai_like/chat/handler.py b/litellm/llms/openai_like/chat/handler.py index 8c548b6b0d6..855c49c320b 100644 --- a/litellm/llms/openai_like/chat/handler.py +++ b/litellm/llms/openai_like/chat/handler.py @@ -5,10 +5,11 @@ For handling OpenAI-like chat completions, like IBM WatsonX, etc. """ import json -from collections.abc import Callable -from typing import Any, Final +from collections.abc import Callable, Mapping, Sequence +from typing import Final, TypedDict import httpx +from typing_extensions import ReadOnly import litellm from litellm import LlmProviders @@ -25,6 +26,23 @@ from ..common_utils import OpenAILikeBase, OpenAILikeError from .transformation import OpenAILikeChatConfig +class _OpenAILikeChatCompletion(TypedDict, total=False): + """The chat-completion JSON body an OpenAI-like provider returns for a non-streamed call.""" + + id: ReadOnly[str] + choices: ReadOnly[Sequence[Mapping[str, object]]] + created: ReadOnly[int] + model: ReadOnly[str] + system_fingerprint: ReadOnly[str] + usage: ReadOnly[Mapping[str, object]] + object: ReadOnly[str] + + +def _fake_streamed_model_response(payload: _OpenAILikeChatCompletion) -> ModelResponse: + """Build the single response a fake-streamed provider call replays as one chunk.""" + return ModelResponse(**payload) + + async def make_call( client: AsyncHTTPHandler | None, api_base: str, @@ -42,9 +60,9 @@ async def make_call( response: Final = await client.post(api_base, headers=headers, data=data, stream=not fake_stream) if streaming_decoder is not None: - completion_stream: Any = streaming_decoder.aiter_bytes(response.aiter_bytes(chunk_size=1024)) + completion_stream = streaming_decoder.aiter_bytes(response.aiter_bytes(chunk_size=1024)) elif fake_stream: - model_response: Final = ModelResponse(**response.json()) + model_response: Final = _fake_streamed_model_response(response.json()) completion_stream = MockResponseIterator(model_response=model_response) else: completion_stream = ModelResponseIterator(streaming_response=response.aiter_lines(), sync_stream=False) @@ -82,7 +100,7 @@ def make_sync_call( if streaming_decoder is not None: completion_stream = streaming_decoder.iter_bytes(response.iter_bytes(chunk_size=1024)) elif fake_stream: - model_response: Final = ModelResponse(**response.json()) + model_response: Final = _fake_streamed_model_response(response.json()) completion_stream = MockResponseIterator(model_response=model_response) else: completion_stream = ModelResponseIterator(streaming_response=response.iter_lines(), sync_stream=True) diff --git a/litellm/llms/runwayml/image_generation/transformation.py b/litellm/llms/runwayml/image_generation/transformation.py index cde65addb65..5913709c8a0 100644 --- a/litellm/llms/runwayml/image_generation/transformation.py +++ b/litellm/llms/runwayml/image_generation/transformation.py @@ -1,8 +1,10 @@ import asyncio import time +from collections.abc import Mapping, Sequence from typing import TYPE_CHECKING, Any, Final import httpx +from typing_extensions import ReadOnly, TypedDict from litellm._logging import verbose_logger from litellm.constants import ( @@ -29,6 +31,16 @@ else: LiteLLMLoggingObj = Any +class _RunwayMLTask(TypedDict, total=False): + """The RunwayML task payload returned by POST /v1/text_to_image and GET /v1/tasks/{id}.""" + + id: ReadOnly[str] + status: ReadOnly[str] + output: ReadOnly[Sequence[str | Mapping[str, str]]] + failure: ReadOnly[str] + failureCode: ReadOnly[str] + + class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): """ Configuration for RunwayML image generation models. @@ -80,7 +92,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): @staticmethod def _transform_runwayml_response_to_openai( - response_data: dict[str, Any], + response_data: _RunwayMLTask, model_response: ImageResponse, ) -> ImageResponse: """ @@ -155,7 +167,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): raise TimeoutError(f"RunwayML task polling timed out after {timeout_secs} seconds") @staticmethod - def _check_task_status(response_data: dict[str, Any]) -> str: + def _check_task_status(response_data: _RunwayMLTask) -> str: """ Check RunwayML task status from response. @@ -227,7 +239,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): response = client.get(url=task_url, headers=headers) response.raise_for_status() - response_data = response.json() + response_data: _RunwayMLTask = response.json() # Check task status status = self._check_task_status(response_data=response_data) @@ -276,7 +288,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): response = await client.get(url=task_url, headers=headers) response.raise_for_status() - response_data = response.json() + response_data: _RunwayMLTask = response.json() # Check task status status = self._check_task_status(response_data=response_data) @@ -322,7 +334,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): } """ try: - response_data = raw_response.json() + response_data: _RunwayMLTask = raw_response.json() except Exception as e: raise self.get_error_class( error_message=f"Error transforming image generation response: {e}", @@ -382,7 +394,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig): We need to poll the task until it completes (status SUCCEEDED) using async polling. """ try: - response_data = raw_response.json() + response_data: _RunwayMLTask = raw_response.json() except Exception as e: raise self.get_error_class( error_message=f"Error transforming image generation response: {e}", diff --git a/litellm/llms/sap/credentials.py b/litellm/llms/sap/credentials.py index d7743d4d337..a2a93b6114a 100644 --- a/litellm/llms/sap/credentials.py +++ b/litellm/llms/sap/credentials.py @@ -8,9 +8,10 @@ from dataclasses import dataclass from datetime import datetime, timedelta, timezone from pathlib import Path from threading import Lock -from typing import Any, Final +from typing import Any, Final, Protocol import httpx +from typing_extensions import NotRequired, ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -33,8 +34,8 @@ def _get_home() -> str: return os.getenv(HOME_PATH_ENV_VAR, DEFAULT_HOME_PATH) -def _get_nested(d: dict[str, Any] | str, path: Sequence[str]) -> Any: - cur: Any = d +def _get_nested(d: object, path: Sequence[str]) -> object: + cur: object = d if isinstance(cur, str): # This shouldn't happen if service keys are pre-parsed correctly try: @@ -54,7 +55,7 @@ def _get_nested(d: dict[str, Any] | str, path: Sequence[str]) -> Any: return cur -def _load_json_env(var_name: str) -> dict[str, Any] | None: +def _load_json_env(var_name: str) -> dict[str, object] | None: raw: Final = os.environ.get(var_name) if not raw: return None @@ -64,7 +65,7 @@ def _load_json_env(var_name: str) -> dict[str, Any] | None: return None -def _str_or_none(value) -> str | None: +def _str_or_none(value: object) -> str | None: try: return str(value) if value is not None else None except Exception: @@ -124,7 +125,7 @@ CREDENTIAL_VALUES: Final[list[CredentialsValue]] = [ ] -def init_conf(profile: str | None = None) -> dict[str, Any]: +def init_conf(profile: str | None = None) -> dict[str, object]: """ Loads config JSON from: 1) $AICORE_CONFIG if set, otherwise @@ -191,7 +192,7 @@ def resolve_resource_group(sources: list[Source]) -> str | None: def _parse_service_key_once( service_key: str | dict | None, -) -> dict[str, Any] | None: +) -> dict[str, object] | None: """ Pre-parse service_key if it's a string to avoid repeated JSON parsing. @@ -348,8 +349,33 @@ def validate_credentials( ) +class _TokenBody(TypedDict): + """Decoded body of the SAP AI Core OAuth2 token response.""" + + access_token: ReadOnly[str] + expires_in: ReadOnly[NotRequired[int]] + + +class _TokenResponse(Protocol): + """The token endpoint's HTTP response, read for the decoded token body it carries.""" + + def json(self) -> _TokenBody: ... + + +def _bearer_token_and_expiry(response: _TokenResponse) -> tuple[str, datetime]: + """Read a token response into the Authorization header value and the token's absolute expiry.""" + payload: Final = response.json() + expires_in: Final = int(payload.get("expires_in", 3600)) + access_token: Final = payload["access_token"] + return f"Bearer {access_token}", datetime.now(timezone.utc) + timedelta(seconds=expires_in) + + def _request_token( - client_id: str, auth_url: str, timeout: float, cert_pair=None, client_secret=None + client_id: str, + auth_url: str, + timeout: float, + cert_pair: tuple[str, str] | None = None, + client_secret: str | None = None, ) -> tuple[str, datetime]: data: Final = {"grant_type": "client_credentials", "client_id": client_id} if client_secret: @@ -361,15 +387,10 @@ def _request_token( with httpx.Client(cert=cert_pair) as raw_client: handler = HTTPHandler(client=raw_client) resp = handler.post(auth_url, data=data, timeout=timeout) - payload = resp.json() - else: - handler = _get_httpx_client() - resp = handler.post(auth_url, data=data, timeout=timeout) - payload = resp.json() - access_token: Final = payload["access_token"] - expires_in: Final = int(payload.get("expires_in", 3600)) - expiry_date: Final = datetime.now(timezone.utc) + timedelta(seconds=expires_in) - return f"Bearer {access_token}", expiry_date + return _bearer_token_and_expiry(resp) + handler = _get_httpx_client() + resp = handler.post(auth_url, data=data, timeout=timeout) + return _bearer_token_and_expiry(resp) except Exception as e: msg: Final = resp.text if resp is not None else getattr(e, "text", str(e)) raise RuntimeError(f"Token request failed: {msg}") from e diff --git a/litellm/llms/vertex_ai/files/transformation.py b/litellm/llms/vertex_ai/files/transformation.py index b7f91bfba0d..b6ad9fbcc04 100644 --- a/litellm/llms/vertex_ai/files/transformation.py +++ b/litellm/llms/vertex_ai/files/transformation.py @@ -12,7 +12,7 @@ from urllib.parse import quote, unquote import httpx from httpx import Headers, Response from openai.types.file_deleted import FileDeleted -from typing_extensions import ReadOnly +from typing_extensions import ReadOnly, Required import litellm from litellm._uuid import uuid @@ -104,6 +104,27 @@ class _VertexBatchRow(TypedDict, total=False): processed_time: ReadOnly[str] +class _VertexEmbeddingVector(TypedDict): + values: ReadOnly[list[float]] + + +class _VertexEmbeddingUsageMetadata(TypedDict, total=False): + promptTokenCount: ReadOnly[int] + + +class _VertexEmbeddingResponse(TypedDict, total=False): + embedding: ReadOnly[Required[_VertexEmbeddingVector]] + usageMetadata: ReadOnly[_VertexEmbeddingUsageMetadata] + tokenCount: ReadOnly[int] + + +class _VertexEmbeddingBatchRow(TypedDict, total=False): + key: ReadOnly[str] + request: ReadOnly[Mapping[str, object]] + status: ReadOnly[Required[str]] + response: ReadOnly[Required[_VertexEmbeddingResponse]] + + class _OpenAIBatchOutputError(TypedDict): code: ReadOnly[str] message: ReadOnly[str] @@ -111,7 +132,7 @@ class _OpenAIBatchOutputError(TypedDict): class _OpenAIBatchOutputResponse(TypedDict): status_code: ReadOnly[int] - request_id: ReadOnly[str] + request_id: ReadOnly[object] body: ReadOnly[Mapping[str, object]] @@ -218,7 +239,7 @@ def _get_litellm_batch_custom_id_from_labels(labels: Mapping[str, object] | None return str(labels.get("litellm_custom_id", "unknown")) -def _is_vertex_embeddings_batch_output_row(vertex_output_row: Mapping[str, Any]) -> bool: +def _is_vertex_embeddings_batch_output_row(vertex_output_row: Mapping[str, object]) -> bool: """ Whether a Vertex batch output row came from an `EmbedContentRequest`. @@ -237,7 +258,7 @@ def _is_vertex_embeddings_batch_output_row(vertex_output_row: Mapping[str, Any]) def _openai_batch_output_row( custom_id: str, - body: Mapping[str, Any] | None = None, + body: Mapping[str, object] | None = None, error_code: str | None = None, error_message: str = "", ) -> _OpenAIBatchOutputRow: @@ -259,7 +280,7 @@ def _openai_batch_output_row( } -def _split_vertex_batch_key(vertex_output_row: Mapping[str, Any]) -> tuple[str, int, int]: +def _split_vertex_batch_key(vertex_output_row: Mapping[str, object]) -> tuple[str, int, int]: """ Resolve `(custom_id, index within that custom_id, group size)` for a Vertex batch output row. @@ -278,7 +299,7 @@ def _split_vertex_batch_key(vertex_output_row: Mapping[str, Any]) -> tuple[str, return unquote(match["custom_id"]), int(match["index"]), int(match["total"]) -def _embedding_prompt_token_count(vertex_response: Mapping[str, Any]) -> int: +def _embedding_prompt_token_count(vertex_response: _VertexEmbeddingResponse) -> int: """ Prompt tokens billed for one Vertex Gemini Embedding batch row. @@ -293,7 +314,7 @@ def _embedding_prompt_token_count(vertex_response: Mapping[str, Any]) -> int: def _vertex_embeddings_rows_to_openai_batch_output_row( custom_id: str, - vertex_output_rows: tuple[Mapping[str, Any], ...], + vertex_output_rows: tuple[_VertexEmbeddingBatchRow, ...], element_indices: tuple[int, ...], element_count: int, model: str | None, @@ -348,7 +369,7 @@ def _vertex_embeddings_rows_to_openai_batch_output_row( def _transform_vertex_embeddings_batch_output_to_openai( - vertex_output_rows: Iterable[Mapping[str, Any]], + vertex_output_rows: Iterable[_VertexEmbeddingBatchRow], model: str | None, ) -> tuple[_OpenAIBatchOutputRow, ...]: """ @@ -388,7 +409,7 @@ def _model_from_managed_gcs_url(url: str) -> str | None: return match.group(1) if match else None -def _is_embeddings_batch_entry(openai_entry: Mapping[str, Any]) -> bool: +def _is_embeddings_batch_entry(openai_entry: Mapping[str, object]) -> bool: """ Whether an OpenAI batch JSONL line targets the embeddings endpoint. @@ -431,7 +452,7 @@ def _vertex_batch_embeddings_key(custom_id: str, index: int, total: int) -> str: return encoded_custom_id if total < 2 else f"{encoded_custom_id}#{index}/{total}" -def _vertex_embeddings_row(key: str | None, embed_content_request: Mapping[str, Any]) -> Mapping[str, Any]: +def _vertex_embeddings_row(key: str | None, embed_content_request: Mapping[str, object]) -> Mapping[str, object]: """ One Vertex Gemini Embedding batch input row. @@ -453,8 +474,8 @@ def _vertex_embeddings_row(key: str | None, embed_content_request: Mapping[str, def _openai_batch_jsonl_entry_to_vertex_embeddings_rows( - openai_entry: Mapping[str, Any], -) -> tuple[Mapping[str, Any], ...]: + openai_entry: Mapping[str, object], +) -> tuple[Mapping[str, object], ...]: """ Transforms a single OpenAI `/v1/embeddings` batch entry into Vertex Gemini Embedding batch rows, one per requested embedding. @@ -512,7 +533,7 @@ def _openai_batch_jsonl_entry_to_vertex_embeddings_rows( def _openai_batch_jsonl_entry_to_vertex_rows( openai_entry: dict[str, Any], map_openai_to_vertex_params: Callable[[dict[str, Any]], dict[str, Any]], -) -> tuple[Mapping[str, Any], ...]: +) -> tuple[Mapping[str, object], ...]: """ Transforms a single OpenAI JSONL batch entry into the Vertex rows it maps to. @@ -533,7 +554,7 @@ def _openai_batch_jsonl_entry_to_vertex_rows( cached_content=None, ) - custom_id: Final = openai_entry.get("custom_id") + custom_id: Final[object] = openai_entry.get("custom_id") if custom_id is not None: if "labels" not in vertex_request_body: vertex_request_body["labels"] = {} diff --git a/litellm/llms/vertex_ai/gemini/transformation.py b/litellm/llms/vertex_ai/gemini/transformation.py index 11c026010ee..e2d62be6a69 100644 --- a/litellm/llms/vertex_ai/gemini/transformation.py +++ b/litellm/llms/vertex_ai/gemini/transformation.py @@ -250,7 +250,7 @@ def _gs_uri_requires_content_type_metadata(url: str) -> bool: def _image_url_payload_may_need_sync_gcs_metadata_fetch( - raw_image_url: Any, + raw_image_url: object, ) -> bool: """ True when this image_url value (content-part image_url or assistant ``images[]`` @@ -326,7 +326,7 @@ def _openai_messages_may_need_sync_gcs_metadata_fetch( def _get_gcs_object_content_type( image_url: str, vertex_project: str | None = None, - vertex_credentials: Any | None = None, + vertex_credentials: object = None, ) -> str | None: """ Resolve content type from GCS object metadata. @@ -479,7 +479,7 @@ def _process_gemini_media( model: str | None = None, video_metadata: dict[str, Any] | None = None, vertex_project: str | None = None, - vertex_credentials: Any | None = None, + vertex_credentials: object = None, ) -> PartType: """ Given a media URL (image, audio, or video), return the appropriate PartType for Gemini @@ -1002,7 +1002,7 @@ def _gemini_convert_messages_with_history( if isinstance(_ss_invocations, list): for invocation in _ss_invocations: # Re-inject toolCall part - tc_part: dict[str, Any] = { + tc_part: dict[str, object] = { "toolCall": { "toolType": invocation.get("tool_type"), "id": invocation.get("id"), @@ -1015,13 +1015,13 @@ def _gemini_convert_messages_with_history( # Re-inject toolResponse part if response is present if "response" in invocation: - tr_dict: dict[str, Any] = { + tr_dict: dict[str, object] = { "id": invocation.get("id"), "response": invocation.get("response"), } if invocation.get("tool_type"): tr_dict["toolType"] = invocation["tool_type"] - tr_part: dict[str, Any] = {"toolResponse": tr_dict} + tr_part: dict[str, object] = {"toolResponse": tr_dict} if "response_thought_signature" in invocation: tr_part["thoughtSignature"] = invocation["response_thought_signature"] assistant_content.append(tr_part) @@ -1090,7 +1090,7 @@ def _pop_and_merge_extra_body(data: RequestBody, optional_params: dict) -> None: data_dict[k] = v -def _has_google_maps_tool(tools: Any | None) -> bool: +def _has_google_maps_tool(tools: object) -> bool: """Return True if any tool object in the list has a 'googleMaps' key.""" if not isinstance(tools, list): return False @@ -1127,7 +1127,7 @@ def _rewrite_mime_type_to_response_format(generation_config: GenerationConfig) - schema = generation_config.pop("response_schema", None) generation_config.pop("response_mime_type", None) - response_format: Final[dict[str, Any]] = {"text": {"mimeType": "APPLICATION_JSON"}} + response_format: Final[dict[str, dict[str, object]]] = {"text": {"mimeType": "APPLICATION_JSON"}} if schema is not None: response_format["text"]["schema"] = schema generation_config["responseFormat"] = response_format @@ -1316,7 +1316,7 @@ async def async_transform_request_body( timeout: float | httpx.Timeout | None, extra_headers: dict | None, optional_params: dict, - logging_obj: litellm.litellm_core_utils.litellm_logging.Logging, + logging_obj: LiteLLMLoggingObj, custom_llm_provider: Literal["vertex_ai", "vertex_ai_beta", "gemini"], litellm_params: dict, vertex_project: str | None, diff --git a/litellm/llms/vertex_ai/vertex_llm_base.py b/litellm/llms/vertex_ai/vertex_llm_base.py index aca257dc095..1942bc850f1 100644 --- a/litellm/llms/vertex_ai/vertex_llm_base.py +++ b/litellm/llms/vertex_ai/vertex_llm_base.py @@ -9,7 +9,7 @@ import json import os import threading from collections.abc import Mapping -from typing import TYPE_CHECKING, Any, Final, Literal +from typing import TYPE_CHECKING, Any, Final, Literal, Protocol from urllib.parse import urlparse import litellm @@ -47,6 +47,21 @@ else: GoogleCredentialsObject = Any +class _VertexCredentialsObject(Protocol): + """Structural view of the google-auth credentials handle that this class caches and refreshes.""" + + @property + def token(self) -> object: ... + + @property + def quota_project_id(self) -> str | None: ... + + @property + def expired(self) -> object: ... + + def refresh(self, request: object) -> None: ... + + class VertexBase: def __init__(self) -> None: super().__init__() @@ -55,7 +70,7 @@ class VertexBase: self._credentials: GoogleCredentialsObject | None = None self._credentials_project_mapping: dict[ tuple[VERTEX_CREDENTIALS_TYPES | None, str | None], - tuple[GoogleCredentialsObject, str | None], + tuple[_VertexCredentialsObject, str | None], ] = {} self.project_id: str | None = None self.async_handler: AsyncHTTPHandler | None = None @@ -109,7 +124,7 @@ class VertexBase: self, credentials: VERTEX_CREDENTIALS_TYPES | None, project_id: str | None, - ) -> tuple[Any, str]: + ) -> tuple[_VertexCredentialsObject | None, str]: if credentials is not None: if isinstance(credentials, str): _is_path: Final = os.path.exists( @@ -209,7 +224,7 @@ class VertexBase: return creds, project_id # Google Auth Helpers -- extracted for mocking purposes in tests - def _credentials_from_identity_pool(self, json_obj, scopes): + def _credentials_from_identity_pool(self, json_obj, scopes) -> _VertexCredentialsObject: try: from google.auth import identity_pool except ImportError: @@ -220,7 +235,7 @@ class VertexBase: creds = creds.with_scopes(scopes) return creds - def _credentials_from_pluggable(self, json_obj, scopes): + def _credentials_from_pluggable(self, json_obj, scopes) -> _VertexCredentialsObject: try: from google.auth import pluggable except ImportError: @@ -231,7 +246,7 @@ class VertexBase: creds = creds.with_scopes(scopes) return creds - def _credentials_from_identity_pool_with_aws(self, json_obj, scopes): + def _credentials_from_identity_pool_with_aws(self, json_obj, scopes) -> _VertexCredentialsObject: try: from google.auth import aws except ImportError: @@ -242,7 +257,7 @@ class VertexBase: creds = creds.with_scopes(scopes) return creds - def _credentials_from_authorized_user(self, json_obj, scopes): + def _credentials_from_authorized_user(self, json_obj, scopes) -> _VertexCredentialsObject: try: import google.oauth2.credentials except ImportError: @@ -250,7 +265,7 @@ class VertexBase: return google.oauth2.credentials.Credentials.from_authorized_user_info(json_obj, scopes=scopes) - def _credentials_from_service_account(self, json_obj, scopes): + def _credentials_from_service_account(self, json_obj, scopes) -> _VertexCredentialsObject: try: import google.oauth2.service_account except ImportError: @@ -258,7 +273,7 @@ class VertexBase: return google.oauth2.service_account.Credentials.from_service_account_info(json_obj, scopes=scopes) - def _credentials_from_default_auth(self, scopes): + def _credentials_from_default_auth(self, scopes) -> tuple[_VertexCredentialsObject, str | None]: try: import google.auth as google_auth except ImportError: @@ -350,7 +365,7 @@ class VertexBase: ) return api_base - def refresh_auth(self, credentials: Any) -> None: + def refresh_auth(self, credentials: _VertexCredentialsObject) -> None: try: from google.auth.transport.requests import ( Request, @@ -426,7 +441,7 @@ class VertexBase: self, credential_cache_key: tuple, project_id: str | None, - ) -> tuple[str, str, "TokenState", Any, str | None] | None: + ) -> tuple[str, str, "TokenState", _VertexCredentialsObject, str | None] | None: """ Look up cached credentials and return usable token info for FRESH or STALE tokens (both are still valid for outbound requests). STALE @@ -449,7 +464,9 @@ class VertexBase: return None return creds.token, resolved_project, token_state, creds, cached_project_id - def _unpack_cached_credentials(self, credential_cache_key: tuple) -> tuple[Any, str | None]: + def _unpack_cached_credentials( + self, credential_cache_key: tuple + ) -> tuple[_VertexCredentialsObject | None, str | None]: """ Return (credentials, project_id) from the cache, or (None, None) if not cached. Handles both tuple and legacy cache formats. @@ -461,7 +478,7 @@ class VertexBase: return cached_entry return cached_entry, cached_entry.quota_project_id or getattr(cached_entry, "project_id", None) - def _get_token_state(self, credentials: Any) -> "TokenState": + def _get_token_state(self, credentials: _VertexCredentialsObject) -> "TokenState": """ Return the token state using google-auth's TokenState enum. @@ -485,7 +502,7 @@ class VertexBase: credentials: VERTEX_CREDENTIALS_TYPES | None, project_id: str | None, credential_cache_key: tuple, - ) -> tuple[Any, str | None]: + ) -> tuple[_VertexCredentialsObject, str | None]: """Load credentials via load_auth (in thread) and cache the result.""" try: _credentials, credential_project_id = await asyncify(self.load_auth)( @@ -505,7 +522,7 @@ class VertexBase: async def _background_refresh_credentials( self, - credentials: Any, + credentials: _VertexCredentialsObject, credential_cache_key: tuple, credential_project_id: str | None, ) -> None: @@ -557,7 +574,7 @@ class VertexBase: def _schedule_background_refresh( self, - credentials: Any, + credentials: _VertexCredentialsObject, credential_cache_key: tuple, credential_project_id: str | None, ) -> None: @@ -575,7 +592,7 @@ class VertexBase: self._background_refresh_credentials(credentials, credential_cache_key, credential_project_id) ) - def _drop_background_refresh_task(_fut: asyncio.Future[Any]) -> None: + def _drop_background_refresh_task(_fut: asyncio.Future[None]) -> None: if self._background_refresh_tasks.get(credential_cache_key) is _fut: self._background_refresh_tasks.pop(credential_cache_key, None) @@ -888,7 +905,7 @@ class VertexBase: # Convert dict credentials to string for caching cache_credentials: Final = json.dumps(credentials) if isinstance(credentials, dict) else credentials credential_cache_key: Final = (cache_credentials, project_id) - _credentials: GoogleCredentialsObject | None = None + _credentials: _VertexCredentialsObject | None = None verbose_logger.debug("Checking cached credentials for project_id: %s", project_id) diff --git a/litellm/main.py b/litellm/main.py index 0c8bff16f81..c4c5bbefc4f 100644 --- a/litellm/main.py +++ b/litellm/main.py @@ -6292,18 +6292,15 @@ def embedding( if headers is not None and headers != {}: optional_params["extra_headers"] = headers - if encoding_format is not None: - optional_params["encoding_format"] = encoding_format + requested_encoding_format: Final = ( + encoding_format + or optional_params.get("encoding_format") + or get_secret_str("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT") + ) + if requested_encoding_format is None or requested_encoding_format.strip().lower() == "none": + optional_params.pop("encoding_format", None) else: - env_fmt: Final = get_secret_str("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT") - if env_fmt is not None and env_fmt.strip().lower() == "none": - optional_params.pop("encoding_format", None) - else: - _default_fmt: Final = optional_params.get("encoding_format") or env_fmt or "float" - if _default_fmt.strip().lower() == "none": - optional_params.pop("encoding_format", None) - else: - optional_params["encoding_format"] = _default_fmt + optional_params["encoding_format"] = requested_encoding_format api_version = None @@ -6952,12 +6949,18 @@ def embedding( aembedding=aembedding, headers=headers, ) - elif custom_llm_provider == "dashscope": - dashscope_key: Final = api_key or litellm.api_key or get_secret_str("DASHSCOPE_API_KEY") + elif custom_llm_provider in ("dashscope", "qwencloud", "qwen_ai_platform"): + from litellm.llms.dashscope.common_utils import ( + missing_dashscope_family_key_message, + resolve_dashscope_family_api_key, + ) + + dashscope_key: Final = resolve_dashscope_family_api_key( + custom_llm_provider=custom_llm_provider, + api_key=api_key or litellm.api_key, + ) if dashscope_key is None: - raise ValueError( - "Missing API key for DashScope. Set DASHSCOPE_API_KEY environment variable or pass api_key parameter." - ) + raise ValueError(missing_dashscope_family_key_message(custom_llm_provider)) if extra_headers is not None and isinstance(extra_headers, dict): headers = extra_headers else: diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 718e6c489fd..1b2001cdadd 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -554,6 +554,7 @@ "supports_vision": true }, "amazon.nova-sonic-v1:0": { + "deprecation_date": "2026-09-14", "input_cost_per_audio_token": 3.4e-06, "input_cost_per_token": 6e-08, "litellm_provider": "bedrock", @@ -1450,6 +1451,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "global.anthropic.claude-fable-5": { "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, @@ -1487,6 +1526,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "global.anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "us.anthropic.claude-fable-5": { "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, @@ -1524,6 +1601,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "us.anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.375e-05, + "cache_creation_input_token_cost_above_1hr": 2.2e-05, + "cache_read_input_token_cost": 2.75e-07, + "input_cost_per_token": 1.1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "eu.anthropic.claude-fable-5": { "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, @@ -1561,6 +1676,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "eu.anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.375e-05, + "cache_creation_input_token_cost_above_1hr": 2.2e-05, + "cache_read_input_token_cost": 2.75e-07, + "input_cost_per_token": 1.1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, "supports_adaptive_thinking": true, @@ -1591,7 +1744,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1627,7 +1780,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1663,7 +1816,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1699,7 +1852,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1735,7 +1888,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1771,7 +1924,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -2064,7 +2217,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2101,7 +2254,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2138,7 +2291,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2175,7 +2328,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2212,7 +2365,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2249,7 +2402,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -3045,6 +3198,7 @@ "prompt_cache_min_tokens": 2048 }, "azure_ai/claude-fable-5": { + "deprecation_date": "2027-12-05", "supports_mid_conversation_system": true, "input_cost_per_token": 1e-05, "output_cost_per_token": 5e-05, @@ -3077,7 +3231,42 @@ "supports_max_reasoning_effort": true, "prompt_cache_min_tokens": 512 }, + "azure_ai/claude-fable-5-1": { + "supports_mid_conversation_system": true, + "input_cost_per_token": 1e-05, + "output_cost_per_token": 5e-05, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512 + }, "azure_ai/claude-opus-5": { + "deprecation_date": "2027-07-08", "supports_mid_conversation_system": true, "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, @@ -3110,6 +3299,7 @@ "prompt_cache_min_tokens": 512 }, "azure_ai/claude-opus-4-8": { + "deprecation_date": "2027-09-01", "supports_mid_conversation_system": true, "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, @@ -3188,6 +3378,7 @@ "prompt_cache_min_tokens": 1024 }, "azure_ai/claude-sonnet-5": { + "deprecation_date": "2027-06-30", "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, @@ -9979,6 +10170,22 @@ "supports_reasoning": true, "supports_tool_choice": true }, + "azure_ai/deepseek-v4-flash-0731": { + "cache_read_input_token_cost": 2.8e-08, + "deprecation_date": "2026-12-03", + "input_cost_per_token": 1.9e-07, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.1e-07, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/deepseek/", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, "azure_ai/embed-v-4-0": { "input_cost_per_token": 1.2e-07, "litellm_provider": "azure_ai", @@ -12287,6 +12494,7 @@ "supports_tool_choice": true }, "cerebras/zai-glm-4.7": { + "deprecation_date": "2026-08-17", "input_cost_per_token": 2.25e-06, "litellm_provider": "cerebras", "max_input_tokens": 128000, @@ -13022,6 +13230,47 @@ "supports_native_structured_output": true, "source": "https://docs.anthropic.com/en/docs/about-claude/models/overview" }, + "claude-fable-5-1": { + "deprecation_date": "2027-09-01", + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "anthropic", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "provider_specific_entry": { + "us": 1.1 + }, + "supports_output_config": true, + "prompt_cache_min_tokens": 512, + "supports_native_structured_output": true, + "source": "https://platform.claude.com/docs/en/models/fable-5-1/overview" + }, "claude-opus-5": { "deprecation_date": "2027-07-24", "cache_creation_input_token_cost": 6.25e-06, @@ -14718,6 +14967,1910 @@ "/v1/images/generations" ] }, + "qwencloud/deepseek-v4-flash": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/deepseek-v4-flash-0731": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/deepseek-v4-pro": { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2.4e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4.8e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/glm-5.1": { + "cache_read_input_token_cost": 2.6e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 202745, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/glm-5.2": { + "cache_read_input_token_cost": 2.8e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 1048576, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/kimi-k2.7-code": { + "cache_read_input_token_cost": 1.9e-07, + "input_cost_per_token": 9.5e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 229376, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen-coder": { + "input_cost_per_token": 3e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.5e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-flash": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-flash-2025-07-28": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-max": { + "input_cost_per_token": 1.6e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 30720, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 6.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-01-25": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-04-28": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-07-14": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-07-28": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-plus-2025-09-11": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-plus-latest": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-turbo": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-turbo-2024-11-01": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-turbo-2025-04-28": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-turbo-latest": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-30b-a3b": { + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-coder-flash": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.2e-07, + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-coder-flash-2025-07-28": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-coder-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.8e-07, + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 6e-07, + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-coder-plus-2025-07-22": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-max-preview": { + "litellm_provider": "qwencloud", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwencloud/qwen3-max": { + "litellm_provider": "qwencloud", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwencloud/qwen3-max-2026-01-23": { + "litellm_provider": "qwencloud", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwencloud/qwen3-next-80b-a3b-instruct": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-next-80b-a3b-thinking": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-vl-235b-a22b-instruct": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.6e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-235b-a22b-thinking": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-32b-instruct": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 6.4e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-32b-thinking": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.87e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 260096, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 2e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 6e-07, + "output_cost_per_token": 4.8e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "qwencloud/qwen3.5-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 3e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3.7-max": { + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2.5e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 7.5e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3.7-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 4e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 2.4e-07, + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 4.8e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3.8-max": { + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwq-plus": { + "input_cost_per_token": 8e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 98304, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-image-2.0": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwencloud/qwen-image-2.0-pro": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwencloud/qwen-image-3.0": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwencloud/qwen-image-3.0-pro": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/deepseek-v4-flash": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/deepseek-v4-flash-0731": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/deepseek-v4-pro": { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2.4e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4.8e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/glm-5.1": { + "cache_read_input_token_cost": 2.6e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 202745, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/glm-5.2": { + "cache_read_input_token_cost": 2.8e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1048576, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/kimi-k2.7-code": { + "cache_read_input_token_cost": 1.9e-07, + "input_cost_per_token": 9.5e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 229376, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen-coder": { + "input_cost_per_token": 3e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.5e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-flash": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-flash-2025-07-28": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-max": { + "input_cost_per_token": 1.6e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 30720, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 6.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-01-25": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-04-28": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-07-14": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-07-28": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-plus-2025-09-11": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-plus-latest": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-turbo": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-turbo-2024-11-01": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-turbo-2025-04-28": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-turbo-latest": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-30b-a3b": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-coder-flash": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.2e-07, + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-coder-flash-2025-07-28": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-coder-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.8e-07, + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 6e-07, + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-coder-plus-2025-07-22": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-max-preview": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-max": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-max-2026-01-23": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-next-80b-a3b-instruct": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-next-80b-a3b-thinking": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-vl-235b-a22b-instruct": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.6e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-235b-a22b-thinking": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-32b-instruct": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 6.4e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-32b-thinking": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.87e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 260096, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 2e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 6e-07, + "output_cost_per_token": 4.8e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3.5-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 3e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3.7-max": { + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2.5e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 7.5e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3.7-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 4e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 2.4e-07, + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 4.8e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3.8-max": { + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwq-plus": { + "input_cost_per_token": 8e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 98304, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-image-2.0": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/qwen-image-2.0-pro": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/qwen-image-3.0": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/qwen-image-3.0-pro": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, "databricks/databricks-bge-large-en": { "cache_creation_input_token_cost": 1.0003e-07, "cache_read_input_token_cost": 1.0003e-07, @@ -15101,6 +17254,62 @@ "supports_tool_choice": true, "supports_vision": true }, + "databricks/databricks-deepseek-v4-flash-0731": { + "cache_creation_input_token_cost": 1.4e-07, + "cache_read_input_token_cost": 2.8e-08, + "input_cost_per_token": 1.4e-07, + "input_dbu_cost_per_token": 2e-06, + "litellm_provider": "databricks", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "metadata": { + "notes": "Input/output cost per token is dbu cost * $0.070. Billing reads the per-token dollar fields; the '*_dbu_cost_per_token' fields are the published Databricks rates, kept for reference. Context/max output are the DeepSeek-published model limits (1M context, 384K max output)." + }, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "output_dbu_cost_per_token": 4e-06, + "source": "https://www.databricks.com/product/pricing/foundation-model-serving", + "supported_modalities": [ + "text" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": false + }, + "databricks/databricks-deepseek-v4-pro-0813": { + "cache_creation_input_token_cost": 1.31999e-06, + "cache_read_input_token_cost": 1.3202e-07, + "input_cost_per_token": 1.31999e-06, + "input_dbu_cost_per_token": 1.8857e-05, + "litellm_provider": "databricks", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "metadata": { + "notes": "Input/output cost per token is dbu cost * $0.070. Billing reads the per-token dollar fields; the '*_dbu_cost_per_token' fields are the published Databricks rates, kept for reference. Context/max output are the DeepSeek-published model limits (1M context, 384K max output)." + }, + "mode": "chat", + "output_cost_per_token": 3.95997e-06, + "output_dbu_cost_per_token": 5.6571e-05, + "source": "https://www.databricks.com/product/pricing/foundation-model-serving", + "supported_modalities": [ + "text" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": false + }, "databricks/databricks-gemini-2-5-flash": { "cache_creation_input_token_cost": 3.0002e-07, "cache_read_input_token_cost": 3.0002e-08, @@ -20631,6 +22840,7 @@ "supports_image_size": false }, "gemini-live-2.5-flash-native-audio": { + "deprecation_date": "2026-12-13", "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "vertex_ai-language-models", @@ -23852,8 +26062,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -23867,7 +26079,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second_4k": 0.6, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -23895,8 +26108,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -23910,7 +26125,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second_4k": 0.6, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -41892,6 +44108,41 @@ "supports_max_reasoning_effort": true, "prompt_cache_min_tokens": 512 }, + "vertex_ai/claude-fable-5-1": { + "regional_endpoint_uplift_multiplier": 1.1, + "supports_mid_conversation_system": true, + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512 + }, "vertex_ai/claude-fable-5@default": { "deprecation_date": "2027-06-08", "regional_endpoint_uplift_multiplier": 1.1, @@ -41927,6 +44178,41 @@ "supports_max_reasoning_effort": true, "prompt_cache_min_tokens": 512 }, + "vertex_ai/claude-fable-5-1@default": { + "regional_endpoint_uplift_multiplier": 1.1, + "supports_mid_conversation_system": true, + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512 + }, "vertex_ai/claude-opus-5": { "deprecation_date": "2027-01-24", "regional_endpoint_uplift_multiplier": 1.1, @@ -43440,7 +45726,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second_4k": 0.6, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -43453,8 +45740,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -43469,7 +45758,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second_4k": 0.6, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -43483,8 +45773,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -52231,14 +54523,14 @@ "supports_vision": true }, "fireworks_ai/deepseek-v4-flash-0731": { - "cache_read_input_token_cost": 2.8e-08, - "input_cost_per_token": 1.4e-07, + "cache_read_input_token_cost": 7e-09, + "input_cost_per_token": 2.2e-07, "litellm_provider": "fireworks_ai", "max_input_tokens": 1048576, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", - "output_cost_per_token": 2.8e-07, + "output_cost_per_token": 6.6e-07, "source": "https://docs.fireworks.ai/serverless/pricing", "supports_function_calling": true, "supports_reasoning": true, @@ -55124,5 +57416,55 @@ "max_tokens": 40960, "mode": "embedding", "source": "https://docs.fireworks.ai/serverless/pricing" + }, + "zai/glm-5.2": { + "cache_creation_input_token_cost": 0, + "cache_read_input_token_cost": 2.6e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "zai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://docs.z.ai/guides/overview/pricing", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "together_ai/Qwen/Qwen3.8-Flash": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "together_ai", + "max_input_tokens": 1000000, + "max_tokens": 1000000, + "mode": "chat", + "output_cost_per_token": 4.7e-07, + "source": "https://docs.together.ai/docs/serverless-models" + }, + "cerebras/gemma-4-31b": { + "input_cost_per_token": 9.9e-07, + "litellm_provider": "cerebras", + "max_input_tokens": 131072, + "max_output_tokens": 40960, + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 1.49e-06, + "source": "https://api.cerebras.ai/public/v1/models/gemma-4-31b", + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "elevenlabs/scribe_v2": { + "input_cost_per_second": 6.11e-05, + "litellm_provider": "elevenlabs", + "mode": "audio_transcription", + "output_cost_per_second": 0.0, + "source": "https://elevenlabs.io/pricing/api", + "supported_endpoints": [ + "/v1/audio/transcriptions" + ] } } diff --git a/litellm/passthrough/main.py b/litellm/passthrough/main.py index 9095cee15a9..c4bd03fb1c3 100644 --- a/litellm/passthrough/main.py +++ b/litellm/passthrough/main.py @@ -6,7 +6,7 @@ from __future__ import annotations import asyncio import contextvars -from collections.abc import AsyncGenerator, AsyncIterator, Coroutine, Generator, Iterator +from collections.abc import AsyncGenerator, AsyncIterator, Awaitable, Coroutine, Generator, Iterator from functools import partial from types import TracebackType from typing import Any, Final, cast @@ -27,19 +27,19 @@ base_llm_http_handler = BaseLLMHTTPHandler() from .utils import BasePassthroughUtils -async def _as_async_generator(iterable: AsyncIterator[bytes]) -> AsyncGenerator[bytes, Any]: +async def _as_async_generator(iterable: AsyncIterator[bytes]) -> AsyncGenerator[bytes, bytes]: async for chunk in iterable: yield chunk -def _as_generator(iterable: Iterator[bytes]) -> Generator[bytes, Any, Any]: +def _as_generator(iterable: Iterator[bytes]) -> Generator[bytes, bytes, None]: yield from iterable -class AsyncPassthroughStreamingResponse(AsyncGenerator[Any, Any]): +class AsyncPassthroughStreamingResponse(AsyncGenerator[bytes, bytes]): def __init__( self, - response: Coroutine[Any, Any, httpx.Response], + response: Awaitable[httpx.Response], litellm_logging_obj: LiteLLMLoggingObj, provider_config: BasePassthroughConfig, ) -> None: @@ -48,7 +48,7 @@ class AsyncPassthroughStreamingResponse(AsyncGenerator[Any, Any]): self._headers = httpx.Headers() self._response_coro = response self._response: httpx.Response - self._iterator: AsyncGenerator[bytes, Any] + self._iterator: AsyncGenerator[bytes, bytes] self._litellm_logging_obj = litellm_logging_obj self._provider_config = provider_config self._raw_bytes: list[bytes] = [] # mutable-ok: instance buffer for streaming chunks @@ -172,7 +172,7 @@ class AsyncPassthroughStreamingResponse(AsyncGenerator[Any, Any]): pass -class PassthroughStreamingResponse(Generator[Any, Any, Any]): +class PassthroughStreamingResponse(Generator[bytes, bytes, None]): def __init__( self, response: httpx.Response, @@ -184,7 +184,7 @@ class PassthroughStreamingResponse(Generator[Any, Any, Any]): self.status_code = response.status_code self._litellm_logging_obj = litellm_logging_obj self._provider_config = provider_config - self._iterator: Generator[bytes, Any, Any] = _as_generator(response.iter_bytes()) + self._iterator: Generator[bytes, bytes, None] = _as_generator(response.iter_bytes()) self._raw_bytes: list[bytes] = [] # mutable-ok: instance buffer for streaming chunks self._flush_scheduled = False @@ -263,7 +263,7 @@ async def allm_passthrough_route( cookies: CookieTypes | None = None, client: HTTPHandler | AsyncHTTPHandler | None = None, **kwargs, -) -> httpx.Response | AsyncGenerator[Any, Any]: +) -> httpx.Response | AsyncGenerator[bytes, bytes]: """ Async: Reranks a list of documents based on their relevance to the query """ @@ -390,10 +390,10 @@ def llm_passthrough_route( **kwargs, ) -> ( httpx.Response - | Coroutine[Any, Any, httpx.Response] - | Coroutine[Any, Any, httpx.Response | AsyncGenerator[Any, Any]] - | Generator[Any, Any, Any] - | AsyncGenerator[Any, Any] + | Coroutine[object, object, httpx.Response] + | Coroutine[object, object, httpx.Response | AsyncGenerator[bytes, bytes]] + | Generator[bytes, bytes, None] + | AsyncGenerator[bytes, bytes] ): """ Pass through requests to the LLM APIs. @@ -592,7 +592,7 @@ async def _async_passthrough_request( is_streaming_request: bool, litellm_logging_obj: LiteLLMLoggingObj, provider_config: BasePassthroughConfig, -) -> httpx.Response | AsyncGenerator[Any, Any]: +) -> httpx.Response | AsyncGenerator[bytes, bytes]: """ Handle async passthrough requests. Uses async client to send request and properly handles streaming. diff --git a/litellm/provider_endpoints_support_backup.json b/litellm/provider_endpoints_support_backup.json index ead26ab65c5..9d6b1e18f59 100644 --- a/litellm/provider_endpoints_support_backup.json +++ b/litellm/provider_endpoints_support_backup.json @@ -671,6 +671,42 @@ "interactions": true } }, + "qwencloud": { + "display_name": "QwenCloud (`qwencloud`)", + "url": "https://docs.litellm.ai/docs/providers/qwencloud", + "endpoints": { + "chat_completions": true, + "messages": true, + "responses": true, + "embeddings": true, + "image_generations": true, + "audio_transcriptions": false, + "audio_speech": false, + "moderations": false, + "batches": false, + "rerank": true, + "a2a": true, + "interactions": true + } + }, + "qwen_ai_platform": { + "display_name": "Qwen AI Platform (`qwen_ai_platform`)", + "url": "https://docs.litellm.ai/docs/providers/qwencloud", + "endpoints": { + "chat_completions": true, + "messages": true, + "responses": true, + "embeddings": true, + "image_generations": true, + "audio_transcriptions": false, + "audio_speech": false, + "moderations": false, + "batches": false, + "rerank": true, + "a2a": true, + "interactions": true + } + }, "databricks": { "display_name": "Databricks (`databricks`)", "url": "https://docs.litellm.ai/docs/providers/databricks", diff --git a/litellm/proxy/_experimental/mcp_server/semantic_tool_filter.py b/litellm/proxy/_experimental/mcp_server/semantic_tool_filter.py index 7ec0f4b5192..dcf1b01bc25 100644 --- a/litellm/proxy/_experimental/mcp_server/semantic_tool_filter.py +++ b/litellm/proxy/_experimental/mcp_server/semantic_tool_filter.py @@ -5,6 +5,7 @@ Filters MCP tools semantically for /chat/completions and /responses endpoints. """ import asyncio +from collections.abc import Mapping, Sequence from typing import TYPE_CHECKING, Any, Final from litellm._logging import verbose_logger @@ -74,7 +75,7 @@ class SemanticMCPToolFilter: self.router_instance = litellm_router_instance self.tool_router: SemanticRouter | None = None self.context_window_error: str | None = None - self._tool_map: dict[str, Any] = {} # MCPTool objects or OpenAI function dicts + self._tool_map: dict[str, object] = {} # MCPTool objects or OpenAI function dicts self._index_sync_lock = asyncio.Lock() async def build_router_from_mcp_registry(self) -> None: @@ -182,11 +183,11 @@ class SemanticMCPToolFilter: return raise - def _has_tools_missing_from_index(self, tools: list[Any]) -> bool: + def _has_tools_missing_from_index(self, tools: Sequence[object]) -> bool: """Allocation-free check for any named tool not yet in the semantic index.""" return any(name and name not in self._tool_map for name in (self._extract_tool_info(t)[0] for t in tools)) - def _tools_missing_from_index(self, tools: list[Any]) -> dict[str, Any]: + def _tools_missing_from_index(self, tools: Sequence[object]) -> Mapping[str, object]: """Map name -> tool for every named tool not yet in the semantic index.""" return { name: tool @@ -194,7 +195,7 @@ class SemanticMCPToolFilter: if name and name not in self._tool_map } - async def _ensure_tools_indexed(self, available_tools: list[Any]) -> None: + async def _ensure_tools_indexed(self, available_tools: Sequence[object]) -> None: """ Index request-time tools the startup build never saw. @@ -385,7 +386,7 @@ class SemanticMCPToolFilter: separator: Final = client_name[-len(canonical) - 1] return separator in ("_", "-") - def _get_tools_by_names(self, tool_names: list[str], available_tools: list[Any]) -> list[Any]: + def _get_tools_by_names(self, tool_names: Sequence[str], available_tools: Sequence[object]) -> list[object]: """ Get tools from available_tools by their names, preserving the semantic router's ordering. @@ -401,14 +402,14 @@ class SemanticMCPToolFilter: # Exact matches win over suffix matches when both are present, and # each incoming tool is returned at most once even if two canonical # names happen to be tail-compatible with the same incoming name. - available_by_name: Final[dict[str, Any]] = {} + available_by_name: Final[dict[str, object]] = {} for tool in available_tools: client_name, _ = self._extract_tool_info(tool) if client_name and client_name not in available_by_name: available_by_name[client_name] = tool - matched: Final[list[Any]] = [] - used_ids: Final[set] = set() + matched: Final[list[object]] = [] + used_ids: Final[set[int]] = set() for canonical in tool_names: tool = available_by_name.get(canonical) if tool is None: @@ -430,7 +431,7 @@ class SemanticMCPToolFilter: used_ids.add(id(tool)) return matched - def extract_user_query(self, messages: list[dict[str, Any]]) -> str: + def extract_user_query(self, messages: Sequence[Mapping[str, object]]) -> str: """ Extract user query from messages for /chat/completions or /responses. diff --git a/litellm/proxy/agent_endpoints/a2a_endpoints.py b/litellm/proxy/agent_endpoints/a2a_endpoints.py index bd02cfdf907..31b05320cd3 100644 --- a/litellm/proxy/agent_endpoints/a2a_endpoints.py +++ b/litellm/proxy/agent_endpoints/a2a_endpoints.py @@ -14,7 +14,7 @@ import json from collections.abc import AsyncGenerator, Mapping from copy import deepcopy from types import MappingProxyType -from typing import TYPE_CHECKING, Any, Final +from typing import TYPE_CHECKING, Any, Final, Protocol from urllib.parse import urlparse from fastapi import APIRouter, Depends, HTTPException, Request, Response @@ -215,11 +215,20 @@ def _enforce_inbound_trace_id(agent: "AgentResponse", request: Request) -> None: ) +class _JsonRpcResponse(Protocol): + def json(self) -> dict[str, object]: ... + + +def _jsonrpc_body(response: _JsonRpcResponse) -> dict[str, object]: + """The decoded JSON-RPC body of ``response``.""" + return response.json() + + async def _forward_jsonrpc( agent_url: str, body: dict[str, object], extra_headers: Mapping[str, str] | None = None, -) -> dict[str, Any]: +) -> dict[str, object]: from litellm.llms.custom_httpx.http_handler import get_async_httpx_client from litellm.types.llms.custom_http import httpxSpecialProvider @@ -230,7 +239,7 @@ async def _forward_jsonrpc( ) resp: Final = await handler.post(agent_url, json=body, headers=headers) try: - result: Final = resp.json() + result: Final = _jsonrpc_body(resp) except Exception: resp.raise_for_status() raise @@ -940,8 +949,8 @@ async def invoke_agent_a2a( ) result = await _forward_jsonrpc(agent_url, forward_body, extra_headers=caller_headers) if method == "agent/getAuthenticatedExtendedCard": - if isinstance(result.get("result"), dict): - card: Final = result["result"] + card: Final = result.get("result") + if isinstance(card, dict): proxy_url: Final = get_custom_url(str(request.base_url), route=f"a2a/{agent_id}") # Rewrite the upstream agent URL in both 0.3 (top-level `url`) # and 1.0 (`supportedInterfaces[0].url`) wire formats so that diff --git a/litellm/proxy/auth/handle_jwt.py b/litellm/proxy/auth/handle_jwt.py index 39e6ca9a369..0795cee7409 100644 --- a/litellm/proxy/auth/handle_jwt.py +++ b/litellm/proxy/auth/handle_jwt.py @@ -14,8 +14,8 @@ import hashlib import os import re import time -from collections.abc import Awaitable, Callable -from typing import Any, Final, Literal, NoReturn, TypeVar, cast +from collections.abc import Awaitable, Callable, Sequence +from typing import Any, Final, Literal, NoReturn, Protocol, TypeVar, cast import httpx import jwt @@ -24,6 +24,7 @@ from cryptography.hazmat.backends import default_backend from cryptography.hazmat.primitives import serialization from fastapi import HTTPException, status from jwt.api_jwk import PyJWK +from typing_extensions import ReadOnly, TypedDict from litellm._logging import verbose_proxy_logger from litellm.litellm_core_utils.dot_notation_indexing import get_nested_value @@ -93,6 +94,47 @@ UNREACHABLE_CACHE_KEY_PREFIX: Final = "litellm_jwks_unreachable_" _CachedValueT = TypeVar("_CachedValueT", bound=JWKKeyValue | str) +class _JWTAuthSettings(Protocol): + """The JWT auth settings block this handler reads back through ``getattr``, when one is configured.""" + + @property + def issuers(self) -> Sequence[JWTIssuerConfig] | None: ... + + @property + def public_key_ttl(self) -> float: ... + + @property + def public_key_stale_ttl(self) -> float: ... + + +class _OIDCDiscoveryBody(TypedDict, total=False): + """Decoded OIDC discovery document, read for the JWKS endpoint it advertises.""" + + jwks_uri: ReadOnly[str] + + +class _OIDCDiscoveryResponse(Protocol): + """The discovery endpoint's HTTP response, read for the decoded document it carries.""" + + def json(self) -> _OIDCDiscoveryBody: ... + + +class _UserInfoResponse(Protocol): + """The OIDC UserInfo endpoint's HTTP response, read for the identity document it carries.""" + + def json(self) -> dict[str, object]: ... + + +def _discovery_document(response: _OIDCDiscoveryResponse) -> _OIDCDiscoveryBody: + """Decode an OIDC discovery response body.""" + return response.json() + + +def _userinfo_document(response: _UserInfoResponse) -> dict[str, object]: + """Decode an OIDC UserInfo response body into its JSON object form.""" + return response.json() + + def jwks_unavailable_exception(error: JWKSUnreachableError) -> ProxyException: return ProxyException( message=( @@ -794,7 +836,7 @@ class JWTHandler: f"JWT Auth: OIDC discovery endpoint {url} returned status {response.status_code}: {response.text}" ) try: - discovery: Final = response.json() + discovery: Final = _discovery_document(response) except Exception as e: raise Exception(f"JWT Auth: Failed to parse OIDC discovery document at {url}: {e}") @@ -806,13 +848,13 @@ class JWTHandler: return jwks_uri def _get_public_key_cache_ttl(self) -> float: - litellm_jwtauth: Final = getattr(self, "litellm_jwtauth", None) + litellm_jwtauth: Final[_JWTAuthSettings | None] = getattr(self, "litellm_jwtauth", None) if litellm_jwtauth is None: return 600 return litellm_jwtauth.public_key_ttl def _get_public_key_stale_ttl(self) -> float: - litellm_jwtauth: Final = getattr(self, "litellm_jwtauth", None) + litellm_jwtauth: Final[_JWTAuthSettings | None] = getattr(self, "litellm_jwtauth", None) if litellm_jwtauth is None: return DEFAULT_JWKS_STALE_TTL return litellm_jwtauth.public_key_stale_ttl @@ -938,7 +980,7 @@ class JWTHandler: if response.status_code != 200: raise Exception(f"OIDC UserInfo endpoint returned status {response.status_code}: {response.text}") - userinfo: Final = response.json() + userinfo: Final = _userinfo_document(response) verbose_proxy_logger.debug("Received OIDC UserInfo: %s", userinfo) # Cache the userinfo response @@ -996,7 +1038,7 @@ class JWTHandler: } def _get_configured_issuer(self, token: str) -> JWTIssuerConfig | None: - litellm_jwtauth: Final = getattr(self, "litellm_jwtauth", None) + litellm_jwtauth: Final[_JWTAuthSettings | None] = getattr(self, "litellm_jwtauth", None) if litellm_jwtauth is None: return None diff --git a/litellm/proxy/common_utils/debug_utils.py b/litellm/proxy/common_utils/debug_utils.py index 3a1d18b48cc..554a6ae8d1a 100644 --- a/litellm/proxy/common_utils/debug_utils.py +++ b/litellm/proxy/common_utils/debug_utils.py @@ -6,9 +6,11 @@ import os import sys import tracemalloc from collections import Counter -from typing import Any, Final +from collections.abc import Mapping, Sequence +from typing import Any, Final, NamedTuple, Protocol, TypedDict from fastapi import APIRouter, Depends, HTTPException, Query +from typing_extensions import ReadOnly from litellm import get_secret_str from litellm._logging import verbose_proxy_logger @@ -194,6 +196,42 @@ async def memory_usage_in_mem_cache_items( } +class _ProcessMemoryInfo(Protocol): + """The resident and virtual sizes psutil reports for a process.""" + + @property + def rss(self) -> int: ... + + @property + def vms(self) -> int: ... + + +class _ProcessHandle(Protocol): + """The psutil process handle members this module reads.""" + + def memory_info(self) -> _ProcessMemoryInfo: ... + + def memory_percent(self) -> float: ... + + +class _ProcessMemoryUsage(NamedTuple): + """Memory usage of a single worker process.""" + + resident_megabytes: float + virtual_megabytes: float + percent: float + + +def _process_memory_usage(process: _ProcessHandle) -> _ProcessMemoryUsage: + """Read resident/virtual megabytes and system memory share for ``process``.""" + memory_info: Final = process.memory_info() + return _ProcessMemoryUsage( + resident_megabytes=memory_info.rss / (1024 * 1024), + virtual_megabytes=memory_info.vms / (1024 * 1024), + percent=process.memory_percent(), + ) + + @router.get("/debug/memory/summary", include_in_schema=False) async def get_memory_summary( _: UserAPIKeyAuth = Depends(user_api_key_auth), @@ -227,10 +265,9 @@ async def get_memory_summary( try: import psutil - process: Final = psutil.Process() - memory_info: Final = process.memory_info() - memory_mb: Final = memory_info.rss / (1024 * 1024) - memory_percent: Final = process.memory_percent() + usage: Final = _process_memory_usage(psutil.Process()) + memory_mb: Final = usage.resident_megabytes + memory_percent: Final = usage.percent process_memory = { "summary": f"{memory_mb:.1f} MB ({memory_percent:.1f}% of system memory)", @@ -252,7 +289,7 @@ async def get_memory_summary( process_memory["error"] = str(e) # Get cache information - caches: Final[dict[str, Any]] = {} + caches: Final[dict[str, object]] = {} total_cache_items = 0 try: @@ -313,7 +350,7 @@ async def get_memory_summary( } -def _get_gc_statistics() -> dict[str, Any]: +def _get_gc_statistics() -> Mapping[str, object]: """Get garbage collector statistics.""" return { "enabled": gc.isenabled(), @@ -341,30 +378,42 @@ def _get_gc_statistics() -> dict[str, Any]: } -def _get_object_type_counts(top_n: int) -> tuple[int, list[dict[str, Any]]]: +class _ObjectTypeCount(TypedDict): + """One row of the tracked-object histogram.""" + + type: ReadOnly[str] + count: ReadOnly[int] + count_readable: ReadOnly[str] + + +def _type_name_counts(objects: Sequence[object]) -> Counter[str]: + """Count ``objects`` by the name of their type.""" + return Counter(type(obj).__name__ for obj in objects) + + +def _get_object_type_counts(top_n: int) -> tuple[int, list[_ObjectTypeCount]]: """Count objects by type and return total count and top N types.""" - type_counts: Final[Counter] = Counter() - total_objects = 0 + type_counts: Final = _type_name_counts(gc.get_objects()) - for obj in gc.get_objects(): - total_objects += 1 - obj_type = type(obj).__name__ - type_counts[obj_type] += 1 - - top_object_types: Final = [ + top_object_types: Final[list[_ObjectTypeCount]] = [ {"type": obj_type, "count": count, "count_readable": f"{count:,}"} for obj_type, count in type_counts.most_common(top_n) ] - return total_objects, top_object_types + return sum(type_counts.values()), top_object_types -def _get_uncollectable_objects_info() -> dict[str, Any]: +def _type_names(objects: Sequence[object]) -> Sequence[str]: + """The type name of each object in ``objects``.""" + return [type(obj).__name__ for obj in objects] + + +def _get_uncollectable_objects_info() -> Mapping[str, object]: """Get information about uncollectable objects (potential memory leaks).""" uncollectable: Final = gc.garbage return { "count": len(uncollectable), - "sample_types": [type(obj).__name__ for obj in uncollectable[:10]], + "sample_types": _type_names(uncollectable[:10]), "warning": ( "If count > 0, you may have reference cycles preventing garbage collection" if len(uncollectable) > 0 @@ -373,9 +422,11 @@ def _get_uncollectable_objects_info() -> dict[str, Any]: } -def _get_cache_memory_stats(user_api_key_cache, llm_router, proxy_logging_obj, redis_usage_cache) -> dict[str, Any]: +def _get_cache_memory_stats( + user_api_key_cache, llm_router, proxy_logging_obj, redis_usage_cache +) -> Mapping[str, object]: """Calculate memory usage for all caches.""" - cache_stats: Final[dict[str, Any]] = {} + cache_stats: Final[dict[str, object]] = {} try: # User API key cache user_cache_size: Final = sys.getsizeof(user_api_key_cache.in_memory_cache.cache_dict) @@ -439,9 +490,9 @@ def _get_cache_memory_stats(user_api_key_cache, llm_router, proxy_logging_obj, r return cache_stats -def _get_router_memory_stats(llm_router) -> dict[str, Any]: +def _get_router_memory_stats(llm_router) -> Mapping[str, object]: """Get memory usage statistics for LiteLLM router.""" - litellm_router_memory: dict[str, Any] = {} + litellm_router_memory: dict[str, object] = {} try: if llm_router is not None: # Model list memory size @@ -505,7 +556,7 @@ def _get_router_memory_stats(llm_router) -> dict[str, Any]: return litellm_router_memory -def _get_process_memory_info(worker_pid: int, include_process_info: bool) -> dict[str, Any] | None: +def _get_process_memory_info(worker_pid: int, include_process_info: bool) -> Mapping[str, object] | None: """Get process-level memory information using psutil.""" if not include_process_info: return None @@ -514,10 +565,10 @@ def _get_process_memory_info(worker_pid: int, include_process_info: bool) -> dic import psutil process: Final = psutil.Process() - memory_info: Final = process.memory_info() - ram_usage_mb: Final = round(memory_info.rss / (1024 * 1024), 2) - virtual_memory_mb: Final = round(memory_info.vms / (1024 * 1024), 2) - memory_percent: Final = round(process.memory_percent(), 2) + usage: Final = _process_memory_usage(process) + ram_usage_mb: Final = round(usage.resident_megabytes, 2) + virtual_memory_mb: Final = round(usage.virtual_megabytes, 2) + memory_percent: Final = round(usage.percent, 2) return { "pid": worker_pid, diff --git a/litellm/proxy/db/db_spend_update_writer.py b/litellm/proxy/db/db_spend_update_writer.py index 202a95ba29b..e6880d521f1 100644 --- a/litellm/proxy/db/db_spend_update_writer.py +++ b/litellm/proxy/db/db_spend_update_writer.py @@ -211,7 +211,7 @@ class DBSpendUpdateWriter: org_id: str | None, # Completion object fields kwargs: dict | None, - completion_response: litellm.ModelResponse | Any | Exception | None, + completion_response: object, start_time: datetime | None, end_time: datetime | None, response_cost: float | None, @@ -323,7 +323,7 @@ class DBSpendUpdateWriter: async def _enqueue_tool_usage_transaction( self, payload: SpendLogsPayload, - completion_response: "litellm.ModelResponse | Any | Exception | None", + completion_response: object, prisma_client: "PrismaClient | None", kwargs: "dict | None" = None, ) -> None: @@ -396,7 +396,7 @@ class DBSpendUpdateWriter: def _enqueue_tool_registry_upsert( self, kwargs: dict | None, - completion_response: Any | None, + completion_response: object, hashed_token: str | None = None, team_id: str | None = None, ) -> None: @@ -849,7 +849,7 @@ class DBSpendUpdateWriter: return # Parse tags from JSON string - tags = [] + tags: Sequence[object] = [] if isinstance(request_tags, str): tags = safe_json_loads(request_tags, default=[]) if not tags: @@ -2260,7 +2260,7 @@ class DBSpendUpdateWriter: verbose_proxy_logger.debug("request_tags is None for request. Skipping incrementing tag spend.") return - request_tags = [] + request_tags: Sequence[str] = [] if isinstance(payload["request_tags"], str): request_tags = json.loads(payload["request_tags"]) elif isinstance(payload["request_tags"], list): diff --git a/litellm/proxy/guardrails/guardrail_hooks/akto/akto.py b/litellm/proxy/guardrails/guardrail_hooks/akto/akto.py index 3716d00774f..2c27531cea1 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/akto/akto.py +++ b/litellm/proxy/guardrails/guardrail_hooks/akto/akto.py @@ -162,10 +162,10 @@ class AktoGuardrail(CustomGuardrail): def build_request_body( inputs: GenericGuardrailAPIInputs, request_data: dict | None = None, - ) -> dict[str, Any]: + ) -> dict[str, object]: """Build the LLM request body from guardrail inputs (messages, model, tools).""" model: Final = inputs.get("model", "") or "" - body: Final[dict[str, Any]] = {"model": model} + body: Final[dict[str, object]] = {"model": model} structured: Final = inputs.get("structured_messages") if structured: @@ -194,7 +194,7 @@ class AktoGuardrail(CustomGuardrail): def build_response_body( inputs: GenericGuardrailAPIInputs, request_data: dict | None = None, - ) -> dict[str, Any]: + ) -> dict[str, object]: """Build the LLM response body, preferring the actual model response if available.""" model_response: Final = request_data.get("response") if request_data else None if model_response is not None and hasattr(model_response, "model_dump"): @@ -224,7 +224,7 @@ class AktoGuardrail(CustomGuardrail): *, status_code: int = 200, include_response: bool = False, - ) -> dict[str, Any]: + ) -> dict[str, object]: """Build the flat MIRRORING payload sent to Akto's HTTP proxy endpoint. All body fields use double-encoding: json.dumps({"body": json.dumps(actual_body)}) diff --git a/litellm/proxy/guardrails/guardrail_hooks/alice/__init__.py b/litellm/proxy/guardrails/guardrail_hooks/alice/__init__.py new file mode 100644 index 00000000000..75ea16f7a88 --- /dev/null +++ b/litellm/proxy/guardrails/guardrail_hooks/alice/__init__.py @@ -0,0 +1,34 @@ +from typing import TYPE_CHECKING, Final + +from litellm.types.guardrails import SupportedGuardrailIntegrations + +from .alice import AliceGuardrail + +if TYPE_CHECKING: + from litellm.types.guardrails import Guardrail, LitellmParams + + +def initialize_guardrail(litellm_params: "LitellmParams", guardrail: "Guardrail"): + import litellm + + _alice_guardrail_callback: Final = AliceGuardrail( + api_key=litellm_params.api_key, + api_base=litellm_params.api_base, + unreachable_fallback=getattr(litellm_params, "unreachable_fallback", "fail_closed"), + guardrail_name=guardrail.get("guardrail_name", ""), + event_hook=litellm_params.mode, + default_on=litellm_params.default_on, + ) + + litellm.logging_callback_manager.add_litellm_callback(_alice_guardrail_callback) + return _alice_guardrail_callback + + +guardrail_initializer_registry: Final = { # mutable-ok: module-level registry, built once and never mutated + SupportedGuardrailIntegrations.ALICE.value: initialize_guardrail, +} + + +guardrail_class_registry: Final = { # mutable-ok: module-level registry, built once and never mutated + SupportedGuardrailIntegrations.ALICE.value: AliceGuardrail, +} diff --git a/litellm/proxy/guardrails/guardrail_hooks/alice/alice.py b/litellm/proxy/guardrails/guardrail_hooks/alice/alice.py new file mode 100644 index 00000000000..27018769909 --- /dev/null +++ b/litellm/proxy/guardrails/guardrail_hooks/alice/alice.py @@ -0,0 +1,369 @@ +# +-------------------------------------------------------------+ +# +# Use Alice for your LLM calls +# https://alice.io/ +# +# +-------------------------------------------------------------+ + +import json +import os +from collections.abc import Mapping +from typing import ( + TYPE_CHECKING, + Any, # noqa: TID251 # **kwargs forwards verbatim to CustomGuardrail.__init__; see ruff-strict.toml + Final, + Literal, + Optional, +) + +import httpx +from typing_extensions import NotRequired, ReadOnly, TypedDict + +from litellm._logging import verbose_proxy_logger +from litellm.exceptions import GuardrailRaisedException, Timeout +from litellm.integrations.custom_guardrail import ( + CustomGuardrail, + log_guardrail_information, +) +from litellm.llms.custom_httpx.http_handler import ( + get_async_httpx_client, + httpxSpecialProvider, +) +from litellm.types.guardrails import GuardrailEventHooks +from litellm.types.utils import GenericGuardrailAPIInputs + +if TYPE_CHECKING: + from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj + +GUARDRAIL_NAME: Final = "alice" + +_DEFAULT_API_BASE: Final = "https://api.alice.io" +_EVALUATE_PATH: Final = "/v2/evaluate/litellm" + +_VERDICT_ALLOW: Final = "ALLOW" +_VERDICT_BLOCK: Final = "BLOCK" +_VERDICT_MASK: Final = "MASK" +_VERDICT_DETECT: Final = "DETECT" +_KNOWN_VERDICTS: Final = frozenset({_VERDICT_ALLOW, _VERDICT_BLOCK, _VERDICT_MASK, _VERDICT_DETECT}) + +_DEFAULT_BLOCK_MESSAGE: Final = "Blocked by your organization's content policy." + +# apply_guardrail selects nothing: it forwards whichever of these came populated and lets Alice +# decide what is worth evaluating. Only skip the call when every one of them is empty — there is +# then genuinely nothing to send. +_SELECTABLE_INPUT_FIELDS: Final = ("texts", "images", "tools", "tool_calls", "structured_messages") + +# Caps on the outbound copy of request_data. A payload deeper or wider than this is malformed +# rather than large, and serializing it would cost more than the evaluation it feeds. +_MAX_DEPTH: Final = 12 +_MAX_ITEMS: Final = 5000 + +# request_data carries the caller's raw credentials under these keys, at any nesting depth — +# a real captured payload puts inbound headers at request_data["proxy_server_request"]["headers"], +# again under ["metadata"]["headers"] / ["litellm_metadata"]["headers"], and again under +# ["metadata"]["requester_metadata"]["headers"], any of which can carry an Authorization or +# x-api-key value. LiteLLM's own spend-log sanitizer excludes `secret_fields` for the same reason +# (spend_tracking_utils._SENSITIVE_REQUEST_BODY_KEYS): `secret_fields.raw_headers` holds the +# caller's Authorization / x-api-key in the clear, and `api_key` can carry a forwarded provider +# credential. Stripping by key name rather than by path means a new nesting path can never +# reintroduce the leak. Posting any of these to a third-party guardrail endpoint would be worse +# than what the proxy already refuses to persist in its own audit trail — so none of them leave +# the process. +_CREDENTIAL_KEYS_TO_STRIP: Final = frozenset( + {"secret_fields", "api_key", "raw_headers", "headers", "provider_specific_header"} +) + + +class AliceReplacement(TypedDict): + """A masked substitution, positional against the texts that were submitted.""" + + index: ReadOnly[NotRequired[int]] + text: ReadOnly[NotRequired[str]] + + +class AliceVerdict(TypedDict): + """Body returned by Alice's LiteLLM evaluate endpoint.""" + + verdict: ReadOnly[NotRequired[str]] + categories: ReadOnly[NotRequired["tuple[str, ...]"]] + correlation_id: ReadOnly[NotRequired[str]] + message: ReadOnly[NotRequired[str]] + replacements: ReadOnly[NotRequired["tuple[AliceReplacement, ...]"]] + + +class AliceGuardrailMissingSecrets(Exception): + """Raised when the Alice API key is not configured.""" + + +class AliceGuardrail(CustomGuardrail): + """ + Alice — policy-based guardrails for prompts and model responses. + + This forwards the hook's arguments as it received them and enforces the verdict that comes + back, with one deliberate exception: any key named `secret_fields`, `api_key`, `raw_headers`, + `headers`, or `provider_specific_header` is dropped from `request_data` at any nesting depth + before it is serialized, and never reaches Alice. Short of that, it selects nothing and + renames nothing: which parts of a conversation are worth evaluating, and how a verdict is + reached, are decided by Alice — so changing either is a change on their side rather than a + LiteLLM upgrade. A batch with nothing selectable at all (no `texts`, `images`, `tools`, + `tool_calls`, or `structured_messages`) still skips the call, since there would be nothing to + send. + + Known limitation: the unified guardrail's `streaming_transform_mode` defaults to + `block_only`, whose streaming path discards any returned text rewrite. A MASK verdict is + therefore a no-op on a streamed response — the original, unmasked text still reaches the + caller — while BLOCK continues to function on both streamed and non-streamed responses. + This is `during_call`'s documented behavior generally, not specific to Alice; configure a + masking-aware `streaming_transform_mode` if that gap matters for your traffic. + + Alice evaluates against policies configured per *application*, and one proxy typically fronts + several, so the application is named on the virtual key rather than in this config: + + curl $PROXY/key/generate -H "Authorization: Bearer $LITELLM_MASTER_KEY" \\ + -d '{"key_alias": "payments-bot", + "metadata": {"alice_app_id": "payments-bot"}}' + + Alice reads that off the authenticated key. Because the proxy strips caller-supplied + `user_api_key_*` from the request before a guardrail sees it, a caller cannot point its own + traffic at an application with laxer policies than the one its key was issued for. + + Configuration example (litellm config YAML): + guardrails: + - guardrail_name: alice + litellm_params: + guardrail: alice + mode: [pre_call, post_call] + api_key: os.environ/ALICE_API_KEY + api_base: https://api.alice.io # optional + unreachable_fallback: fail_closed # optional + """ + + def __init__( + self, + api_key: str | None = None, + api_base: str | None = None, + unreachable_fallback: Literal["fail_closed", "fail_open"] = "fail_closed", + **kwargs: Any, # kwargs-ok: forwarded verbatim to CustomGuardrail.__init__, whose param list is wide and evolving + ) -> None: + self.async_handler = get_async_httpx_client(llm_provider=httpxSpecialProvider.GuardrailCallback) + + alice_api_key: Final = api_key or os.environ.get("ALICE_API_KEY") + if not alice_api_key: + raise AliceGuardrailMissingSecrets( + "Alice API key is required. Set the `ALICE_API_KEY` environment variable or " + "pass `api_key` in the guardrail config." + ) + self.alice_api_key: str = alice_api_key + + base: Final = (api_base or os.environ.get("ALICE_API_BASE") or _DEFAULT_API_BASE).rstrip("/") + self.api_base: str = f"{base}{_EVALUATE_PATH}" + self.unreachable_fallback: Literal["fail_closed", "fail_open"] = unreachable_fallback + + if "supported_event_hooks" not in kwargs: + kwargs["supported_event_hooks"] = [ # mutable-ok: CustomGuardrail.__init__ requires a list here + GuardrailEventHooks.pre_call, + GuardrailEventHooks.during_call, + GuardrailEventHooks.post_call, + ] + + super().__init__(**kwargs) + + @log_guardrail_information + async def apply_guardrail( + self, + inputs: GenericGuardrailAPIInputs, + request_data: dict[str, object], # mutable-ok: overrides CustomGuardrail.apply_guardrail's plain-dict contract + input_type: Literal["request", "response"], + logging_obj: Optional["LiteLLMLoggingObj"] = None, + ) -> GenericGuardrailAPIInputs: + if not any(inputs.get(field) for field in _SELECTABLE_INPUT_FIELDS): + return inputs + + try: + verdict: AliceVerdict = await self._evaluate( + inputs=inputs, request_data=request_data, input_type=input_type + ) + except Timeout as e: + return self._on_unreachable(e, inputs) + except httpx.HTTPStatusError as e: + status_code: Final = getattr(getattr(e, "response", None), "status_code", None) + # Any 5xx is an outage on Alice's side, not our misconfiguration — route the whole + # class through the configured policy. A 4xx (rejected credential, bad request) is + # ours to fix and must never fail open, so it is deliberately left to propagate. + if isinstance(status_code, int) and 500 <= status_code < 600: + return self._on_unreachable(e, inputs) + raise + except httpx.RequestError as e: + return self._on_unreachable(e, inputs) + except (json.JSONDecodeError, UnicodeDecodeError, TypeError) as e: + # A body that cannot be decoded, cannot be parsed as JSON, or parses to something + # other than an object, is as unreachable as a dropped connection: this deployment's + # policy decides, not a raw exception. UnicodeDecodeError is named explicitly because + # it is a sibling of JSONDecodeError under ValueError, not a subclass of it. + return self._on_unreachable(e, inputs) + + return self._enforce(verdict, inputs) + + async def _evaluate( + self, + inputs: GenericGuardrailAPIInputs, + request_data: Mapping[str, object], + input_type: str, + ) -> AliceVerdict: + response: Final = await self.async_handler.post( + url=self.api_base, + json={ # mutable-ok: one-shot HTTP request body, never mutated after construction + "input_type": input_type, + "inputs": _json_safe(inputs), + "request_data": _json_safe(request_data, strip_keys=_CREDENTIAL_KEYS_TO_STRIP), + }, + headers={ # mutable-ok: one-shot HTTP headers, never mutated after construction + "Content-Type": "application/json", + "af-api-key": self.alice_api_key, + }, + ) + response.raise_for_status() + body = response.json() + if not isinstance(body, dict): + raise TypeError("Alice returned a non-object body") + return body + + def _enforce(self, verdict: AliceVerdict, inputs: GenericGuardrailAPIInputs) -> GenericGuardrailAPIInputs: + """Act on the verdict. An answer we cannot read is treated as unavailable, never as a pass.""" + name: Final = verdict.get("verdict") + if name not in _KNOWN_VERDICTS: + return self._on_unreachable(ValueError(f"unrecognized verdict: {name!r}"), inputs) + + if name == _VERDICT_BLOCK: + raise GuardrailRaisedException( + guardrail_name=GUARDRAIL_NAME, + message=verdict.get("message") or _DEFAULT_BLOCK_MESSAGE, + should_wrap_with_default_message=False, + blocked_content=True, + ) + + if name == _VERDICT_DETECT: + # Recorded by Alice and allowed through. The correlation id is what ties this request + # to that record; the evaluated text itself is never logged. + verbose_proxy_logger.warning( + "Alice guardrail: detection recorded, request allowed (correlation_id=%s, categories=%s)", + verdict.get("correlation_id"), + verdict.get("categories"), + ) + return inputs + + if name == _VERDICT_MASK: + self._apply_replacements(verdict, inputs) + + return inputs + + def _apply_replacements(self, verdict: AliceVerdict, inputs: GenericGuardrailAPIInputs) -> None: + """ + Write each replacement onto the text it names. + + Only `texts` is touched. The chat translation layer maps a returned `texts` list back onto + the request positionally, but takes a different branch entirely when `structured_messages` + comes back as a new object — which would drop these edits. + + All-or-nothing: a single out-of-range or malformed replacement blocks the whole verdict + rather than being silently skipped, so content Alice meant to replace can never reach the + model unmasked alongside content that was replaced. + """ + texts: Final = inputs.get("texts") or [] # mutable-ok: empty-list fallback, replaced wholesale below + replacements: Final = verdict.get("replacements") or [] # mutable-ok: empty-list fallback for iteration only + + if not replacements: + raise self._mask_rejected(verdict) + + for replacement in replacements: + index = replacement.get("index") + text = replacement.get("text") + if not (isinstance(index, int) and isinstance(text, str) and 0 <= index < len(texts)): + raise self._mask_rejected(verdict) + texts[index] = text # mutable-ok: item assignment into the local working copy above + + inputs["texts"] = texts + + def _mask_rejected(self, verdict: AliceVerdict) -> GuardrailRaisedException: + """A MASK verdict that cannot be applied in full is refused outright, never partially — + see `_apply_replacements`.""" + return GuardrailRaisedException( + guardrail_name=GUARDRAIL_NAME, + message=verdict.get("message") or _DEFAULT_BLOCK_MESSAGE, + should_wrap_with_default_message=False, + blocked_content=True, + ) + + def _on_unreachable(self, error: Exception, inputs: GenericGuardrailAPIInputs) -> GenericGuardrailAPIInputs: + """Apply the configured policy when Alice cannot be reached or cannot be understood.""" + if self.unreachable_fallback == "fail_open": + verbose_proxy_logger.critical( + "Alice guardrail unreachable, allowing request per unreachable_fallback: %s", + error, + ) + return inputs + raise GuardrailRaisedException( + guardrail_name=GUARDRAIL_NAME, + message="Alice guardrail is unavailable and this request cannot be checked", + should_wrap_with_default_message=False, + ) from error + + @staticmethod + def get_config_model() -> type | None: + from litellm.types.proxy.guardrails.guardrail_hooks.alice import ( + AliceGuardrailConfigModel, + ) + + return AliceGuardrailConfigModel + + +def _json_safe( + value: object, + depth: int = 0, + seen: frozenset[int] = frozenset(), + strip_keys: frozenset[str] = frozenset(), +) -> object: + """ + Copy `value` into something `json.dumps` accepts, dropping only what cannot cross. + + `request_data` carries live Python objects — an OpenTelemetry span among them — so it cannot + be serialized as it stands. What is dropped is decided by a mechanical rule rather than a + field list: a list drifts from what the far side needs, a rule cannot. Serializing naively + raises, and that error would be read as "guardrail unavailable" on every single request. + + `strip_keys` drops a dict key by name at every depth it appears, not just the root — a caller + passes `_CREDENTIAL_KEYS_TO_STRIP` here so a credential nested under any path is caught the + same way a top-level one is, without maintaining a list of paths. The source object is never + mutated: every branch below builds a new container. + """ + if isinstance(value, (str, int, float, bool)) or value is None: + return value + if depth >= _MAX_DEPTH or id(value) in seen: + return None + + nested: Final = seen | {id(value)} # mutable-ok: one-shot set literal, unioned into a frozenset immediately + + if isinstance(value, dict): + out: dict[str, object] = {} # mutable-ok: bounded accumulator local to this call, never escapes as-is + for key, item in list(value.items())[:_MAX_ITEMS]: # mutable-ok: list() only to slice an unordered view + if isinstance(key, str) and key not in strip_keys: + out[key] = _json_safe(item, depth + 1, nested, strip_keys) + return out + + if isinstance(value, (list, tuple, set, frozenset)): + return [ # mutable-ok: return value is a one-shot list, discarded by the caller after use + _json_safe(item, depth + 1, nested, strip_keys) + for item in list(value)[:_MAX_ITEMS] # mutable-ok: list() only to slice an unordered view + ] + + dump: Final = getattr(value, "model_dump", None) + if callable(dump): + try: + return _json_safe(dump(mode="json"), depth + 1, nested, strip_keys) + except Exception: # noqa: BLE001 # a model that will not dump is one we drop + return None + + # Everything json.dumps handles natively — str, int, float, bool, None, dict, list — is + # caught above, and a dict/list subclass is caught by isinstance. So whatever reaches here + # (bytes, datetime, an OpenTelemetry span) cannot cross the wire. + return None diff --git a/litellm/proxy/guardrails/guardrail_hooks/grayswan/grayswan.py b/litellm/proxy/guardrails/guardrail_hooks/grayswan/grayswan.py index 955a868a0d6..48832f8ed5e 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/grayswan/grayswan.py +++ b/litellm/proxy/guardrails/guardrail_hooks/grayswan/grayswan.py @@ -2,9 +2,10 @@ import os import time -from typing import TYPE_CHECKING, Any, Final, Literal, Optional +from typing import TYPE_CHECKING, Any, Final, Literal, Optional, Protocol from fastapi import HTTPException +from typing_extensions import NotRequired, ReadOnly, TypedDict from litellm._logging import verbose_proxy_logger from litellm.integrations.custom_guardrail import ( @@ -27,6 +28,33 @@ if TYPE_CHECKING: GRAYSWAN_BLOCK_ERROR_MSG: Final = "Blocked by Gray Swan Guardrail" +class _GraySwanMonitorResponse(TypedDict): + """Body returned by Gray Swan's `/cygnal/monitor` endpoint.""" + + violation: ReadOnly[NotRequired[float | None]] + violated_rules: ReadOnly[NotRequired[list[object]]] + violated_rule_descriptions: ReadOnly[NotRequired[list[object]]] + mutation: ReadOnly[NotRequired[bool | None]] + ipi: ReadOnly[NotRequired[bool | None]] + + +class _GraySwanMonitorHTTPResponse(Protocol): + def raise_for_status(self) -> object: ... + + def json(self) -> _GraySwanMonitorResponse: ... + + +class _GraySwanMonitorHTTPClient(Protocol): + async def post( + self, + *, + url: str, + headers: dict[str, str], + json: dict[str, object], + timeout: float, + ) -> _GraySwanMonitorHTTPResponse: ... + + class GraySwanGuardrailMissingSecrets(Exception): """Raised when the Gray Swan API key is missing.""" @@ -77,7 +105,9 @@ class GraySwanGuardrail(CustomGuardrail): guardrail_timeout: float | None = 30.0, **kwargs: Any, ) -> None: - self.async_handler = get_async_httpx_client(llm_provider=httpxSpecialProvider.GuardrailCallback) + self.async_handler: _GraySwanMonitorHTTPClient = get_async_httpx_client( + llm_provider=httpxSpecialProvider.GuardrailCallback + ) api_key_value: Final = api_key or os.getenv("GRAYSWAN_API_KEY") if not api_key_value: @@ -266,7 +296,7 @@ class GraySwanGuardrail(CustomGuardrail): # Legacy Test Interface (for backward compatibility) # ------------------------------------------------------------------ - async def run_grayswan_guardrail(self, payload: dict) -> dict[str, Any]: + async def run_grayswan_guardrail(self, payload: dict[str, object]) -> _GraySwanMonitorResponse: """ Run the GraySwan guardrail on a payload. @@ -285,7 +315,7 @@ class GraySwanGuardrail(CustomGuardrail): def _process_grayswan_response( self, - response_json: dict, + response_json: _GraySwanMonitorResponse, data: dict | None = None, hook_type: GuardrailEventHooks | None = None, ) -> None: @@ -385,7 +415,7 @@ class GraySwanGuardrail(CustomGuardrail): # Core GraySwan API interaction # ------------------------------------------------------------------ - async def _call_grayswan_api(self, payload: dict) -> dict[str, Any]: + async def _call_grayswan_api(self, payload: dict[str, object]) -> _GraySwanMonitorResponse: """Call the GraySwan monitoring API.""" headers: Final = self._prepare_headers() @@ -406,7 +436,7 @@ class GraySwanGuardrail(CustomGuardrail): def _process_response_internal( self, - response_json: dict[str, Any], + response_json: _GraySwanMonitorResponse, request_data: dict, inputs: GenericGuardrailAPIInputs, is_output: bool, @@ -534,8 +564,8 @@ class GraySwanGuardrail(CustomGuardrail): dynamic_body: dict, request_data: dict, logging_obj: Optional["LiteLLMLoggingObj"] = None, - ) -> dict[str, Any] | None: - payload: Final[dict[str, Any]] = {"messages": messages} + ) -> dict[str, object] | None: + payload: Final[dict[str, object]] = {"messages": messages} categories: Final = dynamic_body.get("categories") or self.categories if categories: @@ -563,13 +593,13 @@ class GraySwanGuardrail(CustomGuardrail): {**existing_headers, **inbound_headers} if isinstance(existing_headers, dict) else inbound_headers ) if cleaned_litellm_metadata: - sanitized: Final = safe_json_loads(safe_dumps(cleaned_litellm_metadata), default={}) + sanitized: Final[object] = safe_json_loads(safe_dumps(cleaned_litellm_metadata), default={}) if isinstance(sanitized, dict) and sanitized: payload["litellm_metadata"] = sanitized return payload - def _format_violation_message(self, detection_info: Any, is_output: bool = False) -> str: + def _format_violation_message(self, detection_info: object, is_output: bool = False) -> str: """ Format detection info into a user-friendly violation message. diff --git a/litellm/proxy/guardrails/guardrail_hooks/lasso/lasso.py b/litellm/proxy/guardrails/guardrail_hooks/lasso/lasso.py index ea022510309..cf5da27e9ca 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/lasso/lasso.py +++ b/litellm/proxy/guardrails/guardrail_hooks/lasso/lasso.py @@ -8,6 +8,7 @@ import json import os import uuid +from collections.abc import Mapping, Sequence from typing import TYPE_CHECKING, Any, Final, Literal, TypedDict try: @@ -128,7 +129,7 @@ class LassoGuardrail(CustomGuardrail): @staticmethod def _extract_tool_call_fields( - call: Any, + call: object, ) -> tuple[str | None, str | None, dict[str, object] | None]: """Extract (call_id, name, parsed_input) from a tool call. @@ -476,7 +477,7 @@ class LassoGuardrail(CustomGuardrail): def _map_masked_messages_back( self, original_messages: list[dict[str, Any]], - masked_messages: list[dict[str, Any]], + masked_messages: Sequence[Mapping[str, object]], ) -> list[dict[str, object]]: """Map Lasso-format masked messages back onto the original OpenAI-format messages. @@ -638,7 +639,7 @@ class LassoGuardrail(CustomGuardrail): }, ) - def _expand_messages_for_classification(self, messages: list[dict[str, Any]]) -> list[dict[str, Any]]: + def _expand_messages_for_classification(self, messages: list[dict[str, Any]]) -> list[dict[str, object]]: """ Convert raw OpenAI-format messages to Lasso API format with content blocks. @@ -646,7 +647,7 @@ class LassoGuardrail(CustomGuardrail): - role=tool messages → developer role + tool_result block - plain text messages pass through unchanged """ - expanded: Final[list[dict[str, Any]]] = [] + expanded: Final[list[dict[str, object]]] = [] for msg in messages: role = msg.get("role", "") content = msg.get("content") @@ -917,7 +918,7 @@ class LassoGuardrail(CustomGuardrail): def _apply_masking_to_model_response( self, model_response: litellm.ModelResponse, - masked_messages: list[dict[str, Any]], + masked_messages: Sequence[Mapping[str, object]], ) -> None: """Apply masking to the actual model response when mask=True and masked content is available.""" # Index masked tool_use blocks by id for O(1) lookup. diff --git a/litellm/proxy/guardrails/guardrail_hooks/pillar/pillar.py b/litellm/proxy/guardrails/guardrail_hooks/pillar/pillar.py index 78639ce4fd0..7021d41475b 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/pillar/pillar.py +++ b/litellm/proxy/guardrails/guardrail_hooks/pillar/pillar.py @@ -8,11 +8,12 @@ # Standard library imports import json import os -from typing import TYPE_CHECKING, Any, Final, Literal +from typing import TYPE_CHECKING, Any, Final, Literal, Protocol from urllib.parse import quote # Third-party imports from fastapi import HTTPException +from typing_extensions import NotRequired, ReadOnly, TypedDict # LiteLLM imports from litellm import DualCache @@ -42,7 +43,34 @@ if TYPE_CHECKING: MAX_PILLAR_HEADER_VALUE_BYTES: Final = 8 * 1024 -def _encode_json_for_header(data: Any) -> str: +class _PillarProtectResponse(TypedDict): + """Body returned by Pillar's `/api/v1/protect` endpoint.""" + + flagged: ReadOnly[NotRequired[bool]] + session_id: ReadOnly[NotRequired[str]] + scanners: ReadOnly[NotRequired[dict[str, object]]] + evidence: ReadOnly[NotRequired[list[object]]] + masked_session_messages: ReadOnly[NotRequired[list[object]]] + + +class _PillarProtectHTTPResponse(Protocol): + def raise_for_status(self) -> object: ... + + def json(self) -> _PillarProtectResponse: ... + + +class _PillarProtectHTTPClient(Protocol): + async def post( + self, + *, + url: str, + headers: dict[str, str], + json: dict[str, object], + timeout: float, + ) -> _PillarProtectHTTPResponse: ... + + +def _encode_json_for_header(data: object) -> str: """ JSON-serialize and URL-encode data for safe header transmission. """ @@ -50,7 +78,9 @@ def _encode_json_for_header(data: Any) -> str: return quote(json_payload, safe="") -def _truncate_evidence_payload(evidence: Any, max_bytes: int = MAX_PILLAR_HEADER_VALUE_BYTES) -> tuple[Any, str, bool]: +def _truncate_evidence_payload( + evidence: object, max_bytes: int = MAX_PILLAR_HEADER_VALUE_BYTES +) -> tuple[object, str, bool]: """ Truncate evidence payload so the encoded header value stays within max_bytes. @@ -66,12 +96,12 @@ def _truncate_evidence_payload(evidence: Any, max_bytes: int = MAX_PILLAR_HEADER truncated_value: Final = "[truncated]" return truncated_value, _encode_json_for_header(truncated_value), True - truncated: Final[list[Any]] = [] + truncated: Final[list[object]] = [] encoded = _encode_json_for_header(truncated) truncated_flag = False for entry in evidence: - working_entry: Any + working_entry: object if isinstance(entry, dict): working_entry = dict(entry) else: @@ -105,7 +135,7 @@ def _truncate_evidence_payload(evidence: Any, max_bytes: int = MAX_PILLAR_HEADER return truncated, encoded, truncated_flag -def build_pillar_response_headers(metadata_store: dict[str, Any]) -> dict[str, str]: +def build_pillar_response_headers(metadata_store: dict[str, object]) -> dict[str, str]: """ Create URL-safe Pillar response headers and apply truncation metadata. """ @@ -191,7 +221,9 @@ class PillarGuardrail(CustomGuardrail): LiteLLM virtual key context (user_id, team_id, key_alias, etc.) is always automatically passed as X-LiteLLM-* headers to enable application/user tracking. """ - self.async_handler = get_async_httpx_client(llm_provider=httpxSpecialProvider.GuardrailCallback) + self.async_handler: _PillarProtectHTTPClient = get_async_httpx_client( + llm_provider=httpxSpecialProvider.GuardrailCallback + ) self.api_key = api_key or os.environ.get("PILLAR_API_KEY") if self.api_key is None: @@ -686,7 +718,7 @@ class PillarGuardrail(CustomGuardrail): ) return payload - async def _call_pillar_api(self, headers: dict[str, str], payload: dict[str, Any]) -> dict[str, Any]: + async def _call_pillar_api(self, headers: dict[str, str], payload: dict[str, Any]) -> _PillarProtectResponse: """ Call the Pillar API and return the response. @@ -714,7 +746,7 @@ class PillarGuardrail(CustomGuardrail): verbose_proxy_logger.debug("Pillar Guardrail: Analysis complete - flagged=%s, session=%s", flagged, session_id) return res - def _process_pillar_response(self, pillar_response: dict[str, Any], original_data: dict) -> None: + def _process_pillar_response(self, pillar_response: _PillarProtectResponse, original_data: dict) -> None: """ Process the Pillar API response and handle detections based on configuration. @@ -774,7 +806,7 @@ class PillarGuardrail(CustomGuardrail): build_pillar_response_headers(metadata_store) - def _raise_pillar_detection_exception(self, pillar_response: dict[str, Any]) -> None: + def _raise_pillar_detection_exception(self, pillar_response: _PillarProtectResponse) -> None: """ Raise an HTTPException for Pillar security detections. @@ -784,7 +816,7 @@ class PillarGuardrail(CustomGuardrail): Raises: HTTPException: Always raises with security detection details """ - pillar_response_dict: Final = { + pillar_response_dict: Final[dict[str, object]] = { "session_id": pillar_response.get("session_id"), } diff --git a/litellm/proxy/guardrails/guardrail_hooks/semantic_guard/semantic_guard.py b/litellm/proxy/guardrails/guardrail_hooks/semantic_guard/semantic_guard.py index e34beec4d3e..2fbd50b5863 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/semantic_guard/semantic_guard.py +++ b/litellm/proxy/guardrails/guardrail_hooks/semantic_guard/semantic_guard.py @@ -6,7 +6,7 @@ via embedding similarity. Smarter than regex (understands intent), lighter than an LLM call (~20-50ms per request for embedding). """ -from typing import TYPE_CHECKING, Any, Final +from typing import TYPE_CHECKING, Any, Final, Protocol from litellm._logging import verbose_logger from litellm.integrations.custom_guardrail import ( @@ -50,7 +50,7 @@ class SemanticGuardrail(CustomGuardrail): similarity_threshold: float, route_templates: list[str] | None = None, custom_routes_file: str | None = None, - custom_routes: list[dict[str, Any]] | None = None, + custom_routes: list[dict[str, object]] | None = None, on_flagged_action: str = "block", event_hook: GuardrailEventHooks | list[GuardrailEventHooks] | Mode | None = None, default_on: bool = False, @@ -157,7 +157,14 @@ class SemanticGuardrail(CustomGuardrail): return response -def _get_top_route_choice(result: Any) -> Any: +class _RouteChoice(Protocol): + """The semantic-router match this guardrail reads: the route that fired, if any.""" + + @property + def name(self) -> str | None: ... + + +def _get_top_route_choice(result: _RouteChoice | list[_RouteChoice] | None) -> _RouteChoice | None: """Extract the top RouteChoice from SemanticRouter result. SemanticRouter.__call__ can return RouteChoice or List[RouteChoice]. @@ -194,7 +201,7 @@ def _extract_response_text(response: Any) -> str: return "" -def _content_to_text(content: Any) -> str: +def _content_to_text(content: object) -> str: if isinstance(content, str): return content if isinstance(content, list): diff --git a/litellm/proxy/guardrails/guardrail_hooks/tool_permission.py b/litellm/proxy/guardrails/guardrail_hooks/tool_permission.py index 3c5625bc272..a8b33109900 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/tool_permission.py +++ b/litellm/proxy/guardrails/guardrail_hooks/tool_permission.py @@ -1,9 +1,10 @@ import json import re from collections.abc import AsyncGenerator, AsyncIterable, Mapping, Sequence -from typing import Any, Final, Literal +from typing import Any, Final, Literal, TypedDict from fastapi import HTTPException +from typing_extensions import ReadOnly, Required from litellm import ChatCompletionToolParam from litellm._logging import verbose_proxy_logger @@ -51,6 +52,27 @@ def _object_list(value: object) -> Sequence[object] | None: return value if isinstance(value, list) else None +class _ToolPermissionRuleFields(TypedDict, total=False): + """The config-file shape a :class:`ToolPermissionRule` is built from.""" + + id: ReadOnly[Required[str]] + tool_name: ReadOnly[str | None] + tool_type: ReadOnly[str | None] + decision: ReadOnly[Required[Literal["allow", "deny"]]] + allowed_param_patterns: ReadOnly[dict[str, str] | None] + + +def _rule_from_fields(fields: _ToolPermissionRuleFields) -> ToolPermissionRule: + """Validate one config-file rule entry into a :class:`ToolPermissionRule`.""" + return ToolPermissionRule(**fields) + + +def _is_tool_use_block(block: object) -> bool: + """Whether ``block`` is an Anthropic ``tool_use`` content block.""" + fields: Final = _object_mapping(block) + return fields is not None and fields.get("type") == "tool_use" + + class ToolPermissionGuardrail(CustomGuardrail): def __init__( self, @@ -101,7 +123,7 @@ class ToolPermissionGuardrail(CustomGuardrail): compiled_patterns: Final[dict[str, dict[str, re.Pattern]]] = {} for rule_item in rules or []: - rule = rule_item if isinstance(rule_item, ToolPermissionRule) else ToolPermissionRule(**rule_item) + rule = rule_item if isinstance(rule_item, ToolPermissionRule) else _rule_from_fields(rule_item) target_patterns: dict[str, re.Pattern | None] = { "tool_name": None, @@ -440,7 +462,7 @@ class ToolPermissionGuardrail(CustomGuardrail): return is_allowed, None, message @staticmethod - def _get_mapping_value(item: Any, key: str) -> Any: + def _get_mapping_value(item: object, key: str) -> Any: if isinstance(item, dict): return item.get(key) return getattr(item, key, None) @@ -450,7 +472,7 @@ class ToolPermissionGuardrail(CustomGuardrail): return f"legacy_function_call_{choice_index}" def _legacy_function_call_to_tool_call( - self, function_call: Any, choice_index: int + self, function_call: object, choice_index: int ) -> ChatCompletionMessageToolCall | None: if function_call is None: return None @@ -549,7 +571,7 @@ class ToolPermissionGuardrail(CustomGuardrail): def _modify_anthropic_content_with_permission_errors( self, response: object, - content: tuple[Any, ...], + content: tuple[object, ...], denied_tools: tuple[tuple[ChatCompletionMessageToolCall, PermissionError], ...], ) -> None: if not denied_tools or not isinstance(response, dict): @@ -557,27 +579,33 @@ class ToolPermissionGuardrail(CustomGuardrail): verbose_proxy_logger.info("Blocking %s unauthorized tool uses", len(denied_tools)) - error_by_tool_use_id: Final = { # mutable-ok: read-only lookup, never mutated after construction + error_by_tool_use_id: Final[ + Mapping[object, str] + ] = { # mutable-ok: read-only lookup, never mutated after construction tool_call.id: self._create_permission_error_result(tool_call, error).content for tool_call, error in denied_tools } - denied_block_ids: Final = frozenset(error_by_tool_use_id) - def _is_denied(block: object) -> bool: - return isinstance(block, dict) and block.get("type") == "tool_use" and block.get("id") in denied_block_ids + def _denied_message(block: object) -> str | None: + fields: Final = _object_mapping(block) + if fields is None or fields.get("type") != "tool_use": + return None + return error_by_tool_use_id.get(fields.get("id")) - error_messages: Final = tuple(error_by_tool_use_id[block["id"]] for block in content if _is_denied(block)) - kept_blocks: Final = tuple(block for block in content if not _is_denied(block)) + error_messages: Final = tuple( + message for message in (_denied_message(block) for block in content) if message is not None + ) + kept_blocks: Final = tuple(block for block in content if _denied_message(block) is None) new_content: Final = [ # mutable-ok: response content is a JSON array on the wire *kept_blocks, {"type": "text", "text": "\n".join(error_messages)}, # mutable-ok: content block is a JSON object ] response["content"] = new_content # rebind-ok: the guardrail rewrites the provider response in place - if not any(isinstance(block, dict) and block.get("type") == "tool_use" for block in kept_blocks): + if not any(_is_tool_use_block(block) for block in kept_blocks): response["stop_reason"] = "end_turn" # rebind-ok: dropping every tool_use ends the turn - def _get_request_tool_name(self, tool: Any) -> tuple[str | None, str | None]: + def _get_request_tool_name(self, tool: object) -> tuple[str | None, str | None]: tool_type: Final = self._get_mapping_value(tool, "type") if tool_type != "function": return None, tool_type @@ -586,7 +614,7 @@ class ToolPermissionGuardrail(CustomGuardrail): tool_name: Final = self._get_mapping_value(function, "name") return tool_name, tool_type - def _get_legacy_function_name(self, function: Any) -> str | None: + def _get_legacy_function_name(self, function: object) -> str | None: return self._get_mapping_value(function, "name") def _get_named_tool_choice(self, data: dict) -> str | None: diff --git a/litellm/proxy/guardrails/guardrail_hooks/vigil_guard/vigil_guard.py b/litellm/proxy/guardrails/guardrail_hooks/vigil_guard/vigil_guard.py index 6b8148645aa..a5945a39589 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/vigil_guard/vigil_guard.py +++ b/litellm/proxy/guardrails/guardrail_hooks/vigil_guard/vigil_guard.py @@ -433,7 +433,7 @@ class VigilGuardGuardrail(CustomGuardrail): return collected @staticmethod - def _clamp_metadata_value(value: Any) -> _MetadataValue | None: + def _clamp_metadata_value(value: object) -> _MetadataValue | None: if isinstance(value, bool): return None if isinstance(value, str): diff --git a/litellm/proxy/hooks/litellm_skills/main.py b/litellm/proxy/hooks/litellm_skills/main.py index 569ec32c1a0..9edbc6dbf1c 100644 --- a/litellm/proxy/hooks/litellm_skills/main.py +++ b/litellm/proxy/hooks/litellm_skills/main.py @@ -67,6 +67,24 @@ class _ChatMessage(Protocol): def tool_calls(self) -> Sequence[_ChatToolCall] | None: ... +class _ChatChoice(Protocol): + @property + def message(self) -> _ChatMessage: ... + + @property + def finish_reason(self) -> str | None: ... + + +class _ChatCompletion(Protocol): + @property + def choices(self) -> Sequence[_ChatChoice]: ... + + +def _first_choice(response: _ChatCompletion) -> _ChatChoice: + """The first choice of an OpenAI shaped completion response.""" + return response.choices[0] + + class SkillsInjectionHook(CustomLogger): """ Pre/Post-call hook that processes skills from container.skills parameter. @@ -738,8 +756,9 @@ print('No executable skill module found') for iteration in range(self.max_iterations): # OpenAI format response has choices[0].message - assistant_message: _ChatMessage = current_response.choices[0].message - stop_reason: str | None = current_response.choices[0].finish_reason + choice: _ChatChoice = _first_choice(current_response) + assistant_message: _ChatMessage = choice.message + stop_reason: str | None = choice.finish_reason # Build assistant message for conversation history assistant_msg_dict: dict[str, object] = { diff --git a/litellm/proxy/hooks/parallel_request_limiter_v3.py b/litellm/proxy/hooks/parallel_request_limiter_v3.py index 1e65da5b867..63129602082 100644 --- a/litellm/proxy/hooks/parallel_request_limiter_v3.py +++ b/litellm/proxy/hooks/parallel_request_limiter_v3.py @@ -8,7 +8,7 @@ import asyncio import binascii import os import uuid -from collections.abc import Callable, Mapping, Sequence, Set +from collections.abc import Awaitable, Callable, Mapping, Sequence, Set from contextvars import ContextVar from dataclasses import dataclass, field from datetime import datetime @@ -386,6 +386,12 @@ CacheCounterValues: TypeAlias = Sequence[CacheCounterValue | None] ParallelGaugeCacheValue: TypeAlias = dict[str, object] | int | float | str | bytes +class _AsyncLuaScript(Protocol): + """A Lua script registered against the async Redis client, called with KEYS and ARGV.""" + + def __call__(self, *, keys: Sequence[str], args: Sequence[object]) -> Awaitable[list[CacheCounterValue]]: ... + + class RateLimitDescriptorRateLimitObject(TypedDict, total=False): requests_per_unit: int | None tokens_per_unit: int | None @@ -577,6 +583,14 @@ def _parse_output_cap_value(raw_value: object) -> int | None: class _PROXY_MaxParallelRequestsHandler_v3(CustomLogger): + batch_rate_limiter_script: _AsyncLuaScript | None + token_increment_script: _AsyncLuaScript | None + check_and_increment_by_n_script: _AsyncLuaScript | None + window_guarded_token_increment_script: _AsyncLuaScript | None + parallel_acquire_script: _AsyncLuaScript | None + parallel_release_script: _AsyncLuaScript | None + parallel_count_script: _AsyncLuaScript | None + def __init__( self, internal_usage_cache: InternalUsageCache, @@ -3855,7 +3869,7 @@ class _PROXY_MaxParallelRequestsHandler_v3(CustomLogger): expected_window_start = operation.get("expected_window_start") if window_key is None or expected_window_start is None: continue - active_window_start = await self.internal_usage_cache.async_get_cache( + active_window_start: CacheCounterValue | None = await self.internal_usage_cache.async_get_cache( key=window_key, litellm_parent_otel_span=parent_otel_span, local_only=True, @@ -4144,7 +4158,7 @@ class _PROXY_MaxParallelRequestsHandler_v3(CustomLogger): def _collect_tpm_scope_targets( self, standard_logging_metadata: dict[str, Any], - kwargs: Any, + kwargs: object, model_group: str | None, ) -> list[tuple[str, str]]: """ @@ -4301,8 +4315,8 @@ class _PROXY_MaxParallelRequestsHandler_v3(CustomLogger): def _build_success_event_pipeline_operations( self, - kwargs: Any, - response_obj: Any, + kwargs: dict[str, Any], + response_obj: object, rate_limit_type: Literal["output", "input", "total"], ) -> list[RedisPipelineIncrementOperation]: """Build Redis pipeline increment ops for TPM / parallel-request counters.""" diff --git a/litellm/proxy/management_endpoints/model_management_endpoints.py b/litellm/proxy/management_endpoints/model_management_endpoints.py index 012aec38458..14d2332a7eb 100644 --- a/litellm/proxy/management_endpoints/model_management_endpoints.py +++ b/litellm/proxy/management_endpoints/model_management_endpoints.py @@ -543,7 +543,7 @@ def update_db_model(db_model: Deployment, updated_patch: updateDeployment) -> Pr _raise_if_ptu_cost_attribution_disabled(updated_patch.model_info.model_dump(exclude_none=True)) merged_model_name: Final = updated_patch.model_name or db_model.model_name merged_litellm_params: Final = db_model.litellm_params.model_dump(exclude_none=True) - merged_model_info: Final = db_model.model_info.model_dump(exclude_none=True) + merged_model_info: Final[dict[str, object]] = db_model.model_info.model_dump(exclude_none=True) # update litellm params if updated_patch.litellm_params: @@ -1982,7 +1982,7 @@ async def update_model( ### MERGE WITH EXISTING DATA ### merged_dictionary: Final = {} - _mp: Final = model_params.litellm_params.dict() + _mp: Final[dict[str, object]] = model_params.litellm_params.dict() for key, value in _mp.items(): if value is not None: diff --git a/litellm/proxy/management_endpoints/organization_endpoints.py b/litellm/proxy/management_endpoints/organization_endpoints.py index 9198aa35f3f..5e38a016099 100644 --- a/litellm/proxy/management_endpoints/organization_endpoints.py +++ b/litellm/proxy/management_endpoints/organization_endpoints.py @@ -487,12 +487,11 @@ async def new_organization( for m in data.models: await can_user_call_model(m, llm_router=llm_router, user_object=user_object_correct_type) - organization_row: Final = LiteLLM_OrganizationTable( - **data.json(exclude_none=True), - object_permission_id=object_permission_id, - created_by=user_api_key_dict.user_id or litellm_proxy_admin_name, - updated_by=user_api_key_dict.user_id or litellm_proxy_admin_name, - ) + organization_payload: Final = _STR_OBJECT_DICT_ADAPTER.validate_python(data.json(exclude_none=True)) + organization_payload["object_permission_id"] = object_permission_id + organization_payload["created_by"] = user_api_key_dict.user_id or litellm_proxy_admin_name + organization_payload["updated_by"] = user_api_key_dict.user_id or litellm_proxy_admin_name + organization_row: Final = LiteLLM_OrganizationTable.model_validate(organization_payload) for field in LiteLLM_ManagementEndpoint_MetadataFields: if getattr(data, field, None) is not None: @@ -644,7 +643,7 @@ async def update_organization( ) # Transform UI payload to expected format - raw_data: Final = await request.json() + raw_data: Final[dict[str, object]] = await request.json() raw_data_with_flat_budget_fields: Final = handle_nested_budget_structure_in_organization_update_request(raw_data) # Create validated data model @@ -691,7 +690,7 @@ async def update_organization( # Merge metadata from existing organization with updated metadata if updated_organization_row_json.get("metadata") is not None: existing_metadata: Final = existing_organization_row.metadata or {} - updated_metadata: Final = updated_organization_row_json.get("metadata", {}) + updated_metadata: Final[dict[str, object]] = updated_organization_row_json.get("metadata", {}) merged_metadata: Final[Mapping[str, object]] = _update_dictionary( existing_dict=cast( # cast-ok: prisma de-serializes a Json column to the plain python dict it stores "dict[str, object]", existing_metadata diff --git a/litellm/proxy/management_endpoints/ui_sso.py b/litellm/proxy/management_endpoints/ui_sso.py index 613508da22b..606569c5b8b 100644 --- a/litellm/proxy/management_endpoints/ui_sso.py +++ b/litellm/proxy/management_endpoints/ui_sso.py @@ -502,7 +502,7 @@ def _set_nested_metadata_value(metadata: dict[str, object], key_path: str, value placeholder: Final = "\x00" parts = key_path.replace("\\.", placeholder).split(".") parts = [p.replace(placeholder, ".") for p in parts] - current: Any = metadata + current: dict[str, object] = metadata for part in parts[:-1]: existing = current.get(part) if not isinstance(existing, dict): @@ -4076,7 +4076,7 @@ class SSOAuthenticationHandler: ) if resp.status_code == 200: try: - userinfo_raw: Final = resp.json() + userinfo_raw: Final[dict[str, object] | None] = resp.json() if not userinfo_raw: # JSON null (None) or empty dict ({}) — no identity claims. # Treat as failure so id_token fallback can be attempted. @@ -4406,7 +4406,7 @@ class MicrosoftSSOHandler: ) -> tuple[list[str], str | None]: """Helper function to fetch and parse group data from a URL""" response: Final = await async_client.get(url, headers=headers) - response_json: Final = response.json() + response_json: Final[dict[str, object]] = response.json() response_typed: Final = await MicrosoftSSOHandler._cast_graph_api_response_dict(response=response_json) group_ids: Final = MicrosoftSSOHandler._get_group_ids_from_graph_api_response(response=response_typed) return group_ids, response_typed.get("odata_nextLink") diff --git a/litellm/proxy/management_helpers/object_permission_utils.py b/litellm/proxy/management_helpers/object_permission_utils.py index 13080a6cf83..a2fbf80422c 100644 --- a/litellm/proxy/management_helpers/object_permission_utils.py +++ b/litellm/proxy/management_helpers/object_permission_utils.py @@ -286,7 +286,7 @@ async def _resolve_mcp_server_identifiers_to_ids( return resolved -def _rewrite_object_permission_mcp_servers( +def _drop_stale_object_permission_mcp_servers( object_permission: ObjectPermissionDict, identifier_to_server_ids: dict[str, set[str]], ) -> None: @@ -294,16 +294,18 @@ def _rewrite_object_permission_mcp_servers( if not isinstance(mcp_servers, list): return - normalized_servers: Final[list[str]] = [] - for identifier in mcp_servers: - if identifier == SpecialMCPServerNames.no_mcp_servers.value: - normalized_servers.append(SpecialMCPServerNames.no_mcp_servers.value) - continue - normalized_servers.extend(sorted(identifier_to_server_ids.get(identifier, []))) - object_permission["mcp_servers"] = _dedupe_preserving_order(normalized_servers) + # Persist original identifiers, never resolved ids: shared-DB multi-region + # instances each expand a name/alias to their own local server id at read + # time. Only entries resolving to nothing (deleted servers, typos) drop. + kept_servers: Final = [ + identifier + for identifier in mcp_servers + if identifier == SpecialMCPServerNames.no_mcp_servers.value or identifier_to_server_ids.get(identifier) + ] + object_permission["mcp_servers"] = _dedupe_preserving_order(kept_servers) -def _rewrite_object_permission_mcp_tool_permissions( +def _drop_stale_object_permission_mcp_tool_permissions( object_permission: ObjectPermissionDict, identifier_to_server_ids: dict[str, set[str]], ) -> None: @@ -311,31 +313,25 @@ def _rewrite_object_permission_mcp_tool_permissions( if not isinstance(mcp_tool_permissions, dict): return - normalized_tool_permissions: Final[dict[str, list[str]]] = {} - for identifier, tools in mcp_tool_permissions.items(): - if not isinstance(tools, list): - tools = [] - for server_id in sorted(identifier_to_server_ids.get(identifier, [])): - normalized_tool_permissions.setdefault(server_id, []) - normalized_tool_permissions[server_id].extend(tools) - object_permission["mcp_tool_permissions"] = { - server_id: _dedupe_preserving_order(tools) for server_id, tools in normalized_tool_permissions.items() + identifier: _dedupe_preserving_order(tools if isinstance(tools, list) else []) + for identifier, tools in mcp_tool_permissions.items() + if identifier_to_server_ids.get(identifier) } -def _rewrite_object_permission_mcp_identifiers( +def _drop_stale_object_permission_mcp_identifiers( object_permission: ObjectPermissionDict | None, identifier_to_server_ids: dict[str, set[str]], ) -> None: if not object_permission or not isinstance(object_permission, dict): return - _rewrite_object_permission_mcp_servers( + _drop_stale_object_permission_mcp_servers( object_permission=object_permission, identifier_to_server_ids=identifier_to_server_ids, ) - _rewrite_object_permission_mcp_tool_permissions( + _drop_stale_object_permission_mcp_tool_permissions( object_permission=object_permission, identifier_to_server_ids=identifier_to_server_ids, ) @@ -615,7 +611,7 @@ async def validate_key_mcp_servers_against_team( "validate_key_mcp_servers_against_team: ignoring stale MCP server identifiers (no longer in registry or DB): %s", sorted(stale_identifiers), ) - _rewrite_object_permission_mcp_identifiers( + _drop_stale_object_permission_mcp_identifiers( object_permission=object_permission, identifier_to_server_ids=identifier_to_server_ids, ) diff --git a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/vertex_passthrough_logging_handler.py b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/vertex_passthrough_logging_handler.py index ee9a5d94440..49ec18013b5 100644 --- a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/vertex_passthrough_logging_handler.py +++ b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/vertex_passthrough_logging_handler.py @@ -267,7 +267,7 @@ class VertexPassthroughLoggingHandler: model: Final = VertexPassthroughLoggingHandler.extract_model_from_url(url_route) - _json_response: Final = httpx_response.json() + _json_response: Final[dict[str, object]] = httpx_response.json() litellm_prediction_response: ModelResponse | EmbeddingResponse | ImageResponse = ModelResponse() if vertex_image_generation_class.is_image_generation_response(_json_response): @@ -422,7 +422,7 @@ class VertexPassthroughLoggingHandler: - Creates standard logging object - Logs in litellm callbacks """ - kwargs: dict[str, Any] = {} + kwargs: dict[str, object] = {} vertex_location: Final = get_vertex_location_from_url(url_route) if vertex_location is not None: litellm_logging_obj.optional_params["vertex_location"] = vertex_location diff --git a/litellm/proxy/policy_engine/pipeline_executor.py b/litellm/proxy/policy_engine/pipeline_executor.py index 50cb813c6fa..4be0f556ed7 100644 --- a/litellm/proxy/policy_engine/pipeline_executor.py +++ b/litellm/proxy/policy_engine/pipeline_executor.py @@ -6,6 +6,7 @@ pass/fail actions (allow, block, next, modify_response) and data forwarding. """ import time +from collections.abc import Sequence from typing import Any, Final, Literal import litellm @@ -114,11 +115,7 @@ class PipelineExecutor: # Handle terminal actions if action == "allow": - return PipelineExecutionResult( - terminal_action="allow", - step_results=step_results, - modified_data=working_data if working_data != data else None, - ) + return _allow_result(step_results=step_results, working_data=working_data, request_data=data) if action == "block": return PipelineExecutionResult( @@ -138,11 +135,7 @@ class PipelineExecutor: # action == "next" → continue to next step # Ran out of steps without a terminal action → default allow - return PipelineExecutionResult( - terminal_action="allow", - step_results=step_results, - modified_data=working_data if working_data != data else None, - ) + return _allow_result(step_results=step_results, working_data=working_data, request_data=data) @staticmethod async def _run_step( @@ -251,6 +244,45 @@ class PipelineExecutor: return None +def _allow_result( + step_results: Sequence[PipelineStepResult], + working_data: dict, # mutable-ok: same request-payload shape as execute_steps' data + request_data: dict, # mutable-ok: same request-payload shape as execute_steps' data +) -> PipelineExecutionResult: + """Build the terminal-allow result, propagating pipeline modifications without the per-step guardrail override.""" + restored: Final = _restore_request_guardrails(working_data, request_data) + return PipelineExecutionResult( + terminal_action="allow", + step_results=list(step_results), # mutable-ok: PipelineExecutionResult field is a list + modified_data=restored if restored != request_data else None, + ) + + +def _restore_request_guardrails( + working_data: dict, # mutable-ok: same request-payload shape as execute_steps' data + request_data: dict, # mutable-ok: same request-payload shape as execute_steps' data +) -> dict: # mutable-ok: merged back into the request dict, which downstream code mutates + """ + Restore the request's own metadata["guardrails"] activation list. + + _run_step overrides it to [step.guardrail] so should_run_guardrail() allows each + step; letting that override escape via modified_data permanently drops every + independently activated guardrail from later lifecycle stages (post_call, etc.). + """ + working_metadata: Final = working_data.get("metadata") + if not isinstance(working_metadata, dict): + return working_data + request_metadata: Final = request_data.get("metadata") + original_guardrails: Final = request_metadata.get("guardrails") if isinstance(request_metadata, dict) else None + stripped: Final = {k: v for k, v in working_metadata.items() if k != "guardrails"} # mutable-ok: request dict + if original_guardrails is not None: + restored: Final = {**stripped, "guardrails": original_guardrails} # mutable-ok: request dict + return {**working_data, "metadata": restored} # mutable-ok: request dict + if not stripped and not isinstance(request_metadata, dict): + return {k: v for k, v in working_data.items() if k != "metadata"} # mutable-ok: request dict + return {**working_data, "metadata": stripped} # mutable-ok: request dict + + def _pipeline_action_for_outcome(step: PipelineStep, outcome: str) -> str: """ Map pipeline step outcome to the configured action. diff --git a/litellm/proxy/proxy_cli.py b/litellm/proxy/proxy_cli.py index 8ac63ba25c9..86f6853a625 100644 --- a/litellm/proxy/proxy_cli.py +++ b/litellm/proxy/proxy_cli.py @@ -913,13 +913,14 @@ class ProxyInitializationHelpers: envvar="ENFORCE_PRISMA_MIGRATION_CHECK", ) @click.option( - "--use_v2_migration_resolver", - is_flag=True, - default=False, + "--use_v2_migration_resolver/--use_legacy_migration_resolver", + default=True, help=( - "Opt into the v2 migration resolver. Avoids the diff-and-force recovery " - "path that can cause schema thrashing during rolling deploys where two " - "LiteLLM versions contend for the same DB. Default is the v1 resolver." + "Which database migration resolver to run at startup. The default v2 " + "resolver avoids the diff-and-force recovery path that can cause schema " + "thrashing during rolling deploys where two LiteLLM versions contend for " + "the same DB. Pass --use_legacy_migration_resolver, or set " + "USE_V2_MIGRATION_RESOLVER=false, to fall back to v1." ), envvar="USE_V2_MIGRATION_RESOLVER", ) @@ -1310,10 +1311,11 @@ def run_server( else: if not use_v2_migration_resolver: print( - "\033[1;33mLiteLLM Proxy: Using default (v1) migration resolver. " - "If your deployment has seen schema thrashing during rolling " - "deploys, try --use_v2_migration_resolver (safer: avoids the " - "diff-and-force recovery that caused the thrash).\033[0m" + "\033[1;33mLiteLLM Proxy: Using the legacy (v1) migration resolver. " + "The default v2 resolver is safer: it avoids the diff-and-force " + "recovery that caused schema thrashing during rolling deploys. " + "Remove --use_legacy_migration_resolver / " + "USE_V2_MIGRATION_RESOLVER=false to switch back to it.\033[0m" ) try: setup_ok: Final = PrismaManager.setup_database( @@ -1321,10 +1323,10 @@ def run_server( use_v2_resolver=use_v2_migration_resolver, ) except RuntimeError as e: - # Raised on unrecoverable migration errors: the v2 - # resolver's non-idempotent failures and permission - # issues, and any `prisma db push` against a - # partitioned LiteLLM_SpendLogs. + # Raised on unrecoverable migration errors: permission + # failures from either resolver, the v2 resolver's + # non-idempotent failures, and any `prisma db push` + # against a partitioned LiteLLM_SpendLogs. print( f"\033[1;31mLiteLLM Proxy: Database migration cannot proceed. {e}\033[0m", file=sys.stderr, diff --git a/litellm/proxy/public_endpoints/provider_create_fields.json b/litellm/proxy/public_endpoints/provider_create_fields.json index a746e9af326..66f8c2ea36f 100644 --- a/litellm/proxy/public_endpoints/provider_create_fields.json +++ b/litellm/proxy/public_endpoints/provider_create_fields.json @@ -986,6 +986,62 @@ ], "default_model_placeholder": "gpt-3.5-turbo" }, + { + "provider": "QwenCloud", + "provider_display_name": "QwenCloud", + "litellm_provider": "qwencloud", + "credential_fields": [ + { + "key": "api_key", + "label": "QwenCloud API Key", + "placeholder": null, + "tooltip": null, + "required": true, + "field_type": "password", + "options": null, + "default_value": null + }, + { + "key": "api_base", + "label": "API Base", + "placeholder": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1", + "tooltip": "The base URL for QwenCloud. Defaults to https://dashscope-intl.aliyuncs.com/compatible-mode/v1 if not specified.", + "required": true, + "field_type": "text", + "options": null, + "default_value": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1" + } + ], + "default_model_placeholder": "gpt-3.5-turbo" + }, + { + "provider": "Qwen_AI_Platform", + "provider_display_name": "Qwen AI Platform", + "litellm_provider": "qwen_ai_platform", + "credential_fields": [ + { + "key": "api_key", + "label": "Qwen AI Platform API Key", + "placeholder": null, + "tooltip": null, + "required": true, + "field_type": "password", + "options": null, + "default_value": null + }, + { + "key": "api_base", + "label": "API Base", + "placeholder": "https://dashscope.aliyuncs.com/compatible-mode/v1", + "tooltip": "The base URL for Qwen AI Platform. Defaults to https://dashscope.aliyuncs.com/compatible-mode/v1 if not specified.", + "required": true, + "field_type": "text", + "options": null, + "default_value": "https://dashscope.aliyuncs.com/compatible-mode/v1" + } + ], + "default_model_placeholder": "gpt-3.5-turbo" + }, { "provider": "Databricks", "provider_display_name": "Databricks", diff --git a/litellm/proxy/response_api_endpoints/endpoints.py b/litellm/proxy/response_api_endpoints/endpoints.py index aa7595ed13d..5907ffc64eb 100644 --- a/litellm/proxy/response_api_endpoints/endpoints.py +++ b/litellm/proxy/response_api_endpoints/endpoints.py @@ -52,7 +52,7 @@ _TOOL_PAYLOAD_KEYS: Final[Mapping[str, tuple[str, ...]]] = MappingProxyType( "function": ("name", "description", "parameters", "strict"), } ) -_EMPTY_TOOL_PAYLOAD: Final[Mapping[str, Any]] = MappingProxyType({}) +_EMPTY_TOOL_PAYLOAD: Final[Mapping[str, object]] = MappingProxyType({}) def _convert_tool_payload_value(key: str, value: object, *, to_chat: bool) -> object: @@ -105,7 +105,7 @@ def _normalize_tool_dialect( return {**data, **{key: value for key, value in replaceable if key in data}} # mutable-ok: plain body dict -def _is_chat_completions_body(data: Mapping[str, Any]) -> bool: +def _is_chat_completions_body(data: Mapping[str, object]) -> bool: messages: Final = data.get("messages") if isinstance(messages, list) and messages: return True @@ -1373,7 +1373,7 @@ async def _enforce_responses_ws_first_frame_model_auth( request: Request, model: str, user_api_key_dict: UserAPIKeyAuth, - llm_router: Any | None, + llm_router: "Router | None", ) -> None: from litellm.proxy.auth.user_api_key_auth import ( _enforce_key_and_fallback_model_access, @@ -1417,7 +1417,7 @@ async def _enforce_responses_ws_first_frame_model_auth( async def responses_websocket_endpoint( websocket: WebSocket, model: str | None = fastapi.Query(None, description="The model to use for the responses WebSocket session."), - user_api_key_dict=Depends(user_api_key_auth_websocket), + user_api_key_dict: UserAPIKeyAuth = Depends(user_api_key_auth_websocket), ): """ Responses API WebSocket mode endpoint. @@ -1462,7 +1462,7 @@ async def responses_websocket_endpoint( return model, first_message = result - data: dict[str, Any] = { + data: dict[str, object] = { "model": model, "websocket": websocket, } @@ -1471,7 +1471,7 @@ async def responses_websocket_endpoint( # Construct a synthetic Request for pre-call processing headers_list: Final = list(websocket.scope.get("headers") or []) - scope: Final[dict[str, Any]] = { + scope: Final[dict[str, object]] = { "type": "http", "method": "POST", "path": "/v1/responses", diff --git a/litellm/proxy/route_llm_request.py b/litellm/proxy/route_llm_request.py index 91a0c68fd58..3d0bd5e61c9 100644 --- a/litellm/proxy/route_llm_request.py +++ b/litellm/proxy/route_llm_request.py @@ -50,12 +50,12 @@ def _route_user_config_request(data: dict, route_type: str): return ret_val -def _is_a2a_agent_model(model_name: Any) -> bool: +def _is_a2a_agent_model(model_name: object) -> bool: """Check if the model name is for an A2A agent (a2a/ prefix).""" return isinstance(model_name, str) and model_name.startswith("a2a/") -def _raise_if_model_fully_blocked(llm_router: LitellmRouter, model_name: Any, team_id: str | None) -> None: +def _raise_if_model_fully_blocked(llm_router: LitellmRouter, model_name: object, team_id: str | None) -> None: if not isinstance(model_name, str) or not model_name: return if not isinstance(llm_router, litellm.Router): diff --git a/litellm/proxy/video_endpoints/endpoints.py b/litellm/proxy/video_endpoints/endpoints.py index d985a546fa7..66071c05b4f 100644 --- a/litellm/proxy/video_endpoints/endpoints.py +++ b/litellm/proxy/video_endpoints/endpoints.py @@ -1,6 +1,6 @@ #### Video Endpoints ##### -from typing import Any, Final +from typing import Final from fastapi import APIRouter, Depends, File, Form, Request, Response, UploadFile from fastapi.responses import ORJSONResponse @@ -161,7 +161,7 @@ async def video_list( # Read query parameters query_params: Final = dict(request.query_params) - data: Final[dict[str, Any]] = {"query_params": query_params} + data: Final[dict[str, object]] = {"query_params": query_params} # Extract custom_llm_provider from headers, query params, or body custom_llm_provider: Final = ( @@ -246,7 +246,7 @@ async def video_status( ) # Create data with video_id - data: Final[dict[str, Any]] = {"video_id": video_id} + data: Final[dict[str, object]] = {"video_id": video_id} decoded: Final = decode_video_id_with_provider(video_id) provider_from_id: Final = decoded.get("custom_llm_provider") @@ -345,7 +345,7 @@ async def video_content( ) # Create data with video_id - data: Final[dict[str, Any]] = {"video_id": video_id} + data: Final[dict[str, object]] = {"video_id": video_id} decoded: Final = decode_video_id_with_provider(video_id) provider_from_id: Final = decoded.get("custom_llm_provider") @@ -653,7 +653,7 @@ async def video_get_character( ) original_requested_character_id: Final = character_id - data: Final[dict[str, Any]] = {"character_id": character_id} + data: Final[dict[str, object]] = {"character_id": character_id} decoded: Final = decode_character_id_with_provider(character_id) provider_from_id: Final = decoded.get("custom_llm_provider") diff --git a/litellm/rag/main.py b/litellm/rag/main.py index 2dcaa200cc6..7bc1a6a52a3 100644 --- a/litellm/rag/main.py +++ b/litellm/rag/main.py @@ -29,6 +29,7 @@ from litellm.rag.ingestion.openai_ingestion import OpenAIRAGIngestion from litellm.rag.ingestion.s3_vectors_ingestion import S3VectorsRAGIngestion from litellm.rag.ingestion.vertex_ai_ingestion import VertexAIRAGIngestion from litellm.rag.rag_query import RAGQuery +from litellm.types.llms.openai import AllMessageValues from litellm.types.rag import ( RAGIngestOptions, RAGIngestResponse, @@ -204,7 +205,7 @@ def _suppressed_sub_call_billing() -> Iterator[None]: async def _execute_query_pipeline( model: str, - messages: list[Any], + messages: list[AllMessageValues], retrieval_config: dict[str, Any], rerank: dict[str, Any] | None = None, stream: bool = False, @@ -311,7 +312,7 @@ async def _execute_query_pipeline( @client async def aquery( model: str, - messages: list[Any], + messages: list[AllMessageValues], retrieval_config: dict[str, Any], rerank: dict[str, Any] | None = None, stream: bool = False, @@ -358,12 +359,12 @@ async def aquery( @client def query( model: str, - messages: list[Any], + messages: list[AllMessageValues], retrieval_config: dict[str, Any], rerank: dict[str, Any] | None = None, stream: bool = False, **kwargs, -) -> ModelResponse | Coroutine[Any, Any, ModelResponse]: +) -> ModelResponse | Coroutine[None, None, ModelResponse]: """ Query a RAG pipeline. """ @@ -410,7 +411,7 @@ def ingest( file_id: str | None = None, timeout: float | httpx.Timeout | None = None, **kwargs, -) -> RAGIngestResponse | Coroutine[Any, Any, RAGIngestResponse]: +) -> RAGIngestResponse | Coroutine[None, None, RAGIngestResponse]: """ Ingest a document into a vector store. diff --git a/litellm/responses/litellm_completion_transformation/custom_tools.py b/litellm/responses/litellm_completion_transformation/custom_tools.py index cccae06c74b..4aa489d9e50 100644 --- a/litellm/responses/litellm_completion_transformation/custom_tools.py +++ b/litellm/responses/litellm_completion_transformation/custom_tools.py @@ -17,6 +17,7 @@ logic. import json from collections.abc import Mapping, Sequence +from types import MappingProxyType from typing import Final from pydantic import BaseModel, TypeAdapter, ValidationError @@ -28,6 +29,15 @@ from litellm.types.llms.openai import ( _MAX_ARGUMENTS_LEN: Final = 1_000_000 +TOOL_CALL_ITEM_ID_PREFIX_BY_TYPE: Final = MappingProxyType({"function_call": "fc", "custom_tool_call": "ctc"}) + + +def openai_shaped_tool_call_item_id(item_type: str, tool_id: str) -> str: + prefix: Final = TOOL_CALL_ITEM_ID_PREFIX_BY_TYPE.get(item_type) + if prefix is None or not tool_id or tool_id.startswith(prefix): + return tool_id + return f"{prefix}_{tool_id}" + def extract_custom_tool_names(tools: Sequence[object] | None) -> set[str]: """Extract names of tools originally defined as ``type: "custom"``.""" @@ -45,6 +55,21 @@ def is_custom_tool_call(tool_name: str, custom_tool_names: set[str]) -> bool: return tool_name in custom_tool_names +def serialize_tool_call_arguments(raw_arguments: object, default: str = "") -> str: + """Render tool call arguments as the JSON string tool-call schemas require. + + Arguments normally arrive already JSON-encoded, but clients and providers + also send the decoded object. ``str()`` on a dict yields a Python repr with + single quotes, which every downstream JSON parser rejects with errors like + "Expecting ',' delimiter". + """ + if isinstance(raw_arguments, str): + return raw_arguments or default + if raw_arguments is None: + return default + return json.dumps(raw_arguments, default=str) + + def unwrap_custom_tool_arguments(arguments: str) -> str: """Extract the raw content string from JSON-wrapped arguments. @@ -88,7 +113,7 @@ def build_tool_call_item_kwargs( item_type: Final = "custom_tool_call" if custom else "function_call" kwargs: Final[dict[str, str]] = { "type": item_type, - "id": call_id, + "id": openai_shaped_tool_call_item_id(item_type, call_id), "call_id": call_id, "name": name, "status": status, diff --git a/litellm/responses/litellm_completion_transformation/streaming_iterator.py b/litellm/responses/litellm_completion_transformation/streaming_iterator.py index 8b1eeb30306..db1c3acbefb 100644 --- a/litellm/responses/litellm_completion_transformation/streaming_iterator.py +++ b/litellm/responses/litellm_completion_transformation/streaming_iterator.py @@ -8,6 +8,7 @@ from litellm.main import stream_chunk_builder from litellm.responses.litellm_completion_transformation.custom_tools import ( build_tool_call_item_kwargs, extract_custom_tool_names, + serialize_tool_call_arguments, ) from litellm.responses.litellm_completion_transformation.transformation import ( LiteLLMCompletionResponsesConfig, @@ -113,6 +114,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._pending_tool_events: list[BaseLiteLLMOpenAIResponseObject] = [] self._tool_output_index_by_call_id: dict[str, int] = {} self._tool_args_by_call_id: dict[str, str] = {} + self._tool_item_id_by_call_id: dict[str, str] = {} # mutable-ok: filled per call id as tool call events stream self._tool_call_id_by_index: dict[int, str] = {} self._ambiguous_tool_call_indexes: set[int] = set() self._next_tool_output_index: int = 1 # output_index=0 reserved for the message item @@ -213,10 +215,10 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): fn_args_delta = "" if isinstance(fn, dict): fn_name = str(fn.get("name") or "") - fn_args_delta = str(fn.get("arguments") or "") + fn_args_delta = serialize_tool_call_arguments(fn.get("arguments")) else: fn_name = str(getattr(fn, "name", "") or "") - fn_args_delta = str(getattr(fn, "arguments", "") or "") + fn_args_delta = serialize_tool_call_arguments(getattr(fn, "arguments", "")) tool_name, tool_namespace = self._responses_namespace_tool_call_fields(fn_name) output_index = self._get_or_assign_tool_output_index(call_id) @@ -226,6 +228,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._sequence_number += 1 names = self._custom_tool_names item_kwargs = build_tool_call_item_kwargs(call_id, tool_name, "", "in_progress", names) + self._tool_item_id_by_call_id[call_id] = item_kwargs["id"] if tool_namespace: item_kwargs["namespace"] = tool_namespace event = OutputItemAddedEvent( @@ -247,7 +250,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._sequence_number += 1 delta_event: BaseLiteLLMOpenAIResponseObject = FunctionCallArgumentsDeltaEvent( type=ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DELTA, - item_id=call_id, + item_id=self._tool_item_id_by_call_id.get(call_id, call_id), output_index=output_index, delta=delta_chunk, ) @@ -284,10 +287,10 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): fn_args = "" if isinstance(fn, dict): fn_name = str(fn.get("name") or "") - fn_args = str(fn.get("arguments") or "") + fn_args = serialize_tool_call_arguments(fn.get("arguments")) else: fn_name = str(getattr(fn, "name", "") or "") - fn_args = str(getattr(fn, "arguments", "") or "") + fn_args = serialize_tool_call_arguments(getattr(fn, "arguments", "")) tool_name, tool_namespace = self._responses_namespace_tool_call_fields(fn_name) # Track if this is a new tool call that wasn't streamed @@ -299,6 +302,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._sequence_number += 1 names = self._custom_tool_names item_kwargs = build_tool_call_item_kwargs(call_id, tool_name, "", "in_progress", names) + self._tool_item_id_by_call_id[call_id] = item_kwargs["id"] if tool_namespace: item_kwargs["namespace"] = tool_namespace event = OutputItemAddedEvent( @@ -324,7 +328,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._sequence_number += 1 delta_event = FunctionCallArgumentsDeltaEvent( type=ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DELTA, - item_id=call_id, + item_id=self._tool_item_id_by_call_id.get(call_id, call_id), output_index=output_index, delta=delta_chunk, ) @@ -334,7 +338,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._sequence_number += 1 done_event = FunctionCallArgumentsDoneEvent( type=ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DONE, - item_id=call_id, + item_id=self._tool_item_id_by_call_id.get(call_id, call_id), output_index=output_index, arguments=final_args, ) @@ -344,6 +348,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._sequence_number += 1 names = self._custom_tool_names item_kwargs = build_tool_call_item_kwargs(call_id, tool_name, final_args, "completed", names) + item_kwargs["id"] = self._tool_item_id_by_call_id.setdefault(call_id, item_kwargs["id"]) if tool_namespace: item_kwargs["namespace"] = tool_namespace item_done_event = OutputItemDoneEvent( diff --git a/litellm/responses/litellm_completion_transformation/transformation.py b/litellm/responses/litellm_completion_transformation/transformation.py index 4a416f61e1b..5f3e88bb12f 100644 --- a/litellm/responses/litellm_completion_transformation/transformation.py +++ b/litellm/responses/litellm_completion_transformation/transformation.py @@ -93,6 +93,8 @@ from .custom_tools import ( convert_custom_tool_to_function_tool, extract_custom_tool_names, is_custom_tool_call, + openai_shaped_tool_call_item_id, + serialize_tool_call_arguments, unwrap_custom_tool_arguments, validated_allowed_callers, ) @@ -1010,7 +1012,7 @@ class LiteLLMCompletionResponsesConfig: type=cast(Literal["function"], tool_use_type), function=ChatCompletionToolCallFunctionChunk( name=str(function.get("name", "")), - arguments=str(function.get("arguments", "{}")), + arguments=serialize_tool_call_arguments(function.get("arguments"), "{}"), ), index=index, ) @@ -1539,7 +1541,7 @@ class LiteLLMCompletionResponsesConfig: type=cast(Literal["function"], _tool_use_definition.get("type") or "function"), function=ChatCompletionToolCallFunctionChunk( name=function.get("name") or "", - arguments=str(function.get("arguments") or ""), + arguments=serialize_tool_call_arguments(function.get("arguments")), ), index=0, ) @@ -1589,7 +1591,7 @@ class LiteLLMCompletionResponsesConfig: type="function", function=ChatCompletionToolCallFunctionChunk( name=f"{namespace}__{raw_name}" if qualify else raw_name, - arguments=str(raw_arguments or ""), + arguments=serialize_tool_call_arguments(raw_arguments), ), index=0, ) @@ -2024,7 +2026,7 @@ class LiteLLMCompletionResponsesConfig: function_definition = tool.function tool_name = function_definition.name or "" tool_id = tool.id or "" - tool_arguments = function_definition.get("arguments") or "" + tool_arguments = serialize_tool_call_arguments(function_definition.get("arguments")) # Check if this is a custom tool if is_custom_tool_call(tool_name, custom_tool_names): @@ -2033,7 +2035,7 @@ class LiteLLMCompletionResponsesConfig: custom_item = CustomToolCallOutputItem( type="custom_tool_call", call_id=tool_id, - id=tool_id, + id=openai_shaped_tool_call_item_id("custom_tool_call", tool_id), name=tool_name, input=input_str, status=function_definition.get("status") or "completed", @@ -2064,7 +2066,7 @@ class LiteLLMCompletionResponsesConfig: name=tool_name, arguments=tool_arguments, call_id=tool_id, - id=tool_id, + id=openai_shaped_tool_call_item_id("function_call", tool_id), type="function_call", status=function_definition.get("status") or "completed", ) @@ -2501,8 +2503,7 @@ class LiteLLMCompletionResponsesConfig: choice=choice, ) message_output_items.extend(image_generation_items) - else: - # Regular message output + elif choice.message.content is not None: message_output_items.append( GenericResponseOutputItem( type="message", @@ -2559,7 +2560,7 @@ class LiteLLMCompletionResponsesConfig: type="function", function=Function( name=tool_call.get("name") or "", - arguments=tool_call.get("arguments") or "", + arguments=serialize_tool_call_arguments(tool_call.get("arguments")), ), ) diff --git a/litellm/responses/mcp/litellm_proxy_mcp_handler.py b/litellm/responses/mcp/litellm_proxy_mcp_handler.py index 197d0c02ba8..367915156d1 100644 --- a/litellm/responses/mcp/litellm_proxy_mcp_handler.py +++ b/litellm/responses/mcp/litellm_proxy_mcp_handler.py @@ -399,7 +399,7 @@ class LiteLLM_Proxy_MCP_Handler: @staticmethod async def _process_mcp_tools_without_openai_transform( - user_api_key_auth: Any, + user_api_key_auth: "UserAPIKeyAuth | None", mcp_tools_with_litellm_proxy: Sequence[Mapping[str, object]], litellm_trace_id: str | None = None, mcp_auth_header: str | None = None, @@ -636,7 +636,7 @@ class LiteLLM_Proxy_MCP_Handler: async def _execute_tool_calls( tool_server_map: dict[str, str], tool_calls: Sequence[object], - user_api_key_auth: Any, + user_api_key_auth: "UserAPIKeyAuth | None", mcp_auth_header: str | None = None, mcp_server_auth_headers: dict[str, dict[str, str]] | None = None, oauth2_headers: dict[str, str] | None = None, diff --git a/litellm/router.py b/litellm/router.py index edff8294c3e..471a1116f44 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -8165,6 +8165,52 @@ class Router: if backend_value is not None: model_info[field] = backend_value + @staticmethod + def _inherit_builtin_base_rates_for_off_peak( + model_info: dict, # mutable-ok: cost-map entry filled in place + backend_model: str, + custom_llm_provider: str | None, + ) -> None: + """Fill missing pricing fields on a deployment entry that only sets + ``off_peak_pricing``, from the backend model's built-in cost map entry. + + Cost lookup selects the deployment-scoped entry over the shared backend + entry only when the deployment entry carries a base pricing field, and + ``off_peak_pricing`` is deliberately kept off the shared entry, so a + deployment spelling out only its off-peak schedule would otherwise + never receive the discount. The backend model's entire canonical cost + map entry is copied, field by field, so threshold, tiered, + service-tier, cache, character, and per-second rates as well as + companion billing fields like ``web_search_billing_unit`` and the + regional uplift multipliers all carry over, and peak-hour billing + through the deployment entry matches the shared backend entry exactly. + The raw ``litellm.model_cost`` entry is the copy source rather than + ``get_model_info``'s view of it, since that view synthesizes zero flat + token rates for backends without one and storing those would mark a + tiered-only backend explicitly priced free. Values are deep-copied to + keep the builtin entry isolated. User-specified fields always win; + no-op when any base pricing field is already set or the backend model + has no canonical entry. + """ + if not model_info.get("off_peak_pricing"): + return + if any( + model_info.get(field) is not None + for field in ("input_cost_per_token", "input_cost_per_second", "tiered_pricing") + ): + return + try: + backend_info: Final = litellm.get_model_info(model=backend_model, custom_llm_provider=custom_llm_provider) + except Exception: # noqa: BLE001 # get_model_info raises plain Exception for an unmapped backend model + return + backend_entry: Final = litellm.model_cost.get(backend_info.get("key") or "") + if not isinstance(backend_entry, dict): + return + for field, backend_value in backend_entry.items(): + if model_info.get(field) is not None or backend_value is None: + continue + model_info[field] = copy.deepcopy(backend_value) + @staticmethod def _inherit_builtin_tiered_output_rate( model_info: dict, backend_model: str, custom_llm_provider: str | None @@ -8253,6 +8299,11 @@ class Router: if deployment.litellm_params.get(field) is not None: _model_info[field] = deployment.litellm_params[field] + Router._inherit_builtin_base_rates_for_off_peak( + model_info=_model_info, + backend_model=deployment.litellm_params.model, + custom_llm_provider=deployment.litellm_params.custom_llm_provider, + ) if _model_info.get("input_cost_per_token") is not None: Router._inherit_builtin_cache_pricing( model_info=_model_info, @@ -8994,6 +9045,11 @@ class Router: if field_value is not None: _model_info_dict[field] = field_value + Router._inherit_builtin_base_rates_for_off_peak( + model_info=_model_info_dict, + backend_model=deployment.litellm_params.model, + custom_llm_provider=deployment.litellm_params.custom_llm_provider, + ) if _model_info_dict.get("input_cost_per_token") is not None: Router._inherit_builtin_cache_pricing( model_info=_model_info_dict, @@ -9249,6 +9305,11 @@ class Router: field_value = deployment.litellm_params.get(field) if field_value is not None: model_info[field] = field_value + Router._inherit_builtin_base_rates_for_off_peak( + model_info=model_info, + backend_model=deployment.litellm_params.model, + custom_llm_provider=deployment.litellm_params.custom_llm_provider, + ) if model_info.get("input_cost_per_token") is not None: Router._inherit_builtin_cache_pricing( model_info=model_info, diff --git a/litellm/router_strategy/budget_limiter.py b/litellm/router_strategy/budget_limiter.py index d57d7da0410..a8d51f95e45 100644 --- a/litellm/router_strategy/budget_limiter.py +++ b/litellm/router_strategy/budget_limiter.py @@ -20,6 +20,7 @@ anthropic: import asyncio import builtins +from collections.abc import Mapping from datetime import datetime, timedelta, timezone from typing import Any, Final @@ -54,19 +55,19 @@ class _LiteLLMParamsDictView: __slots__ = ("_params",) - def __init__(self, params: dict[str, Any]): + def __init__(self, params: Mapping[str, object]): self._params = params - def __getattr__(self, key: str) -> Any: + def __getattr__(self, key: str) -> object: return self._params.get(key) - def __getitem__(self, key: str) -> Any: + def __getitem__(self, key: str) -> object: return self._params.get(key) def __contains__(self, key: str) -> bool: return key in self._params - def get(self, key: str, default: Any = None) -> Any: + def get(self, key: str, default: object = None) -> object: return self._params.get(key, default) def keys(self): @@ -84,10 +85,10 @@ class _LiteLLMParamsDictView: def __len__(self) -> int: return len(self._params) - def dict(self) -> dict[str, Any]: + def dict(self) -> builtins.dict[str, object]: return dict(self._params) - def model_dump(self) -> builtins.dict[str, Any]: + def model_dump(self) -> builtins.dict[str, object]: return dict(self._params) diff --git a/litellm/router_strategy/complexity_router/complexity_router.py b/litellm/router_strategy/complexity_router/complexity_router.py index be7653902a7..577cee0920d 100644 --- a/litellm/router_strategy/complexity_router/complexity_router.py +++ b/litellm/router_strategy/complexity_router/complexity_router.py @@ -282,7 +282,7 @@ def _response_cost_or_none(response: ModelResponse) -> float | None: return float(cost) -def _effective_turn_off_message_logging(request_kwargs: Mapping[str, Any] | None) -> bool | None: +def _effective_turn_off_message_logging(request_kwargs: Mapping[str, object] | None) -> bool | None: from litellm.litellm_core_utils.initialize_dynamic_callback_params import ( initialize_standard_callback_dynamic_params, ) @@ -1925,7 +1925,7 @@ class ComplexityRouter(CustomLogger): ceiling_severity: Final = self._active_tier_severity(hard_ceiling) if hard_ceiling is not None else None best_model: str | None = None best_score = float("-inf") - candidate_scores: Final[list[dict[str, Any]]] = [] + candidate_scores: Final[list[dict[str, object]]] = [] for model in candidates: if floor_severity is not None and all( self._active_tier_severity(model_tier) < floor_severity diff --git a/litellm/secret_managers/hashicorp_secret_manager.py b/litellm/secret_managers/hashicorp_secret_manager.py index e2662d96b52..8f677b54700 100644 --- a/litellm/secret_managers/hashicorp_secret_manager.py +++ b/litellm/secret_managers/hashicorp_secret_manager.py @@ -1,7 +1,9 @@ import os -from typing import Any, Final +from collections.abc import Mapping +from typing import Final, Protocol import httpx +from typing_extensions import ReadOnly, TypedDict import litellm from litellm._logging import verbose_logger @@ -17,6 +19,72 @@ from litellm.proxy._types import KeyManagementSystem from .base_secret_manager import BaseSecretManager, raise_if_unsafe_secret_name +class _VaultAuthData(TypedDict): + """The ``auth`` block Vault returns from a login endpoint.""" + + client_token: ReadOnly[str] + lease_duration: ReadOnly[int] + + +class _VaultLoginResponse(TypedDict): + """Body of a Vault ``/v1/auth/.../login`` response.""" + + auth: ReadOnly[_VaultAuthData] + + +class _VaultSecretTarget(TypedDict): + """Resolved coordinates of one Vault KV v2 secret.""" + + url: ReadOnly[str] + data_key: ReadOnly[str] + secret_name: ReadOnly[str] + + +class _VaultSecretDataBlock(TypedDict, total=False): + """The inner ``data`` block of a Vault KV v2 read body.""" + + data: ReadOnly[Mapping[str, object]] + + +class _VaultSecretReadResponse(TypedDict, total=False): + """Body of a Vault KV v2 secret read, narrowed to the nesting this module walks.""" + + data: ReadOnly[_VaultSecretDataBlock] + + +class _VaultLoginResponseSource(Protocol): + """A Vault login call's HTTP response, read for the auth block it carries.""" + + def json(self) -> _VaultLoginResponse: ... + + +class _VaultSecretReadSource(Protocol): + """A Vault KV v2 read response, read for the nested secret data it carries.""" + + def json(self) -> _VaultSecretReadResponse: ... + + +class _JsonObjectSource(Protocol): + """A Vault response whose body is a JSON object nothing further is assumed about.""" + + def json(self) -> dict[str, object]: ... + + +def _vault_login_body(response: _VaultLoginResponseSource) -> _VaultLoginResponse: + """Decode the body of a Vault login response.""" + return response.json() + + +def _vault_secret_read_body(response: _VaultSecretReadSource) -> _VaultSecretReadResponse: + """Decode the body of a Vault KV v2 secret read response.""" + return response.json() + + +def _json_object_body(response: _JsonObjectSource) -> dict[str, object]: + """Decode a Vault response body as a plain JSON object.""" + return response.json() + + class HashicorpSecretManager(BaseSecretManager): def __init__(self): from litellm.proxy.proxy_server import CommonProxyErrors, premium_user @@ -130,7 +198,8 @@ class HashicorpSecretManager(BaseSecretManager): ) resp.raise_for_status() - auth_data: Final = resp.json()["auth"] + login_response: Final = _vault_login_body(resp) + auth_data: Final = login_response["auth"] token: Final = auth_data["client_token"] _lease_duration: Final = auth_data["lease_duration"] @@ -191,8 +260,10 @@ class HashicorpSecretManager(BaseSecretManager): json=self._get_tls_cert_auth_body(), ) resp.raise_for_status() - token: Final = resp.json()["auth"]["client_token"] - _lease_duration: Final = resp.json()["auth"]["lease_duration"] + token_response: Final = _vault_login_body(resp) + token: Final = token_response["auth"]["client_token"] + lease_response: Final = _vault_login_body(resp) + _lease_duration: Final = lease_response["auth"]["lease_duration"] verbose_logger.debug("Successfully obtained Vault token via TLS cert auth.") self.cache.set_cache(key="hcp_vault_token", value=token, ttl=_lease_duration) return token @@ -205,9 +276,9 @@ class HashicorpSecretManager(BaseSecretManager): def get_url( self, secret_name: str, - namespace: str | None = None, - mount_name: str | None = None, - path_prefix: str | None = None, + namespace: object = None, + mount_name: object = None, + path_prefix: object = None, ) -> str: """ Constructs the Vault URL for KV v2 secrets. @@ -238,7 +309,7 @@ class HashicorpSecretManager(BaseSecretManager): _url += secret_name return _url - def _sanitize_plain_value(self, value: str | int | None) -> str | None: + def _sanitize_plain_value(self, value: object) -> str | None: if value is None: return None value_str: Final = str(value).strip() @@ -246,23 +317,23 @@ class HashicorpSecretManager(BaseSecretManager): return None return value_str - def _sanitize_path_component(self, value: str | int | None) -> str | None: + def _sanitize_path_component(self, value: object) -> str | None: sanitized_value = self._sanitize_plain_value(value) if sanitized_value is None: return None sanitized_value = sanitized_value.strip("/") return sanitized_value or None - def _extract_secret_manager_settings(self, optional_params: dict | None) -> dict[str, Any]: + def _extract_secret_manager_settings(self, optional_params: dict | None) -> dict[str, object]: if not isinstance(optional_params, dict): return {} candidate: Final = optional_params.get("secret_manager_settings") - source: Final = candidate if isinstance(candidate, dict) else optional_params + source: Final[Mapping[str, object]] = candidate if isinstance(candidate, dict) else optional_params allowed_keys: Final = {"namespace", "mount", "path_prefix", "data"} return {k: source[k] for k in allowed_keys if k in source} - def _build_secret_target(self, secret_name: str, optional_params: dict | None) -> dict[str, Any]: + def _build_secret_target(self, secret_name: str, optional_params: dict | None) -> _VaultSecretTarget: settings: Final = self._extract_secret_manager_settings(optional_params) namespace: Final = settings.get("namespace", self.vault_namespace) @@ -331,7 +402,7 @@ class HashicorpSecretManager(BaseSecretManager): response.raise_for_status() # For KV v2, the secret is in response.json()["data"]["data"] - json_resp: Final = response.json() + json_resp: Final = _json_object_body(response) _value: Final = self._get_secret_value_from_json_response(json_resp) self.cache.set_cache(secret_name, _value) return _value @@ -362,7 +433,7 @@ class HashicorpSecretManager(BaseSecretManager): response.raise_for_status() # For KV v2, the secret is in response.json()["data"]["data"] - json_resp: Final = response.json() + json_resp: Final = _json_object_body(response) _value: Final = self._get_secret_value_from_json_response(json_resp) self.cache.set_cache(secret_name, _value) return _value @@ -379,7 +450,7 @@ class HashicorpSecretManager(BaseSecretManager): optional_params: dict | None = None, timeout: float | httpx.Timeout | None = None, tags: dict | list | None = None, - ) -> dict[str, Any]: + ) -> dict[str, object]: """ Writes a secret to Vault KV v2 using an async HTTPX client. @@ -413,7 +484,7 @@ class HashicorpSecretManager(BaseSecretManager): json=data, ) response.raise_for_status() - return response.json() + return _json_object_body(response) except Exception as e: verbose_logger.exception("Error writing secret to Hashicorp Vault: %s", e) return {"status": "error", "message": str(e)} @@ -500,7 +571,7 @@ class HashicorpSecretManager(BaseSecretManager): headers=self._get_request_headers(), ) response.raise_for_status() - json_resp: Final = response.json() + json_resp: Final = _vault_secret_read_body(response) # Use data_key from target to get the correct value data_key: Final = new_target["data_key"] new_secret_value_from_vault: Final = json_resp.get("data", {}).get("data", {}).get(data_key, None) diff --git a/litellm/setup_wizard.py b/litellm/setup_wizard.py index d6b3dfa3285..dd147aaccee 100644 --- a/litellm/setup_wizard.py +++ b/litellm/setup_wizard.py @@ -53,11 +53,12 @@ PROVIDERS: Final[list[dict]] = [ { "id": "anthropic", "name": "Anthropic", - "description": "Claude Fable 5, Opus 5, Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 5, Sonnet 4.6, Haiku 4.5", + "description": "Claude Fable 5.1, Fable 5, Opus 5, Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 5, Sonnet 4.6, Haiku 4.5", "env_key": "ANTHROPIC_API_KEY", "key_hint": "sk-ant-...", "test_model": "claude-haiku-4-5-20251001", "models": [ + "claude-fable-5-1", "claude-fable-5", "claude-opus-5", "claude-sonnet-5", diff --git a/litellm/types/guardrails.py b/litellm/types/guardrails.py index 4cf4fa62eff..c17103da890 100644 --- a/litellm/types/guardrails.py +++ b/litellm/types/guardrails.py @@ -136,6 +136,7 @@ class SupportedGuardrailIntegrations(Enum): HEADROOM = "headroom" COMPRESR = "compresr" STRAIKER = "straiker" + ALICE = "alice" class Role(Enum): diff --git a/litellm/types/llms/openai.py b/litellm/types/llms/openai.py index fcade835cce..32d88da0085 100644 --- a/litellm/types/llms/openai.py +++ b/litellm/types/llms/openai.py @@ -327,6 +327,10 @@ class BatchGuardrailReport(BaseModel): """Every record that was redacted or dropped, in file order.""" +_JsonValue: TypeAlias = object +"""Alias for ``object``, usable inside model bodies that declare a field named ``object``.""" + + BATCH_GUARDRAIL_RESPONSE_FIELD: Final = "litellm_batch_guardrail" @@ -1191,7 +1195,7 @@ class ShellToolParam(TypedDict, total=False): type: Required[Literal["shell"] | str] """The type of tool. Use ``\"shell\"``.""" - environment: Required[dict[str, Any]] + environment: Required[dict[str, object]] """Environment config: ``type`` (e.g. ``\"container_auto\"``, ``\"container_reference\"``, ``\"local\"``), optional ``container_id``, ``network_policy``, ``domain_secrets``, ``skills``.""" @@ -1308,7 +1312,7 @@ class ResponseAPIUsage(BaseLiteLLMOpenAIResponseObject): @field_validator("cost", mode="before") @classmethod - def parse_cost(cls, v: Any) -> float | None: + def parse_cost(cls, v: object) -> object: """Normalise cost: accept either a float or a dict with a ``total_cost`` key.""" if isinstance(v, dict): return v.get("total_cost") @@ -1805,7 +1809,7 @@ class ErrorEventError(BaseLiteLLMOpenAIResponseObject): type: str # e.g., 'invalid_request_error' code: str # e.g., 'context_length_exceeded' message: str - param: str | dict[str, Any] | None = None + param: str | dict[str, object] | None = None class ErrorEvent(BaseLiteLLMOpenAIResponseObject): @@ -2418,7 +2422,7 @@ class OpenAIVideoObject(BaseModel): expires_at: int | None = None """Unix timestamp (seconds) for when the downloadable assets expire, if set.""" - error: dict[str, Any] | None = None + error: dict[str, _JsonValue] | None = None """Error payload that explains why generation failed, if applicable.""" progress: int | None = None @@ -2436,15 +2440,15 @@ class OpenAIVideoObject(BaseModel): model: str | None = None """The video generation model that produced the job.""" - _hidden_params: dict[str, Any] = {} + _hidden_params: dict[str, _JsonValue] = {} def __contains__(self, key) -> bool: return hasattr(self, key) - def get(self, key, default=None): + def get(self, key, default=None) -> _JsonValue: return getattr(self, key, default) - def __getitem__(self, key): + def __getitem__(self, key) -> _JsonValue: return getattr(self, key) def json(self, **kwargs): diff --git a/litellm/types/proxy/guardrails/guardrail_hooks/alice.py b/litellm/types/proxy/guardrails/guardrail_hooks/alice.py new file mode 100644 index 00000000000..73d31673dab --- /dev/null +++ b/litellm/types/proxy/guardrails/guardrail_hooks/alice.py @@ -0,0 +1,21 @@ +from pydantic import Field + +from .base import GuardrailConfigModel + + +class AliceGuardrailConfigModel(GuardrailConfigModel): + api_key: str | None = Field( + default=None, + description=("The API key for Alice. If not provided, the `ALICE_API_KEY` environment variable is checked."), + ) + api_base: str | None = Field( + default=None, + description=( + "The API base URL for Alice. If not provided, the `ALICE_API_BASE` environment " + "variable is checked, then `https://api.alice.io`." + ), + ) + + @staticmethod + def ui_friendly_name() -> str: + return "Alice" diff --git a/litellm/types/router.py b/litellm/types/router.py index ab6c807ba20..e0957383aac 100644 --- a/litellm/types/router.py +++ b/litellm/types/router.py @@ -369,7 +369,7 @@ class GenericLiteLLMParams(CredentialLiteLLMParams, CustomPricingLiteLLMParams): @model_validator(mode="before") @classmethod - def preprocess_input_data(cls, data: Any) -> Any: + def preprocess_input_data(cls, data: object) -> object: """ Pre-process input data before validation: 1. Filter out reserved Python keywords ('self', 'params', '__class__') to prevent @@ -627,6 +627,11 @@ class AlertingConfig(BaseModel): alerting_threshold: float | None = 300 +def _resolved_annotations(model_class: type[object]) -> Mapping[str, object]: + """Resolve a class's annotations, keeping each resolved annotation opaque.""" + return get_type_hints(model_class) + + class ModelGroupInfo(BaseModel): model_group: str providers: list[str] @@ -655,7 +660,7 @@ class ModelGroupInfo(BaseModel): configurable_clientside_auth_params: CONFIGURABLE_CLIENTSIDE_AUTH_PARAMS = None def __init__(self, **data) -> None: - for field_name, field_type in get_type_hints(self.__class__).items(): + for field_name, field_type in _resolved_annotations(self.__class__).items(): if field_type is bool and data.get(field_name) is None: data[field_name] = False super().__init__(**data) diff --git a/litellm/types/utils.py b/litellm/types/utils.py index 55a32989b1c..3a0883b6607 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -193,6 +193,38 @@ class AgenticLoopParams(TypedDict, total=False): """The LLM provider name (e.g., 'bedrock', 'anthropic')""" +class OffPeakWindow(TypedDict, total=False): + """One off-peak rule: UTC time-of-day windows, optionally restricted to weekdays. + + hours_utc is a "HH:MM-HH:MM" string in UTC, or a list of them; a window may wrap past + midnight and an equal-ended window covers the whole day. weekdays is a list of days the + rule applies on, as ISO-8601 numbers (1 = Monday .. 7 = Sunday) or English day names; + omitted means every day. The weekday is read on the calendar named by the block's + weekday_timezone. + """ + + hours_utc: ReadOnly[str | Sequence[str]] + weekdays: ReadOnly[Sequence[int | str]] + + +class OffPeakPricing(TypedDict, total=False): + """Time-windowed off-peak rates for providers that discount by time of day (e.g. DeepSeek). + + hours_utc is a "HH:MM-HH:MM" string in UTC, or a list of them for multiple daily windows, + applying on every day of the week; a window may wrap past midnight. windows adds + day-of-week-qualified rules (e.g. weekend-only whole-day off-peak), matched as a union + with hours_utc. weekday_timezone names the IANA calendar weekdays are read on, defaulting + to UTC. Any rate left unset falls back to the standard rate. + """ + + hours_utc: ReadOnly[str | Sequence[str]] + windows: ReadOnly[Sequence[OffPeakWindow]] + weekday_timezone: ReadOnly[str] + input_cost_per_token: ReadOnly[float] + output_cost_per_token: ReadOnly[float] + cache_read_input_token_cost: ReadOnly[float] + + class ModelInfoBase(ProviderSpecificModelInfo, total=False): key: Required[str] # the key in litellm.model_cost which is returned @@ -225,6 +257,7 @@ class ModelInfoBase(ProviderSpecificModelInfo, total=False): # Smallest prefix this model will actually cache, whatever caching mechanism its provider uses. # Absent means the provider-agnostic default applies; see MINIMUM_PROMPT_CACHE_TOKEN_COUNT. prompt_cache_min_tokens: int | None + off_peak_pricing: ReadOnly[OffPeakPricing | None] # time-windowed off-peak rates input_cost_per_character: float | None # only for vertex ai models input_cost_per_audio_token: float | None input_cost_per_token_above_128k_tokens: float | None # only for vertex ai models @@ -3486,17 +3519,22 @@ class CustomPricingLiteLLMParams(MirroredPricingParams): return {k: v for k, v in model_info.items() if k not in cls.model_fields} -SHARED_BACKEND_MODEL_INFO_FIELDS: Final[frozenset[str]] = frozenset( - ModelInfoBase.__required_keys__ | ModelInfoBase.__optional_keys__ -) - frozenset(CustomPricingLiteLLMParams.model_fields) +DEPLOYMENT_SCOPED_PRICING_FIELDS: Final[frozenset[str]] = frozenset({"off_peak_pricing"}) + +SHARED_BACKEND_MODEL_INFO_FIELDS: Final[frozenset[str]] = ( + frozenset(ModelInfoBase.__required_keys__ | ModelInfoBase.__optional_keys__) + - frozenset(CustomPricingLiteLLMParams.model_fields) + - DEPLOYMENT_SCOPED_PRICING_FIELDS +) def shared_backend_model_info(model_info: dict[str, Any]) -> dict[str, Any]: """Return only the fields safe to register under a shared ``{provider}/{model}`` key in ``litellm.model_cost``: cost-map schema fields (``ModelInfoBase``) minus - per-deployment pricing overrides. Per-deployment metadata (``id``, - ``access_via_team_ids``, arbitrary custom keys) never belongs on the shared key; - it stays under the deployment's unique model id. + per-deployment pricing overrides and deployment-scoped pricing blocks such as + ``off_peak_pricing``. Per-deployment metadata (``id``, ``access_via_team_ids``, + arbitrary custom keys) never belongs on the shared key; it stays under the + deployment's unique model id. """ return {k: v for k, v in model_info.items() if k in SHARED_BACKEND_MODEL_INFO_FIELDS} @@ -3771,6 +3809,8 @@ class LlmProviders(str, Enum): CODESTRAL = "codestral" TEXT_COMPLETION_CODESTRAL = "text-completion-codestral" DASHSCOPE = "dashscope" + QWENCLOUD = "qwencloud" + QWEN_AI_PLATFORM = "qwen_ai_platform" MODELSCOPE = "modelscope" MOONSHOT = "moonshot" PUBLICAI = "publicai" diff --git a/litellm/utils.py b/litellm/utils.py index ab011f4123d..2fdc4a45148 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -2869,10 +2869,9 @@ def _update_dictionary(existing_dict: dict, new_dict: dict) -> dict: elif isinstance(v, dict): existing_nested_dict = existing_dict.get(k) if isinstance(existing_nested_dict, dict): - existing_nested_dict.update(v) - existing_dict[k] = existing_nested_dict + existing_dict[k] = {**existing_nested_dict, **v} # mutable-ok: copy-on-write merge else: - existing_dict[k] = v + existing_dict[k] = dict(v) # mutable-ok: detached copy, never the caller's dict by reference else: existing_dict[k] = v @@ -3561,10 +3560,10 @@ def get_optional_params_embeddings( non_default_params=non_default_params, optional_params={}, kwargs=kwargs ) elif custom_llm_provider == "vertex_ai" or custom_llm_provider == "gemini": - # OpenAI SDKs (and litellm's own client) send encoding_format="float" - # by default; float lists are exactly what the vertex API returns, so - # the param is a no-op — don't reject the provider default. Other - # values (e.g. "base64") stay on the unsupported-param path below. + # OpenAI SDKs send encoding_format="float" by default; float lists are + # exactly what the vertex API returns, so the param is a no-op and the + # provider default is not rejected. Other values (e.g. "base64") stay + # on the unsupported-param path below. if non_default_params.get("encoding_format") == "float": non_default_params.pop("encoding_format") supported_params = get_supported_openai_params( @@ -5860,6 +5859,7 @@ def _get_model_info_helper( cache_creation_input_token_cost_above_1hr=_model_info.get( "cache_creation_input_token_cost_above_1hr", None ), + off_peak_pricing=_model_info.get("off_peak_pricing", None), input_cost_per_character=_model_info.get("input_cost_per_character", None), input_cost_per_token_above_128k_tokens=_model_info.get("input_cost_per_token_above_128k_tokens", None), input_cost_per_token_above_200k_tokens=_model_info.get("input_cost_per_token_above_200k_tokens", None), @@ -6586,11 +6586,11 @@ def validate_environment( keys_in_environment = True else: missing_keys.append("WANDB_API_KEY") - elif custom_llm_provider == "dashscope": - if "DASHSCOPE_API_KEY" in os.environ: + elif custom_llm_provider in ("dashscope", "qwencloud", "qwen_ai_platform"): + if f"{custom_llm_provider.upper()}_API_KEY" in os.environ or "DASHSCOPE_API_KEY" in os.environ: keys_in_environment = True else: - missing_keys.append("DASHSCOPE_API_KEY") + missing_keys.append(f"{custom_llm_provider.upper()}_API_KEY") elif custom_llm_provider == "modelscope": if "MODELSCOPE_API_KEY" in os.environ: keys_in_environment = True @@ -8152,6 +8152,11 @@ class ProviderConfigManager: LlmProviders.NEBIUS: (lambda: litellm.NebiusConfig(), False), LlmProviders.WANDB: (lambda: litellm.WandbConfig(), False), LlmProviders.DASHSCOPE: (lambda: litellm.DashScopeChatConfig(), False), + LlmProviders.QWENCLOUD: (lambda: litellm.QwenCloudChatConfig(), False), + LlmProviders.QWEN_AI_PLATFORM: ( + lambda: litellm.QwenAIPlatformChatConfig(), + False, + ), LlmProviders.MODELSCOPE: (lambda: litellm.ModelScopeChatConfig(), False), LlmProviders.MOONSHOT: (lambda: litellm.MoonshotChatConfig(), False), LlmProviders.DOCKER_MODEL_RUNNER: ( @@ -8366,12 +8371,16 @@ class ProviderConfigManager: ) return VolcEngineEmbeddingConfig() - elif litellm.LlmProviders.DASHSCOPE == provider: - from litellm.llms.dashscope.embed.transformation import ( - DashScopeEmbeddingConfig, + elif provider in ( + litellm.LlmProviders.DASHSCOPE, + litellm.LlmProviders.QWENCLOUD, + litellm.LlmProviders.QWEN_AI_PLATFORM, + ): + from litellm.llms.dashscope.common_utils import ( + get_dashscope_family_embedding_config, ) - return DashScopeEmbeddingConfig() + return get_dashscope_family_embedding_config(provider.value) elif litellm.LlmProviders.OVHCLOUD == provider: return litellm.OVHCloudEmbeddingConfig() elif litellm.LlmProviders.SNOWFLAKE == provider: @@ -8444,12 +8453,16 @@ class ProviderConfigManager: return litellm.VoyageRerankConfig() elif litellm.LlmProviders.WATSONX == provider: return litellm.IBMWatsonXRerankConfig() - elif litellm.LlmProviders.DASHSCOPE == provider: - from litellm.llms.dashscope.rerank.transformation import ( - DashScopeRerankConfig, + elif provider in ( + litellm.LlmProviders.DASHSCOPE, + litellm.LlmProviders.QWENCLOUD, + litellm.LlmProviders.QWEN_AI_PLATFORM, + ): + from litellm.llms.dashscope.common_utils import ( + get_dashscope_family_rerank_config, ) - return DashScopeRerankConfig() + return get_dashscope_family_rerank_config(provider.value) return litellm.CohereRerankConfig() @staticmethod @@ -9122,12 +9135,16 @@ class ProviderConfigManager: ) return get_openrouter_image_generation_config(model) - elif LlmProviders.DASHSCOPE == provider: - from litellm.llms.dashscope.image_generation import ( - get_dashscope_image_generation_config, + elif provider in ( + LlmProviders.DASHSCOPE, + LlmProviders.QWENCLOUD, + LlmProviders.QWEN_AI_PLATFORM, + ): + from litellm.llms.dashscope.common_utils import ( + get_dashscope_family_image_generation_config, ) - return get_dashscope_image_generation_config(model) + return get_dashscope_family_image_generation_config(provider.value) elif LlmProviders.MODELSCOPE == provider: from litellm.llms.modelscope.image_generation import ( get_modelscope_image_generation_config, diff --git a/litellm/vector_stores/vector_store_registry.py b/litellm/vector_stores/vector_store_registry.py index 22d27bc3266..b71d6784873 100644 --- a/litellm/vector_stores/vector_store_registry.py +++ b/litellm/vector_stores/vector_store_registry.py @@ -112,7 +112,9 @@ class VectorStoreRegistry: Dynamically extracts all parameters defined in VECTOR_STORE_OPENAI_PARAMS. """ # Get the list of supported param names from the Literal type - supported_params: Final = get_args(VECTOR_STORE_OPENAI_PARAMS) + supported_params: Final = tuple( + param for param in get_args(VECTOR_STORE_OPENAI_PARAMS) if isinstance(param, str) + ) # Extract only the params that exist in the tool kwargs: Final = {param: tool.get(param) for param in supported_params if param in tool} @@ -503,7 +505,7 @@ class VectorStoreRegistry: vector_stores_from_db.append(_litellm_managed_vector_store) return vector_stores_from_db - def get_credentials_for_vector_store(self, vector_store_id: str) -> dict[str, Any]: + def get_credentials_for_vector_store(self, vector_store_id: str) -> dict[str, object]: """ Get the credentials for a vector store diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 718e6c489fd..1b2001cdadd 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -554,6 +554,7 @@ "supports_vision": true }, "amazon.nova-sonic-v1:0": { + "deprecation_date": "2026-09-14", "input_cost_per_audio_token": 3.4e-06, "input_cost_per_token": 6e-08, "litellm_provider": "bedrock", @@ -1450,6 +1451,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "global.anthropic.claude-fable-5": { "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, @@ -1487,6 +1526,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "global.anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "us.anthropic.claude-fable-5": { "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, @@ -1524,6 +1601,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "us.anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.375e-05, + "cache_creation_input_token_cost_above_1hr": 2.2e-05, + "cache_read_input_token_cost": 2.75e-07, + "input_cost_per_token": 1.1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "eu.anthropic.claude-fable-5": { "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, @@ -1561,6 +1676,44 @@ "supports_parallel_tool_use_config": true, "prompt_cache_min_tokens": 512 }, + "eu.anthropic.claude-fable-5-1": { + "cache_creation_input_token_cost": 1.375e-05, + "cache_creation_input_token_cost_above_1hr": 2.2e-05, + "cache_read_input_token_cost": 2.75e-07, + "input_cost_per_token": 1.1e-05, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, + "bedrock_output_config_effort_ceiling": "xhigh", + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 512 + }, "anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, "supports_adaptive_thinking": true, @@ -1591,7 +1744,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1627,7 +1780,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1663,7 +1816,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1699,7 +1852,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1735,7 +1888,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -1771,7 +1924,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "supports_parallel_tool_use_config": true, @@ -2064,7 +2217,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2101,7 +2254,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2138,7 +2291,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2175,7 +2328,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2212,7 +2365,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -2249,7 +2402,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, + "supports_native_structured_output": false, "supports_max_reasoning_effort": true, "supports_output_config": true, "bedrock_output_config_effort_ceiling": "xhigh", @@ -3045,6 +3198,7 @@ "prompt_cache_min_tokens": 2048 }, "azure_ai/claude-fable-5": { + "deprecation_date": "2027-12-05", "supports_mid_conversation_system": true, "input_cost_per_token": 1e-05, "output_cost_per_token": 5e-05, @@ -3077,7 +3231,42 @@ "supports_max_reasoning_effort": true, "prompt_cache_min_tokens": 512 }, + "azure_ai/claude-fable-5-1": { + "supports_mid_conversation_system": true, + "input_cost_per_token": 1e-05, + "output_cost_per_token": 5e-05, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512 + }, "azure_ai/claude-opus-5": { + "deprecation_date": "2027-07-08", "supports_mid_conversation_system": true, "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, @@ -3110,6 +3299,7 @@ "prompt_cache_min_tokens": 512 }, "azure_ai/claude-opus-4-8": { + "deprecation_date": "2027-09-01", "supports_mid_conversation_system": true, "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, @@ -3188,6 +3378,7 @@ "prompt_cache_min_tokens": 1024 }, "azure_ai/claude-sonnet-5": { + "deprecation_date": "2027-06-30", "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, @@ -9979,6 +10170,22 @@ "supports_reasoning": true, "supports_tool_choice": true }, + "azure_ai/deepseek-v4-flash-0731": { + "cache_read_input_token_cost": 2.8e-08, + "deprecation_date": "2026-12-03", + "input_cost_per_token": 1.9e-07, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.1e-07, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/deepseek/", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, "azure_ai/embed-v-4-0": { "input_cost_per_token": 1.2e-07, "litellm_provider": "azure_ai", @@ -12287,6 +12494,7 @@ "supports_tool_choice": true }, "cerebras/zai-glm-4.7": { + "deprecation_date": "2026-08-17", "input_cost_per_token": 2.25e-06, "litellm_provider": "cerebras", "max_input_tokens": 128000, @@ -13022,6 +13230,47 @@ "supports_native_structured_output": true, "source": "https://docs.anthropic.com/en/docs/about-claude/models/overview" }, + "claude-fable-5-1": { + "deprecation_date": "2027-09-01", + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "anthropic", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_mid_conversation_system": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "provider_specific_entry": { + "us": 1.1 + }, + "supports_output_config": true, + "prompt_cache_min_tokens": 512, + "supports_native_structured_output": true, + "source": "https://platform.claude.com/docs/en/models/fable-5-1/overview" + }, "claude-opus-5": { "deprecation_date": "2027-07-24", "cache_creation_input_token_cost": 6.25e-06, @@ -14718,6 +14967,1910 @@ "/v1/images/generations" ] }, + "qwencloud/deepseek-v4-flash": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/deepseek-v4-flash-0731": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/deepseek-v4-pro": { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2.4e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4.8e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/glm-5.1": { + "cache_read_input_token_cost": 2.6e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 202745, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/glm-5.2": { + "cache_read_input_token_cost": 2.8e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 1048576, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/kimi-k2.7-code": { + "cache_read_input_token_cost": 1.9e-07, + "input_cost_per_token": 9.5e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 229376, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen-coder": { + "input_cost_per_token": 3e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.5e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-flash": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-flash-2025-07-28": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-max": { + "input_cost_per_token": 1.6e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 30720, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 6.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-01-25": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-04-28": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-07-14": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-plus-2025-07-28": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-plus-2025-09-11": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-plus-latest": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen-turbo": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-turbo-2024-11-01": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-turbo-2025-04-28": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-turbo-latest": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwencloud", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-30b-a3b": { + "litellm_provider": "qwencloud", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-coder-flash": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.2e-07, + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-coder-flash-2025-07-28": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-coder-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.8e-07, + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 6e-07, + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-coder-plus-2025-07-22": { + "litellm_provider": "qwencloud", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3-max-preview": { + "litellm_provider": "qwencloud", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwencloud/qwen3-max": { + "litellm_provider": "qwencloud", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwencloud/qwen3-max-2026-01-23": { + "litellm_provider": "qwencloud", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwencloud/qwen3-next-80b-a3b-instruct": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-next-80b-a3b-thinking": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3-vl-235b-a22b-instruct": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.6e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-235b-a22b-thinking": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-32b-instruct": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 6.4e-07, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-32b-thinking": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.87e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwen3-vl-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 260096, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 2e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 6e-07, + "output_cost_per_token": 4.8e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "qwencloud/qwen3.5-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 3e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3.7-max": { + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2.5e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 7.5e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwencloud/qwen3.7-plus": { + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 4e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 2.4e-07, + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 4.8e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwencloud/qwen3.8-max": { + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "qwencloud", + "max_input_tokens": 991808, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwencloud/qwq-plus": { + "input_cost_per_token": 8e-07, + "litellm_provider": "qwencloud", + "max_input_tokens": 98304, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://www.qwencloud.com/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwencloud/qwen-image-2.0": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwencloud/qwen-image-2.0-pro": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwencloud/qwen-image-3.0": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwencloud/qwen-image-3.0-pro": { + "litellm_provider": "qwencloud", + "mode": "image_generation", + "source": "https://www.qwencloud.com/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/deepseek-v4-flash": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/deepseek-v4-flash-0731": { + "cache_read_input_token_cost": 4e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/deepseek-v4-pro": { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2.4e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 4.8e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/glm-5.1": { + "cache_read_input_token_cost": 2.6e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 202745, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/glm-5.2": { + "cache_read_input_token_cost": 2.8e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1048576, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/kimi-k2.7-code": { + "cache_read_input_token_cost": 1.9e-07, + "input_cost_per_token": 9.5e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 229376, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen-coder": { + "input_cost_per_token": 3e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.5e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-flash": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-flash-2025-07-28": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 5e-08, + "output_cost_per_token": 4e-07, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 2e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-max": { + "input_cost_per_token": 1.6e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 30720, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 6.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-01-25": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-04-28": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-07-14": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-plus-2025-07-28": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-plus-2025-09-11": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-plus-latest": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 1.2e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_reasoning_token": 1.2e-05, + "output_cost_per_token": 3.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen-turbo": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-turbo-2024-11-01": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-turbo-2025-04-28": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-turbo-latest": { + "input_cost_per_token": 5e-08, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_reasoning_token": 5e-07, + "output_cost_per_token": 2e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-30b-a3b": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 129024, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-coder-flash": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.2e-07, + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-coder-flash-2025-07-28": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 2.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 8e-07, + "output_cost_per_token": 4e-06, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 1.6e-06, + "output_cost_per_token": 9.6e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-coder-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "cache_read_input_token_cost": 1.8e-07, + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 6e-07, + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-coder-plus-2025-07-22": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 997952, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1e-06, + "output_cost_per_token": 5e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 1.8e-06, + "output_cost_per_token": 9e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 256000.0 + ] + }, + { + "input_cost_per_token": 6e-06, + "output_cost_per_token": 6e-05, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-max-preview": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-max": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-max-2026-01-23": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 258048, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "tiered_pricing": [ + { + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 2.4e-06, + "output_cost_per_token": 1.2e-05, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "range": [ + 128000.0, + 252000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3-next-80b-a3b-instruct": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-next-80b-a3b-thinking": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 262144, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3-vl-235b-a22b-instruct": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.6e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-235b-a22b-thinking": { + "input_cost_per_token": 4e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-32b-instruct": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 6.4e-07, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-32b-thinking": { + "input_cost_per_token": 1.6e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.87e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/model-pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwen3-vl-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 260096, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 2e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 3e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 6e-07, + "output_cost_per_token": 4.8e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3.5-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4e-07, + "output_cost_per_token": 2.4e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "input_cost_per_token": 5e-07, + "output_cost_per_token": 3e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3.7-max": { + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2.5e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 7.5e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen3.7-plus": { + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "tiered_pricing": [ + { + "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 4e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 0, + 256000.0 + ] + }, + { + "cache_read_input_token_cost": 2.4e-07, + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 4.8e-06, + "range": [ + 256000.0, + 1000000.0 + ] + } + ] + }, + "qwen_ai_platform/qwen3.8-max": { + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 991808, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "qwen_ai_platform/qwq-plus": { + "input_cost_per_token": 8e-07, + "litellm_provider": "qwen_ai_platform", + "max_input_tokens": 98304, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "qwen_ai_platform/qwen-image-2.0": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/qwen-image-2.0-pro": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/qwen-image-3.0": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, + "qwen_ai_platform/qwen-image-3.0-pro": { + "litellm_provider": "qwen_ai_platform", + "mode": "image_generation", + "source": "https://www.alibabacloud.com/help/en/model-studio/models", + "supported_endpoints": [ + "/v1/images/generations" + ] + }, "databricks/databricks-bge-large-en": { "cache_creation_input_token_cost": 1.0003e-07, "cache_read_input_token_cost": 1.0003e-07, @@ -15101,6 +17254,62 @@ "supports_tool_choice": true, "supports_vision": true }, + "databricks/databricks-deepseek-v4-flash-0731": { + "cache_creation_input_token_cost": 1.4e-07, + "cache_read_input_token_cost": 2.8e-08, + "input_cost_per_token": 1.4e-07, + "input_dbu_cost_per_token": 2e-06, + "litellm_provider": "databricks", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "metadata": { + "notes": "Input/output cost per token is dbu cost * $0.070. Billing reads the per-token dollar fields; the '*_dbu_cost_per_token' fields are the published Databricks rates, kept for reference. Context/max output are the DeepSeek-published model limits (1M context, 384K max output)." + }, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "output_dbu_cost_per_token": 4e-06, + "source": "https://www.databricks.com/product/pricing/foundation-model-serving", + "supported_modalities": [ + "text" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": false + }, + "databricks/databricks-deepseek-v4-pro-0813": { + "cache_creation_input_token_cost": 1.31999e-06, + "cache_read_input_token_cost": 1.3202e-07, + "input_cost_per_token": 1.31999e-06, + "input_dbu_cost_per_token": 1.8857e-05, + "litellm_provider": "databricks", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "metadata": { + "notes": "Input/output cost per token is dbu cost * $0.070. Billing reads the per-token dollar fields; the '*_dbu_cost_per_token' fields are the published Databricks rates, kept for reference. Context/max output are the DeepSeek-published model limits (1M context, 384K max output)." + }, + "mode": "chat", + "output_cost_per_token": 3.95997e-06, + "output_dbu_cost_per_token": 5.6571e-05, + "source": "https://www.databricks.com/product/pricing/foundation-model-serving", + "supported_modalities": [ + "text" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": false + }, "databricks/databricks-gemini-2-5-flash": { "cache_creation_input_token_cost": 3.0002e-07, "cache_read_input_token_cost": 3.0002e-08, @@ -20631,6 +22840,7 @@ "supports_image_size": false }, "gemini-live-2.5-flash-native-audio": { + "deprecation_date": "2026-12-13", "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "vertex_ai-language-models", @@ -23852,8 +26062,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -23867,7 +26079,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second_4k": 0.6, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -23895,8 +26108,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -23910,7 +26125,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://ai.google.dev/gemini-api/docs/video", + "output_cost_per_second_4k": 0.6, + "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_modalities": [ "text" ], @@ -41892,6 +44108,41 @@ "supports_max_reasoning_effort": true, "prompt_cache_min_tokens": 512 }, + "vertex_ai/claude-fable-5-1": { + "regional_endpoint_uplift_multiplier": 1.1, + "supports_mid_conversation_system": true, + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512 + }, "vertex_ai/claude-fable-5@default": { "deprecation_date": "2027-06-08", "regional_endpoint_uplift_multiplier": 1.1, @@ -41927,6 +44178,41 @@ "supports_max_reasoning_effort": true, "prompt_cache_min_tokens": 512 }, + "vertex_ai/claude-fable-5-1@default": { + "regional_endpoint_uplift_multiplier": 1.1, + "supports_mid_conversation_system": true, + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_1hr": 2e-05, + "cache_read_input_token_cost": 2.5e-07, + "input_cost_per_token": 1e-05, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "thinking_always_on": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_forced_tool_use": false, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512 + }, "vertex_ai/claude-opus-5": { "deprecation_date": "2027-01-24", "regional_endpoint_uplift_multiplier": 1.1, @@ -43440,7 +45726,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second_4k": 0.6, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -43453,8 +45740,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -43469,7 +45758,8 @@ "max_tokens": 1024, "mode": "video_generation", "output_cost_per_second": 0.4, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second_4k": 0.6, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -43483,8 +45773,10 @@ "max_input_tokens": 1024, "max_tokens": 1024, "mode": "video_generation", - "output_cost_per_second": 0.15, - "source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate", + "output_cost_per_second": 0.1, + "output_cost_per_second_1080p": 0.12, + "output_cost_per_second_4k": 0.3, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text" ], @@ -52231,14 +54523,14 @@ "supports_vision": true }, "fireworks_ai/deepseek-v4-flash-0731": { - "cache_read_input_token_cost": 2.8e-08, - "input_cost_per_token": 1.4e-07, + "cache_read_input_token_cost": 7e-09, + "input_cost_per_token": 2.2e-07, "litellm_provider": "fireworks_ai", "max_input_tokens": 1048576, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", - "output_cost_per_token": 2.8e-07, + "output_cost_per_token": 6.6e-07, "source": "https://docs.fireworks.ai/serverless/pricing", "supports_function_calling": true, "supports_reasoning": true, @@ -55124,5 +57416,55 @@ "max_tokens": 40960, "mode": "embedding", "source": "https://docs.fireworks.ai/serverless/pricing" + }, + "zai/glm-5.2": { + "cache_creation_input_token_cost": 0, + "cache_read_input_token_cost": 2.6e-07, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "zai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://docs.z.ai/guides/overview/pricing", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "together_ai/Qwen/Qwen3.8-Flash": { + "input_cost_per_token": 1.5e-07, + "litellm_provider": "together_ai", + "max_input_tokens": 1000000, + "max_tokens": 1000000, + "mode": "chat", + "output_cost_per_token": 4.7e-07, + "source": "https://docs.together.ai/docs/serverless-models" + }, + "cerebras/gemma-4-31b": { + "input_cost_per_token": 9.9e-07, + "litellm_provider": "cerebras", + "max_input_tokens": 131072, + "max_output_tokens": 40960, + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 1.49e-06, + "source": "https://api.cerebras.ai/public/v1/models/gemma-4-31b", + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "elevenlabs/scribe_v2": { + "input_cost_per_second": 6.11e-05, + "litellm_provider": "elevenlabs", + "mode": "audio_transcription", + "output_cost_per_second": 0.0, + "source": "https://elevenlabs.io/pricing/api", + "supported_endpoints": [ + "/v1/audio/transcriptions" + ] } } diff --git a/model_prices_and_context_window.schema.json b/model_prices_and_context_window.schema.json index 3f6d3b4f910..9e370e5406a 100644 --- a/model_prices_and_context_window.schema.json +++ b/model_prices_and_context_window.schema.json @@ -662,6 +662,9 @@ "supports_embedding_image_input": { "type": "boolean" }, + "supports_forced_tool_use": { + "type": "boolean" + }, "supports_function_calling": { "type": "boolean" }, diff --git a/provider_endpoints_support.json b/provider_endpoints_support.json index 7c7d508856f..ebc220b3496 100644 --- a/provider_endpoints_support.json +++ b/provider_endpoints_support.json @@ -724,6 +724,42 @@ "interactions": true } }, + "qwencloud": { + "display_name": "QwenCloud (`qwencloud`)", + "url": "https://docs.litellm.ai/docs/providers/qwencloud", + "endpoints": { + "chat_completions": true, + "messages": true, + "responses": true, + "embeddings": true, + "image_generations": true, + "audio_transcriptions": false, + "audio_speech": false, + "moderations": false, + "batches": false, + "rerank": true, + "a2a": true, + "interactions": true + } + }, + "qwen_ai_platform": { + "display_name": "Qwen AI Platform (`qwen_ai_platform`)", + "url": "https://docs.litellm.ai/docs/providers/qwencloud", + "endpoints": { + "chat_completions": true, + "messages": true, + "responses": true, + "embeddings": true, + "image_generations": true, + "audio_transcriptions": false, + "audio_speech": false, + "moderations": false, + "batches": false, + "rerank": true, + "a2a": true, + "interactions": true + } + }, "databricks": { "display_name": "Databricks (`databricks`)", "url": "https://docs.litellm.ai/docs/providers/databricks", diff --git a/pyproject.toml b/pyproject.toml index f33bf68822a..2866e27e84c 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "litellm" -version = "1.100.0" +version = "1.101.0" description = "Library to easily interface with LLM API providers" readme = "README.md" requires-python = ">=3.10, <3.15" @@ -67,8 +67,8 @@ proxy = [ "azure-identity>=1.25.2,<2.0", "azure-storage-blob>=12.28.0,<13.0", "mcp>=1.28.1,<2.0", - "litellm-proxy-extras==0.4.91", - "litellm-enterprise==0.1.62", + "litellm-proxy-extras==0.4.92", + "litellm-enterprise==0.1.63", "RestrictedPython>=8.5,<9.0", "rich>=13.9.4,<14.0", "InquirerPy>=0.3.4,<1.0", @@ -319,7 +319,7 @@ members = ["enterprise", "litellm-proxy-extras"] profile = "black" [tool.commitizen] -version = "1.100.0" +version = "1.101.0" version_files = [ "pyproject.toml:^version", ] diff --git a/ruff-strict-budget.json b/ruff-strict-budget.json index 569c23cd03f..9b1cc977a64 100644 --- a/ruff-strict-budget.json +++ b/ruff-strict-budget.json @@ -1,6 +1,6 @@ { "ANN001": { - "limit": 2991 + "limit": 2985 }, "ANN002": { "limit": 71 @@ -9,13 +9,13 @@ "limit": 809 }, "ANN201": { - "limit": 2002 + "limit": 2001 }, "ANN202": { - "limit": 841 + "limit": 835 }, "ANN204": { - "limit": 694 + "limit": 693 }, "ANN205": { "limit": 112 @@ -24,7 +24,7 @@ "limit": 133 }, "ANN401": { - "limit": 387 + "limit": 307 }, "ASYNC230": { "limit": 11 @@ -231,7 +231,7 @@ "limit": 5 }, "TID251": { - "limit": 1084 + "limit": 1073 }, "TRY002": { "limit": 524 @@ -240,13 +240,13 @@ "limit": 96 }, "TRY201": { - "limit": 405 + "limit": 403 }, "TRY203": { - "limit": 113 + "limit": 111 }, "TRY300": { - "limit": 855 + "limit": 854 }, "UP028": { "limit": 2 diff --git a/ruff-strict.toml b/ruff-strict.toml index 7afc5da71ee..ae092bdde7d 100644 --- a/ruff-strict.toml +++ b/ruff-strict.toml @@ -26,6 +26,10 @@ external = [ # caught a real mismatch, confirming Any is correct here, not a shortcut. "litellm/litellm_core_utils/litellm_logging.py" = ["ANN401"] "litellm/utils.py" = ["ANN401"] +# `**kwargs` forwards verbatim to CustomGuardrail.__init__, whose param list is wide and +# grows over time; typing it concretely (`object`) broke that forwarding call outright — +# basedpyright turned every named param into a reportArgumentType error. Any is correct here. +"litellm/proxy/guardrails/guardrail_hooks/alice/alice.py" = ["ANN401"] [lint.mccabe] max-complexity = 15 diff --git a/test-quality-budget.json b/test-quality-budget.json index ee33eb581d6..d834c581609 100644 --- a/test-quality-budget.json +++ b/test-quality-budget.json @@ -3,7 +3,7 @@ "limit": 733 }, "TQ002": { - "limit": 742 + "limit": 741 }, "TQ003": { "limit": 62 @@ -21,6 +21,6 @@ "limit": 117 }, "TQ008": { - "limit": 11139 + "limit": 11135 } } diff --git a/tests/code_coverage_tests/recursive_detector.py b/tests/code_coverage_tests/recursive_detector.py index 37e940460f6..790956156b0 100644 --- a/tests/code_coverage_tests/recursive_detector.py +++ b/tests/code_coverage_tests/recursive_detector.py @@ -63,6 +63,7 @@ IGNORE_FUNCTIONS = [ "json_unrewritable_labels", # max depth set (MAX_STRUCTURED_CONTENT_SCAN_DEPTH); returns the None sentinel at the cap so the caller blocks. "_flatten_form_field", # bounded by the nesting depth of the already-parsed request body (a finite JSON tree, no cycles possible). "_flatten_form_data_field", # bounded by the nesting depth of the already-parsed request body (a finite JSON tree, no cycles possible). + "_json_safe", # max depth set (_MAX_DEPTH) plus a seen-ids cycle guard for self-referential input. ] diff --git a/tests/e2e/mcp/test_mcp_key_access_e2e.py b/tests/e2e/mcp/test_mcp_key_access_e2e.py index 88ab5666084..68005ae3f6a 100644 --- a/tests/e2e/mcp/test_mcp_key_access_e2e.py +++ b/tests/e2e/mcp/test_mcp_key_access_e2e.py @@ -30,6 +30,34 @@ def _key(client: McpClient, resources: ResourceManager, *, mcp_servers: list[str return key +class TestMcpKeyGrantByAlias: + def test_alias_grant_persists_verbatim_and_lists_tools( + self, + client: McpClient, + resources: ResourceManager, + ) -> None: + """A key granted an MCP server by its alias must store the alias, not the + resolved server_id: in a shared-DB multi-region deployment each instance + derives a different id for the same config server, so only the alias + grants access on every region. The same key must still see the server's + tools, proving the alias grant is honored at request time.""" + server_id = register_datadog_mcp(client, resources) + client.await_registered(server_id) + alias = next(row.alias for row in client.registered_servers() if row.server_id == server_id) + assert alias, f"registered server {server_id} has no alias to grant by" + + key = _key(client, resources, mcp_servers=[alias]) + + stored = client.proxy.key_info(key).object_permission + assert stored is not None and stored.mcp_servers == [alias], ( + f"alias grant was rewritten before persisting (expected [{alias!r}]): " + f"{stored.mcp_servers if stored else None}. A stored server_id is region-local " + f"and breaks the grant on every other instance sharing this database" + ) + + _ = client.await_tool(key, server_id, SEARCH_LOGS_TOOL) + + class TestMcpKeyWithoutAccessIsDenied: @pytest.mark.covers("mcp.list_tools.api_key.denied_without_permission") def test_list_tools_denied_without_permission( diff --git a/tests/e2e/models.py b/tests/e2e/models.py index 6d9ccad9a24..967144dfb16 100644 --- a/tests/e2e/models.py +++ b/tests/e2e/models.py @@ -114,6 +114,7 @@ class KeyInfo(BaseModel): budget_id: str | None = None litellm_budget_table: LiteLLMBudgetTable | None = None budget_limits: list[BudgetWindowState] | None = None + object_permission: ObjectPermission | None = None class KeyInfoResponse(BaseModel): diff --git a/tests/e2e/ui/helpers/premium.ts b/tests/e2e/ui/helpers/premium.ts new file mode 100644 index 00000000000..28bc2e58bbc --- /dev/null +++ b/tests/e2e/ui/helpers/premium.ts @@ -0,0 +1,20 @@ +import * as fs from "fs"; +import { ADMIN_STORAGE_PATH } from "../constants"; + +/** + * Whether the proxy under test is licensed, read from the admin session JWT's `premium_user` + * claim. That is the same value the dashboard reads to enable premium-gated controls, so it + * describes the proxy Playwright is pointed at rather than the environment the runner happens + * to have, which are not the same machine when E2E_UI_BASE_URL points elsewhere. + */ +export function proxyIsPremium(): boolean { + const storage = JSON.parse(fs.readFileSync(ADMIN_STORAGE_PATH, "utf-8")) as { + cookies?: { name: string; value: string }[]; + }; + const token = storage.cookies?.find((cookie) => cookie.name === "token")?.value; + const payload = token?.split(".")[1]; + if (!payload) { + return false; + } + return JSON.parse(Buffer.from(payload, "base64url").toString("utf-8")).premium_user === true; +} diff --git a/tests/e2e/ui/helpers/traffic.ts b/tests/e2e/ui/helpers/traffic.ts index ebd3c9a417f..25eb671fd0e 100644 --- a/tests/e2e/ui/helpers/traffic.ts +++ b/tests/e2e/ui/helpers/traffic.ts @@ -4,12 +4,16 @@ import { APIRequestContext, expect } from "@playwright/test"; export const CHAT_MODEL_A = "fake-openai-gpt-4"; export const CHAT_MODEL_B = "fake-anthropic-claude"; +/** The deployment each of those models routes to, as spend logs and usage breakdowns name it. */ +export const DEPLOYMENT_MODEL_A = "openai/fake-gpt-4"; +export const DEPLOYMENT_MODEL_B = "openai/fake-claude"; + /** The only completion text fixtures/mock_llm_server/server.py ever returns. */ export const MOCK_RESPONSE_TEXT = "This is a mock response."; export const masterKey = (): string => process.env.LITELLM_MASTER_KEY || "sk-1234"; -const rootPath = (): string => process.env.SERVER_ROOT_PATH ?? ""; +export const rootPath = (): string => process.env.SERVER_ROOT_PATH ?? ""; interface ChatOptions { model: string; @@ -114,13 +118,54 @@ export async function waitForSpendLogByPrompt( const isoDay = (d: Date): string => d.toISOString().slice(0, 10); +interface DailyActivityKey { + metrics?: { api_requests?: number }; +} + +interface DailyActivityPage { + results?: { breakdown?: { api_keys?: Record } }[]; + metadata?: { total_pages?: number }; +} + +const requestsOnPage = (body: DailyActivityPage, keyToken: string): number => + (body.results ?? []).reduce((sum, day) => sum + (day.breakdown?.api_keys?.[keyToken]?.metrics?.api_requests ?? 0), 0); + +/** + * The route paginates its per-key breakdown. Reading only the first page finds a key while the + * database is small and stops finding it once a run has generated more keys than one page holds, + * which reads as "the rollup is not running" when the rollup is fine. + */ +async function keyRequestsInDailyActivity( + request: APIRequestContext, + query: string, + keyToken: string, + page = 1, + seen = 0, +): Promise { + const res = await request.get(`${rootPath()}/user/daily/activity?${query}&page=${page}`, { + headers: { Authorization: `Bearer ${masterKey()}` }, + }); + if (!res.ok()) { + return seen; + } + const body = (await res.json()) as DailyActivityPage; + const total = seen + requestsOnPage(body, keyToken); + return page >= (body.metadata?.total_pages ?? 1) + ? total + : keyRequestsInDailyActivity(request, query, keyToken, page + 1, total); +} + /** * The Usage page reads /user/daily/activity, a rollup written by a background job, and fetches it once * on mount. Navigating before the rollup lands leaves a stale render that never refreshes. + * + * The rollup lands request by request, so waiting only for the key to appear leaves a caller that + * sent several requests reading a partial count. Pass `minRequests` to wait for all of them. */ export async function waitForKeyInDailyActivity( request: APIRequestContext, keyToken: string, + minRequests = 1, timeoutMs = 120_000, ): Promise { const now = new Date(); @@ -129,25 +174,17 @@ export async function waitForKeyInDailyActivity( const query = `start_date=${isoDay(start)}&end_date=${isoDay(now)}`; const deadline = Date.now() + timeoutMs; - let lastStatus = 0; - while (Date.now() < deadline) { - const res = await request.get(`${rootPath()}/user/daily/activity?${query}`, { - headers: { Authorization: `Bearer ${masterKey()}` }, - }); - lastStatus = res.status(); - if (res.ok()) { - const body = await res.json(); - const seen = (body?.results ?? []).some( - (day: { breakdown?: { api_keys?: Record } }) => keyToken in (day.breakdown?.api_keys ?? {}), + for (;;) { + const seen = await keyRequestsInDailyActivity(request, query, keyToken); + if (seen >= minRequests) { + return; + } + if (Date.now() >= deadline) { + throw new Error( + `key ${keyToken} reached ${seen} of ${minRequests} requests in /user/daily/activity across every page; ` + + "the daily spend rollup may not be running", ); - if (seen) { - return; - } } await new Promise((r) => setTimeout(r, 3_000)); } - throw new Error( - `key ${keyToken} never appeared in /user/daily/activity (last status ${lastStatus}); ` + - "the daily spend rollup may not be running", - ); } diff --git a/tests/e2e/ui/tests/budgets/budgets.spec.ts b/tests/e2e/ui/tests/budgets/budgets.spec.ts new file mode 100644 index 00000000000..1ad1e488d25 --- /dev/null +++ b/tests/e2e/ui/tests/budgets/budgets.spec.ts @@ -0,0 +1,133 @@ +import { test, expect, type Page as PlaywrightPage } from "@playwright/test"; +import { ADMIN_STORAGE_PATH } from "../../constants"; +import { Page } from "../../fixtures/pages"; +import { navigateToPage, dismissFeedbackPopup } from "../../helpers/navigation"; +import { masterKey } from "../../helpers/traffic"; + +interface StoredBudget { + budget_id: string; + max_budget: number | null; + tpm_limit: number | null; + rpm_limit: number | null; + budget_duration: string | null; +} + +/** A different route from the one the table renders from, so a row that only lives in its cache fails here. */ +async function findBudget(page: PlaywrightPage, budgetId: string): Promise { + const res = await page.request.get("/budget/list", { + headers: { Authorization: `Bearer ${masterKey()}` }, + }); + expect(res.ok(), `GET /budget/list (${res.status()})`).toBe(true); + return ((await res.json()) as StoredBudget[]).find((row) => row.budget_id === budgetId); +} + +async function createBudgetViaApi(page: PlaywrightPage, budget: Partial): Promise { + const res = await page.request.post("/budget/new", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: budget, + }); + expect(res.ok(), `POST /budget/new failed (${res.status()}): ${await res.text()}`).toBe(true); +} + +async function searchForBudget(page: PlaywrightPage, budgetId: string): Promise { + await page.getByPlaceholder("Search by budget ID").fill(budgetId); +} + +test.describe("Budgets", () => { + test.use({ storageState: ADMIN_STORAGE_PATH }); + + test("Create a budget with rate limits and a spend cap", async ({ page }) => { + const budgetId = `e2e-budget-create-${Date.now()}`; + + await navigateToPage(page, Page.Budgets); + await dismissFeedbackPopup(page); + + await page.getByRole("button", { name: "Create Budget" }).click(); + + const modal = page.getByRole("dialog", { name: "Create Budget" }); + await expect(modal).toBeVisible({ timeout: 10_000 }); + + await modal.getByRole("textbox", { name: "Budget ID" }).fill(budgetId); + await modal.getByRole("spinbutton", { name: "Max Tokens per minute" }).fill("5000"); + await modal.getByRole("spinbutton", { name: "Max Requests per minute" }).fill("60"); + + await modal.getByRole("button", { name: "Optional Settings" }).click(); + await modal.getByRole("spinbutton", { name: "Max Budget (USD)" }).fill("25.5"); + await modal.getByRole("combobox", { name: "Reset Budget" }).click(); + await page.getByRole("option", { name: "weekly" }).click(); + + await modal.getByRole("button", { name: "Create Budget" }).click(); + await expect(modal).not.toBeVisible({ timeout: 10_000 }); + + await searchForBudget(page, budgetId); + const row = page.getByRole("row").filter({ hasText: budgetId }); + await expect(row).toBeVisible({ timeout: 10_000 }); + await expect(row).toContainText("$25.50"); + + const stored = await findBudget(page, budgetId); + expect(stored, `budget ${budgetId} readable from /budget/list`).toBeTruthy(); + expect(stored?.max_budget, "spend cap persisted").toBe(25.5); + expect(stored?.tpm_limit, "TPM limit persisted").toBe(5000); + expect(stored?.rpm_limit, "RPM limit persisted").toBe(60); + expect(stored?.budget_duration, "reset window persisted").toBe("7d"); + }); + + test("Raising a budget's spend cap leaves its rate limits alone", async ({ page }) => { + const budgetId = `e2e-budget-edit-${Date.now()}`; + await createBudgetViaApi(page, { budget_id: budgetId, max_budget: 10, tpm_limit: 1000, rpm_limit: 20 }); + + await navigateToPage(page, Page.Budgets); + await dismissFeedbackPopup(page); + + await searchForBudget(page, budgetId); + await expect(page.getByRole("row").filter({ hasText: budgetId })).toBeVisible({ timeout: 10_000 }); + + await page.getByTestId(`budget-actions-${budgetId}`).click(); + await page.getByTestId("budget-action-edit").click(); + + const modal = page.getByRole("dialog", { name: "Edit Budget" }); + await expect(modal).toBeVisible({ timeout: 10_000 }); + + await modal.getByRole("button", { name: "Optional Settings" }).click(); + await modal.getByRole("spinbutton", { name: "Max Budget (USD)" }).fill("99"); + await modal.getByRole("button", { name: "Save", exact: true }).click(); + await expect(modal).not.toBeVisible({ timeout: 10_000 }); + + await expect(page.getByRole("row").filter({ hasText: budgetId })).toContainText("$99.00", { timeout: 10_000 }); + + // Not hypothetical: the edit form posts the whole budget, so a field it fails to + // seed from the existing row goes to the server as null and silently clears. + const stored = await findBudget(page, budgetId); + expect(stored?.max_budget, "spend cap raised").toBe(99); + expect(stored?.tpm_limit, "TPM limit untouched by a spend-cap edit").toBe(1000); + expect(stored?.rpm_limit, "RPM limit untouched by a spend-cap edit").toBe(20); + }); + + test("Delete a budget", async ({ page }) => { + const budgetId = `e2e-budget-delete-${Date.now()}`; + await createBudgetViaApi(page, { budget_id: budgetId, max_budget: 5 }); + + await navigateToPage(page, Page.Budgets); + await dismissFeedbackPopup(page); + + await searchForBudget(page, budgetId); + await expect(page.getByRole("row").filter({ hasText: budgetId })).toBeVisible({ timeout: 10_000 }); + + await page.getByTestId(`budget-actions-${budgetId}`).click(); + await page.getByTestId("budget-action-delete").click(); + + const modal = page.getByRole("dialog", { name: "Delete Budget?" }); + await expect(modal).toBeVisible({ timeout: 5_000 }); + await modal.getByRole("button", { name: "Delete", exact: true }).click(); + + await expect(page.getByRole("row").filter({ hasText: budgetId })).toHaveCount(0, { timeout: 10_000 }); + + // The row disappearing is a cache invalidation; the budget is gone when the route stops serving it. + await expect + .poll(async () => await findBudget(page, budgetId), { + message: `budget ${budgetId} still readable from /budget/list after delete`, + timeout: 15_000, + }) + .toBeUndefined(); + }); +}); diff --git a/tests/e2e/ui/tests/guardrails/guardrails.spec.ts b/tests/e2e/ui/tests/guardrails/guardrails.spec.ts index 77ff020510b..1e43c7a2b22 100644 --- a/tests/e2e/ui/tests/guardrails/guardrails.spec.ts +++ b/tests/e2e/ui/tests/guardrails/guardrails.spec.ts @@ -1,11 +1,213 @@ -import { test, expect } from "@playwright/test"; +import { test, expect, type Page as PlaywrightPage } from "@playwright/test"; import { ADMIN_STORAGE_PATH, E2E_TEAM_NO_ADMIN_ID } from "../../constants"; import { Page } from "../../fixtures/pages"; import { navigateToPage, dismissFeedbackPopup, clickTeamId } from "../../helpers/navigation"; +import { CHAT_MODEL_A, MOCK_RESPONSE_TEXT, masterKey } from "../../helpers/traffic"; + +interface StoredGuardrail { + guardrail_id: string; + guardrail_name: string | null; +} + +async function listGuardrails(page: PlaywrightPage): Promise { + const res = await page.request.get("/v2/guardrails/list", { + headers: { Authorization: `Bearer ${masterKey()}` }, + }); + expect(res.ok(), `GET /v2/guardrails/list (${res.status()})`).toBe(true); + return ((await res.json()) as { guardrails: StoredGuardrail[] }).guardrails; +} + +async function findGuardrail(page: PlaywrightPage, name: string): Promise { + return (await listGuardrails(page)).find((row) => row.guardrail_name === name); +} + +const createdGuardrails: string[] = []; + +async function createKeywordGuardrailViaApi(page: PlaywrightPage, name: string, keyword: string): Promise { + const res = await page.request.post("/guardrails", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { + guardrail: { + guardrail_name: name, + litellm_params: { + guardrail: "litellm_content_filter", + mode: "pre_call", + default_on: false, + blocked_words: [{ keyword, action: "BLOCK" }], + }, + }, + }, + }); + expect(res.ok(), `POST /guardrails failed (${res.status()}): ${await res.text()}`).toBe(true); + createdGuardrails.push(name); + const guardrail = await findGuardrail(page, name); + expect(guardrail?.guardrail_id, `guardrail ${name} has an id`).toBeTruthy(); + return guardrail!.guardrail_id; +} + +async function openKeywordsStep(page: PlaywrightPage, name: string) { + await page.getByRole("button", { name: "Add New Guardrail" }).click(); + await page.getByRole("menuitem", { name: "Add Provider Guardrail" }).click(); + + const wizard = page.getByRole("dialog", { name: "Create guardrail" }); + await expect(wizard).toBeVisible({ timeout: 10_000 }); + + await wizard.getByRole("textbox", { name: "Guardrail Name" }).fill(name); + await wizard.getByRole("combobox", { name: "Guardrail Provider" }).click(); + // The content filter runs inside the proxy, so this is the one provider a test can + // configure end to end without standing up a third-party moderation service. + await page.getByRole("option", { name: /LiteLLM Content Filter/ }).click(); + + for (const step of ["Topics", "Patterns", "Keywords"]) { + await wizard.getByRole("button", { name: "Next" }).click(); + await expect(wizard).toContainText(step, { timeout: 10_000 }); + } + return wizard; +} test.describe("Guardrails", () => { test.use({ storageState: ADMIN_STORAGE_PATH }); + test.afterEach(async ({ page }) => { + // Guardrails live in the database and show up in the table and the playground list, so a run + // that leaves them behind changes what the next run sees. + for (const name of createdGuardrails.splice(0)) { + const guardrail = await findGuardrail(page, name); + if (guardrail) { + const deleted = await page.request.delete(`/guardrails/${guardrail.guardrail_id}`, { + headers: { Authorization: `Bearer ${masterKey()}` }, + }); + expect(deleted.ok(), `DELETE /guardrails/${guardrail.guardrail_id} (${deleted.status()})`).toBe(true); + } + } + }); + + test("A guardrail created through the wizard blocks the keyword it was given", async ({ page }) => { + const stamp = Date.now(); + const guardrailName = `e2e-guardrail-create-${stamp}`; + // Unique per run so a concurrent test's prompt can never trip this guardrail, or vice versa. + const bannedKeyword = `e2ebanned${stamp}`; + + await navigateToPage(page, Page.Guardrails); + await dismissFeedbackPopup(page); + + createdGuardrails.push(guardrailName); + const wizard = await openKeywordsStep(page, guardrailName); + + await wizard.getByRole("button", { name: "Add keyword" }).click(); + const keywordModal = page.getByRole("dialog", { name: "Add blocked keyword" }); + await expect(keywordModal).toBeVisible({ timeout: 10_000 }); + await keywordModal.getByPlaceholder("Enter sensitive keyword or phrase").fill(bannedKeyword); + await keywordModal.getByRole("button", { name: "Add", exact: true }).click(); + await expect(keywordModal).not.toBeVisible({ timeout: 10_000 }); + + await wizard.getByRole("button", { name: "Next" }).click(); + await wizard.getByRole("button", { name: "Create Guardrail" }).click(); + await expect(wizard).not.toBeVisible({ timeout: 15_000 }); + + await expect(page.getByRole("row").filter({ hasText: guardrailName })).toBeVisible({ timeout: 15_000 }); + expect(await findGuardrail(page, guardrailName), "guardrail readable from /v2/guardrails/list").toBeTruthy(); + + // A row in the table only proves the record was written. The point of a guardrail is that it + // refuses traffic, so drive a request through it. + // + // Polled: a guardrail written through /guardrails reaches the request path on the proxy's + // periodic refresh, so the first call after creation can still be served unguarded. The + // assertion is unchanged, it just allows that refresh to land. + let blockedBody = ""; + await expect + .poll( + async () => { + const res = await page.request.post("/v1/chat/completions", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { + model: CHAT_MODEL_A, + messages: [{ role: "user", content: `please tell me about ${bannedKeyword}` }], + guardrails: [guardrailName], + }, + }); + blockedBody = await res.text(); + return res.status(); + }, + { message: "a prompt carrying the banned keyword is refused", timeout: 60_000 }, + ) + .toBe(400); + expect(blockedBody).toContain(bannedKeyword); + + const allowed = await page.request.post("/v1/chat/completions", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { + model: CHAT_MODEL_A, + messages: [{ role: "user", content: "hello there" }], + guardrails: [guardrailName], + }, + }); + expect(allowed.status(), "a clean prompt still gets through the same guardrail").toBe(200); + expect((await allowed.json()).choices?.[0]?.message?.content).toContain(MOCK_RESPONSE_TEXT); + }); + + test("The Test Playground reports the verdict for the text it is given", async ({ page }) => { + const stamp = Date.now(); + const guardrailName = `e2e-guardrail-play-${stamp}`; + const bannedKeyword = `e2eplay${stamp}`; + await createKeywordGuardrailViaApi(page, guardrailName, bannedKeyword); + + await navigateToPage(page, Page.Guardrails); + await dismissFeedbackPopup(page); + + await page.getByRole("tab", { name: "Test Playground" }).click(); + // Every tab on this page stays mounted, so the other tabs' search boxes match too. + const playground = page.getByRole("tabpanel", { name: "Test Playground" }); + await playground.getByPlaceholder("Search guardrails...").fill(guardrailName); + await playground.getByText(guardrailName, { exact: true }).click(); + + const input = playground.getByPlaceholder("Enter text to test with guardrails..."); + await input.fill(`this sentence contains ${bannedKeyword}`); + await playground.getByRole("button", { name: /^Test 1 guardrail$/ }).click(); + + // The playground is where an admin checks a guardrail before rolling it out, so the + // verdict it prints has to be the one the gateway would give. + await expect(playground.getByText(`${guardrailName} - Error`)).toBeVisible({ timeout: 20_000 }); + await expect(playground.getByText(new RegExp(`Content blocked.*${bannedKeyword}`))).toBeVisible({ + timeout: 10_000, + }); + + await input.fill("this sentence is perfectly ordinary"); + await playground.getByRole("button", { name: /^Test 1 guardrail$/ }).click(); + + await expect(playground.getByText(`${guardrailName} - Error`)).toHaveCount(0, { timeout: 20_000 }); + await expect(playground.getByText("this sentence is perfectly ordinary").last()).toBeVisible({ timeout: 10_000 }); + }); + + test("Delete a guardrail", async ({ page }) => { + const stamp = Date.now(); + const guardrailName = `e2e-guardrail-delete-${stamp}`; + const guardrailId = await createKeywordGuardrailViaApi(page, guardrailName, `e2edelete${stamp}`); + + await navigateToPage(page, Page.Guardrails); + await dismissFeedbackPopup(page); + + await expect(page.getByRole("row").filter({ hasText: guardrailName })).toBeVisible({ timeout: 15_000 }); + + await page.getByTestId(`guardrail-actions-${guardrailId}`).click(); + await page.getByTestId("guardrail-action-delete").click(); + + const modal = page.getByRole("dialog"); + await expect(modal).toBeVisible({ timeout: 5_000 }); + await modal.getByRole("button", { name: "Delete", exact: true }).click(); + + await expect(page.getByRole("row").filter({ hasText: guardrailName })).toHaveCount(0, { timeout: 15_000 }); + + // The RC checklist deletes then reloads, because a row vanishing from the table has + // fooled us before; assert against the route the reload would read. + await expect + .poll(async () => await findGuardrail(page, guardrailName), { + message: `guardrail ${guardrailName} still listed after delete`, + timeout: 15_000, + }) + .toBeUndefined(); + }); + test("Create a Presidio guardrail, see it in team settings, and delete it", async ({ page }) => { const guardrailName = `e2e-presidio-${Date.now()}`; diff --git a/tests/e2e/ui/tests/logs/logsFilters.spec.ts b/tests/e2e/ui/tests/logs/logsFilters.spec.ts new file mode 100644 index 00000000000..7174f339296 --- /dev/null +++ b/tests/e2e/ui/tests/logs/logsFilters.spec.ts @@ -0,0 +1,152 @@ +import { test, expect, type APIRequestContext, type Locator, type Page as PlaywrightPage } from "@playwright/test"; +import { ADMIN_STORAGE_PATH } from "../../constants"; +import { navigateToPage, dismissFeedbackPopup } from "../../helpers/navigation"; +import { Page } from "../../fixtures/pages"; +import { + CHAT_MODEL_A, + CHAT_MODEL_B, + createVirtualKey, + sendChatCompletion, + waitForSpendLog, +} from "../../helpers/traffic"; + +/** + * Every test mints its own key and asserts against request ids it generated, so a filter that + * quietly does nothing shows up as the other key's row still being on screen, and concurrent + * specs' traffic cannot decide the outcome. + */ + +const uniqueSuffix = (): string => `${Date.now()}-${Math.random().toString(36).slice(2, 8)}`; + +/** Every tab stays mounted, so the DOM holds four tables at once; scope to the visible one. */ +const requestLogsRows = (page: PlaywrightPage): Locator => + page.locator("table").filter({ visible: true }).first().locator("tbody tr"); + +const visibleTestId = (page: PlaywrightPage, id: string): Locator => page.getByTestId(id).filter({ visible: true }); + +async function openLogs(page: PlaywrightPage): Promise { + await navigateToPage(page, Page.Logs); + await dismissFeedbackPopup(page); + await expect(visibleTestId(page, "datatable-search")).toBeVisible({ timeout: 20_000 }); +} + +async function openFilterDrawer(page: PlaywrightPage): Promise { + await visibleTestId(page, "datatable-filters-trigger").click(); + const drawer = page.getByRole("dialog", { name: "Filters" }); + await expect(drawer).toBeVisible({ timeout: 10_000 }); + return drawer; +} + +/** Picks a value in one of the drawer's searchable comboboxes and applies the filter. */ +async function applyComboboxFilter( + page: PlaywrightPage, + drawer: Locator, + comboboxLabel: string, + value: string, +): Promise { + await drawer.getByRole("combobox", { name: comboboxLabel }).click(); + await page.keyboard.type(value); + await page.getByRole("option", { name: value, exact: true }).first().click(); + await drawer.getByRole("button", { name: "Apply Filters" }).click(); + await expect(drawer).not.toBeVisible({ timeout: 10_000 }); +} + +/** A request the key is not entitled to make, so the proxy refuses it and logs the refusal. */ +async function sendDeniedCompletion(request: APIRequestContext, apiKey: string): Promise { + const res = await request.post("/v1/chat/completions", { + headers: { Authorization: `Bearer ${apiKey}`, "Content-Type": "application/json" }, + data: { model: CHAT_MODEL_B, messages: [{ role: "user", content: "denied" }] }, + }); + expect(res.status(), "a model outside the key's allow-list is refused").toBe(403); +} + +test.describe("Logs page filters", () => { + test.use({ storageState: ADMIN_STORAGE_PATH }); + + test("the Key Alias filter narrows the table to that key's requests", async ({ page, request }) => { + const suffix = uniqueSuffix(); + const mine = await createVirtualKey(request, { key_alias: `e2e-logs-mine-${suffix}` }); + const theirs = await createVirtualKey(request, { key_alias: `e2e-logs-theirs-${suffix}` }); + + const myRequestId = await sendChatCompletion(request, { + model: CHAT_MODEL_A, + prompt: `logs-filter-mine-${suffix}`, + apiKey: mine.key, + }); + const theirRequestId = await sendChatCompletion(request, { + model: CHAT_MODEL_A, + prompt: `logs-filter-theirs-${suffix}`, + apiKey: theirs.key, + }); + await waitForSpendLog(request, myRequestId); + await waitForSpendLog(request, theirRequestId); + + await openLogs(page); + const drawer = await openFilterDrawer(page); + await applyComboboxFilter(page, drawer, "Search a key alias", mine.alias!); + + await expect(requestLogsRows(page).filter({ hasText: myRequestId })).toHaveCount(1, { timeout: 30_000 }); + // The filter is only doing its job if the other key's request is gone, not merely if ours is present. + await expect(requestLogsRows(page).filter({ hasText: theirRequestId })).toHaveCount(0, { timeout: 10_000 }); + }); + + test("the Status filter narrows the table to the refused request", async ({ page, request }) => { + const suffix = uniqueSuffix(); + const alias = `e2e-logs-status-${suffix}`; + const scoped = await createVirtualKey(request, { key_alias: alias, models: [CHAT_MODEL_A] }); + + const servedRequestId = await sendChatCompletion(request, { + model: CHAT_MODEL_A, + prompt: `logs-filter-served-${suffix}`, + apiKey: scoped.key, + }); + await sendDeniedCompletion(request, scoped.key); + await waitForSpendLog(request, servedRequestId); + + await openLogs(page); + const drawer = await openFilterDrawer(page); + await drawer.getByRole("combobox", { name: "Search a key alias" }).click(); + await page.keyboard.type(alias); + await page.getByRole("option", { name: alias, exact: true }).first().click(); + // The Status field labels its group, not the trigger, so it is addressed by the value it shows. + await drawer.getByRole("combobox").filter({ hasText: "All Statuses" }).click(); + await page.getByRole("option", { name: "Failure", exact: true }).click(); + await drawer.getByRole("button", { name: "Apply Filters" }).click(); + await expect(drawer).not.toBeVisible({ timeout: 10_000 }); + + // Both requests were made by this key, so a Status filter that does nothing leaves the served one on screen. + await expect(requestLogsRows(page)).toHaveCount(1, { timeout: 30_000 }); + await expect(requestLogsRows(page)).toContainText("Failure"); + await expect(requestLogsRows(page).filter({ hasText: servedRequestId })).toHaveCount(0); + }); + + test("Reset Filters brings back the rows a filter hid", async ({ page, request }) => { + const suffix = uniqueSuffix(); + const mine = await createVirtualKey(request, { key_alias: `e2e-logs-reset-mine-${suffix}` }); + const theirs = await createVirtualKey(request, { key_alias: `e2e-logs-reset-theirs-${suffix}` }); + + const myRequestId = await sendChatCompletion(request, { + model: CHAT_MODEL_A, + prompt: `logs-reset-mine-${suffix}`, + apiKey: mine.key, + }); + const theirRequestId = await sendChatCompletion(request, { + model: CHAT_MODEL_A, + prompt: `logs-reset-theirs-${suffix}`, + apiKey: theirs.key, + }); + await waitForSpendLog(request, myRequestId); + await waitForSpendLog(request, theirRequestId); + + await openLogs(page); + const drawer = await openFilterDrawer(page); + await applyComboboxFilter(page, drawer, "Search a key alias", mine.alias!); + await expect(requestLogsRows(page).filter({ hasText: theirRequestId })).toHaveCount(0, { timeout: 30_000 }); + + // A filter you cannot clear is a page that looks empty forever, which is how it reads to a user. + await page.getByRole("button", { name: "Reset Filters" }).filter({ visible: true }).click(); + + await expect(requestLogsRows(page).filter({ hasText: theirRequestId })).toHaveCount(1, { timeout: 30_000 }); + await expect(requestLogsRows(page).filter({ hasText: myRequestId })).toHaveCount(1, { timeout: 10_000 }); + }); +}); diff --git a/tests/e2e/ui/tests/modelsPage/addModel.spec.ts b/tests/e2e/ui/tests/modelsPage/addModel.spec.ts index 073d3c0b79c..de25ec1aac5 100644 --- a/tests/e2e/ui/tests/modelsPage/addModel.spec.ts +++ b/tests/e2e/ui/tests/modelsPage/addModel.spec.ts @@ -5,6 +5,11 @@ import { navigateToPage } from "../../helpers/navigation"; import { Page } from "../../fixtures/pages"; import { captureRequestBody, readBack } from "../../helpers/roundTrip"; import { sendChatCompletion } from "../../helpers/traffic"; +import { proxyIsPremium } from "../../helpers/premium"; + +/** Four probes 13s apart span 39s, one PROXY_CONFIG_RELOAD_INTERVAL_SECONDS (30s) plus margin. */ +const CREDENTIAL_PROBE_SUCCESSES = 4; +const CREDENTIAL_PROBE_SPACING_MS = 13_000; /** The mock LLM as the proxy reaches it: same host locally, a sidecar in the deployed stack. */ const MOCK_LLM_BASE = `http://127.0.0.1:${process.env.MOCK_LLM_PORT ?? "8090"}/v1`; @@ -35,7 +40,10 @@ async function selectProvider(page: PlaywrightPage, providerName: string) { const providerDropdown = page.getByRole("combobox", { name: "Provider", exact: true }); await providerDropdown.click(); await providerDropdown.fill(providerName); - await page.getByRole("option").filter({ hasText: exactly(providerName) }).click(); + await page + .getByRole("option") + .filter({ hasText: exactly(providerName) }) + .click(); await expect(providerDropdown).toHaveValue(providerName); } @@ -78,6 +86,9 @@ test.describe("Add Model", () => { }); test("Edit team model TPM and RPM limits", async ({ page }) => { + // /model/new refuses a team-scoped deployment on an unlicensed proxy, so there this fails in + // setup on a product gate rather than on a regression in the edit it covers. + test.skip(!proxyIsPremium(), "proxy under test is unlicensed — team-scoped models are premium"); const masterKey = users[Role.ProxyAdmin].password; const modelName = `e2e-team-model-${Date.now()}`; @@ -226,8 +237,11 @@ test.describe("Add Model", () => { }); expect(createCred.ok(), `POST /credentials failed (${createCred.status()}): ${await createCred.text()}`).toBe(true); - // Multi-instance stacks propagate a new credential to the probe-serving instances on a periodic - // sync; consecutive successes guard against a load balancer alternating synced and stale replicas + // The proxy's periodic credential refresh prunes its in-memory list against a database snapshot + // it took before this credential landed, so a credential that resolves right after POST + // /credentials can stop resolving until the refresh after that. Successes spanning a whole + // PROXY_CONFIG_RELOAD_INTERVAL_SECONDS prove it survived a refresh, after which it stays. + // Resolution fails open onto the ambient key, so losing it reads as a confusing upstream 404. let consecutiveProbeSuccesses = 0; await expect .poll( @@ -249,11 +263,12 @@ test.describe("Add Model", () => { return consecutiveProbeSuccesses; }, { - message: `stored credential ${credentialName} never became usable for a connection test`, - timeout: 60_000, + message: `stored credential ${credentialName} never stayed usable across a config reload`, + intervals: [0, CREDENTIAL_PROBE_SPACING_MS], + timeout: 110_000, }, ) - .toBeGreaterThanOrEqual(3); + .toBeGreaterThanOrEqual(CREDENTIAL_PROBE_SUCCESSES); try { await navigateToPage(page, Page.Models); @@ -458,7 +473,7 @@ test.describe("Add Model", () => { await page.waitForLoadState("networkidle"); await page.getByPlaceholder("Search model names").fill("cohere"); - + // Clearer failure than timing out on a row assertion when the table is empty. await expect(page.getByTestId("pagination-range")).toHaveText(/Showing \d+-\d+ of \d+/, { timeout: 15_000, @@ -466,10 +481,7 @@ test.describe("Add Model", () => { // Pin to one row carrying both the name and the team, so the sibling test's // team-less cohere row can't satisfy it. - const teamCohereRow = page - .getByRole("row") - .filter({ hasText: "cohere/" }) - .filter({ hasText: E2E_TEAM_CRUD_ID }); + const teamCohereRow = page.getByRole("row").filter({ hasText: "cohere/" }).filter({ hasText: E2E_TEAM_CRUD_ID }); await expect(teamCohereRow).toHaveCount(1, { timeout: 15_000 }); } finally { await deleteTeamScopedCohereModels(); diff --git a/tests/e2e/ui/tests/proxy-admin/keys.spec.ts b/tests/e2e/ui/tests/proxy-admin/keys.spec.ts index f5bee68f245..deb7ae70d07 100644 --- a/tests/e2e/ui/tests/proxy-admin/keys.spec.ts +++ b/tests/e2e/ui/tests/proxy-admin/keys.spec.ts @@ -1,15 +1,17 @@ import { test, expect, type Page as PlaywrightPage } from "@playwright/test"; import { ADMIN_STORAGE_PATH, - E2E_DELETE_KEY_ALIAS, E2E_REGENERATE_KEY_ALIAS, E2E_UPDATE_LIMITS_KEY_ALIAS, E2E_INTERNAL_USER_KEY_ALIAS, E2E_TEAM_CRUD_ALIAS, + E2E_TEAM_CRUD_ID, } from "../../constants"; import { Page } from "../../fixtures/pages"; import { navigateToPage, dismissFeedbackPopup } from "../../helpers/navigation"; import { captureRequestBody, readBack } from "../../helpers/roundTrip"; +import { masterKey } from "../../helpers/traffic"; +import { proxyIsPremium } from "../../helpers/premium"; /** * Looks a key up by alias, undefined when none carries it. `return_full_object=true` is what makes @@ -23,6 +25,17 @@ async function findKeyByAlias(page: PlaywrightPage, alias: string): Promise row.key_alias === alias); } +/** A key this test owns, so deleting it costs the suite nothing on a retry or a second run. */ +async function createDeletableKey(page: PlaywrightPage): Promise { + const alias = `e2e-delete-key-${Date.now()}`; + const res = await page.request.post("/key/generate", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { key_alias: alias, team_id: E2E_TEAM_CRUD_ID }, + }); + expect(res.ok(), `POST /key/generate failed (${res.status()}): ${await res.text()}`).toBe(true); + return alias; +} + test.describe("Proxy Admin - Keys", () => { test.use({ storageState: ADMIN_STORAGE_PATH }); @@ -67,6 +80,9 @@ test.describe("Proxy Admin - Keys", () => { }); test("Regenerate key", async ({ page }) => { + // The Regenerate Key button renders disabled when the proxy is unlicensed, so without one this + // fails on a product gate rather than on a regression. + test.skip(!proxyIsPremium(), "proxy under test is unlicensed — Regenerate Key is premium-gated"); await navigateToPage(page, Page.ApiKeys); await dismissFeedbackPopup(page); @@ -143,12 +159,16 @@ test.describe("Proxy Admin - Keys", () => { }); test("Delete key", async ({ page }) => { + // Deleting the seeded key leaves nothing for the next attempt, so the retries CI runs with are + // guaranteed to fail and the suite cannot run twice against one database. Bring our own. + const alias = await createDeletableKey(page); + await navigateToPage(page, Page.ApiKeys); await dismissFeedbackPopup(page); - const keyRow = page.getByRole("row").filter({ hasText: E2E_DELETE_KEY_ALIAS }); + const keyRow = page.getByRole("row").filter({ hasText: alias }); await expect(keyRow).toBeVisible({ timeout: 10_000 }); - await keyRow.getByRole("button", { name: E2E_DELETE_KEY_ALIAS }).click(); + await keyRow.getByRole("button", { name: alias }).click(); await expect(page.getByText("Back to Keys")).toBeVisible({ timeout: 10_000 }); @@ -157,7 +177,7 @@ test.describe("Proxy Admin - Keys", () => { const modal = page.getByRole("dialog", { name: "Delete Key" }); await expect(modal).toBeVisible({ timeout: 5_000 }); - await modal.locator("input").fill(E2E_DELETE_KEY_ALIAS); + await modal.locator("input").fill(alias); const deleteButton = modal.getByRole("button", { name: "Delete", exact: true }); await expect(deleteButton).toBeEnabled(); @@ -167,8 +187,8 @@ test.describe("Proxy Admin - Keys", () => { // The key is gone when the management API stops returning it, not when the toast says so. await expect - .poll(async () => await findKeyByAlias(page, E2E_DELETE_KEY_ALIAS), { - message: `key ${E2E_DELETE_KEY_ALIAS} still readable from /key/list after delete`, + .poll(async () => await findKeyByAlias(page, alias), { + message: `key ${alias} still readable from /key/list after delete`, timeout: 15_000, }) .toBeUndefined(); diff --git a/tests/e2e/ui/tests/proxy-admin/teamSettings.spec.ts b/tests/e2e/ui/tests/proxy-admin/teamSettings.spec.ts new file mode 100644 index 00000000000..e71945a4ccd --- /dev/null +++ b/tests/e2e/ui/tests/proxy-admin/teamSettings.spec.ts @@ -0,0 +1,162 @@ +import { test, expect, type Page as PlaywrightPage } from "@playwright/test"; +import { ADMIN_STORAGE_PATH } from "../../constants"; +import { Page } from "../../fixtures/pages"; +import { navigateToPage, dismissFeedbackPopup, clickTeamId } from "../../helpers/navigation"; +import { readBack } from "../../helpers/roundTrip"; +import { CHAT_MODEL_A, MOCK_RESPONSE_TEXT, masterKey } from "../../helpers/traffic"; + +interface TeamInfo { + team_id: string; + team_alias: string; + models: string[]; + max_budget: number | null; + tpm_limit: number | null; + rpm_limit: number | null; + metadata: Record | null; + members_with_roles: { user_id?: string; role?: string }[]; +} + +/** + * Each test owns a team it created, rather than editing a seeded one, so a save that clobbers a + * field cannot take another spec's fixture down with it. + */ +async function createTeam(page: PlaywrightPage, alias: string, members: string[] = []): Promise { + const res = await page.request.post("/team/new", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { + team_alias: alias, + models: [CHAT_MODEL_A], + members_with_roles: members.map((user_id) => ({ user_id, role: "user" })), + }, + }); + expect(res.ok(), `POST /team/new failed (${res.status()}): ${await res.text()}`).toBe(true); + return (await res.json()).team_id as string; +} + +/** + * A member of this test's own, not one of the seeded users. Putting a seeded user on an extra team + * changes what every spec that asserts on their memberships sees. + */ +async function createMember(page: PlaywrightPage, userId: string): Promise { + const res = await page.request.post("/user/new", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { user_id: userId, user_role: "internal_user", auto_create_key: false }, + }); + expect(res.ok(), `POST /user/new failed (${res.status()}): ${await res.text()}`).toBe(true); + return userId; +} + +async function teamInfo(page: PlaywrightPage, teamId: string): Promise { + const body = await readBack<{ team_info: TeamInfo }>(page, `/team/info?team_id=${encodeURIComponent(teamId)}`); + return body.team_info; +} + +async function openTeamSettings(page: PlaywrightPage, teamId: string): Promise { + await navigateToPage(page, Page.Teams); + await dismissFeedbackPopup(page); + await clickTeamId(page, teamId); + await page.getByRole("tab", { name: "Settings" }).click(); + await page.getByRole("button", { name: "Edit Settings" }).click(); + await expect(page.getByRole("button", { name: "Save Changes" })).toBeVisible({ timeout: 10_000 }); +} + +test.describe("Proxy Admin - Team settings", () => { + test.use({ storageState: ADMIN_STORAGE_PATH }); + + test("Setting a team's spend cap and rate limits leaves its models and members alone", async ({ page }) => { + const stamp = Date.now(); + const alias = `e2e-team-limits-${stamp}`; + const member = await createMember(page, `e2e-team-limits-member-${stamp}`); + const teamId = await createTeam(page, alias, [member]); + const before = await teamInfo(page, teamId); + + await openTeamSettings(page, teamId); + + await page.getByRole("spinbutton", { name: "Max Budget (USD)" }).fill("42.5"); + await page.getByRole("spinbutton", { name: "Tokens per minute Limit (TPM)" }).fill("7000"); + await page.getByRole("spinbutton", { name: "Requests per minute Limit (RPM)" }).fill("70"); + await page.getByRole("button", { name: "Save Changes" }).click(); + + await expect + .poll( + async () => { + const team = await teamInfo(page, teamId); + return [team.max_budget, team.tpm_limit, team.rpm_limit]; + }, + { message: "team limits did not persist", timeout: 20_000 }, + ) + .toEqual([42.5, 7000, 70]); + + // The Settings form posts the whole team. A field it fails to seed goes back as null, and + // the toast still says success, so pin the fields this edit had no business touching. + const after = await teamInfo(page, teamId); + expect(after.models, "model access untouched by a limits edit").toEqual(before.models); + expect( + after.members_with_roles.map((member) => member.user_id).sort(), + "membership untouched by a limits edit", + ).toEqual(before.members_with_roles.map((member) => member.user_id).sort()); + }); + + test("A model alias added on the Settings tab serves traffic under the alias name", async ({ page }) => { + const stamp = Date.now(); + const alias = `e2e-team-alias-${stamp}`; + const modelAlias = `e2e-alias-${stamp}`; + const teamId = await createTeam(page, alias); + + await openTeamSettings(page, teamId); + + await page.getByRole("textbox", { name: "Alias Name" }).fill(modelAlias); + await page.getByRole("combobox", { name: "Select target model" }).click(); + await page.getByRole("option", { name: CHAT_MODEL_A, exact: true }).first().click(); + await page.getByRole("button", { name: "Add Alias" }).click(); + await page.getByRole("button", { name: "Save Changes" }).click(); + + await expect + .poll(async () => (await teamInfo(page, teamId)).models, { message: "team lost its models", timeout: 20_000 }) + .toEqual([CHAT_MODEL_A]); + + const keyRes = await page.request.post("/key/generate", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { team_id: teamId, key_alias: `e2e-alias-key-${stamp}` }, + }); + expect(keyRes.ok(), `POST /key/generate failed (${keyRes.status()})`).toBe(true); + const teamKey = (await keyRes.json()).key as string; + + // An alias the team can see but cannot call is the actual complaint; the readback alone + // would pass for an alias the router never resolves. + const served = await page.request.post("/v1/chat/completions", { + headers: { Authorization: `Bearer ${teamKey}`, "Content-Type": "application/json" }, + data: { model: modelAlias, messages: [{ role: "user", content: "ping" }] }, + }); + expect(served.status(), `a team key calling ${modelAlias} is served`).toBe(200); + expect((await served.json()).choices?.[0]?.message?.content).toContain(MOCK_RESPONSE_TEXT); + }); + + test("Team metadata added as key-value pairs survives a reload", async ({ page }) => { + const stamp = Date.now(); + const alias = `e2e-team-metadata-${stamp}`; + const metadataValue = `cost-center-${stamp}`; + const teamId = await createTeam(page, alias); + + await openTeamSettings(page, teamId); + + await page.getByRole("button", { name: "Add Key-Value Pair" }).click(); + await page.getByPlaceholder("Key", { exact: true }).last().fill("owner"); + await page.getByPlaceholder("Value", { exact: true }).last().fill(metadataValue); + await page.getByRole("button", { name: "Save Changes" }).click(); + + await expect + .poll(async () => (await teamInfo(page, teamId)).metadata?.owner, { + message: "team metadata did not persist", + timeout: 20_000, + }) + .toBe(metadataValue); + + // Reopening the form is the step that catches metadata the page writes but cannot read back. + await page.reload(); + await page.getByRole("tab", { name: "Settings" }).click(); + await page.getByRole("button", { name: "Edit Settings" }).click(); + await expect(page.getByPlaceholder("Key", { exact: true })).toHaveValue("owner", { timeout: 15_000 }); + await expect(page.getByPlaceholder("Value", { exact: true })).toHaveValue(metadataValue); + }); +}); diff --git a/tests/e2e/ui/tests/proxy-admin/teams.spec.ts b/tests/e2e/ui/tests/proxy-admin/teams.spec.ts index 7383b452162..303e4488e09 100644 --- a/tests/e2e/ui/tests/proxy-admin/teams.spec.ts +++ b/tests/e2e/ui/tests/proxy-admin/teams.spec.ts @@ -1,14 +1,9 @@ import { test, expect, type Page as PlaywrightPage } from "@playwright/test"; -import { - ADMIN_STORAGE_PATH, - E2E_TEAM_CRUD_ID, - E2E_TEAM_DELETE_ALIAS, - E2E_TEAM_NO_ADMIN_ID, - E2E_TEAM_ORG_ID, -} from "../../constants"; +import { ADMIN_STORAGE_PATH, E2E_TEAM_CRUD_ID, E2E_TEAM_NO_ADMIN_ID, E2E_TEAM_ORG_ID } from "../../constants"; import { Page } from "../../fixtures/pages"; import { navigateToPage, dismissFeedbackPopup, clickTeamId } from "../../helpers/navigation"; import { readBack } from "../../helpers/roundTrip"; +import { masterKey } from "../../helpers/traffic"; /** GET /team/list returns a bare array of teams, each carrying team_alias/team_id. */ async function findTeamByAlias(page: PlaywrightPage, alias: string): Promise | undefined> { @@ -25,6 +20,17 @@ async function teamMemberEmails(page: PlaywrightPage, teamId: string): Promise member.user_email ?? "").filter(Boolean); } +/** A team this test owns, so deleting it costs the suite nothing on a retry or a second run. */ +async function createDeletableTeam(page: PlaywrightPage): Promise { + const alias = `e2e-delete-team-${Date.now()}`; + const res = await page.request.post("/team/new", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { team_alias: alias, models: ["fake-openai-gpt-4"] }, + }); + expect(res.ok(), `POST /team/new failed (${res.status()}): ${await res.text()}`).toBe(true); + return alias; +} + test.describe("Proxy Admin - Teams", () => { test.use({ storageState: ADMIN_STORAGE_PATH }); @@ -121,10 +127,14 @@ test.describe("Proxy Admin - Teams", () => { }); test("Delete a team", async ({ page }) => { + // Deleting the seeded team leaves nothing for the next attempt, so the retries CI runs with are + // guaranteed to fail and the suite cannot run twice against one database. Bring our own. + const alias = await createDeletableTeam(page); + await navigateToPage(page, Page.Teams); await dismissFeedbackPopup(page); - const teamRow = page.locator("tr", { hasText: E2E_TEAM_DELETE_ALIAS }).first(); + const teamRow = page.locator("tr", { hasText: alias }).first(); await expect(teamRow).toBeVisible({ timeout: 10_000 }); // Actions live in a kebab menu: open it, then click "Delete team". await teamRow.locator('[data-testid^="team-actions-"]').click(); @@ -132,15 +142,15 @@ test.describe("Proxy Admin - Teams", () => { const modal = page.getByRole("dialog", { name: "Delete Team?" }); await expect(modal).toBeVisible({ timeout: 5_000 }); - await modal.locator("input").fill(E2E_TEAM_DELETE_ALIAS); + await modal.locator("input").fill(alias); await modal.getByRole("button", { name: /Force Delete|Delete/i }).click(); await expect(teamRow).not.toBeVisible({ timeout: 10_000 }); // A row vanishing is local state, which happens whether or not the delete landed. await expect - .poll(async () => await findTeamByAlias(page, E2E_TEAM_DELETE_ALIAS), { - message: `team ${E2E_TEAM_DELETE_ALIAS} still readable from /team/list after delete`, + .poll(async () => await findTeamByAlias(page, alias), { + message: `team ${alias} still readable from /team/list after delete`, timeout: 15_000, }) .toBeUndefined(); diff --git a/tests/e2e/ui/tests/team-admin/teamAdmin.spec.ts b/tests/e2e/ui/tests/team-admin/teamAdmin.spec.ts index 26a6fa50b4b..f3c031f0172 100644 --- a/tests/e2e/ui/tests/team-admin/teamAdmin.spec.ts +++ b/tests/e2e/ui/tests/team-admin/teamAdmin.spec.ts @@ -35,7 +35,44 @@ async function findKeyByAlias(page: PlaywrightPage, alias: string): Promise row.key_alias === alias); } +/** A member this test adds itself, so removing it costs the suite nothing on a retry or a re-run. */ +async function addRemovableMember(page: PlaywrightPage, registerForCleanup: string[]): Promise { + const userId = `e2e-removable-${Date.now()}`; + // Claimed before the call: /user/new can persist the user and still answer non-2xx, and the id is + // ours either way, so registering it up front is what no failure path can skip. + registerForCleanup.push(userId); + const created = await page.request.post("/user/new", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { user_id: userId, user_role: "internal_user", auto_create_key: false }, + }); + expect(created.ok(), `POST /user/new failed (${created.status()}): ${await created.text()}`).toBe(true); + + const added = await page.request.post("/team/member_add", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { team_id: E2E_TEAM_CRUD_ID, member: { user_id: userId, role: "user" } }, + }); + expect(added.ok(), `POST /team/member_add failed (${added.status()}): ${await added.text()}`).toBe(true); + return userId; +} + test.describe("Team Admin", () => { + const createdMembers: string[] = []; + + test.afterEach(async ({ page }) => { + // Runs on the failure path too, which a call at the end of the test body would not. Ids are + // claimed before the user is created, so the delete is attempted unconditionally and only its + // own 404 counts as never persisted; any other answer is a cleanup failure worth reporting + // rather than a reason to leave the user behind. + for (const userId of createdMembers.splice(0)) { + const deleted = await page.request.post("/user/delete", { + headers: { Authorization: `Bearer ${masterKey()}` }, + data: { user_ids: [userId] }, + }); + const settled = deleted.ok() || deleted.status() === 404; + expect(settled, `POST /user/delete for ${userId} (${deleted.status()}): ${await deleted.text()}`).toBe(true); + } + }); + test.use({ storageState: TEAM_ADMIN_STORAGE_PATH }); test("Team admin can see all team keys including internal user keys", async ({ page }) => { @@ -95,6 +132,10 @@ test.describe("Team Admin", () => { }); test("Team admin can remove a member from their team", async ({ page }) => { + // Removing the seeded member leaves nothing for the next attempt, so the retries CI runs with + // are guaranteed to fail and the suite cannot run twice against one database. Bring our own. + const memberId = await addRemovableMember(page, createdMembers); + await navigateToPage(page, Page.Teams); await dismissFeedbackPopup(page); @@ -102,9 +143,9 @@ test.describe("Team Admin", () => { await page.getByRole("tab", { name: "Members" }).click(); - // Seeded members appear in the roster by user_id (members_with_roles has no - // email), so match the row on the user_id rather than the email. - const row = page.locator("tr", { hasText: "e2e-removable-member" }).first(); + // Members appear in the roster by user_id (members_with_roles has no email), so match + // the row on the user_id rather than the email. + const row = page.locator("tr", { hasText: memberId }).first(); await expect(row).toBeVisible({ timeout: 10_000 }); await row.getByTestId("delete-member").click(); @@ -117,7 +158,7 @@ test.describe("Team Admin", () => { // Removing the wrong member is exactly what a success toast hides, so pin both halves. expect(remove.team_id, "delete targets the team being viewed").toBe(E2E_TEAM_CRUD_ID); expect([remove.user_id, remove.user_email], "delete identifies the member whose row was clicked").toContain( - "e2e-removable-member", + memberId, ); await expect(page.getByText("Team member removed successfully").first()).toBeVisible({ timeout: 10_000 }); @@ -128,7 +169,7 @@ test.describe("Team Admin", () => { message: "removed member is still on the team", timeout: 15_000, }) - .not.toContain("e2e-removable-member"); + .not.toContain(memberId); }); test("Team admin sees all team models in the Playground model dropdown", async ({ page, request }) => { diff --git a/tests/e2e/ui/tests/usage/usageActivityTabs.spec.ts b/tests/e2e/ui/tests/usage/usageActivityTabs.spec.ts new file mode 100644 index 00000000000..2ee5ae3e392 --- /dev/null +++ b/tests/e2e/ui/tests/usage/usageActivityTabs.spec.ts @@ -0,0 +1,135 @@ +import { test, expect, type Locator, type Page as PlaywrightPage } from "@playwright/test"; +import { ADMIN_STORAGE_PATH } from "../../constants"; +import { navigateToPage, dismissFeedbackPopup } from "../../helpers/navigation"; +import { Page } from "../../fixtures/pages"; +import { + CHAT_MODEL_A, + CHAT_MODEL_B, + DEPLOYMENT_MODEL_A, + DEPLOYMENT_MODEL_B, + createVirtualKey, + masterKey, + rootPath, + sendChatCompletion, + waitForKeyInDailyActivity, + waitForSpendLog, +} from "../../helpers/traffic"; + +/** + * Covers the per-entity breakdowns on /ui/usage. The page-level totals move with every other spec's + * traffic, so each assertion is scoped to a key this test minted and to the requests it sent. + */ + +/** Each breakdown renders one expandable card per entity, named " $x.xx N requests". */ +const entityCard = (page: PlaywrightPage, tab: string, name: string): Locator => + page.getByRole("tabpanel", { name: tab }).getByRole("button", { name: new RegExp(`^${name}\\s`) }); + +async function openUsageTab(page: PlaywrightPage, tab: string): Promise { + await navigateToPage(page, Page.NewUsage); + await dismissFeedbackPopup(page); + await page.getByRole("tab", { name: tab }).click(); + const panel = page.getByRole("tabpanel", { name: tab }); + await expect(panel).toBeVisible({ timeout: 30_000 }); + return panel; +} + +/** Sends `count` completions on one model and waits for each to reach the spend log. */ +async function sendTraffic( + request: Parameters[0], + apiKey: string, + model: string, + count: number, + label: string, +): Promise { + for (let i = 0; i < count; i++) { + const requestId = await sendChatCompletion(request, { model, prompt: `${label} ${i}`, apiKey }); + await waitForSpendLog(request, requestId); + } +} + +test.describe("Usage page activity tabs", () => { + test.use({ storageState: ADMIN_STORAGE_PATH }); + + test("Key Activity breaks a key's traffic down by model", async ({ page, request }) => { + const alias = `e2e-usage-keyact-${Date.now()}`; + const { key, token } = await createVirtualKey(request, { key_alias: alias }); + + // An uneven split, so a breakdown that lumps everything into one row or attributes to the + // wrong model cannot land on these numbers by accident. + await sendTraffic(request, key, CHAT_MODEL_A, 2, alias); + await sendTraffic(request, key, CHAT_MODEL_B, 1, alias); + await waitForKeyInDailyActivity(request, token, 3); + + await openUsageTab(page, "Key Activity"); + + const card = entityCard(page, "Key Activity", alias); + await expect(card, `${alias} missing from Key Activity`).toBeVisible({ timeout: 30_000 }); + await expect(card).toContainText("3 requests"); + + // Every key gets a card, and the page opens the first one. Scope to this key's own section, + // which the collapsible renders as the trigger's next sibling. + await card.click(); + const details = card.locator("xpath=following-sibling::*[1]"); + const successfulFor = (model: string) => + details.getByRole("row").filter({ hasText: model }).getByRole("cell").nth(2); // Model | Spend | Successful | Failed | Tokens + + await expect(successfulFor(DEPLOYMENT_MODEL_A)).toHaveText("2", { timeout: 20_000 }); + await expect(successfulFor(DEPLOYMENT_MODEL_B)).toHaveText("1"); + }); + + test("Model Activity can name its models by deployment instead of by public name", async ({ page, request }) => { + const alias = `e2e-usage-modelact-${Date.now()}`; + const { key, token } = await createVirtualKey(request, { key_alias: alias }); + await sendTraffic(request, key, CHAT_MODEL_A, 1, alias); + await waitForKeyInDailyActivity(request, token); + + const panel = await openUsageTab(page, "Model Activity"); + + await expect(entityCard(page, "Model Activity", CHAT_MODEL_A), `${CHAT_MODEL_A} missing`).toBeVisible({ + timeout: 30_000, + }); + // Nothing is published under the deployment's name, so its absence here is what makes the + // toggle below a real change of key rather than a relabelled button. + await expect(entityCard(page, "Model Activity", DEPLOYMENT_MODEL_A)).toHaveCount(0); + + // Admins reconcile provider bills against the deployment, not the name their users call. + await panel.getByRole("button", { name: "Litellm Model Name" }).click(); + await expect(entityCard(page, "Model Activity", DEPLOYMENT_MODEL_A)).toBeVisible({ timeout: 20_000 }); + }); + + test("Filter by user narrows Key Activity to that user's keys", async ({ page, request }) => { + const stamp = Date.now(); + const email = `e2e-usage-owner-${stamp}@test.local`; + const ownedAlias = `e2e-usage-owned-${stamp}`; + const otherAlias = `e2e-usage-other-${stamp}`; + + const userRes = await request.post(`${rootPath()}/user/new`, { + headers: { Authorization: `Bearer ${masterKey()}`, "Content-Type": "application/json" }, + data: { user_email: email, user_role: "internal_user", auto_create_key: false }, + }); + expect(userRes.ok(), `POST /user/new failed (${userRes.status()})`).toBe(true); + const userId = (await userRes.json()).user_id as string; + + const owned = await createVirtualKey(request, { key_alias: ownedAlias, user_id: userId }); + const other = await createVirtualKey(request, { key_alias: otherAlias }); + await sendTraffic(request, owned.key, CHAT_MODEL_A, 1, ownedAlias); + await sendTraffic(request, other.key, CHAT_MODEL_A, 1, otherAlias); + await waitForKeyInDailyActivity(request, owned.token); + await waitForKeyInDailyActivity(request, other.token); + + await openUsageTab(page, "Key Activity"); + await expect(entityCard(page, "Key Activity", otherAlias)).toBeVisible({ timeout: 30_000 }); + + await page.getByRole("combobox", { name: "Search users by email" }).click(); + await page.keyboard.type(email); + await page + .getByRole("option", { name: new RegExp(email) }) + .first() + .click(); + + // The filter earns its place only by dropping the other key; the owned key showing up + // proves nothing on a page that already listed every key. + await expect(entityCard(page, "Key Activity", otherAlias)).toHaveCount(0, { timeout: 30_000 }); + await expect(entityCard(page, "Key Activity", ownedAlias)).toBeVisible({ timeout: 20_000 }); + }); +}); diff --git a/tests/e2e/ui/tests/usage/usagePage.spec.ts b/tests/e2e/ui/tests/usage/usagePage.spec.ts index f61ab018e1b..3d057cfa2c9 100644 --- a/tests/e2e/ui/tests/usage/usagePage.spec.ts +++ b/tests/e2e/ui/tests/usage/usagePage.spec.ts @@ -1,10 +1,11 @@ -import { test, expect, type Locator, type Page as PlaywrightPage } from "@playwright/test"; +import { test, expect, type APIRequestContext, type Locator, type Page as PlaywrightPage } from "@playwright/test"; import { ADMIN_STORAGE_PATH } from "../../constants"; import { navigateToPage, dismissFeedbackPopup } from "../../helpers/navigation"; import { Page } from "../../fixtures/pages"; import { - CHAT_MODEL_A, createVirtualKey, + masterKey, + rootPath, sendChatCompletion, waitForKeyInDailyActivity, waitForSpendLog, @@ -27,9 +28,105 @@ async function openUsage(page: PlaywrightPage): Promise { return card; } +/** The upstream fixtures/config.yml points its models at, so the mock server answers this too. */ +const MOCK_DEPLOYMENT = "openai/fake-gpt-4"; + +/** A deployment whose traffic costs real money, so the key that used it outranks the $0 crowd. */ +async function createPricedDeployment( + request: APIRequestContext, + label: string, + registerForCleanup: string[], +): Promise<{ modelName: string }> { + const modelName = `e2e-usage-priced-${label}`; + // Claimed before the call: /model/new can persist the deployment and still answer non-2xx, so a + // name recorded up front is the only registration no response shape can skip. + registerForCleanup.push(modelName); + const res = await request.post("/model/new", { + headers: { Authorization: `Bearer ${masterKey()}`, "Content-Type": "application/json" }, + data: { + model_name: modelName, + litellm_params: { + model: MOCK_DEPLOYMENT, + api_base: `http://127.0.0.1:${process.env.MOCK_LLM_PORT ?? "8090"}/v1`, + api_key: "fake-key", + input_cost_per_token: 0.01, + output_cost_per_token: 0.01, + }, + }, + }); + expect(res.ok(), `POST /model/new failed (${res.status()}): ${await res.text()}`).toBe(true); + + // /model/new returns once the row is written, but the router only picks the deployment up on its + // next refresh, so sending traffic straight away can still get "no healthy deployments". A ping + // that fails writes no spend log, so retrying it costs the ranking this test asserts nothing. + await expect + .poll( + async () => { + const ping = await request.post(`${rootPath()}/v1/chat/completions`, { + headers: { Authorization: `Bearer ${masterKey()}`, "Content-Type": "application/json" }, + data: { model: modelName, messages: [{ role: "user", content: "readiness ping" }] }, + }); + return ping.ok(); + }, + { message: `deployment ${modelName} never became routable`, timeout: 60_000 }, + ) + .toBe(true); + + return { modelName }; +} + test.describe("Usage page", () => { test.use({ storageState: ADMIN_STORAGE_PATH }); + const pricedDeployments: string[] = []; + + test.afterEach(async ({ request }) => { + // A deployment left behind keeps its custom pricing, so it goes on changing what later runs + // route and what they cost. Runs on the failure path too, which the test body would not. + // Resolved by name rather than by a returned id, so a create that persisted without answering + // 2xx is still cleaned up. /model/info serves the router, and /model/new answers 2xx even when + // its in-request router reload failed, so the search-backed listing is what covers a deployment + // that reached the database only. Absent from both means it never persisted. + const names = pricedDeployments.splice(0); + if (names.length === 0) return; + const auth = { Authorization: `Bearer ${masterKey()}`, "Content-Type": "application/json" }; + + type Lookup = + | { readonly listed: true; readonly id: string | undefined } + | { readonly listed: false; readonly status: number }; + + const idIn = async (path: string, name: string): Promise => { + const listed = await request.get(path, { headers: auth }); + if (!listed.ok()) return { listed: false, status: listed.status() }; + const deployments = ((await listed.json()).data ?? []) as { + model_name?: string; + model_info?: { id?: string }; + }[]; + return { listed: true, id: deployments.find((d) => d.model_name === name)?.model_info?.id }; + }; + + const remove = async (name: string, id: string) => { + const deleted = await request.post(`${rootPath()}/model/delete`, { headers: auth, data: { id } }); + expect(deleted.ok(), `POST /model/delete for ${name} (${deleted.status()})`).toBe(true); + }; + + for (const name of names) { + const fromRouter = await idIn(`${rootPath()}/model/info`, name); + if (fromRouter.listed && fromRouter.id !== undefined) { + await remove(name, fromRouter.id); + continue; + } + const search = encodeURIComponent(name); + const fromDb = await idIn(`${rootPath()}/v2/model/info?search=${search}`, name); + expect( + fromDb.listed, + `GET /v2/model/info?search=${search} (${fromDb.listed ? 200 : fromDb.status}), so ${name} could not be checked`, + ).toBe(true); + if (!fromDb.listed || fromDb.id === undefined) continue; + await remove(name, fromDb.id); + } + }); + test("Top Virtual Keys lists a key that served traffic, toggles views, and opens key info", async ({ page, request, @@ -39,8 +136,13 @@ test.describe("Usage page", () => { key_alias: alias, }); + // Top Virtual Keys ranks by spend, and every mock deployment costs $0, so once a run has more + // keys than the list shows, whether this one makes the cut is down to how ties happen to sort. + // Give it a priced deployment of its own so it earns its place. + const { modelName } = await createPricedDeployment(request, alias, pricedDeployments); + const requestId = await sendChatCompletion(request, { - model: CHAT_MODEL_A, + model: modelName, prompt: `usage ping for ${alias}`, apiKey: key, }); diff --git a/tests/litellm-proxy-extras/test_setup_database_fail_fast.py b/tests/litellm-proxy-extras/test_setup_database_fail_fast.py new file mode 100644 index 00000000000..ef447315a8c --- /dev/null +++ b/tests/litellm-proxy-extras/test_setup_database_fail_fast.py @@ -0,0 +1,571 @@ +"""Regression tests for ProxyExtrasDBManager's v2 migration resolver. + +v2 is the proxy CLI default; v1 stays reachable via the `use_v2_resolver` +kwarg, which still defaults to False for direct callers. +""" + +import subprocess +from unittest.mock import patch + +import pytest + +from litellm_proxy_extras.utils import ( + _PRISMA_ATTEMPTS, + ProxyExtrasDBManager, + _max_migration_timestamp, + _migration_timestamp, +) + + +def _fake_migrate_deploy_failure(returncode: int, stderr: str): + def _run(*args, **kwargs): + raise subprocess.CalledProcessError( + returncode=returncode, + cmd=args[0], + stderr=stderr, + output="", + ) + + return _run + + +def test_v2_p3018_permission_error_raises_runtime_error(monkeypatch, tmp_path): + """v2: a permission failure during migrate deploy raises RuntimeError.""" + monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") + monkeypatch.setattr( + ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None + ) + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + + stderr = ( + "Error: P3018\nMigration name: 20250326162113_baseline\n" + "Database error code: 42501\npermission denied for schema public" + ) + with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): + with pytest.raises(RuntimeError, match="permission"): + ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + +def test_v2_non_idempotent_p3009_raises_runtime_error(monkeypatch, tmp_path): + """v2: a non-idempotent migration failure raises (no silent recovery).""" + monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") + monkeypatch.setattr( + ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None + ) + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + + stderr = ( + "Error: P3009\nMigration `20260101000000_genuinely_broken` failed\n" + 'Reason: syntax error at or near "BRKN" LINE 42' + ) + with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): + with pytest.raises(RuntimeError, match="cannot be auto-recovered"): + ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + +def test_strip_prisma_query_params_removes_connection_limit(): + """DATABASE_URLs with Prisma-specific params should be parseable by psycopg.""" + url = "postgresql://u:p@h:5432/db?connection_limit=100&pool_timeout=60&sslmode=require" + stripped = ProxyExtrasDBManager._strip_prisma_query_params(url) + assert "connection_limit" not in stripped + assert "pool_timeout" not in stripped + assert "sslmode=require" in stripped + + +def test_strip_prisma_query_params_passthrough_no_query(): + """URLs without query strings are returned unchanged.""" + url = "postgresql://u:p@h:5432/db" + assert ProxyExtrasDBManager._strip_prisma_query_params(url) == url + + +def test_migration_timestamp_extracts_leading_digits(): + assert _migration_timestamp("20260101000000_add_foo") == 20260101000000 + assert _migration_timestamp("20250326162113_baseline") == 20250326162113 + + +def test_migration_timestamp_returns_zero_on_malformed(): + assert _migration_timestamp("0_init") == 0 + assert _migration_timestamp("not_a_migration") == 0 + + +def test_max_migration_timestamp(): + names = {"20250326000000_a", "20260415000000_b", "20251115000000_c"} + assert _max_migration_timestamp(names) == 20260415000000 + + +def test_max_migration_timestamp_empty_set(): + assert _max_migration_timestamp(set()) == 0 + + +def test_v1_default_still_calls_resolve_all_migrations(monkeypatch, tmp_path): + """v1 (default) continues to call _resolve_all_migrations on the happy path. + + This is the existing buggy behavior — we're not fixing it in v1, only + offering v2 as opt-in. This test pins the default so that a future + inadvertent default flip is caught. + """ + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + + # Stub `prisma migrate deploy` to claim success with pending migrations + # applied, which is the code path that triggers the legacy post-migration + # sanity check (a call to _resolve_all_migrations). + class FakeResult: + stdout = "Applied migration.\n" + stderr = "" + + def fake_run(cmd, *args, **kwargs): + return FakeResult() + + resolve_called = {"n": 0} + + def fake_resolve(*args, **kwargs): + resolve_called["n"] += 1 + + monkeypatch.setattr("subprocess.run", fake_run) + monkeypatch.setattr(ProxyExtrasDBManager, "_resolve_all_migrations", fake_resolve) + + ok = ProxyExtrasDBManager.setup_database(use_migrate=True) # v2 flag NOT set + assert ok is True + assert resolve_called["n"] == 1, "v1 default should still invoke the legacy path" + + +def test_v2_db_push_wraps_subprocess_error_as_runtime_error(monkeypatch, tmp_path): + """v2: a failing `prisma db push` must raise RuntimeError, not leak + CalledProcessError past proxy_cli.py's `except RuntimeError`.""" + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + + stderr = "db push error" + with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): + with pytest.raises(RuntimeError, match="prisma db push failed"): + ProxyExtrasDBManager.setup_database(use_migrate=False, use_v2_resolver=True) + + +def test_v2_warn_ahead_of_head_swallows_db_errors(monkeypatch, tmp_path): + """_warn_if_db_ahead_of_head must never raise — it's informational. + + Non-connection DB errors (e.g. InsufficientPrivilege from a user + without SELECT on _prisma_migrations) must be caught, not propagated. + """ + import psycopg + + monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + + class _FakeConn: + def __enter__(self): + return self + + def __exit__(self, *a): + return False + + def execute(self, *a, **kw): + # Simulate an InsufficientPrivilege (subclass of DatabaseError). + raise psycopg.errors.InsufficientPrivilege("permission denied") + + connects = {"n": 0} + + def _fake_connect(*a, **kw): + connects["n"] += 1 + return _FakeConn() + + monkeypatch.setattr("psycopg.connect", _fake_connect) + + assert ProxyExtrasDBManager._warn_if_db_ahead_of_head(str(tmp_path)) is None + assert connects["n"] == 1, "the failing query must actually have been reached" + + +def test_v2_resolve_specific_migration_failure_raises_runtime_error( + monkeypatch, tmp_path +): + """If marking a migration as applied fails inside P3009 idempotent + recovery, the subprocess error must be re-raised as RuntimeError so + proxy_cli.py catches it cleanly (instead of leaking CalledProcessError).""" + monkeypatch.setattr( + ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None + ) + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + monkeypatch.setattr( + ProxyExtrasDBManager, "_roll_back_migration", lambda *a, **kw: None + ) + + # First call: migrate deploy -> P3009 idempotent error. + # Recovery path tries _resolve_specific_migration; that also raises. + def _failing_resolve(*a, **kw): + raise subprocess.CalledProcessError( + returncode=1, + cmd="prisma migrate resolve --applied", + stderr="resolve failed", + output="", + ) + + monkeypatch.setattr( + ProxyExtrasDBManager, "_resolve_specific_migration", _failing_resolve + ) + + stderr = ( + "Error: P3009\nMigration `20260101000000_some_migration` failed\n" + "relation already exists" + ) + with patch("subprocess.run", side_effect=_fake_migrate_deploy_failure(1, stderr)): + with pytest.raises( + RuntimeError, match=r"Failed to mark migration .* as applied" + ): + ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + +def test_v2_does_not_call_resolve_all_migrations(monkeypatch, tmp_path): + """v2 must never call _resolve_all_migrations — that's the bug it fixes.""" + monkeypatch.setattr( + ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None + ) + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + + class FakeResult: + stdout = "Applied migration.\n" + stderr = "" + + monkeypatch.setattr("subprocess.run", lambda *a, **kw: FakeResult()) + + resolve_called = {"n": 0} + monkeypatch.setattr( + ProxyExtrasDBManager, + "_resolve_all_migrations", + lambda *a, **kw: resolve_called.__setitem__("n", resolve_called["n"] + 1), + ) + + ok = ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + assert ok is True + assert resolve_called["n"] == 0, "v2 must not invoke the diff-and-force recovery" + + +_DEADLOCK_STDERR = ( + "Error: ERROR: deadlock detected\n" + "DETAIL: Process 277 waits for ExclusiveLock on advisory lock " + "[17556,0,72707369,1]; blocked by process 278.\n" + "Process 278 waits for ShareLock on virtual transaction 3/1041; " + "blocked by process 277." +) + + +class _DeployApplied: + stdout = "All migrations have been successfully applied." + stderr = "" + returncode = 0 + + +def _deploy_only(deploy_side_effect): + """subprocess.run stand-in that only intercepts `prisma migrate deploy`. + + Scoped by argv so the Prisma toolchain check cannot consume the mock first. + """ + deploys = {"n": 0} + + def _run(*args, **kwargs): + cmd = args[0] if args else kwargs.get("args", []) + if list(cmd)[-2:] == ["migrate", "deploy"]: + deploys["n"] += 1 + return deploy_side_effect(deploys["n"], cmd) + return _DeployApplied() + + return _run, deploys + + +def _prepare_v2_resolver(monkeypatch, tmp_path): + monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@localhost:9/x") + monkeypatch.setattr( + ProxyExtrasDBManager, "_warn_if_db_ahead_of_head", lambda _: None + ) + monkeypatch.setattr(ProxyExtrasDBManager, "_get_prisma_dir", lambda: str(tmp_path)) + (tmp_path / "schema.prisma").write_text("// stub") + monkeypatch.setattr("time.sleep", lambda *_a, **_k: None) + + +def test_v2_retries_transient_advisory_lock_deadlock(monkeypatch, tmp_path): + """v2: replicas racing `migrate deploy` deadlock on Prisma's advisory + lock, which is transient and must be retried rather than kill the boot.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + if n == 1: + raise subprocess.CalledProcessError( + returncode=1, cmd=cmd, stderr=_DEADLOCK_STDERR, output="" + ) + return _DeployApplied() + + run, deploys = _deploy_only(_side_effect) + with patch("subprocess.run", side_effect=run): + ok = ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + assert ok is True + assert deploys["n"] == 2, "the deadlocked deploy must be retried, not raised" + + +def test_v2_persistent_advisory_lock_deadlock_eventually_raises(monkeypatch, tmp_path): + """v2: the deadlock retry is bounded, so a deadlock that never clears + still raises instead of looping or reporting success.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + raise subprocess.CalledProcessError( + returncode=1, cmd=cmd, stderr=_DEADLOCK_STDERR, output="" + ) + + run, deploys = _deploy_only(_side_effect) + with patch("subprocess.run", side_effect=run): + with pytest.raises(RuntimeError, match="after 4 attempts"): + ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + assert deploys["n"] == 4 + + +@pytest.mark.parametrize( + "stderr", + [ + "Error: P1001: Can't reach database server at `db`:`5432`", + "Error: P1002: The database server was reached but timed out.", + ], +) +def test_v2_retries_transient_database_connectivity_errors(monkeypatch, tmp_path, stderr): + """v2: a database not accepting connections yet is retried, not fatal.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + if n == 1: + raise subprocess.CalledProcessError( + returncode=1, cmd=cmd, stderr=stderr, output="" + ) + return _DeployApplied() + + run, deploys = _deploy_only(_side_effect) + with patch("subprocess.run", side_effect=run): + ok = ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + assert ok is True + assert deploys["n"] == 2, "an unreachable database must be retried, not raised" + + +def test_v2_unreachable_database_still_fails_after_the_retries(monkeypatch, tmp_path): + """v2: a genuinely unreachable database still raises once the attempts + are spent, rather than passing as a successful migration.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + raise subprocess.CalledProcessError( + returncode=1, + cmd=cmd, + stderr="Error: P1001: Can't reach database server at `db`:`5432`", + output="", + ) + + run, deploys = _deploy_only(_side_effect) + with patch("subprocess.run", side_effect=run): + with pytest.raises(RuntimeError, match="after 4 attempts"): + ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + assert deploys["n"] == 4 + + +def test_v2_exhausted_retries_report_the_prisma_error(monkeypatch, tmp_path, caplog): + """v2: retrying must not swallow Prisma's stderr, which is captured and is + the only place the cause appears for an operator or a boot-log grep.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + stderr = "Error: P1001: Can't reach database server at `wrong`:`5432`" + + def _side_effect(n, cmd): + raise subprocess.CalledProcessError( + returncode=1, cmd=cmd, stderr=stderr, output="" + ) + + run, _ = _deploy_only(_side_effect) + with caplog.at_level("INFO", logger="litellm_proxy_extras"): + with patch("subprocess.run", side_effect=run): + with pytest.raises(RuntimeError) as exc_info: + ProxyExtrasDBManager.setup_database( + use_migrate=True, use_v2_resolver=True + ) + + assert "P1001" in str(exc_info.value) + assert "P1001" in caplog.text + + +def test_v2_db_push_retries_transient_failures(monkeypatch, tmp_path): + """v2: `prisma db push` retries a transient failure like v1 did. + + Reached from the migrations Job (USE_PRISMA_DB_PUSH=true), not from the + proxy CLI, whose --use_prisma_db_push has its own loop in prisma_client. + """ + _prepare_v2_resolver(monkeypatch, tmp_path) + + pushes = {"n": 0} + + def _run(*args, **kwargs): + cmd = list(args[0] if args else kwargs.get("args", [])) + if cmd[-3:] != ["db", "push", "--accept-data-loss"]: + return _DeployApplied() + pushes["n"] += 1 + if pushes["n"] == 1: + raise subprocess.CalledProcessError( + returncode=1, + cmd=cmd, + stderr="Error: P1001: Can't reach database server at `db`:`5432`", + output="", + ) + return _DeployApplied() + + monkeypatch.setattr( + ProxyExtrasDBManager, "spend_logs_is_partitioned", lambda: False + ) + with patch("subprocess.run", side_effect=_run): + ok = ProxyExtrasDBManager.setup_database(use_migrate=False, use_v2_resolver=True) + + assert ok is True + assert pushes["n"] == 2 + + +def test_v2_db_push_retries_are_bounded_and_report_the_prisma_error( + monkeypatch, tmp_path +): + """v2: a database that never comes back stops after _PRISMA_ATTEMPTS and + surfaces the prisma error, rather than retrying the boot forever.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + pushes = {"n": 0} + + def _run(*args, **kwargs): + cmd = list(args[0] if args else kwargs.get("args", [])) + if cmd[-3:] != ["db", "push", "--accept-data-loss"]: + return _DeployApplied() + pushes["n"] += 1 + raise subprocess.CalledProcessError( + returncode=1, + cmd=cmd, + stderr="Error: P1001: Can't reach database server at `db`:`5432`", + output="", + ) + + monkeypatch.setattr( + ProxyExtrasDBManager, "spend_logs_is_partitioned", lambda: False + ) + with patch("subprocess.run", side_effect=_run): + with pytest.raises(RuntimeError) as exc: + ProxyExtrasDBManager.setup_database( + use_migrate=False, use_v2_resolver=True + ) + + assert pushes["n"] == _PRISMA_ATTEMPTS + assert "P1001" in str(exc.value) + + +def _db_push_only(push_side_effect): + """subprocess.run stand-in that only intercepts `prisma db push`.""" + pushes = {"n": 0} + + def _run(*args, **kwargs): + cmd = list(args[0] if args else kwargs.get("args", [])) + if cmd[-3:] != ["db", "push", "--accept-data-loss"]: + return _DeployApplied() + pushes["n"] += 1 + return push_side_effect(pushes["n"], cmd) + + return _run, pushes + + +def _timed_out_for_real(): + """Capture what subprocess.run really puts on a TimeoutExpired. + + Under text=True it still leaves stderr as bytes, unlike CalledProcessError, + so hardcoding a str here would test a shape production never sees. Derived + at import, before any test patches subprocess.run. + """ + try: + subprocess.run( + ["sh", "-c", "echo 'Error: P1001 unreachable' >&2; sleep 5"], + timeout=0.2, + check=True, + capture_output=True, + text=True, + ) + except subprocess.TimeoutExpired as e: + return e + raise AssertionError("the helper command was supposed to time out") + + +_TIMEOUT_TEMPLATE = _timed_out_for_real() + + +def _real_timeout_expired(cmd): + return subprocess.TimeoutExpired( + cmd=cmd, + timeout=_TIMEOUT_TEMPLATE.timeout, + output=_TIMEOUT_TEMPLATE.stdout, + stderr=_TIMEOUT_TEMPLATE.stderr, + ) + + +def test_v2_db_push_retries_a_timeout(monkeypatch, tmp_path): + """v2: a `prisma db push` that times out is retried, not turned into a + TypeError by classifying its bytes stderr as if it were text.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + if n == 1: + raise _real_timeout_expired(cmd) + return _DeployApplied() + + monkeypatch.setattr( + ProxyExtrasDBManager, "spend_logs_is_partitioned", lambda: False + ) + run, pushes = _db_push_only(_side_effect) + with patch("subprocess.run", side_effect=run): + ok = ProxyExtrasDBManager.setup_database(use_migrate=False, use_v2_resolver=True) + + assert ok is True + assert pushes["n"] == 2 + + +def test_v2_db_push_timeouts_are_bounded(monkeypatch, tmp_path): + """v2: a `prisma db push` that never stops timing out gives up as a + RuntimeError, which is the only exception proxy_cli.py exits cleanly on.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + raise _real_timeout_expired(cmd) + + monkeypatch.setattr( + ProxyExtrasDBManager, "spend_logs_is_partitioned", lambda: False + ) + run, pushes = _db_push_only(_side_effect) + with patch("subprocess.run", side_effect=run): + with pytest.raises(RuntimeError, match=r"prisma db push failed after \d+"): + ProxyExtrasDBManager.setup_database(use_migrate=False, use_v2_resolver=True) + + assert pushes["n"] == _PRISMA_ATTEMPTS + + +def test_v2_unclassified_failure_is_not_treated_as_transient(monkeypatch, tmp_path): + """v2: an unrecognised deploy failure still raises on the first attempt.""" + _prepare_v2_resolver(monkeypatch, tmp_path) + + def _side_effect(n, cmd): + raise subprocess.CalledProcessError( + returncode=1, + cmd=cmd, + stderr="Error: relation \"LiteLLM_SpendLogs\" does not exist", + output="", + ) + + run, deploys = _deploy_only(_side_effect) + with patch("subprocess.run", side_effect=run): + with pytest.raises(RuntimeError, match="cannot be auto-recovered"): + ProxyExtrasDBManager.setup_database(use_migrate=True, use_v2_resolver=True) + + assert deploys["n"] == 1 + + diff --git a/tests/llm_translation/reasoning_effort_grid/grid_spec.py b/tests/llm_translation/reasoning_effort_grid/grid_spec.py index 485f78ed8c7..aa9f66f6665 100644 --- a/tests/llm_translation/reasoning_effort_grid/grid_spec.py +++ b/tests/llm_translation/reasoning_effort_grid/grid_spec.py @@ -163,6 +163,13 @@ _CAPS_NONE: FrozenSet[str] = frozenset() ANTHROPIC_DIRECT_MODELS: Tuple[ModelEntry, ...] = ( + ModelEntry( + alias="claude-fable-5-1", + model="anthropic/claude-fable-5-1", + mode="adaptive", + required_env=_ANTHROPIC_REQ, + caps=_CAPS_XHIGH_MAX, + ), ModelEntry( alias="claude-fable-5", model="anthropic/claude-fable-5", @@ -222,6 +229,19 @@ ANTHROPIC_DIRECT_MODELS: Tuple[ModelEntry, ...] = ( AZURE_AI_MODELS: Tuple[ModelEntry, ...] = ( + ModelEntry( + alias="azure-claude-fable-5-1", + model="azure_ai/claude-fable-5-1", + mode="adaptive", + required_env=_AZURE_FOUNDRY_REQ, + caps=_CAPS_XHIGH_MAX, + fail_reason=( + "claude-fable-5-1 has no deployment on the CI Microsoft Foundry " + "resource yet, so Foundry returns DeploymentNotFound and this cell " + "stays loud in CI. Remove this fail_reason once the deployment " + "exists." + ), + ), ModelEntry( alias="azure-claude-fable-5", model="azure_ai/claude-fable-5", @@ -268,6 +288,20 @@ AZURE_AI_MODELS: Tuple[ModelEntry, ...] = ( VERTEX_AI_MODELS: Tuple[ModelEntry, ...] = ( + ModelEntry( + alias="vertex-claude-fable-5-1", + model="vertex_ai/claude-fable-5-1", + mode="adaptive", + extra_params=(("vertex_location", "global"),), + required_env=_VERTEX_REQ, + caps=_CAPS_XHIGH_MAX, + fail_reason=( + "claude-fable-5-1 availability on the CI Vertex project is not yet " + "confirmed for this brand-new release, so this cell stays loud in " + "CI until verified. Remove this fail_reason once the model is " + "confirmed available on the global Vertex endpoint." + ), + ), ModelEntry( alias="vertex-claude-fable-5", model="vertex_ai/claude-fable-5", @@ -332,6 +366,22 @@ VERTEX_AI_MODELS: Tuple[ModelEntry, ...] = ( BEDROCK_CONVERSE_MODELS: Tuple[ModelEntry, ...] = ( + ModelEntry( + alias="bedrock-claude-fable-5-1", + model="bedrock/converse/us.anthropic.claude-fable-5-1", + mode="adaptive", + extra_params=(("aws_region_name", "us-east-1"),), + required_env=_BEDROCK_REQ, + caps=_CAPS_XHIGH_MAX, + bedrock_effort_ceiling="xhigh", + unavailable_error="is not available for this account", + fail_reason=( + "claude-fable-5-1 access on the CI Bedrock account is not yet " + "confirmed for this brand-new release, so this cell stays loud in " + "CI until verified. Remove this fail_reason once the model is " + "enabled for the account." + ), + ), ModelEntry( alias="bedrock-claude-fable-5", model="bedrock/converse/us.anthropic.claude-fable-5", diff --git a/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py b/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py index 517e3173b8c..714d544ecd6 100644 --- a/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py +++ b/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py @@ -201,8 +201,8 @@ async def test_reasoning_effort_grid( def test_grid_cell_count() -> None: - assert len(_PARAMS) == 31 * 11, ( - f"expected 341 cells (31 provider x model combos x 11 efforts), " + assert len(_PARAMS) == 35 * 11, ( + f"expected 385 cells (35 provider x model combos x 11 efforts), " f"got {len(_PARAMS)}" ) diff --git a/tests/llm_translation/test_litellm_proxy_provider.py b/tests/llm_translation/test_litellm_proxy_provider.py index 1cb805bf9ba..8630259877d 100644 --- a/tests/llm_translation/test_litellm_proxy_provider.py +++ b/tests/llm_translation/test_litellm_proxy_provider.py @@ -5,6 +5,7 @@ from io import BytesIO from unittest.mock import AsyncMock +import httpx import litellm from litellm import completion, embedding import pytest @@ -92,44 +93,54 @@ async def test_litellm_gateway_from_sdk_embedding(is_async): litellm.set_verbose = True litellm._turn_on_debug() + captured_bodies = [] + + def handler(request: httpx.Request) -> httpx.Response: + captured_bodies.append(json.loads(request.content)) + return httpx.Response( + 200, + json={ + "object": "list", + "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], + "model": "my-vllm-model", + "usage": {"prompt_tokens": 2, "total_tokens": 2}, + }, + ) + if is_async: from openai import AsyncOpenAI - openai_client = AsyncOpenAI(api_key="fake-key") - mock_method = AsyncMock() - patch_target = openai_client.embeddings.create + openai_client = AsyncOpenAI( + api_key="fake-key", + http_client=httpx.AsyncClient(transport=httpx.MockTransport(handler)), + ) + response = await litellm.aembedding( + model="litellm_proxy/my-vllm-model", + input="Hello world", + client=openai_client, + api_base="my-custom-api-base", + ) else: from openai import OpenAI - openai_client = OpenAI(api_key="fake-key") - mock_method = MagicMock() - patch_target = openai_client.embeddings.create + openai_client = OpenAI( + api_key="fake-key", + http_client=httpx.Client(transport=httpx.MockTransport(handler)), + ) + response = litellm.embedding( + model="litellm_proxy/my-vllm-model", + input="Hello world", + client=openai_client, + api_base="my-custom-api-base", + ) - with patch.object(patch_target.__self__, patch_target.__name__, new=mock_method): - try: - if is_async: - await litellm.aembedding( - model="litellm_proxy/my-vllm-model", - input="Hello world", - client=openai_client, - api_base="my-custom-api-base", - ) - else: - litellm.embedding( - model="litellm_proxy/my-vllm-model", - input="Hello world", - client=openai_client, - api_base="my-custom-api-base", - ) - except Exception as e: - print(e) + request_body = captured_bodies[0] + print("Request body - {}".format(request_body)) - mock_method.assert_called_once() - - print("Call KWARGS - {}".format(mock_method.call_args.kwargs)) - - assert "Hello world" == mock_method.call_args.kwargs["input"] - assert "my-vllm-model" == mock_method.call_args.kwargs["model"] + assert "Hello world" == request_body["input"] + assert "my-vllm-model" == request_body["model"] + assert "encoding_format" not in request_body + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] @pytest.mark.parametrize("is_async", [False, True]) diff --git a/tests/llm_translation/test_nvidia_nim.py b/tests/llm_translation/test_nvidia_nim.py index 7ee4f347f72..d5942e674d0 100644 --- a/tests/llm_translation/test_nvidia_nim.py +++ b/tests/llm_translation/test_nvidia_nim.py @@ -63,27 +63,39 @@ def test_embedding_nvidia_nim(): litellm.set_verbose = True from openai import OpenAI + captured_bodies = [] + + def handler(request: httpx.Request) -> httpx.Response: + captured_bodies.append(json.loads(request.content)) + return httpx.Response( + 200, + json={ + "object": "list", + "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], + "model": "nvidia/nv-embedqa-e5-v5", + "usage": {"prompt_tokens": 6, "total_tokens": 6}, + }, + ) + client = OpenAI( api_key="fake-api-key", + http_client=httpx.Client(transport=httpx.MockTransport(handler)), ) - with patch.object(client.embeddings.with_raw_response, "create") as mock_client: - try: - litellm.embedding( - model="nvidia_nim/nvidia/nv-embedqa-e5-v5", - input="What is the meaning of life?", - input_type="passage", - dimensions=1024, - client=client, - ) - except Exception as e: - print(e) - mock_client.assert_called_once() - request_body = mock_client.call_args.kwargs - print("request_body: ", request_body) - assert request_body["input"] == "What is the meaning of life?" - assert request_body["model"] == "nvidia/nv-embedqa-e5-v5" - assert request_body["extra_body"]["input_type"] == "passage" - assert request_body["dimensions"] == 1024 + response = litellm.embedding( + model="nvidia_nim/nvidia/nv-embedqa-e5-v5", + input="What is the meaning of life?", + input_type="passage", + dimensions=1024, + client=client, + ) + request_body = captured_bodies[0] + print("request_body: ", request_body) + assert request_body["input"] == "What is the meaning of life?" + assert request_body["model"] == "nvidia/nv-embedqa-e5-v5" + assert request_body["input_type"] == "passage" + assert request_body["dimensions"] == 1024 + assert "encoding_format" not in request_body + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] def test_chat_completion_nvidia_nim_with_tools(): diff --git a/tests/llm_translation/test_prompt_factory.py b/tests/llm_translation/test_prompt_factory.py index a90a3df584e..7b03736920b 100644 --- a/tests/llm_translation/test_prompt_factory.py +++ b/tests/llm_translation/test_prompt_factory.py @@ -1446,9 +1446,17 @@ def test_convert_to_anthropic_tool_invoke_sanitizes_invalid_ids(): def test_convert_to_anthropic_tool_invoke_server_tool(): """ - Test that server_tool_use (srvtoolu_) is reconstructed as server_tool_use. + Test that a server tool call (srvtoolu_) with no stored result is replayed + as a regular tool_use block. - Fixes: https://github.com/BerriAI/litellm/issues/17737 + A server_tool_use block is only valid when paired with its result block, so + an unpaired one must degrade to tool_use for Anthropic to accept the replay. + A paired call still becomes server_tool_use, covered by + test_convert_to_anthropic_tool_invoke_with_web_search_results. + + Context: https://github.com/BerriAI/litellm/issues/17737 (original + server_tool_use reconstruction) and LIT-6622 / PR #39144 (unpaired calls + degrade instead of 400ing at Anthropic). """ tool_calls = [ { @@ -1464,7 +1472,7 @@ def test_convert_to_anthropic_tool_invoke_server_tool(): result = convert_to_anthropic_tool_invoke(tool_calls) assert len(result) == 1 - assert result[0]["type"] == "server_tool_use" # NOT tool_use + assert result[0]["type"] == "tool_use" assert result[0]["id"] == "srvtoolu_01ABC123" assert result[0]["name"] == "web_search" assert result[0]["input"] == {"query": "elephant weight"} diff --git a/tests/local_testing/test_basic_python_version.py b/tests/local_testing/test_basic_python_version.py index fb06ed6b69d..506c58d26b4 100644 --- a/tests/local_testing/test_basic_python_version.py +++ b/tests/local_testing/test_basic_python_version.py @@ -305,14 +305,16 @@ def _run_proxy_server_smoke_test(extra_proxy_args=None): def test_litellm_proxy_server_config_no_general_settings(): - """Exercises the default (v1) migration resolver.""" + """Exercises the default (v2) migration resolver.""" _run_proxy_server_smoke_test() -def test_litellm_proxy_server_config_no_general_settings_v2_resolver(): - """Exercises the opt-in v2 migration resolver. +def test_litellm_proxy_server_config_no_general_settings_legacy_resolver(): + """Exercises the legacy (v1) migration resolver against a real database. - Runs in a separate CI job against a local Postgres to avoid collisions - with the v1 variant when they share a database. + v2 is the default, so the no-arg test above already covers it. This one is + the only place the v1 opt-out gets real-DB migration plus proxy-boot + coverage, and it runs in a separate CI job against its own Postgres to + avoid collisions with the default variant. """ - _run_proxy_server_smoke_test(extra_proxy_args=["--use_v2_migration_resolver"]) + _run_proxy_server_smoke_test(extra_proxy_args=["--use_legacy_migration_resolver"]) diff --git a/tests/local_testing/test_embedding.py b/tests/local_testing/test_embedding.py index aed2849f056..ee2ac14f498 100644 --- a/tests/local_testing/test_embedding.py +++ b/tests/local_testing/test_embedding.py @@ -3,6 +3,8 @@ import os import re import traceback +import httpx + import openai import pytest from dotenv import load_dotenv @@ -1255,56 +1257,42 @@ def test_jina_ai_img_embeddings(input_data, expected_payload_input): assert sent_data["input"] == expected_payload_input -def test_encoding_format_defaults_to_float_for_openai_sdk(monkeypatch): +def test_encoding_format_omitted_by_default_for_openai_sdk(monkeypatch): """ - When encoding_format is not provided, LiteLLM sends `float` for OpenAI-path embeddings. + When encoding_format is not provided, LiteLLM leaves it out of the upstream request. Optional global override: `LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT`. """ monkeypatch.delenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", raising=False) - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - # Create a mock client instance - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance + captured_bodies = [] - # Mock the embeddings.with_raw_response.create method - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", + def handler(request: httpx.Request) -> httpx.Response: + captured_bodies.append(json.loads(request.content)) + return httpx.Response( + 200, + json={ "object": "list", + "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], + "model": "text-embedding-ada-002", "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } - ) - mock_response.headers = {} - - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response + }, ) - # Call the embedding function without encoding_format - response = embedding( - model="text-embedding-ada-002", - input="Hello world", - ) + client = openai.OpenAI( + api_key="sk-test", http_client=httpx.Client(transport=httpx.MockTransport(handler)) + ) - # Get the call arguments to verify what was sent to OpenAI SDK - call_args = mock_client_instance.embeddings.with_raw_response.create.call_args - assert ( - call_args is not None - ), "OpenAI SDK embeddings.create should have been called" + response = embedding( + model="text-embedding-ada-002", + input="Hello world", + api_key="sk-test", + client=client, + ) - call_kwargs = call_args[1] # Get kwargs - - assert "encoding_format" in call_kwargs - assert ( - call_kwargs["encoding_format"] == "float" - ), "encoding_format should default to float when not provided by user" - - print("✅ PASS: encoding_format='float' is correctly passed to OpenAI SDK") + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] + assert "encoding_format" not in captured_bodies[0], ( + "encoding_format should be omitted from the upstream request when not provided by user" + ) def test_encoding_format_explicit_value_preserved(): diff --git a/tests/local_testing/test_exceptions.py b/tests/local_testing/test_exceptions.py index 8370046446d..e6392cda406 100644 --- a/tests/local_testing/test_exceptions.py +++ b/tests/local_testing/test_exceptions.py @@ -5,7 +5,7 @@ import traceback from typing import Any import httpx -from openai import AsyncOpenAI, AuthenticationError, BadRequestError, OpenAIError, RateLimitError +from openai import AsyncAzureOpenAI, AsyncOpenAI, AuthenticationError, AzureOpenAI, BadRequestError, OpenAIError, RateLimitError from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, HTTPHandler @@ -895,7 +895,12 @@ def _pre_call_utils( ): if call_type == "embedding": data["input"] = "Hello world!" - mapped_target: Any = client.embeddings.with_raw_response + if isinstance(client, (AzureOpenAI, AsyncAzureOpenAI)): + mapped_target: Any = client.embeddings.with_raw_response + patched_attr = "create" + else: + mapped_target = client + patched_attr = "post" if sync_mode: original_function = litellm.embedding else: @@ -905,6 +910,7 @@ def _pre_call_utils( if streaming is True: data["stream"] = True mapped_target = client.chat.completions.with_raw_response # type: ignore + patched_attr = "create" if sync_mode: original_function = litellm.completion else: @@ -914,12 +920,13 @@ def _pre_call_utils( if streaming is True: data["stream"] = True mapped_target = client.completions.with_raw_response # type: ignore + patched_attr = "create" if sync_mode: original_function = litellm.text_completion else: original_function = litellm.atext_completion - return data, original_function, mapped_target + return data, original_function, mapped_target, patched_attr def _pre_call_utils_httpx( @@ -1003,7 +1010,7 @@ async def test_exception_with_headers(sync_mode, provider, model, call_type, str ) data = {"model": model} - data, original_function, mapped_target = _pre_call_utils( + data, original_function, mapped_target, patched_attr = _pre_call_utils( call_type=call_type, data=data, client=openai_client, @@ -1049,7 +1056,7 @@ async def test_exception_with_headers(sync_mode, provider, model, call_type, str with patch.object( mapped_target, - "create", + patched_attr, side_effect=_return_exception, ): new_retry_after_mock_client = MagicMock(return_value=-1) diff --git a/tests/local_testing/test_router.py b/tests/local_testing/test_router.py index 370c43f8f44..c714bb4f9a7 100644 --- a/tests/local_testing/test_router.py +++ b/tests/local_testing/test_router.py @@ -2032,8 +2032,8 @@ def test_router_dynamic_cooldown_correct_retry_after_time(): raise exception with patch.object( - openai_client.embeddings.with_raw_response, - "create", + openai_client, + "post", side_effect=_return_exception, ): new_retry_after_mock_client = MagicMock(return_value=-1) diff --git a/tests/pass_through_unit_tests/test_websearch_interception_e2e.py b/tests/pass_through_unit_tests/test_websearch_interception_e2e.py index 091ea106b91..fd95b7fa8f2 100644 --- a/tests/pass_through_unit_tests/test_websearch_interception_e2e.py +++ b/tests/pass_through_unit_tests/test_websearch_interception_e2e.py @@ -937,6 +937,14 @@ async def test_pre_request_hook_modifies_request_body(): print("✅ WebSearchInterceptionLogger initialized") + mock_router = MagicMock() + mock_router.search_tools = [ + { + "search_tool_name": "test-search-tool", + "litellm_params": {"search_provider": "tavily"}, + } + ] + # Track what actually gets sent to the API captured_request = {} @@ -987,6 +995,9 @@ async def test_pre_request_hook_modifies_request_body(): with patch( "litellm.llms.anthropic.experimental_pass_through.messages.handler.anthropic_messages_handler", side_effect=mock_anthropic_messages_handler, + ), patch( # test-quality-ok: the hook imports this process-global router at call time; no injection seam exists to register search_tools + "litellm.proxy.proxy_server.llm_router", + mock_router, ): print( diff --git a/tests/test_litellm/integrations/test_s3.py b/tests/test_litellm/integrations/test_s3.py index 7e997870852..58b15b79e76 100644 --- a/tests/test_litellm/integrations/test_s3.py +++ b/tests/test_litellm/integrations/test_s3.py @@ -2,26 +2,27 @@ from datetime import datetime from unittest.mock import MagicMock, patch import litellm +from litellm.constants import MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES, MAX_S3_OBJECT_KEY_BYTES from litellm.integrations.s3 import S3Logger TEST_KMS_KEY_ARN = "arn:aws:kms:us-east-1:111122223333:key/test-key-id" -def _standard_logging_payload() -> dict: +def _standard_logging_payload(response_id: str = "chatcmpl-test-id") -> dict: return { - "id": "chatcmpl-test-id", + "id": response_id, "metadata": {"user_api_key_team_alias": None}, } -def _log_event_kwargs() -> dict: +def _log_event_kwargs(response_id: str = "chatcmpl-test-id") -> dict: return { "litellm_params": {"metadata": {}}, - "standard_logging_object": _standard_logging_payload(), + "standard_logging_object": _standard_logging_payload(response_id), } -def _run_log_event(callback_params: dict) -> MagicMock: +def _run_log_event(callback_params: dict, response_id: str = "chatcmpl-test-id") -> MagicMock: original = litellm.s3_callback_params litellm.s3_callback_params = callback_params try: @@ -30,8 +31,8 @@ def _run_log_event(callback_params: dict) -> MagicMock: mock_boto3_client.return_value = mock_s3_client logger = S3Logger() logger.log_event( - kwargs=_log_event_kwargs(), - response_obj={}, + kwargs=_log_event_kwargs(response_id), + response_obj={"id": response_id}, start_time=datetime(2026, 7, 30, 12, 0, 0), end_time=datetime(2026, 7, 30, 12, 0, 1), print_verbose=lambda *args, **kwargs: None, @@ -154,3 +155,30 @@ def test_non_string_key_id_is_dropped_and_valid_algorithm_is_kept(): put_object_kwargs = mock_s3_client.put_object.call_args.kwargs assert put_object_kwargs["ServerSideEncryption"] == "aws:kms" assert "SSEKMSKeyId" not in put_object_kwargs + + +def test_put_object_key_and_filename_are_bounded_for_an_oversized_response_id(): + """The sync logger bounds both the key and the Content-Disposition filename.""" + mock_s3_client = _run_log_event( + {"s3_bucket_name": "test-bucket", "s3_region_name": "us-west-2", "s3_path": "logs"}, + response_id="resp_" + "A" * 1100, + ) + + put_object_kwargs = mock_s3_client.put_object.call_args.kwargs + assert len(put_object_kwargs["Key"].encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES + assert put_object_kwargs["Key"].startswith("logs/2026-07-30/time-12-00-00-000000_resp_") + filename = put_object_kwargs["ContentDisposition"].removeprefix('inline; filename="').removesuffix('"') + assert len(filename.encode("utf-8")) <= MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES + + +def test_put_object_keeps_the_configured_path_intact_when_only_the_id_has_to_shrink(): + """A long configured s3_path survives whole when the id can be shortened instead.""" + long_path = "litellm-prod-logs/" + "t" * 921 + mock_s3_client = _run_log_event( + {"s3_bucket_name": "test-bucket", "s3_region_name": "us-west-2", "s3_path": long_path}, + response_id="resp_" + "B" * 100, + ) + + key = mock_s3_client.put_object.call_args.kwargs["Key"] + assert key.startswith(long_path + "/2026-07-30/") + assert len(key.encode("utf-8")) == MAX_S3_OBJECT_KEY_BYTES diff --git a/tests/test_litellm/integrations/test_s3_v2.py b/tests/test_litellm/integrations/test_s3_v2.py index 51671d5101e..a037284d7c1 100644 --- a/tests/test_litellm/integrations/test_s3_v2.py +++ b/tests/test_litellm/integrations/test_s3_v2.py @@ -1170,6 +1170,294 @@ def test_create_s3_batch_logging_element_flat_key_for_arn_response_id(): assert file_segment.endswith("model-invocation-job_gl18r6skk9yy.json") +# -------------------------------------------------------------- +# object keys bounded to S3's 1024 UTF-8 byte limit +# -------------------------------------------------------------- +def _oversized_response_id() -> str: + return "resp_" + "A" * 1100 + + +def test_s3_object_key_at_the_byte_limit_is_left_alone(): + """A key that still fits is left byte-identical.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + start_time = datetime(2026, 8, 24, 6, 18, 41, 948021) + fixed_len = len("input/2026-08-24/.json") + file_name = "x" * (MAX_S3_OBJECT_KEY_BYTES - fixed_len) + + key = get_s3_object_key(s3_path="input", prefix="", start_time=start_time, s3_file_name=file_name) + + assert key == f"input/2026-08-24/{file_name}.json" + assert len(key.encode("utf-8")) == MAX_S3_OBJECT_KEY_BYTES + + +def test_s3_object_key_is_bounded_for_oversized_response_id(): + """An oversized Responses API id is shortened to a readable head plus a digest.""" + import hashlib + + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + start_time = datetime(2026, 8, 24, 6, 18, 41, 948021) + file_name = f"time-06-18-41-948021_{_oversized_response_id()}" + + key = get_s3_object_key(s3_path="input", prefix="DefaultTeamProd/", start_time=start_time, s3_file_name=file_name) + + assert len(key.encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES + assert key.startswith("input/DefaultTeamProd/2026-08-24/time-06-18-41-948021_resp_") + assert key.endswith(f"_{hashlib.sha256(file_name.encode('utf-8')).hexdigest()}.json") + + +@pytest.mark.parametrize( + "s3_path,prefix", + [ + ("input", ""), + ("a" * 900, ""), + ("input", "team-" + "b" * 900 + "/"), + ("c" * 600, "team-" + "d" * 600 + "/key-" + "e" * 600 + "/"), + # many short segments, so the trim lands exactly on the budget edge + ("", "ssss/" * 200), + ], +) +def test_s3_object_key_is_bounded_for_long_paths_and_aliases(s3_path: str, prefix: str): + """Long paths, team aliases and key aliases stay within the cap.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + key = get_s3_object_key( + s3_path=s3_path, + prefix=prefix, + start_time=datetime(2026, 8, 24, 6, 18, 41, 948021), + s3_file_name=f"time-06-18-41-948021_{_oversized_response_id()}", + ) + + assert len(key.encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES + assert key.endswith(".json") + assert "/2026-08-24/" in key or key.startswith("2026-08-24/") + assert "/" not in key.rsplit("2026-08-24/", 1)[1] + + +def test_s3_object_key_trimmed_prefixes_stay_distinct_per_operator(): + """Prefixes that differ only past the trim point keep separate folders.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + start_time = datetime(2026, 8, 24, 6, 18, 41, 948021) + keys = [ + get_s3_object_key( + s3_path="input", + prefix="team-" + "b" * 1000 + suffix + "/", + start_time=start_time, + s3_file_name=f"time-06-18-41-948021_{_oversized_response_id()}", + ) + for suffix in ("-one", "-two") + ] + + assert keys[0] != keys[1] + assert all(key.startswith("input/team-" + "b" * 900) for key in keys) + assert all(len(key.encode("utf-8")) == MAX_S3_OBJECT_KEY_BYTES for key in keys) + + +def test_s3_object_key_bounded_prefix_never_splits_a_multibyte_character(): + """A multibyte prefix is trimmed on a character boundary.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + s3_path = "\u65e5\u672c\u8a9e" * 200 + + key = get_s3_object_key( + s3_path=s3_path, + prefix="\u30c1\u30fc\u30e0" * 200 + "/", + start_time=datetime(2026, 8, 24, 6, 18, 41, 948021), + s3_file_name=f"time-06-18-41-948021_{_oversized_response_id()}", + ) + + assert len(key.encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES + assert key.startswith(s3_path[:100]) + assert "\ufffd" not in key + + +def test_s3_object_key_stays_unique_for_ids_sharing_a_head(): + """Ids sharing a visible head still get distinct keys.""" + from litellm.integrations.s3 import get_s3_object_key + + start_time = datetime(2026, 8, 24, 6, 18, 41, 948021) + keys = { + get_s3_object_key( + s3_path="input", + prefix="", + start_time=start_time, + s3_file_name=f"time-06-18-41-948021_{_oversized_response_id()}{suffix}", + ) + for suffix in ("first", "second", "third") + } + + assert len(keys) == 3 + + +def test_s3_object_key_bounding_matches_the_documented_layout(): + """The bounded key is `//_.json`.""" + import hashlib + + from litellm.integrations.s3 import get_s3_object_key + + file_name = f"time-06-18-41-948021_{_oversized_response_id()}" + + key = get_s3_object_key( + s3_path="input", + prefix="team/", + start_time=datetime(2026, 8, 24, 6, 18, 41, 948021), + s3_file_name=file_name, + ) + + digest = hashlib.sha256(file_name.encode("utf-8")).hexdigest() + assert key == f"input/team/2026-08-24/{file_name[:64]}_{digest}.json" + + +def test_s3_object_key_keeps_the_configured_prefix_when_only_the_id_overflows(): + """A 940 byte configured prefix survives whole when only the id overflows.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + prefix = "team-" + "b" * 934 + "/" + + key = get_s3_object_key( + s3_path="", + prefix=prefix, + start_time=datetime(2026, 8, 24, 6, 18, 41, 948021), + s3_file_name=f"time-06-18-41-948021_{_oversized_response_id()}", + ) + + assert key.startswith(prefix + "2026-08-24/") + assert len(key.encode("utf-8")) == MAX_S3_OBJECT_KEY_BYTES + + +def test_s3_object_key_spends_the_whole_budget_when_the_prefix_must_be_trimmed(): + """A trimmed prefix keeps every byte the budget allows, not whole segments.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + s3_path = "p" * 400 + "/" + "q" * 600 + + key = get_s3_object_key( + s3_path=s3_path, + prefix="", + start_time=datetime(2026, 8, 24, 6, 18, 41, 948021), + s3_file_name="time-06-18-41-948021_abc", + ) + + assert len(key.encode("utf-8")) == MAX_S3_OBJECT_KEY_BYTES + assert key.startswith("p" * 400 + "/" + "q" * 500) + + +def test_s3_object_key_keeps_a_single_segment_path_as_far_as_it_fits(): + """A path with no separator is kept as far as it fits, never dropped to the bucket root.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + from litellm.integrations.s3 import get_s3_object_key + + key = get_s3_object_key( + s3_path="a" * 1050, + prefix="", + start_time=datetime(2026, 8, 24, 6, 18, 41, 948021), + s3_file_name="time-06-18-41-948021_chatcmpl-xyz", + ) + + assert len(key.encode("utf-8")) == MAX_S3_OBJECT_KEY_BYTES + assert key.startswith("a" * 900) + + +def test_create_s3_batch_logging_element_bounds_key_and_keeps_full_response_id(): + """The batch element bounds the key and keeps the full response id in the payload.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + + logger = S3Logger(s3_use_team_prefix=True, s3_use_key_prefix=True) + response_id = _oversized_response_id() + payload = StandardLoggingPayload( + id=response_id, + metadata={"user_api_key_team_alias": "DefaultTeamProd", "user_api_key_alias": "prod-key"}, + messages=[], + ) + + result = logger.create_s3_batch_logging_element(datetime(2026, 8, 24, 6, 18, 41, 948021), payload) + + assert result is not None + assert len(result.s3_object_key.encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES + assert result.s3_object_key.startswith("DefaultTeamProd/prod-key/2026-08-24/") + assert result.payload["id"] == response_id + + +def test_s3_object_download_filename_is_bounded_for_oversized_response_id(): + """The Content-Disposition filename is bounded too, or the PUT fails with MetadataTooLarge.""" + from litellm.constants import MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES + from litellm.integrations.s3 import get_s3_object_download_filename + + file_name = get_s3_object_download_filename(datetime(2026, 8, 24, 6, 18, 41, 948021), _oversized_response_id()) + + assert len(file_name.encode("utf-8")) <= MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES + assert file_name.startswith("time-2026-08-24T06-18-41-948021_resp_") + assert file_name.endswith(".json") + + +def test_s3_object_download_filenames_stay_distinct_when_shortened(): + """Shortened filenames stay distinct.""" + from litellm.integrations.s3 import get_s3_object_download_filename + + start_time = datetime(2026, 8, 24, 6, 18, 41, 948021) + file_names = { + get_s3_object_download_filename(start_time, _oversized_response_id() + suffix) + for suffix in ("first", "second", "third") + } + + assert len(file_names) == 3 + + +def test_s3_object_download_filename_short_id_is_unchanged(): + """An ordinary response id keeps the filename it had before.""" + from litellm.integrations.s3 import get_s3_object_download_filename + + file_name = get_s3_object_download_filename(datetime(2026, 8, 24, 6, 18, 41, 948021), "resp_abc123") + + assert file_name == "time-2026-08-24T06-18-41-948021_resp_abc123.json" + + +def test_create_s3_batch_logging_element_bounds_the_download_filename(): + """The batch element carries a bounded Content-Disposition filename.""" + from litellm.constants import MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES + + logger = S3Logger() + payload = StandardLoggingPayload(id=_oversized_response_id(), metadata={}, messages=[]) + + result = logger.create_s3_batch_logging_element(datetime(2026, 8, 24, 6, 18, 41, 948021), payload) + + assert result is not None + assert len(result.s3_object_download_filename.encode("utf-8")) <= MAX_S3_OBJECT_DOWNLOAD_FILENAME_BYTES + + +@pytest.mark.asyncio +async def test_audit_log_object_key_is_bounded_for_a_long_configured_path(): + """Audit log keys are bounded by the same builder.""" + from litellm.constants import MAX_S3_OBJECT_KEY_BYTES + + logger = S3Logger() + logger.s3_path = "audit-archive/" + "z" * 1100 + + await logger.async_log_audit_log_event({"id": "1a4f7bd0-6f1e-4d0a-9b3c-9f2e1d5a7c88"}) + + assert len(logger.log_queue) == 1 + assert len(logger.log_queue[0].s3_object_key.encode("utf-8")) <= MAX_S3_OBJECT_KEY_BYTES + assert logger.log_queue[0].s3_object_key.startswith("audit-archive/" + "z" * 900) + + +def test_s3_object_download_filename_drops_characters_that_break_the_header(): + """A quote or separator in the response id cannot escape the quoted header value.""" + from litellm.integrations.s3 import get_s3_object_download_filename + + file_name = get_s3_object_download_filename(datetime(2026, 8, 24, 6, 18, 41, 948021), 'resp_a"b/c') + + assert file_name == "time-2026-08-24T06-18-41-948021_resp_a_b_c.json" + + # -------------------------------------------------------------- # params_source / s3_callback_params_override (audit-log decoupling) # -------------------------------------------------------------- diff --git a/tests/test_litellm/integrations/websearch_interception/test_websearch_interception_handler.py b/tests/test_litellm/integrations/websearch_interception/test_websearch_interception_handler.py index ec4bc1f49eb..f39f41a6d12 100644 --- a/tests/test_litellm/integrations/websearch_interception/test_websearch_interception_handler.py +++ b/tests/test_litellm/integrations/websearch_interception/test_websearch_interception_handler.py @@ -14,7 +14,7 @@ from litellm.integrations.websearch_interception.handler import ( ) from litellm.llms.base_llm.search.transformation import SearchResponse from litellm.proxy._types import LiteLLM_ObjectPermissionTable, LiteLLM_TeamTable, ProxyException, UserAPIKeyAuth -from litellm.types.utils import CallTypes, LlmProviders +from litellm.types.utils import LlmProviders def test_initialize_from_proxy_config(): @@ -230,124 +230,6 @@ async def test_execute_search_passes_selected_search_tool_litellm_params(monkeyp assert forwarded_kwargs["max_retries"] == 2 -@pytest.mark.asyncio -@pytest.mark.parametrize( - ("search_tools", "error"), - [ - pytest.param(None, "was not found", id="router-not-configured"), - pytest.param( - [{"search_tool_name": "other-search", "litellm_params": {"search_provider": "tavily"}}], - "was not found", - id="requested-tool-not-configured", - ), - pytest.param( - [{"search_tool_name": "parallel-search", "litellm_params": "not-a-mapping"}], - "does not define a valid search provider", - id="invalid-parameters", - ), - pytest.param( - [{"search_tool_name": "parallel-search", "litellm_params": {}}], - "does not define a valid search provider", - id="missing-provider", - ), - pytest.param( - [{"search_tool_name": "parallel-search", "litellm_params": {"search_provider": " "}}], - "does not define a valid search provider", - id="whitespace-provider", - ), - pytest.param( - [{"search_tool_name": "parallel-search", "litellm_params": {"search_provider": 123}}], - "does not define a valid search provider", - id="invalid-provider", - ), - ], -) -async def test_execute_search_rejects_invalid_explicit_search_tool(monkeypatch, search_tools, error): - import litellm - from litellm.proxy import proxy_server - - logger = WebSearchInterceptionLogger(search_tool_name="parallel-search") - router = None if search_tools is None else MagicMock(search_tools=search_tools) - mock_asearch = AsyncMock() - - monkeypatch.setattr(proxy_server, "llm_router", router) - monkeypatch.setattr(litellm, "asearch", mock_asearch) - - with pytest.raises(ValueError, match=f"Configured search tool 'parallel-search' {error}"): - await logger._execute_search("what is litellm") - - mock_asearch.assert_not_awaited() - - -@pytest.mark.asyncio -async def test_execute_search_honors_explicit_parallel_search_tool(monkeypatch): - import litellm - from litellm.proxy import proxy_server - - logger = WebSearchInterceptionLogger(search_tool_name="parallel-search") - router = MagicMock( - search_tools=[ - { - "search_tool_name": "other-search", - "litellm_params": {"search_provider": "tavily", "api_key": "other-key"}, - }, - { - "search_tool_name": "parallel-search", - "litellm_params": {"search_provider": "parallel_ai", "api_key": "parallel-key"}, - }, - ], - ) - mock_asearch = AsyncMock(return_value=SearchResponse(object="search", results=[])) - - monkeypatch.setattr(proxy_server, "llm_router", router) - monkeypatch.setattr(litellm, "asearch", mock_asearch) - - await logger._execute_search("what is litellm") - - mock_asearch.assert_awaited_once_with( - query="what is litellm", - search_provider="parallel_ai", - api_key="parallel-key", - ) - - -@pytest.mark.asyncio -@pytest.mark.parametrize( - ("search_tools", "expected_search_kwargs"), - [ - pytest.param(None, {"search_provider": "perplexity"}, id="router-not-configured"), - pytest.param( - [ - { - "search_tool_name": "first-search", - "litellm_params": {"search_provider": "tavily", "api_key": "first-key"}, - }, - { - "search_tool_name": "parallel-search", - "litellm_params": {"search_provider": "parallel_ai", "api_key": "parallel-key"}, - }, - ], - {"search_provider": "tavily", "api_key": "first-key"}, - id="first-configured-tool", - ), - ], -) -async def test_execute_search_preserves_implicit_provider_selection(monkeypatch, search_tools, expected_search_kwargs): - import litellm - from litellm.proxy import proxy_server - - logger = WebSearchInterceptionLogger() - router = None if search_tools is None else MagicMock(search_tools=search_tools) - mock_asearch = AsyncMock(return_value=SearchResponse(object="search", results=[])) - - monkeypatch.setattr(proxy_server, "llm_router", router) - monkeypatch.setattr(litellm, "asearch", mock_asearch) - - await logger._execute_search("what is litellm") - - mock_asearch.assert_awaited_once_with(query="what is litellm", **expected_search_kwargs) - - @pytest.mark.asyncio async def test_execute_search_attributes_spend_to_the_calling_key(monkeypatch): """An intercepted search is billed and logged against the key that made the LLM request. @@ -515,72 +397,6 @@ async def test_execute_search_enforces_team_search_tool_permission(monkeypatch): mock_asearch.assert_not_awaited() -@pytest.mark.asyncio -@pytest.mark.parametrize( - ("call_type", "web_search_tool"), - [ - pytest.param( - CallTypes.acompletion, - {"type": "web_search_20250305", "name": "web_search"}, - id="chat-completion", - ), - pytest.param(CallTypes.responses, {"type": "web_search"}, id="responses"), - pytest.param(CallTypes.aresponses, {"type": "web_search"}, id="async-responses"), - pytest.param( - CallTypes.anthropic_messages, - {"type": "web_search_20250305", "name": "web_search"}, - id="anthropic-messages", - ), - ], -) -async def test_deployment_hook_dispatcher_propagates_missing_explicit_search_tool( - monkeypatch, call_type, web_search_tool -): - import litellm - from litellm.proxy import proxy_server - from litellm.utils import async_pre_call_deployment_hook - - logger = WebSearchInterceptionLogger(enabled_providers=["bedrock"], search_tool_name="parallel-search") - mock_asearch = AsyncMock() - kwargs = { - "model": "bedrock/claude-sonnet-4", - "tools": [web_search_tool], - "custom_llm_provider": "bedrock", - } - - monkeypatch.setattr( - proxy_server, - "llm_router", - MagicMock(search_tools=[{"search_tool_name": "other-search", "litellm_params": {"search_provider": "tavily"}}]), - ) - monkeypatch.setattr(litellm, "callbacks", [logger]) - monkeypatch.setattr(litellm, "asearch", mock_asearch) - - with pytest.raises(ValueError, match="Configured search tool 'parallel-search' was not found"): - await async_pre_call_deployment_hook(kwargs=kwargs, call_type=call_type.value) - - assert kwargs["tools"] == [web_search_tool] - mock_asearch.assert_not_awaited() - - -@pytest.mark.asyncio -async def test_deployment_hook_skips_explicit_tool_validation_for_non_search_responses(monkeypatch): - from litellm.proxy import proxy_server - - logger = WebSearchInterceptionLogger(enabled_providers=["bedrock"], search_tool_name="parallel-search") - monkeypatch.setattr(proxy_server, "llm_router", MagicMock(search_tools=[])) - - result = await logger.async_pre_call_deployment_hook( - kwargs={ - "tools": [{"type": "function", "name": "calculator"}], - "custom_llm_provider": "bedrock", - }, - call_type=CallTypes.aresponses, - ) - - assert result is None - - @pytest.mark.asyncio async def test_async_pre_call_deployment_hook_provider_from_top_level_kwargs(): """Test that async_pre_call_deployment_hook finds custom_llm_provider at top-level kwargs. diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py index 3c4121977de..0e1c832ebf5 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py @@ -32,6 +32,8 @@ from litellm.litellm_core_utils.llm_cost_calc.utils import ( TokenTypeCostBreakdown, _calculate_input_cost, _get_token_base_cost, + _is_off_peak, + _is_within_off_peak_window, calculate_cache_writing_cost, generic_cost_per_token, get_token_type_cost_breakdown, @@ -409,6 +411,377 @@ def test_get_token_base_cost_picks_highest_crossed_tier(): assert prompt_base_cost == 9e-6 +def test_is_within_off_peak_window_same_day(): + from datetime import datetime, timezone + + window = "09:00-17:00" + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 12, 0, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 8, 59, tzinfo=timezone.utc)) is False + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 9, 0, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 17, 0, tzinfo=timezone.utc)) is False + + +def test_is_within_off_peak_window_wraps_midnight(): + from datetime import datetime, timezone + + window = "16:30-00:30" + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 0, 15, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 16, 30, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 0, 30, tzinfo=timezone.utc)) is False + assert _is_within_off_peak_window(window, datetime(2026, 1, 1, 12, 0, tzinfo=timezone.utc)) is False + + +def test_is_within_off_peak_window_equal_start_and_end_covers_whole_day(): + """An equal start and end is the natural way to spell off-peak all day. It used to take the + non-wrap branch, where start <= now < end can never hold, so it matched nothing and billed at + standard rates around the clock without raising or logging anything.""" + from datetime import datetime, timezone + + for window in ("00:00-00:00", "10:00-10:00"): + for hour in range(24): + assert ( + _is_within_off_peak_window(window, datetime(2026, 1, 1, hour, 0, tzinfo=timezone.utc)) is True + ), f"{window} should cover {hour:02d}:00" + + +def test_is_within_off_peak_window_multiple_windows(): + from datetime import datetime, timezone + + # Providers like DeepSeek V4 have more than one daily peak/off-peak window. + windows = ["01:00-05:00", "13:00-16:00"] + assert _is_within_off_peak_window(windows, datetime(2026, 1, 1, 3, 0, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(windows, datetime(2026, 1, 1, 14, 30, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window(windows, datetime(2026, 1, 1, 9, 0, tzinfo=timezone.utc)) is False + # a malformed entry in the list is ignored, valid entries still match + assert _is_within_off_peak_window(["bad", "13:00-16:00"], datetime(2026, 1, 1, 14, 0, tzinfo=timezone.utc)) is True + assert _is_within_off_peak_window([], datetime(2026, 1, 1, 14, 0, tzinfo=timezone.utc)) is False + + +def test_is_within_off_peak_window_normalizes_timezone_aware_input(): + from datetime import datetime, timedelta, timezone + + # A caller may pass a non-UTC aware datetime; the window is UTC and must be + # evaluated in UTC, not against the caller's wall-clock. 09:00 at UTC+8 is + # 01:00 UTC, inside the 01:00-05:00 window. + tz_plus_8 = timezone(timedelta(hours=8)) + assert _is_within_off_peak_window("01:00-05:00", datetime(2026, 1, 1, 9, 0, tzinfo=tz_plus_8)) is True + assert _is_within_off_peak_window("01:00-05:00", datetime(2026, 1, 1, 12, 0, tzinfo=tz_plus_8)) is True + # 06:00 at UTC+8 is 22:00 UTC the previous day, outside the window + assert _is_within_off_peak_window("01:00-05:00", datetime(2026, 1, 1, 6, 0, tzinfo=tz_plus_8)) is False + + +def test_is_within_off_peak_window_malformed_returns_false(): + from datetime import datetime, timezone + + now = datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc) + assert _is_within_off_peak_window("not-a-window", now) is False + assert _is_within_off_peak_window("16:30", now) is False + assert _is_within_off_peak_window("25:00-26:00", now) is False + + +def test_is_off_peak_weekday_qualified_windows_deepseek_schedule(): + """DeepSeek since 2026-08-23: peak is 01:00-04:00 and 06:00-10:00 UTC on weekdays only, with + weekends off-peak around the clock. The weekday axis is not a filter on one window set; on + two days of seven the off-peak window becomes the whole day, so the schedule needs two + day-qualified rules. The weekend instants inside would-be peak hours are the ones a + time-only implementation bills wrong.""" + from datetime import datetime, timezone + + deepseek = { + "windows": [ + {"hours_utc": ["00:00-01:00", "04:00-06:00", "10:00-00:00"], "weekdays": [1, 2, 3, 4, 5]}, + {"hours_utc": "00:00-00:00", "weekdays": [6, 7]}, + ], + } + peak_instants = [ + datetime(2026, 8, 24, 1, 30, tzinfo=timezone.utc), + datetime(2026, 8, 26, 7, 0, tzinfo=timezone.utc), + datetime(2026, 8, 28, 9, 59, tzinfo=timezone.utc), + ] + off_peak_instants = [ + datetime(2026, 8, 23, 1, 30, tzinfo=timezone.utc), + datetime(2026, 8, 29, 2, 0, tzinfo=timezone.utc), + datetime(2026, 8, 30, 8, 0, tzinfo=timezone.utc), + datetime(2026, 8, 26, 5, 0, tzinfo=timezone.utc), + datetime(2026, 8, 28, 16, 30, tzinfo=timezone.utc), + datetime(2026, 8, 24, 0, 30, tzinfo=timezone.utc), + ] + for when in peak_instants: + assert _is_off_peak(deepseek, when) is False, f"{when.isoformat()} should bill peak" + for when in off_peak_instants: + assert _is_off_peak(deepseek, when) is True, f"{when.isoformat()} should bill off-peak" + + +def test_is_off_peak_weekday_timezone_reads_vendor_calendar(): + """The UTC and Asia/Shanghai calendars only disagree about the date over 16:00-24:00 UTC, so + a window in that stretch is the one place a vendor-local weekday differs from a UTC one: + 2026-08-28T16:30Z is Friday in UTC but already Saturday in Beijing.""" + from datetime import datetime, timezone + + shanghai_saturday = { + "weekday_timezone": "Asia/Shanghai", + "windows": [{"hours_utc": "16:00-17:00", "weekdays": [6]}], + } + assert _is_off_peak(shanghai_saturday, datetime(2026, 8, 28, 16, 30, tzinfo=timezone.utc)) is True + assert _is_off_peak(shanghai_saturday, datetime(2026, 8, 29, 16, 30, tzinfo=timezone.utc)) is False + + +def test_is_off_peak_weekdays_default_utc_calendar_and_accept_names(): + from datetime import datetime, timezone + + named_weekend = {"windows": [{"hours_utc": "00:00-00:00", "weekdays": ["Sat", "sunday"]}]} + assert _is_off_peak(named_weekend, datetime(2026, 8, 29, 12, 0, tzinfo=timezone.utc)) is True + assert _is_off_peak(named_weekend, datetime(2026, 8, 28, 12, 0, tzinfo=timezone.utc)) is False + + utc_friday = {"windows": [{"hours_utc": "16:00-17:00", "weekdays": [5]}]} + assert _is_off_peak(utc_friday, datetime(2026, 8, 28, 16, 30, tzinfo=timezone.utc)) is True + assert _is_off_peak(utc_friday, datetime(2026, 8, 29, 16, 30, tzinfo=timezone.utc)) is False + + +def test_is_off_peak_naive_current_time_read_as_utc(): + from datetime import datetime + + block = {"windows": [{"hours_utc": "16:00-17:00", "weekdays": [5]}]} + assert _is_off_peak(block, datetime(2026, 8, 28, 16, 30)) is True + assert _is_off_peak(block, datetime(2026, 8, 29, 16, 30)) is False + + +def test_is_off_peak_invalid_weekday_timezone_falls_back_to_utc(): + from datetime import datetime, timezone + + block = {"weekday_timezone": "Not/AZone", "windows": [{"hours_utc": "16:00-17:00", "weekdays": [5]}]} + assert _is_off_peak(block, datetime(2026, 8, 28, 16, 30, tzinfo=timezone.utc)) is True + assert _is_off_peak(block, datetime(2026, 8, 29, 16, 30, tzinfo=timezone.utc)) is False + + +def test_is_off_peak_ignores_malformed_weekday_rules(): + from datetime import datetime, timezone + + when = datetime(2026, 8, 29, 12, 0, tzinfo=timezone.utc) + assert _is_off_peak({"windows": [{"hours_utc": "00:00-00:00", "weekdays": []}]}, when) is False + assert _is_off_peak({"windows": [{"hours_utc": "00:00-00:00", "weekdays": [0, 8, "noday", True]}]}, when) is False + assert _is_off_peak({"windows": [{"weekdays": [6]}]}, when) is False + assert _is_off_peak({"windows": [{"hours_utc": 1630}]}, when) is False + assert _is_off_peak({"windows": ["00:00-00:00"]}, when) is False + assert _is_off_peak({"windows": "00:00-00:00"}, when) is False + assert _is_off_peak({"hours_utc": 1630}, when) is False + assert _is_off_peak({}, when) is False + + +def test_is_off_peak_flat_hours_and_windows_are_a_union(): + from datetime import datetime, timezone + + block = { + "hours_utc": "04:00-06:00", + "windows": [{"hours_utc": "00:00-00:00", "weekdays": [7]}], + } + assert _is_off_peak(block, datetime(2026, 8, 28, 5, 0, tzinfo=timezone.utc)) is True + assert _is_off_peak(block, datetime(2026, 8, 30, 20, 0, tzinfo=timezone.utc)) is True + assert _is_off_peak(block, datetime(2026, 8, 28, 20, 0, tzinfo=timezone.utc)) is False + + +def test_get_token_base_cost_weekend_only_off_peak_rate(): + from datetime import datetime, timezone + from typing import cast + + from litellm.types.utils import ModelInfo + + model_info = cast( + ModelInfo, + { + "input_cost_per_token": 1e-6, + "output_cost_per_token": 2e-6, + "off_peak_pricing": { + "windows": [ + {"hours_utc": ["00:00-01:00", "04:00-06:00", "10:00-00:00"], "weekdays": [1, 2, 3, 4, 5]}, + {"hours_utc": "00:00-00:00", "weekdays": [6, 7]}, + ], + "input_cost_per_token": 5e-7, + "output_cost_per_token": 1e-6, + }, + }, + ) + usage = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150) + + saturday_peak_hours = _get_token_base_cost( + model_info, usage, current_time=datetime(2026, 8, 29, 2, 0, tzinfo=timezone.utc) + ) + assert saturday_peak_hours[:2] == (5e-7, 1e-6) + + monday_same_hours = _get_token_base_cost( + model_info, usage, current_time=datetime(2026, 8, 24, 2, 0, tzinfo=timezone.utc) + ) + assert monday_same_hours[:2] == (1e-6, 2e-6) + + +def test_get_token_base_cost_applies_off_peak_pricing(): + from datetime import datetime, timezone + from typing import cast + + from litellm.types.utils import ModelInfo + + model_info = cast( + ModelInfo, + { + "input_cost_per_token": 1e-6, + "output_cost_per_token": 2e-6, + "cache_read_input_token_cost": 1e-7, + "off_peak_pricing": { + "hours_utc": "16:30-00:30", + "input_cost_per_token": 5e-7, + "output_cost_per_token": 1e-6, + "cache_read_input_token_cost": 5e-8, + }, + }, + ) + usage = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150) + + off_peak = _get_token_base_cost(model_info, usage, current_time=datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc)) + assert off_peak[0] == 5e-7 + assert off_peak[1] == 1e-6 + assert off_peak[4] == 5e-8 + + peak = _get_token_base_cost(model_info, usage, current_time=datetime(2026, 1, 1, 12, 0, tzinfo=timezone.utc)) + assert peak[0] == 1e-6 + assert peak[1] == 2e-6 + assert peak[4] == 1e-7 + + +def test_get_token_base_cost_non_mapping_off_peak_block_bills_standard_rates(): + """A truthy non-mapping off_peak_pricing value (a bare string or a list in + YAML) must bill standard rates rather than raising, matching how every + other malformed piece of the block behaves. + """ + from datetime import datetime, timezone + from typing import cast + + from litellm.types.utils import ModelInfo + + usage = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150) + when = datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc) + + for malformed_block in ("16:00-19:00", ["16:00-19:00"], 5e-7, True): + model_info = cast( + ModelInfo, + { + "input_cost_per_token": 1e-6, + "output_cost_per_token": 2e-6, + "off_peak_pricing": malformed_block, + }, + ) + result = _get_token_base_cost(model_info, usage, current_time=when) + assert result[0] == 1e-6 + assert result[1] == 2e-6 + + +def test_get_token_base_cost_off_peak_falls_back_to_standard_when_unset(): + from datetime import datetime, timezone + from typing import cast + + from litellm.types.utils import ModelInfo + + model_info = cast( + ModelInfo, + { + "input_cost_per_token": 1e-6, + "output_cost_per_token": 2e-6, + "off_peak_pricing": {"hours_utc": "16:30-00:30", "input_cost_per_token": 5e-7}, + }, + ) + usage = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150) + + result = _get_token_base_cost(model_info, usage, current_time=datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc)) + assert result[0] == 5e-7 + assert result[1] == 2e-6 + + +def test_get_token_base_cost_off_peak_wins_over_threshold(): + from datetime import datetime, timezone + from typing import cast + + from litellm.types.utils import ModelInfo + + model_info = cast( + ModelInfo, + { + "input_cost_per_token": 1e-6, + "output_cost_per_token": 2e-6, + "input_cost_per_token_above_200k_tokens": 3e-6, + "output_cost_per_token_above_200k_tokens": 4e-6, + "off_peak_pricing": { + "hours_utc": "16:30-00:30", + "input_cost_per_token": 5e-7, + "output_cost_per_token": 1e-6, + }, + }, + ) + usage = Usage(prompt_tokens=250000, completion_tokens=250000, total_tokens=500000) + + off_peak = _get_token_base_cost(model_info, usage, current_time=datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc)) + assert off_peak[0] == 5e-7 + assert off_peak[1] == 1e-6 + + peak = _get_token_base_cost(model_info, usage, current_time=datetime(2026, 1, 1, 12, 0, tzinfo=timezone.utc)) + assert peak[0] == 3e-6 + assert peak[1] == 4e-6 + + +def test_get_model_info_propagates_off_peak_fields(): + model_name = "test-off-peak-model" + off_peak_pricing = { + "hours_utc": "16:30-00:30", + "input_cost_per_token": 5e-7, + "output_cost_per_token": 1e-6, + "cache_read_input_token_cost": 5e-8, + } + litellm.register_model( + { + model_name: { + "litellm_provider": "openai", + "mode": "chat", + "input_cost_per_token": 1e-6, + "output_cost_per_token": 2e-6, + "off_peak_pricing": off_peak_pricing, + } + } + ) + info = litellm.get_model_info(model=model_name) + assert info["off_peak_pricing"] == off_peak_pricing + + +def test_get_token_base_cost_off_peak_wins_over_tiered_pricing(): + """Tiered pricing resolves base rates on its own path and returns early, so off-peak has to + be applied there too or a model carrying both would silently bill the tier rate all day.""" + from datetime import datetime, timezone + + model_name = "litellm-test-off-peak-tiered" + litellm.register_model( + { + model_name: { + "litellm_provider": "openai", + "mode": "chat", + "tiered_pricing": [ + {"range": [0, 128000], "input_cost_per_token": 3e-6, "output_cost_per_token": 6e-6}, + ], + "off_peak_pricing": { + "hours_utc": "16:30-00:30", + "input_cost_per_token": 5e-7, + "output_cost_per_token": 1e-6, + }, + } + } + ) + info = litellm.get_model_info(model=model_name) + usage = Usage(prompt_tokens=1_000, completion_tokens=100, total_tokens=1_100) + + inside = _get_token_base_cost(info, usage, current_time=datetime(2026, 1, 1, 18, 0, tzinfo=timezone.utc)) + assert inside[:2] == (5e-7, 1e-6) + + outside = _get_token_base_cost(info, usage, current_time=datetime(2026, 1, 1, 12, 0, tzinfo=timezone.utc)) + assert outside[:2] == (3e-6, 6e-6) + + def test_generic_cost_per_token_gpt54_above_272k_tokens(_local_model_cost_map): """GPT-5.4/5.4-pro: prompts >272K input tokens priced at 2x input, 1.5x output.""" model = "gpt-5.4" diff --git a/tests/test_litellm/litellm_core_utils/prompt_templates/test_litellm_core_utils_prompt_templates_factory.py b/tests/test_litellm/litellm_core_utils/prompt_templates/test_litellm_core_utils_prompt_templates_factory.py index 72d26f31c60..64c96b575c5 100644 --- a/tests/test_litellm/litellm_core_utils/prompt_templates/test_litellm_core_utils_prompt_templates_factory.py +++ b/tests/test_litellm/litellm_core_utils/prompt_templates/test_litellm_core_utils_prompt_templates_factory.py @@ -3627,3 +3627,67 @@ def test_convert_gemini_tool_call_result_answers_tool_reference_only_result(): ) assert result == {"function_response": {"name": "ToolSearch", "response": {"content": ""}}} + + +def test_convert_to_anthropic_tool_invoke_degrades_unpaired_server_tool_use(): + """A replayed srvtoolu_ call whose server tool result is not available + (e.g. the Responses bridge replays items without provider_specific_fields) + must become a plain client tool_use so the client's tool_result can pair + with it. A dangling server_tool_use makes Anthropic 400 the request with + "unexpected `tool_use_id` found in `tool_result` blocks".""" + from litellm.litellm_core_utils.prompt_templates.factory import convert_to_anthropic_tool_invoke + + result = convert_to_anthropic_tool_invoke( + tool_calls=[ + { + "id": "srvtoolu_01Unpaired", + "type": "function", + "function": {"name": "web_search", "arguments": '{"query": "zig version"}'}, + } + ], + web_search_results=None, + tool_results=None, + ) + + assert result == [ + { + "type": "tool_use", + "id": "srvtoolu_01Unpaired", + "name": "web_search", + "input": {"query": "zig version"}, + } + ] + + +def test_convert_to_anthropic_tool_invoke_keeps_paired_server_tool_use(): + """When the paired server tool result is available, the srvtoolu_ call is + still reconstructed as server_tool_use followed by its result block.""" + from litellm.litellm_core_utils.prompt_templates.factory import convert_to_anthropic_tool_invoke + + server_result = { + "type": "web_search_tool_result", + "tool_use_id": "srvtoolu_01Paired", + "content": [{"type": "web_search_result", "url": "https://ziglang.org", "title": "Zig"}], + } + + result = convert_to_anthropic_tool_invoke( + tool_calls=[ + { + "id": "srvtoolu_01Paired", + "type": "function", + "function": {"name": "web_search", "arguments": '{"query": "zig version"}'}, + } + ], + web_search_results=[server_result], + tool_results=None, + ) + + assert result == [ + { + "type": "server_tool_use", + "id": "srvtoolu_01Paired", + "name": "web_search", + "input": {"query": "zig version"}, + }, + server_result, + ] diff --git a/tests/test_litellm/litellm_core_utils/test_fallback_generalizations.py b/tests/test_litellm/litellm_core_utils/test_fallback_generalizations.py index 0587628e2fe..b1e8163b91d 100644 --- a/tests/test_litellm/litellm_core_utils/test_fallback_generalizations.py +++ b/tests/test_litellm/litellm_core_utils/test_fallback_generalizations.py @@ -366,7 +366,7 @@ def test_shipped_rules_stack_adaptive_and_mid_conversation_flags(shipped_cost_ma def test_shipped_rules_flag_unmapped_fable_as_always_on_thinking(shipped_cost_map): """An unmapped Fable/Mythos id picks up ``thinking_always_on`` from the claude-always-on-thinking rule, while other unmapped Claudes stay unflagged.""" - model = "claude-fable-5-1" + model = "claude-fable-6-1" assert model not in litellm.model_cost info = litellm.get_model_info(model, custom_llm_provider="anthropic") assert info["thinking_always_on"] is True @@ -419,7 +419,7 @@ def test_shipped_rules_cover_new_families_like_fable_at_5_plus(shipped_cost_map) """Both version gates accept any claude-- id at major 5 or higher, bare major or major-minor, so a new family shaped like claude-fable-5 gets adaptive thinking and mid-conversation system support without a cost-map entry.""" - model = "claude-fable-5-1" + model = "claude-fable-6-1" assert model not in litellm.model_cost info = litellm.get_model_info(model, custom_llm_provider="anthropic") assert info["supports_mid_conversation_system"] is True diff --git a/tests/test_litellm/litellm_core_utils/test_streaming_handler.py b/tests/test_litellm/litellm_core_utils/test_streaming_handler.py index 7f54fbfb4c2..4c99bce2b0f 100644 --- a/tests/test_litellm/litellm_core_utils/test_streaming_handler.py +++ b/tests/test_litellm/litellm_core_utils/test_streaming_handler.py @@ -4692,3 +4692,82 @@ async def test_async_stream_assembled_response_keeps_vertex_traffic_type(logging assembled = litellm.stream_chunk_builder(chunks=received, messages=[{"role": "user", "content": "hi"}]) assert assembled is not None assert assembled._hidden_params["provider_specific_fields"]["traffic_type"] == "ON_DEMAND_FLEX" + + +class TestStableStreamingResponseId: + """ + All chunks of one streamed response must share the same top-level id + (OpenAI streaming contract). Providers streaming via GenericStreamingChunk + (e.g. GigaChat) do not propagate an upstream response id, so + CustomStreamWrapper must pin the id from the first chunk it creates, + mirroring the existing `created` pinning (issue #11437). + + Clients such as goose merge streamed deltas into one assistant message by + chunk id; per-chunk ids split a single reply into many messages. + """ + + def test_generic_chunks_share_one_id(self): + def _generic_chunks(): + return iter( + [ + { + "text": "Hello", + "tool_use": None, + "is_finished": False, + "finish_reason": "", + "usage": None, + "index": 0, + }, + { + "text": " world", + "tool_use": None, + "is_finished": False, + "finish_reason": "", + "usage": None, + "index": 0, + }, + { + "text": "", + "tool_use": None, + "is_finished": True, + "finish_reason": "stop", + "usage": { + "prompt_tokens": 1, + "completion_tokens": 2, + "total_tokens": 3, + }, + "index": 0, + }, + ] + ) + + wrapper = CustomStreamWrapper( + completion_stream=_generic_chunks(), + model="gigachat/GigaChat-2-Max", + logging_obj=MagicMock(), + custom_llm_provider="gigachat", + ) + ids = [chunk.id for chunk in wrapper if chunk.id] + assert ids, "no chunks emitted" + assert len(set(ids)) == 1, f"chunk ids differ across one stream: {ids}" + + def test_creator_pins_id_from_first_chunk(self): + wrapper = CustomStreamWrapper( + completion_stream=iter([]), + model="gigachat/GigaChat-2-Max", + logging_obj=MagicMock(), + custom_llm_provider="gigachat", + ) + first = wrapper.model_response_creator() + assert wrapper.response_id == first.id + assert wrapper.model_response_creator().id == first.id + + def test_provider_supplied_id_still_wins(self): + wrapper = CustomStreamWrapper( + completion_stream=iter([]), + model="gigachat/GigaChat-2-Max", + logging_obj=MagicMock(), + custom_llm_provider="gigachat", + ) + wrapper.response_id = "chatcmpl-from-provider" + assert wrapper.model_response_creator().id == "chatcmpl-from-provider" diff --git a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py index 25e2c3cda80..2a955845861 100644 --- a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py +++ b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py @@ -6176,9 +6176,10 @@ def test_is_anthropic_usage_object_rejects_responses_api_usage(): [ # always-on-thinking models reject thinking.type=disabled with a 400 ("claude-fable-5", True), + ("claude-fable-5-1", True), ("claude-mythos-5", True), # unmapped future family member -> claude-always-on-thinking fallback rule - ("claude-fable-5-1", True), + ("claude-fable-6-1", True), # adaptive-capable models that ACCEPT disabled must keep it verbatim ("claude-opus-5", False), ("claude-sonnet-5", False), @@ -6207,3 +6208,149 @@ def test_disabled_thinking_omitted_only_for_always_on_models( assert "thinking" not in request else: assert request["thinking"] == {"type": "disabled"} + + +@pytest.mark.parametrize( + "tool_choice", + ["required", {"type": "required"}, {"type": "function", "function": {"name": "get_weather"}}], +) +def test_forced_tool_choice_raises_clean_error_on_fable_5_1_without_drop_params( + local_model_cost_map, tool_choice, monkeypatch +): + """Fable 5.1 400s on tool_choice type any/tool (thinking is always on and a + forced call would skip it); without drop_params the caller gets a clean + client-side 400 that explains the workaround, not a provider error.""" + monkeypatch.setattr(litellm, "drop_params", False) + config = AnthropicConfig() + + with pytest.raises(litellm.utils.UnsupportedParamsError, match="forced tool use"): + config.map_openai_params( + non_default_params={"tool_choice": tool_choice}, + optional_params={}, + model="claude-fable-5-1", + drop_params=False, + ) + + +@pytest.mark.parametrize( + "tool_choice", + ["required", {"type": "required"}, {"type": "function", "function": {"name": "get_weather"}}], +) +def test_forced_tool_choice_downgraded_to_auto_on_fable_5_1_with_drop_params( + local_model_cost_map, tool_choice +): + config = AnthropicConfig() + + result = config.map_openai_params( + non_default_params={"tool_choice": tool_choice}, + optional_params={}, + model="claude-fable-5-1", + drop_params=True, + ) + + assert result["tool_choice"] == {"type": "auto"} + + +def test_forced_tool_choice_downgrade_keeps_parallel_tool_calls_flag(local_model_cost_map): + config = AnthropicConfig() + + result = config.map_openai_params( + non_default_params={"tool_choice": "required", "parallel_tool_calls": False}, + optional_params={}, + model="claude-fable-5-1", + drop_params=True, + ) + + assert result["tool_choice"] == {"type": "auto", "disable_parallel_tool_use": True} + + +@pytest.mark.parametrize("tool_choice, expected_type", [("auto", "auto"), ("none", "none")]) +def test_unforced_tool_choice_forwarded_on_fable_5_1( + local_model_cost_map, tool_choice, expected_type, monkeypatch +): + monkeypatch.setattr(litellm, "drop_params", False) + config = AnthropicConfig() + + result = config.map_openai_params( + non_default_params={"tool_choice": tool_choice}, + optional_params={}, + model="claude-fable-5-1", + drop_params=False, + ) + + assert result["tool_choice"]["type"] == expected_type + + +@pytest.mark.parametrize("model", ["claude-fable-5", "claude-opus-5", "claude-sonnet-5"]) +def test_forced_tool_choice_forwarded_on_models_that_support_it( + local_model_cost_map, model, monkeypatch +): + monkeypatch.setattr(litellm, "drop_params", False) + config = AnthropicConfig() + + result = config.map_openai_params( + non_default_params={"tool_choice": "required"}, + optional_params={}, + model=model, + drop_params=True, + ) + + assert result["tool_choice"] == {"type": "any"} + + +def test_forced_tool_choice_gating_driven_by_model_map_flag(local_model_cost_map, monkeypatch): + """The gate must read ``supports_forced_tool_use`` from the model map, not + the model name: a flagged entry gates a model whose name says nothing.""" + monkeypatch.setitem(litellm.model_cost, "claude-zeta-9", {"supports_forced_tool_use": False}) + config = AnthropicConfig() + + result = config.map_openai_params( + non_default_params={"tool_choice": "required"}, + optional_params={}, + model="claude-zeta-9", + drop_params=True, + ) + + assert result["tool_choice"] == {"type": "auto"} + + +def test_anthropic_drop_params_keeps_format_only_output_config(monkeypatch): + """``drop_params=True`` must not consume ``output_config.format``: the drop + gate is an effort gate and ``format`` is a structured-output field.""" + monkeypatch.setattr(litellm, "drop_params", True) + config = AnthropicConfig() + schema_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"z": {"type": "integer"}}}, + } + + result = config.transform_request( + model="claude-3-haiku-20240307", + messages=[{"role": "user", "content": "Hello"}], + optional_params={"output_config": {"format": schema_format}}, + litellm_params={}, + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} + + +def test_anthropic_drop_params_reduces_mixed_output_config_to_format(monkeypatch): + """``drop_params=True`` drops the effort key on unsupported models but keeps + ``format`` so structured outputs still reach the provider.""" + monkeypatch.setattr(litellm, "drop_params", True) + config = AnthropicConfig() + schema_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"z": {"type": "integer"}}}, + } + + result = config.transform_request( + model="claude-3-haiku-20240307", + messages=[{"role": "user", "content": "Hello"}], + optional_params={"output_config": {"effort": "low", "format": schema_format}}, + litellm_params={}, + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} diff --git a/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py b/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py index cea299280f8..a122d97a0f0 100644 --- a/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py +++ b/tests/test_litellm/llms/bedrock/chat/invoke_transformations/test_bedrock_chat_invoke_transformations_anthropic_claude3_transformation.py @@ -428,30 +428,58 @@ def test_output_config_forwarded_for_bedrock_chat_invoke_request(): def test_output_config_format_converted_for_bedrock_chat_invoke_request(): - """Bedrock Invoke chat path consumes ``output_config.format`` before forwarding.""" + """Bedrock Invoke chat path inlines ``output_config.format`` for models + without native structured-output support and keeps the effort key.""" config = AmazonAnthropicClaudeConfig() schema = { "type": "object", "properties": {"answer": {"type": "string"}}, } - result = config.transform_request( + with patch( # test-quality-ok: pin non-native path + "litellm.llms.bedrock.common_utils._bedrock_model_supports", + side_effect=lambda _model, key: key == "supports_output_config", + ): + result = config.transform_request( + model="anthropic.claude-opus-4-7", + messages=[{"role": "user", "content": "test"}], + optional_params={ + "max_tokens": 100, + "output_config": { + "effort": "xhigh", + "format": {"type": "json_schema", "schema": schema}, + }, + }, + litellm_params={}, + headers={}, + ) + + assert result.get("output_config") == {"effort": "xhigh"} + last_content = result["messages"][0]["content"] + assert json.loads(last_content[-1]["text"]) == schema + + +def test_output_config_format_forwarded_for_bedrock_chat_invoke_request(): + """Bedrock Invoke chat path forwards ``output_config.format`` alongside effort + for models with native structured-output support (Claude Opus 4.7).""" + schema_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"answer": {"type": "string"}}}, + } + + result = AmazonAnthropicClaudeConfig().transform_request( model="anthropic.claude-opus-4-7", messages=[{"role": "user", "content": "test"}], optional_params={ "max_tokens": 100, - "output_config": { - "effort": "xhigh", - "format": {"type": "json_schema", "schema": schema}, - }, + "output_config": {"effort": "xhigh", "format": schema_format}, }, litellm_params={}, headers={}, ) - assert result.get("output_config") == {"effort": "xhigh"} - last_content = result["messages"][0]["content"] - assert json.loads(last_content[-1]["text"]) == schema + assert result.get("output_config") == {"effort": "xhigh", "format": schema_format} + assert "answer" not in json.dumps(result["messages"]) @pytest.mark.parametrize( @@ -488,7 +516,7 @@ def test_bedrock_chat_invoke_checks_output_config_support_with_bedrock_provider( optional_params = {"max_tokens": 100, "output_config": {"effort": "high"}} with patch( - "litellm.llms.bedrock.chat.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ) as mock_supports_factory: result = config.transform_request( @@ -499,11 +527,7 @@ def test_bedrock_chat_invoke_checks_output_config_support_with_bedrock_provider( headers={}, ) - mock_supports_factory.assert_called_once_with( - model="us.anthropic.claude-opus-4-7", - custom_llm_provider="bedrock", - key="supports_output_config", - ) + mock_supports_factory.assert_called_once_with("us.anthropic.claude-opus-4-7", "supports_output_config") assert result["output_config"] == {"effort": "high"} @@ -542,3 +566,80 @@ def test_output_format_removed_from_bedrock_invoke_request(): assert ( "output_format" not in result ), f"output_format should be removed for Bedrock Invoke, got keys: {result.keys()}" + + +def test_bedrock_chat_invoke_forwards_output_config_format_natively(local_model_cost_map): + """Regression: ``output_config.format`` is forwarded verbatim on models Bedrock + enforces structured outputs for, instead of being inlined as prompt text.""" + import json + + config = AmazonAnthropicClaudeConfig() + schema_format = { + "type": "json_schema", + "schema": { + "type": "object", + "properties": {"zebra_count": {"type": "integer"}}, + "required": ["zebra_count"], + "additionalProperties": False, + }, + } + + result = config.transform_request( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + messages=[{"role": "user", "content": "say hello"}], + optional_params={ + "max_tokens": 100, + "output_config": {"format": schema_format}, + }, + litellm_params={}, + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} + assert "zebra_count" not in json.dumps(result["messages"]) + + +def test_bedrock_chat_invoke_drop_params_keeps_native_output_config_format(local_model_cost_map, monkeypatch): + """``drop_params=True`` must not eat ``output_config.format`` before the + native-forwarding router runs (Sonnet 4.5 has no effort flags).""" + import litellm + + monkeypatch.setattr(litellm, "drop_params", True) + schema_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"zebra_count": {"type": "integer"}}}, + } + + result = AmazonAnthropicClaudeConfig().transform_request( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + messages=[{"role": "user", "content": "say hello"}], + optional_params={"max_tokens": 100, "output_config": {"format": schema_format}}, + litellm_params={}, + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} + + +def test_bedrock_chat_invoke_drop_params_still_inlines_for_non_native(local_model_cost_map, monkeypatch): + """``drop_params=True`` on a model without native structured-output support + still reaches the inline-schema fallback instead of losing the schema.""" + import litellm + + monkeypatch.setattr(litellm, "drop_params", True) + schema = {"type": "object", "properties": {"zebra_count": {"type": "integer"}}} + + result = AmazonAnthropicClaudeConfig().transform_request( + model="anthropic.claude-3-haiku-20240307-v1:0", + messages=[{"role": "user", "content": "say hello"}], + optional_params={ + "max_tokens": 100, + "output_config": {"format": {"type": "json_schema", "schema": schema}}, + }, + litellm_params={}, + headers={}, + ) + + assert "output_config" not in result + last_content = result["messages"][-1]["content"] + assert json.loads(last_content[-1]["text"]) == schema diff --git a/tests/test_litellm/llms/bedrock/chat/test_converse_transformation.py b/tests/test_litellm/llms/bedrock/chat/test_converse_transformation.py index 63f895e1819..37ca801a7a7 100644 --- a/tests/test_litellm/llms/bedrock/chat/test_converse_transformation.py +++ b/tests/test_litellm/llms/bedrock/chat/test_converse_transformation.py @@ -6449,3 +6449,62 @@ def test_disabled_thinking_omitted_for_always_on_models_converse( assert "thinking" not in additional else: assert additional.get("thinking") == {"type": "disabled"} + +@pytest.mark.parametrize( + "model", + ["anthropic.claude-fable-5-1", "us.anthropic.claude-fable-5-1"], +) +@pytest.mark.parametrize( + "tool_choice", + ["required", {"type": "function", "function": {"name": "get_weather"}}], +) +def test_forced_tool_choice_downgraded_to_auto_on_fable_5_1_converse( + local_model_cost_map, model, tool_choice +): + config = AmazonConverseConfig() + + result = config.map_tool_choice_values( + model=model, tool_choice=tool_choice, drop_params=True + ) + + assert result == {"auto": {}} + + +@pytest.mark.parametrize( + "tool_choice", + ["required", {"type": "function", "function": {"name": "get_weather"}}], +) +def test_forced_tool_choice_raises_clean_error_on_fable_5_1_converse( + local_model_cost_map, tool_choice, monkeypatch +): + monkeypatch.setattr(litellm, "drop_params", False) + config = AmazonConverseConfig() + + with pytest.raises(litellm.utils.UnsupportedParamsError, match="forced tool use"): + config.map_tool_choice_values( + model="anthropic.claude-fable-5-1", tool_choice=tool_choice, drop_params=False + ) + + +@pytest.mark.parametrize("tool_choice", ["auto", "none"]) +def test_unforced_tool_choice_unaffected_on_fable_5_1_converse(local_model_cost_map, tool_choice): + config = AmazonConverseConfig() + + result = config.map_tool_choice_values( + model="anthropic.claude-fable-5-1", tool_choice=tool_choice, drop_params=True + ) + + assert result == ({"auto": {}} if tool_choice == "auto" else None) + + +def test_forced_tool_choice_forwarded_on_converse_models_that_support_it( + local_model_cost_map, monkeypatch +): + monkeypatch.setattr(litellm, "drop_params", False) + config = AmazonConverseConfig() + + result = config.map_tool_choice_values( + model="anthropic.claude-fable-5", tool_choice="required", drop_params=False + ) + + assert result == {"any": {}} diff --git a/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py b/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py index 8d07d38b1b6..09ebc1a3c95 100644 --- a/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py +++ b/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py @@ -935,7 +935,7 @@ def test_bedrock_messages_strips_output_config(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=False, ): result = cfg.transform_anthropic_messages_request( @@ -970,7 +970,7 @@ def test_bedrock_messages_preserves_output_config_for_claude_4_6(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -1003,7 +1003,7 @@ def test_bedrock_messages_checks_output_config_support_with_bedrock_provider(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ) as mock_supports_factory: result = cfg.transform_anthropic_messages_request( @@ -1014,11 +1014,7 @@ def test_bedrock_messages_checks_output_config_support_with_bedrock_provider(): headers={}, ) - mock_supports_factory.assert_called_with( - model="us.anthropic.claude-opus-4-7", - custom_llm_provider="bedrock", - key="supports_output_config", - ) + mock_supports_factory.assert_called_with("us.anthropic.claude-opus-4-7", "supports_output_config") assert result["output_config"] == {"effort": "high"} @@ -1038,7 +1034,7 @@ def test_bedrock_messages_forwards_output_config(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -1054,27 +1050,29 @@ def test_bedrock_messages_forwards_output_config(): def test_bedrock_messages_forwards_output_config_with_output_format(): - """``output_config`` is forwarded; ``output_format`` is converted to inline schema.""" + """Legacy ``output_format`` is forwarded as ``output_config.format`` on models + that support native structured outputs, alongside the effort key.""" from unittest.mock import patch from litellm.types.router import GenericLiteLLMParams cfg = AmazonAnthropicClaudeMessagesConfig() messages = [{"role": "user", "content": [{"type": "text", "text": "Hello"}]}] + schema_format = { + "type": "json_schema", + "schema": { + "type": "object", + "properties": {"answer": {"type": "string"}}, + }, + } optional_params = { "max_tokens": 4096, "output_config": {"effort": "low"}, - "output_format": { - "type": "json_schema", - "schema": { - "type": "object", - "properties": {"answer": {"type": "string"}}, - }, - }, + "output_format": schema_format, } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -1085,12 +1083,14 @@ def test_bedrock_messages_forwards_output_config_with_output_format(): headers={}, ) - assert result.get("output_config") == {"effort": "low"} + assert result.get("output_config") == {"effort": "low", "format": schema_format} assert "output_format" not in result + assert "answer" not in json.dumps(result["messages"]) def test_bedrock_messages_converts_output_config_format_to_inline_schema(): - """``output_config.format`` is consumed so Bedrock does not see an unknown nested key.""" + """Without native structured-output support, ``output_config.format`` falls back + to the inline schema so Bedrock does not see an unknown nested key.""" from unittest.mock import patch from litellm.types.router import GenericLiteLLMParams @@ -1110,8 +1110,8 @@ def test_bedrock_messages_converts_output_config_format_to_inline_schema(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", - return_value=True, + "litellm.llms.bedrock.common_utils._bedrock_model_supports", + side_effect=lambda _model, key: key == "supports_output_config", ): result = cfg.transform_anthropic_messages_request( model="anthropic.claude-opus-4-7", @@ -1146,7 +1146,7 @@ def test_bedrock_messages_normalizes_output_config_effort_for_opus( cfg = AmazonAnthropicClaudeMessagesConfig() with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -1184,8 +1184,8 @@ def test_bedrock_messages_does_not_mutate_callers_messages_when_embedding_schema } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", - return_value=True, + "litellm.llms.bedrock.common_utils._bedrock_model_supports", + side_effect=lambda _model, key: key == "supports_output_config", ): result = cfg.transform_anthropic_messages_request( model="anthropic.claude-opus-4-7", @@ -1229,7 +1229,7 @@ def test_bedrock_messages_does_not_mutate_callers_output_config(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): cfg.transform_anthropic_messages_request( @@ -1271,7 +1271,7 @@ def test_bedrock_messages_strips_output_config_with_output_format(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=False, ): result = cfg.transform_anthropic_messages_request( @@ -1332,7 +1332,7 @@ def test_bedrock_messages_drop_params_keeps_output_config_for_4_7(): litellm.drop_params = True try: with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -1375,7 +1375,7 @@ def test_bedrock_messages_maps_reasoning_effort_for_adaptive_model( } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -1482,7 +1482,7 @@ def test_bedrock_messages_explicit_output_config_wins_over_reasoning_effort(): } with patch( - "litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation._supports_factory", + "litellm.llms.bedrock.common_utils._bedrock_model_supports", return_value=True, ): result = cfg.transform_anthropic_messages_request( @@ -3104,3 +3104,149 @@ async def test_bedrock_sse_wrapper_dispatches_logging_on_client_disconnect(): break await asyncio.sleep(0.01) assert logging_obj.completion_start_time is not None + + +def test_bedrock_messages_forwards_output_config_format_natively(local_model_cost_map): + """Regression: on a model Bedrock enforces structured outputs for (Claude + Sonnet 4.5), ``output_config.format`` must be forwarded verbatim, not + silently rewritten into inline prompt text.""" + from litellm.types.router import GenericLiteLLMParams + + cfg = AmazonAnthropicClaudeMessagesConfig() + schema_format = { + "type": "json_schema", + "schema": { + "type": "object", + "properties": { + "zebra_count": {"type": "integer"}, + "is_tuesday": {"type": "boolean"}, + }, + "required": ["zebra_count", "is_tuesday"], + "additionalProperties": False, + }, + } + + result = cfg.transform_anthropic_messages_request( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + messages=[{"role": "user", "content": [{"type": "text", "text": "say hello"}]}], + anthropic_messages_optional_request_params={ + "max_tokens": 100, + "output_config": {"format": schema_format}, + }, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} + assert "zebra_count" not in json.dumps(result["messages"]) + + +def test_bedrock_messages_inlines_schema_for_claude_5(local_model_cost_map): + """Bedrock rejects ``output_config.format`` for the Claude 5 family, so the + schema falls back to the inline-text path instead of a deterministic 400.""" + from litellm.types.router import GenericLiteLLMParams + + cfg = AmazonAnthropicClaudeMessagesConfig() + schema = { + "type": "object", + "properties": {"zebra_count": {"type": "integer"}}, + } + + result = cfg.transform_anthropic_messages_request( + model="us.anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": [{"type": "text", "text": "say hello"}]}], + anthropic_messages_optional_request_params={ + "max_tokens": 100, + "output_config": {"format": {"type": "json_schema", "schema": schema}}, + }, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + + assert "output_config" not in result + last_content = result["messages"][-1]["content"] + assert json.loads(last_content[-1]["text"]) == schema + + +def test_bedrock_messages_legacy_output_format_wins_over_output_config_format(local_model_cost_map): + """When a request carries both schema forms, the legacy top-level + ``output_format`` keeps winning, matching the pre-existing precedence.""" + from litellm.types.router import GenericLiteLLMParams + + cfg = AmazonAnthropicClaudeMessagesConfig() + legacy_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"legacy_field": {"type": "string"}}}, + } + newer_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"newer_field": {"type": "string"}}}, + } + + result = cfg.transform_anthropic_messages_request( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + messages=[{"role": "user", "content": [{"type": "text", "text": "say hello"}]}], + anthropic_messages_optional_request_params={ + "max_tokens": 100, + "output_format": legacy_format, + "output_config": {"format": newer_format}, + }, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + + assert result.get("output_config") == {"format": legacy_format} + assert "output_format" not in result + assert "newer_field" not in json.dumps(result) + + +def test_bedrock_messages_drop_params_keeps_native_output_config_format(local_model_cost_map, monkeypatch): + """``drop_params=True`` must not strip a natively forwarded + ``output_config.format`` on models without effort support (Sonnet 4.5).""" + import litellm + from litellm.types.router import GenericLiteLLMParams + + monkeypatch.setattr(litellm, "drop_params", True) + cfg = AmazonAnthropicClaudeMessagesConfig() + schema_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"zebra_count": {"type": "integer"}}}, + } + + result = cfg.transform_anthropic_messages_request( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + messages=[{"role": "user", "content": [{"type": "text", "text": "say hello"}]}], + anthropic_messages_optional_request_params={ + "max_tokens": 100, + "output_config": {"format": schema_format}, + }, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} + + +def test_bedrock_messages_strips_effort_but_keeps_format_for_sonnet_4_5(local_model_cost_map): + """Sonnet 4.5 has native structured-output support but no effort support, so + a mixed ``output_config`` keeps ``format`` and drops ``effort``.""" + from litellm.types.router import GenericLiteLLMParams + + cfg = AmazonAnthropicClaudeMessagesConfig() + schema_format = { + "type": "json_schema", + "schema": {"type": "object", "properties": {"zebra_count": {"type": "integer"}}}, + } + + result = cfg.transform_anthropic_messages_request( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + messages=[{"role": "user", "content": [{"type": "text", "text": "say hello"}]}], + anthropic_messages_optional_request_params={ + "max_tokens": 4096, + "output_config": {"format": schema_format, "effort": "high"}, + }, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + + assert result.get("output_config") == {"format": schema_format} diff --git a/tests/test_litellm/llms/bedrock/test_bedrock_common_utils.py b/tests/test_litellm/llms/bedrock/test_bedrock_common_utils.py index afd5e83ca52..9302dc01abe 100644 --- a/tests/test_litellm/llms/bedrock/test_bedrock_common_utils.py +++ b/tests/test_litellm/llms/bedrock/test_bedrock_common_utils.py @@ -522,6 +522,47 @@ def test_merge_bedrock_aws_request_params_keeps_caller_credentials_without_stati assert merged["aws_region_name"] == "us-west-2" +def test_strip_unsupported_output_config_keeps_format_drops_effort(local_model_cost_map): + """On a model with neither effort flag, only the ``format`` key survives.""" + from litellm.llms.bedrock.common_utils import ( + strip_unsupported_bedrock_invoke_output_config_keys, + ) + + schema_format = {"type": "json_schema", "schema": {"type": "object"}} + body = {"output_config": {"effort": "high", "format": schema_format}} + + strip_unsupported_bedrock_invoke_output_config_keys( + model="anthropic.claude-3-haiku-20240307-v1:0", + request_body=body, + ) + + assert body["output_config"] == {"format": schema_format} + + +def test_apply_structured_output_prefers_legacy_output_format(local_model_cost_map): + """The legacy ``output_format`` wins over ``output_config.format`` when a + request carries both, matching the pre-existing precedence.""" + from litellm.llms.bedrock.common_utils import ( + apply_bedrock_invoke_structured_output, + ) + + legacy = {"type": "json_schema", "schema": {"type": "object", "properties": {"a": {"type": "string"}}}} + newer = {"type": "json_schema", "schema": {"type": "object", "properties": {"b": {"type": "string"}}}} + body = { + "messages": [{"role": "user", "content": "hi"}], + "output_format": legacy, + "output_config": {"format": newer}, + } + + apply_bedrock_invoke_structured_output( + model="us.anthropic.claude-sonnet-4-5-20250929-v1:0", + request_body=body, + ) + + assert body["output_config"] == {"format": legacy} + assert "output_format" not in body + + def test_sign_aws_request_assumes_role_with_external_id(monkeypatch): """A trust policy requiring sts:ExternalId must be satisfied when signing batch API requests.""" import datetime diff --git a/tests/test_litellm/llms/dashscope/test_qwen_brand_aliases.py b/tests/test_litellm/llms/dashscope/test_qwen_brand_aliases.py new file mode 100644 index 00000000000..064d9d58f0c --- /dev/null +++ b/tests/test_litellm/llms/dashscope/test_qwen_brand_aliases.py @@ -0,0 +1,331 @@ +import math + +import pytest + +import litellm +from litellm import completion, get_llm_provider +from litellm.llms.dashscope.chat.transformation import DashScopeChatConfig +from litellm.llms.dashscope.cost_calculator import ( + cost_per_token as dashscope_cost_per_token, +) +from litellm.llms.dashscope.embed.transformation import DashScopeEmbeddingConfig +from litellm.llms.dashscope.image_generation.transformation import ( + DashScopeImageGenerationConfig, +) +from litellm.llms.dashscope.qwen_ai_platform import ( + QWEN_AI_PLATFORM_API_BASE, + QWEN_AI_PLATFORM_IMAGE_API_BASE, + QWEN_AI_PLATFORM_RERANK_API_BASE, + QwenAIPlatformChatConfig, + QwenAIPlatformEmbeddingConfig, + QwenAIPlatformImageGenerationConfig, + QwenAIPlatformRerankConfig, +) +from litellm.llms.dashscope.qwencloud import ( + QWENCLOUD_API_BASE, + QWENCLOUD_IMAGE_API_BASE, + QWENCLOUD_RERANK_API_BASE, + QwenCloudChatConfig, + QwenCloudEmbeddingConfig, + QwenCloudImageGenerationConfig, + QwenCloudRerankConfig, +) +from litellm.llms.dashscope.rerank.transformation import DashScopeRerankConfig +from litellm.types.utils import LlmProviders, Usage +from litellm.utils import ProviderConfigManager + +DASHSCOPE_FAMILY_ENV_VARS = [ + "DASHSCOPE_API_KEY", + "DASHSCOPE_API_BASE", + "DASHSCOPE_API_BASE_RERANK", + "DASHSCOPE_API_BASE_IMAGE", + "QWENCLOUD_API_KEY", + "QWENCLOUD_API_BASE", + "QWENCLOUD_API_BASE_RERANK", + "QWENCLOUD_API_BASE_IMAGE", + "QWEN_AI_PLATFORM_API_KEY", + "QWEN_AI_PLATFORM_API_BASE", + "QWEN_AI_PLATFORM_API_BASE_RERANK", + "QWEN_AI_PLATFORM_API_BASE_IMAGE", +] + +BRAND_CASES = [ + pytest.param( + { + "provider": "qwencloud", + "enum": LlmProviders.QWENCLOUD, + "key_env": "QWENCLOUD_API_KEY", + "base_env": "QWENCLOUD_API_BASE", + "default_base": QWENCLOUD_API_BASE, + "default_rerank_base": QWENCLOUD_RERANK_API_BASE, + "default_image_base": QWENCLOUD_IMAGE_API_BASE, + "chat_config": QwenCloudChatConfig, + "embedding_config": QwenCloudEmbeddingConfig, + "rerank_config": QwenCloudRerankConfig, + "image_config": QwenCloudImageGenerationConfig, + }, + id="qwencloud", + ), + pytest.param( + { + "provider": "qwen_ai_platform", + "enum": LlmProviders.QWEN_AI_PLATFORM, + "key_env": "QWEN_AI_PLATFORM_API_KEY", + "base_env": "QWEN_AI_PLATFORM_API_BASE", + "default_base": QWEN_AI_PLATFORM_API_BASE, + "default_rerank_base": QWEN_AI_PLATFORM_RERANK_API_BASE, + "default_image_base": QWEN_AI_PLATFORM_IMAGE_API_BASE, + "chat_config": QwenAIPlatformChatConfig, + "embedding_config": QwenAIPlatformEmbeddingConfig, + "rerank_config": QwenAIPlatformRerankConfig, + "image_config": QwenAIPlatformImageGenerationConfig, + }, + id="qwen_ai_platform", + ), +] + + +@pytest.fixture(autouse=True) +def clear_dashscope_family_env(monkeypatch): + for env_var in DASHSCOPE_FAMILY_ENV_VARS: + monkeypatch.delenv(env_var, raising=False) + + +class TestQwenBrandProviderResolution: + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_get_llm_provider_resolves_brand_default_base(self, brand): + model, provider, api_key, api_base = get_llm_provider(f"{brand['provider']}/qwen-max", api_key="sk-explicit") + assert model == "qwen-max" + assert provider == brand["provider"] + assert api_key == "sk-explicit" + assert api_base == brand["default_base"] + + def test_dashscope_resolution_unchanged(self): + model, provider, api_key, api_base = get_llm_provider("dashscope/qwen-max", api_key="sk-explicit") + assert model == "qwen-max" + assert provider == "dashscope" + assert api_base == "https://dashscope.aliyuncs.com/compatible-mode/v1" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_brand_env_key_wins_over_dashscope_key(self, monkeypatch, brand): + monkeypatch.setenv(brand["key_env"], "sk-brand") + monkeypatch.setenv("DASHSCOPE_API_KEY", "sk-dashscope") + _, _, api_key, _ = get_llm_provider(f"{brand['provider']}/qwen-max") + assert api_key == "sk-brand" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_dashscope_key_is_fallback(self, monkeypatch, brand): + monkeypatch.setenv("DASHSCOPE_API_KEY", "sk-dashscope") + _, _, api_key, _ = get_llm_provider(f"{brand['provider']}/qwen-max") + assert api_key == "sk-dashscope" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_dashscope_api_base_does_not_leak_into_brand(self, monkeypatch, brand): + monkeypatch.setenv("DASHSCOPE_API_BASE", "https://legacy.example.com/v1") + _, _, _, api_base = get_llm_provider(f"{brand['provider']}/qwen-max", api_key="sk-explicit") + assert api_base == brand["default_base"] + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_brand_api_base_env_wins(self, monkeypatch, brand): + monkeypatch.setenv(brand["base_env"], "https://brand.example.com/v1") + _, _, _, api_base = get_llm_provider(f"{brand['provider']}/qwen-max", api_key="sk-explicit") + assert api_base == "https://brand.example.com/v1" + + +class TestQwenBrandConfigDispatch: + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_chat_config(self, brand): + config = ProviderConfigManager.get_provider_chat_config("qwen-max", brand["enum"]) + assert isinstance(config, brand["chat_config"]) + assert isinstance(config, DashScopeChatConfig) + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_embedding_config(self, brand): + config = ProviderConfigManager.get_provider_embedding_config(model="text-embedding-v3", provider=brand["enum"]) + assert isinstance(config, brand["embedding_config"]) + assert isinstance(config, DashScopeEmbeddingConfig) + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_rerank_config(self, brand): + config = ProviderConfigManager.get_provider_rerank_config( + model="gte-rerank-v2", + provider=brand["enum"], + api_base=None, + present_version_params=[], + ) + assert isinstance(config, brand["rerank_config"]) + assert isinstance(config, DashScopeRerankConfig) + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_image_generation_config(self, brand): + config = ProviderConfigManager.get_provider_image_generation_config(model="qwen-image", provider=brand["enum"]) + assert isinstance(config, brand["image_config"]) + assert isinstance(config, DashScopeImageGenerationConfig) + + +class TestQwenBrandDefaultUrls: + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_chat_complete_url(self, brand): + url = brand["chat_config"]().get_complete_url( + api_base=None, + api_key="sk-test", + model="qwen-max", + optional_params={}, + litellm_params={}, + ) + assert url == f"{brand['default_base']}/chat/completions" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_embedding_complete_url(self, brand): + url = brand["embedding_config"]().get_complete_url( + api_base=None, + api_key="sk-test", + model="text-embedding-v3", + optional_params={}, + litellm_params={}, + ) + assert url == f"{brand['default_base']}/embeddings" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_embedding_ignores_dashscope_api_base(self, monkeypatch, brand): + monkeypatch.setenv("DASHSCOPE_API_BASE", "https://legacy.example.com/v1") + url = brand["embedding_config"]().get_complete_url( + api_base=None, + api_key="sk-test", + model="text-embedding-v3", + optional_params={}, + litellm_params={}, + ) + assert url == f"{brand['default_base']}/embeddings" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_rerank_complete_url(self, brand): + url = brand["rerank_config"]().get_complete_url(api_base=None, model="gte-rerank-v2") + assert url == brand["default_rerank_base"] + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_rerank_env_override(self, monkeypatch, brand): + monkeypatch.setenv(f"{brand['base_env']}_RERANK", "https://rerank.example.com/v1/reranks") + url = brand["rerank_config"]().get_complete_url(api_base=None, model="gte-rerank-v2") + assert url == "https://rerank.example.com/v1/reranks" + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_image_generation_complete_url(self, brand): + url = brand["image_config"]().get_complete_url( + api_base=None, + api_key="sk-test", + model="qwen-image", + optional_params={}, + litellm_params={}, + ) + assert url == brand["default_image_base"] + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_image_generation_ignores_chat_compatible_api_base(self, brand): + url = brand["image_config"]().get_complete_url( + api_base=brand["default_base"], + api_key="sk-test", + model="qwen-image", + optional_params={}, + litellm_params={}, + ) + assert url == brand["default_image_base"] + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_validate_environment_requires_key(self, brand): + with pytest.raises(ValueError, match="DASHSCOPE_API_KEY"): + brand["embedding_config"]().validate_environment( + headers={}, + model="text-embedding-v3", + messages=[], + optional_params={}, + litellm_params={}, + api_key=None, + api_base=None, + ) + + +class TestQwenBrandCostParity: + @pytest.fixture(autouse=True) + def setup_model_cost_map(self, monkeypatch): + monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True") + monkeypatch.setattr(litellm, "model_cost", litellm.get_model_cost_map(url="")) + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_get_model_info(self, brand): + model_info = litellm.get_model_info(f"{brand['provider']}/qwen-max") + dashscope_info = litellm.get_model_info("dashscope/qwen-max") + assert model_info["litellm_provider"] == brand["provider"] + assert model_info["input_cost_per_token"] == dashscope_info["input_cost_per_token"] + assert model_info["output_cost_per_token"] == dashscope_info["output_cost_per_token"] + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_flat_pricing_matches_dashscope(self, brand): + usage = Usage(prompt_tokens=1000, completion_tokens=500) + brand_costs = dashscope_cost_per_token(model="qwen-max", usage=usage, custom_llm_provider=brand["provider"]) + dashscope_costs = dashscope_cost_per_token(model="qwen-max", usage=usage) + assert brand_costs == dashscope_costs + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_tiered_pricing_matches_dashscope(self, brand): + usage = Usage(prompt_tokens=300000, completion_tokens=300000) + brand_costs = dashscope_cost_per_token(model="qwen-flash", usage=usage, custom_llm_provider=brand["provider"]) + dashscope_costs = dashscope_cost_per_token(model="qwen-flash", usage=usage) + assert brand_costs == dashscope_costs + tier_2 = litellm.get_model_info(f"{brand['provider']}/qwen-flash")["tiered_pricing"][1] + assert math.isclose(brand_costs[0], 300000 * tier_2["input_cost_per_token"], rel_tol=1e-10) + + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_public_cost_per_token_routes_to_dashscope_calculator(self, brand): + brand_costs = litellm.cost_per_token( + model=f"{brand['provider']}/qwen-max", + prompt_tokens=1000, + completion_tokens=500, + custom_llm_provider=brand["provider"], + ) + dashscope_costs = litellm.cost_per_token( + model="dashscope/qwen-max", + prompt_tokens=1000, + completion_tokens=500, + custom_llm_provider="dashscope", + ) + assert brand_costs == dashscope_costs + + +class TestQwenBrandCompletionMock: + @pytest.mark.respx() + @pytest.mark.parametrize("brand", BRAND_CASES) + def test_completion_hits_brand_default_host(self, respx_mock, brand, monkeypatch): + monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) + respx_mock.post(f"{brand['default_base']}/chat/completions").respond( + json={ + "id": "chatcmpl-123", + "object": "chat.completion", + "created": 1677652288, + "model": "qwen-turbo", + "choices": [ + { + "index": 0, + "message": {"role": "assistant", "content": "Hey from LiteLLM!"}, + "finish_reason": "stop", + } + ], + "usage": { + "prompt_tokens": 9, + "completion_tokens": 12, + "total_tokens": 21, + }, + }, + status_code=200, + ) + + response = completion( + model=f"{brand['provider']}/qwen-turbo", + messages=[{"role": "user", "content": "say hey from LiteLLM"}], + api_key="fake-brand-key", + ) + + assert response.choices[0].message.content == "Hey from LiteLLM!" + request = respx_mock.calls[0].request + assert request.url == f"{brand['default_base']}/chat/completions" + assert request.headers["Authorization"] == "Bearer fake-brand-key" diff --git a/tests/test_litellm/llms/databricks/test_databricks_cost_calculator.py b/tests/test_litellm/llms/databricks/test_databricks_cost_calculator.py index 29ad8ee4b6e..e72642f7a04 100644 --- a/tests/test_litellm/llms/databricks/test_databricks_cost_calculator.py +++ b/tests/test_litellm/llms/databricks/test_databricks_cost_calculator.py @@ -62,6 +62,8 @@ PUBLISHED_DBU_PER_MILLION: Final = { "databricks/databricks-gemini-2-5-pro": ("22.321", "178.571", "22.321", "2.232"), "databricks/databricks-gemini-2-5-flash": ("5.357", "44.643", "5.357", "0.536"), "databricks/databricks-kimi-k3": ("42.857", "214.286", "42.857", "4.286"), + "databricks/databricks-deepseek-v4-flash-0731": ("2.000", "4.000", "2.000", "0.400"), + "databricks/databricks-deepseek-v4-pro-0813": ("18.857", "56.571", "18.857", "1.886"), "databricks/databricks-glm-5-2": ("20.000", "62.857", "20.000", "3.714"), } PROMOTIONAL_DISCOUNT: Final = 0.80 diff --git a/tests/test_litellm/llms/mistral/ocr/test_mistral_ocr_cost.py b/tests/test_litellm/llms/mistral/ocr/test_mistral_ocr_cost.py index 890df597933..a0e1616d4b2 100644 --- a/tests/test_litellm/llms/mistral/ocr/test_mistral_ocr_cost.py +++ b/tests/test_litellm/llms/mistral/ocr/test_mistral_ocr_cost.py @@ -24,6 +24,9 @@ OCR3_MODEL = "mistral/mistral-ocr-2512" OCR3_COST_PER_PAGE = 0.002 OCR3_ANNOTATION_COST_PER_PAGE = 0.003 +AZURE_DOC_AI_MODEL = "azure_ai/mistral-document-ai-2512" +AZURE_DOC_AI_COST_PER_PAGE = 0.003 + def _ocr_response(model: str, pages_processed: int) -> OCRResponse: return OCRResponse( @@ -33,6 +36,14 @@ def _ocr_response(model: str, pages_processed: int) -> OCRResponse: ) +def _annotated_ocr_response(model: str, pages_processed: int | None, annotation_pages: int) -> OCRResponse: + return OCRResponse( + pages=[], + model=model, + usage_info=OCRUsageInfo(pages_processed=pages_processed, pages_processed_annotation=annotation_pages), + ) + + @pytest.mark.parametrize("model", ["mistral-ocr-4-0", "mistral-ocr-latest"]) def test_model_info_ocr4_price(model: str) -> None: info = litellm.get_model_info(model=f"mistral/{model}", custom_llm_provider="mistral") @@ -79,3 +90,46 @@ def test_ocr3_cost_scales_with_pages(local_model_cost_map, pages_processed: int) call_type="ocr", ) assert cost == pytest.approx(OCR3_COST_PER_PAGE * pages_processed) + + +def test_ocr3_bills_ocr_and_annotation_pages_at_their_own_rates(local_model_cost_map) -> None: + cost = completion_cost( + completion_response=_annotated_ocr_response("mistral-ocr-2512", 2, 3), + model=OCR3_MODEL, + custom_llm_provider="mistral", + call_type="ocr", + ) + assert cost == pytest.approx(2 * OCR3_COST_PER_PAGE + 3 * OCR3_ANNOTATION_COST_PER_PAGE) + + +def test_ocr3_bills_annotation_only_response(local_model_cost_map) -> None: + cost = completion_cost( + completion_response=_annotated_ocr_response("mistral-ocr-2512", 0, 3), + model=OCR3_MODEL, + custom_llm_provider="mistral", + call_type="ocr", + ) + assert cost == pytest.approx(3 * OCR3_ANNOTATION_COST_PER_PAGE) + + +def test_ocr3_bills_annotation_pages_when_pages_processed_missing(local_model_cost_map) -> None: + cost = completion_cost( + completion_response=_annotated_ocr_response("mistral-ocr-2512", None, 4), + model=OCR3_MODEL, + custom_llm_provider="mistral", + call_type="ocr", + ) + assert cost == pytest.approx(4 * OCR3_ANNOTATION_COST_PER_PAGE) + + +def test_azure_doc_ai_annotation_pages_fall_back_to_ocr_rate(local_model_cost_map) -> None: + info = litellm.get_model_info(model=AZURE_DOC_AI_MODEL, custom_llm_provider="azure_ai") + assert info.get("annotation_cost_per_page") is None + assert info["ocr_cost_per_page"] == AZURE_DOC_AI_COST_PER_PAGE + cost = completion_cost( + completion_response=_annotated_ocr_response("mistral-document-ai-2512", 0, 1), + model=AZURE_DOC_AI_MODEL, + custom_llm_provider="azure_ai", + call_type="ocr", + ) + assert cost == pytest.approx(AZURE_DOC_AI_COST_PER_PAGE) diff --git a/tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py b/tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py index 3ef5e39fc5f..9737d63cc26 100644 --- a/tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py +++ b/tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py @@ -871,6 +871,134 @@ class TestCacheControlPreservationForCustomEndpoint: assert all("cache_control" not in m for m in body["messages"]) +class TestToolChoiceWithoutToolsDropped: + def setup_method(self): + self.config = OpenAIGPTConfig() + + @staticmethod + def _pi_compact_summarization_messages(): + return [ + { + "role": "system", + "content": "You are a context summarization assistant. Your task is to read a conversation between a user and an AI assistant, then produce a structured summary following the exact format specified.", + }, + { + "role": "user", + "content": [ + { + "type": "text", + "text": "\n[User]: Reply with exactly: ok-1\n\n[Assistant]: ok-1\n\n\nThe messages above are a conversation to summarize.", + } + ], + }, + ] + + def _transform(self, optional_params, config=None, model="gpt-5.6-sol"): + return (config or self.config).transform_request( + model=model, + messages=self._pi_compact_summarization_messages(), + optional_params=optional_params, + litellm_params={"custom_llm_provider": "openai", "api_base": None}, + headers={}, + ) + + def test_pi_compact_shape_drops_tool_choice_none_without_tools(self): + body = self._transform( + { + "stream": True, + "stream_options": {"include_usage": True}, + "store": False, + "max_completion_tokens": 13107, + "tool_choice": "none", + } + ) + assert "tool_choice" not in body + assert "tools" not in body + assert body["model"] == "gpt-5.6-sol" + assert body["stream"] is True + assert body["stream_options"] == {"include_usage": True} + assert body["store"] is False + assert body["max_completion_tokens"] == 13107 + + def test_drops_tool_choice_auto_without_tools(self): + body = self._transform({"tool_choice": "auto"}) + assert "tool_choice" not in body + + def test_drops_named_function_tool_choice_without_tools(self): + body = self._transform( + {"tool_choice": {"type": "function", "function": {"name": "get_weather"}}} + ) + assert "tool_choice" not in body + + def test_drops_tool_choice_but_keeps_empty_tools_array(self): + body = self._transform({"tools": [], "tool_choice": "none"}) + assert "tool_choice" not in body + assert body["tools"] == [] + + def test_gpt5_config_drops_tool_choice_without_tools(self): + body = self._transform({"tool_choice": "none"}, config=OpenAIGPT5Config()) + assert "tool_choice" not in body + + @pytest.mark.parametrize( + "tool_choice", + [ + "none", + "auto", + "required", + {"type": "function", "function": {"name": "get_weather"}}, + ], + ) + def test_preserves_tool_choice_when_tools_present(self, tool_choice): + tools = [ + { + "type": "function", + "function": {"name": "get_weather", "parameters": {}}, + } + ] + body = self._transform({"tools": tools, "tool_choice": tool_choice}) + assert body["tool_choice"] == tool_choice + assert body["tools"] == tools + + def test_preserves_tool_choice_with_legacy_functions(self): + functions = [{"name": "get_weather", "parameters": {}}] + body = self._transform({"functions": functions, "tool_choice": "auto"}) + assert body["tool_choice"] == "auto" + assert body["functions"] == functions + + def test_preserves_function_call_without_functions(self): + body = self._transform({"function_call": "none"}) + assert body["function_call"] == "none" + + @pytest.mark.asyncio + async def test_async_transform_drops_tool_choice_without_tools(self): + body = await self.config.async_transform_request( + model="gpt-5.6-sol", + messages=self._pi_compact_summarization_messages(), + optional_params={"stream": True, "tool_choice": "none"}, + litellm_params={"custom_llm_provider": "openai", "api_base": None}, + headers={}, + ) + assert "tool_choice" not in body + + @pytest.mark.asyncio + async def test_async_transform_preserves_tool_choice_when_tools_present(self): + tools = [ + { + "type": "function", + "function": {"name": "get_weather", "parameters": {}}, + } + ] + body = await self.config.async_transform_request( + model="gpt-5.6-sol", + messages=self._pi_compact_summarization_messages(), + optional_params={"tools": tools, "tool_choice": "auto"}, + litellm_params={"custom_llm_provider": "openai", "api_base": None}, + headers={}, + ) + assert body["tool_choice"] == "auto" + assert body["tools"] == tools + + class TestToolMessageImageHoisting: """transform_request moves tool-message images into a following user message (OpenAI-compatible APIs only accept text in role:"tool" messages).""" @@ -1038,3 +1166,118 @@ class TestOpenAIPromptCacheBreakpointChatPath: assert request["messages"][1]["content"] == [{"type": "text", "text": "hi", "prompt_cache_breakpoint": self.EXPLICIT}] assert request["extra_body"] == {"prompt_cache_options": self.EXPLICIT} assert "prompt_cache_options" not in request + + +class TestToolSchemaCombinatorFlatteningForOpenAI: + """ + Regression tests for LIT-6488: OpenAI's chat completions validator rejects + tool parameters carrying a top-level anyOf/oneOf/allOf for every model + family, GPT-5 included, unlike the Responses API. + """ + + def setup_method(self): + self.config = OpenAIGPTConfig() + + @pytest.fixture(autouse=True) + def _clean_openai_base_env(self, monkeypatch): + monkeypatch.delenv("OPENAI_BASE_URL", raising=False) + monkeypatch.delenv("OPENAI_API_BASE", raising=False) + monkeypatch.setattr(litellm, "api_base", None, raising=False) + + @staticmethod + def _anyof_tool(): + return { + "type": "function", + "function": { + "name": "automation_update", + "description": "Update an automation", + "parameters": { + "type": "object", + "anyOf": [ + { + "properties": {"id": {"type": "string"}, "enabled": {"type": "boolean"}}, + "required": ["id", "enabled"], + }, + { + "properties": {"id": {"type": "string"}, "schedule": {"type": "string"}}, + "required": ["id", "schedule"], + }, + ], + "properties": {"id": {"type": "string"}}, + "required": ["id"], + }, + }, + } + + def _transform(self, config, model, litellm_params, tools): + return config.transform_request( + model=model, + messages=[{"role": "user", "content": "hi"}], + optional_params={"tools": tools}, + litellm_params=litellm_params, + headers={}, + ) + + def test_flattens_top_level_anyof_for_hosted_openai(self): + request = self._transform( + self.config, "gpt-4o", {"custom_llm_provider": "openai", "api_base": None}, [self._anyof_tool()] + ) + parameters = request["tools"][0]["function"]["parameters"] + assert "anyOf" not in parameters + assert parameters["type"] == "object" + assert set(parameters["properties"]) == {"id", "enabled", "schedule"} + assert parameters["required"] == ["id"] + assert request["tools"][0]["function"]["name"] == "automation_update" + + def test_gpt5_family_flattens_on_chat_completions(self): + request = self._transform( + OpenAIGPT5Config(), "gpt-5.6", {"custom_llm_provider": "openai", "api_base": None}, [self._anyof_tool()] + ) + assert "anyOf" not in request["tools"][0]["function"]["parameters"] + + def test_custom_api_base_keeps_union(self): + tool = self._anyof_tool() + request = self._transform( + self.config, + "gpt-4o", + {"custom_llm_provider": "openai", "api_base": "http://localhost:8000/v1"}, + [tool], + ) + assert request["tools"][0]["function"]["parameters"] == self._anyof_tool()["function"]["parameters"] + + def test_non_openai_provider_keeps_union(self): + request = self._transform( + self.config, "some-oss-model", {"custom_llm_provider": "groq", "api_base": None}, [self._anyof_tool()] + ) + assert request["tools"][0]["function"]["parameters"] == self._anyof_tool()["function"]["parameters"] + + def test_caller_tool_dict_is_not_mutated(self): + tool = self._anyof_tool() + self._transform(self.config, "gpt-4o", {"custom_llm_provider": "openai", "api_base": None}, [tool]) + assert tool == self._anyof_tool() + + def test_clean_object_schema_passes_through_as_same_object(self): + tool = { + "type": "function", + "function": { + "name": "lookup", + "parameters": {"type": "object", "properties": {"id": {"type": "string"}}, "required": ["id"]}, + }, + } + request = self._transform( + self.config, "gpt-4o", {"custom_llm_provider": "openai", "api_base": None}, [tool] + ) + assert request["tools"][0] is tool + + @pytest.mark.asyncio + async def test_async_transform_request_flattens_for_hosted_openai(self): + request = await self.config.async_transform_request( + model="gpt-4o", + messages=[{"role": "user", "content": "hi"}], + optional_params={"tools": [self._anyof_tool()]}, + litellm_params={"custom_llm_provider": "openai", "api_base": None}, + headers={}, + ) + parameters = request["tools"][0]["function"]["parameters"] + assert "anyOf" not in parameters + assert set(parameters["properties"]) == {"id", "enabled", "schedule"} diff --git a/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py b/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py index 1a90db7c1fe..b0ffd1845fe 100644 --- a/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py +++ b/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py @@ -220,6 +220,111 @@ class TestOpenAIResponsesAPIConfig: assert result["input"] == input_clean + def test_transform_drops_foreign_tool_call_item_ids(self): + """Replayed tool call items whose ids are not OpenAI-shaped (e.g. + Anthropic toolu_/srvtoolu_ ids after a router fallback) must be sent + without an id: OpenAI 400s foreign ids ("Expected an ID that begins + with 'fc'") but accepts the items with no id at all. Genuine fc_/ctc_ + ids and non-tool-call items pass through untouched.""" + replayed_input = [ + {"role": "user", "content": [{"type": "input_text", "text": "hi"}]}, + { + "type": "function_call", + "id": "toolu_01Foreign", + "call_id": "toolu_01Foreign", + "name": "get_weather", + "arguments": '{"city": "SF"}', + }, + {"type": "function_call_output", "call_id": "toolu_01Foreign", "output": "sunny"}, + { + "type": "custom_tool_call", + "id": "srvtoolu_01Foreign", + "call_id": "srvtoolu_01Foreign", + "name": "apply_patch", + "input": "patch", + }, + { + "type": "function_call", + "id": "fc_genuine", + "call_id": "call_genuine", + "name": "get_weather", + "arguments": "{}", + }, + {"type": "message", "id": "msg_1", "role": "assistant", "content": []}, + ] + + result = self.config.transform_responses_api_request( + model=self.model, + input=replayed_input, + response_api_optional_request_params={}, + litellm_params={}, + headers={}, + ) + + assert "id" not in result["input"][1] + assert result["input"][1]["call_id"] == "toolu_01Foreign" + assert "id" not in result["input"][3] + assert result["input"][3]["call_id"] == "srvtoolu_01Foreign" + assert result["input"][4]["id"] == "fc_genuine" + assert result["input"][5]["id"] == "msg_1" + assert replayed_input[1]["id"] == "toolu_01Foreign" + assert replayed_input[3]["id"] == "srvtoolu_01Foreign" + + def test_transform_keeps_foreign_tool_call_item_ids_for_other_providers(self): + """Providers reusing this config that do not enforce OpenAI's id + shapes must keep replayed ids untouched.""" + from litellm.types.utils import LlmProviders + + class _OpenRouterLikeConfig(OpenAIResponsesAPIConfig): + @property + def custom_llm_provider(self) -> LlmProviders: + return LlmProviders.OPENROUTER + + replayed_input = [ + { + "type": "function_call", + "id": "toolu_01Foreign", + "call_id": "toolu_01Foreign", + "name": "get_weather", + "arguments": "{}", + } + ] + + result = _OpenRouterLikeConfig().transform_responses_api_request( + model="openrouter/some-model", + input=replayed_input, + response_api_optional_request_params={}, + litellm_params={}, + headers={}, + ) + + assert result["input"][0]["id"] == "toolu_01Foreign" + + def test_transform_compact_drops_foreign_tool_call_item_ids(self): + """The compact request path replays input the same way, so it must + apply the same id drop.""" + replayed_input = [ + { + "type": "function_call", + "id": "toolu_01Foreign", + "call_id": "toolu_01Foreign", + "name": "get_weather", + "arguments": "{}", + } + ] + + _url, data = self.config.transform_compact_response_api_request( + model=self.model, + input=replayed_input, + response_api_optional_request_params={}, + api_base="https://api.openai.com/v1/responses", + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + + assert "id" not in data["input"][0] + assert data["input"][0]["call_id"] == "toolu_01Foreign" + def test_transform_streaming_response(self): """Test streaming response transformation""" # Test with a text delta event diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/conftest.py b/tests/test_litellm/proxy/_experimental/mcp_server/conftest.py index b477bf3f406..2ccba2b2055 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/conftest.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/conftest.py @@ -2,6 +2,30 @@ import os import pytest +from litellm.proxy._experimental.mcp_server.mcp_server_manager import ( + global_mcp_server_manager, +) + + +@pytest.fixture(autouse=True) +def _hermetic_mcp_server_registry(): + """Restore the singleton ``global_mcp_server_manager``'s registry state around every + test, so entries seeded by one test never leak into another on a shared shard.""" + saved_registry = dict(global_mcp_server_manager.registry) + saved_config_servers = dict(global_mcp_server_manager.config_mcp_servers) + saved_tool_mapping = dict(global_mcp_server_manager.tool_name_to_mcp_server_name_mapping) + saved_oauth_slots = global_mcp_server_manager._oauth_discovery_slots + try: + yield + finally: + global_mcp_server_manager.registry.clear() + global_mcp_server_manager.registry.update(saved_registry) + global_mcp_server_manager.config_mcp_servers.clear() + global_mcp_server_manager.config_mcp_servers.update(saved_config_servers) + global_mcp_server_manager.tool_name_to_mcp_server_name_mapping.clear() + global_mcp_server_manager.tool_name_to_mcp_server_name_mapping.update(saved_tool_mapping) + global_mcp_server_manager._oauth_discovery_slots = saved_oauth_slots + @pytest.fixture(autouse=True) def _hermetic_server_root_path(): diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server_identity_env.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server_identity_env.py index ac7082c2668..1c65adac4c6 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server_identity_env.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server_identity_env.py @@ -23,6 +23,12 @@ MGMT_MODULE = "litellm.proxy.management_endpoints.mcp_management_endpoints" @contextlib.contextmanager def _env_and_reload(**env): saved = {key: os.environ.get(key) for key in env} + utils_module = importlib.import_module(UTILS_MODULE) + mgmt_module = importlib.import_module(MGMT_MODULE) + # Restore pre-reload module attributes afterwards instead of reloading again: + # a reload re-creates the module's classes, breaking exception identity for + # modules that imported them earlier + snapshots = {module: dict(vars(module)) for module in (utils_module, mgmt_module)} def _apply_env(values): for key, value in values.items(): @@ -32,8 +38,8 @@ def _env_and_reload(**env): os.environ[key] = value def _reload(): - utils = importlib.reload(importlib.import_module(UTILS_MODULE)) - mgmt = importlib.reload(importlib.import_module(MGMT_MODULE)) + utils = importlib.reload(utils_module) + mgmt = importlib.reload(mgmt_module) return utils, mgmt try: @@ -41,7 +47,10 @@ def _env_and_reload(**env): yield _reload() finally: _apply_env(saved) - _reload() + for module, snapshot in snapshots.items(): + for key in [key for key in vars(module) if key not in snapshot]: + delattr(module, key) + vars(module).update(snapshot) def test_defaults_used_when_env_unset(): diff --git a/tests/test_litellm/proxy/guardrails/guardrail_hooks/test_alice.py b/tests/test_litellm/proxy/guardrails/guardrail_hooks/test_alice.py new file mode 100644 index 00000000000..fd2e86ccde8 --- /dev/null +++ b/tests/test_litellm/proxy/guardrails/guardrail_hooks/test_alice.py @@ -0,0 +1,614 @@ +import json +import os +from copy import deepcopy +from unittest.mock import AsyncMock + +import httpx +import pytest +from httpx import Request, Response + +import litellm +from litellm.exceptions import GuardrailRaisedException +from litellm.proxy.guardrails.guardrail_hooks.alice.alice import ( + GUARDRAIL_NAME, + AliceGuardrail, + AliceGuardrailMissingSecrets, + _json_safe, +) +from litellm.proxy.guardrails.init_guardrails import init_guardrails_v2 + + +def _guardrail(**overrides: object) -> AliceGuardrail: + params: dict[str, object] = {"api_key": "test-key", "guardrail_name": "alice", "event_hook": "pre_call"} + params.update(overrides) + return AliceGuardrail(**params) + + +def _verdict(payload: dict[str, object], status_code: int = 200) -> Response: + return Response( + status_code=status_code, + json=payload, + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + + +def test_alice_guardrail_config(monkeypatch: pytest.MonkeyPatch): + """Should register through init_guardrails_v2 like any other provider.""" + monkeypatch.setattr(litellm, "guardrail_name_config_map", {}) + monkeypatch.setenv("ALICE_API_KEY", "test-key") + + init_guardrails_v2( + all_guardrails=[ + { + "guardrail_name": "alice", + "litellm_params": {"guardrail": "alice", "mode": "pre_call", "default_on": True}, + } + ], + config_file_path="", + ) + + registered = [cb for cb in litellm.callbacks if isinstance(cb, AliceGuardrail)] + assert len(registered) == 1 + assert registered[0].guardrail_name == "alice" + + +class TestAliceGuardrailInitialization: + def setup_method(self): + for key in ("ALICE_API_KEY", "ALICE_API_BASE"): + os.environ.pop(key, None) + + def test_missing_api_key_raises(self): + with pytest.raises(AliceGuardrailMissingSecrets, match="API key"): + AliceGuardrail(guardrail_name="alice", event_hook="pre_call") + + def test_reads_credentials_from_environment(self, monkeypatch: pytest.MonkeyPatch): + monkeypatch.setenv("ALICE_API_KEY", "env-key") + monkeypatch.setenv("ALICE_API_BASE", "https://env.alice.test") + + guardrail = AliceGuardrail(guardrail_name="alice", event_hook="pre_call") + + assert guardrail.alice_api_key == "env-key" + assert guardrail.api_base == "https://env.alice.test/v2/evaluate/litellm" + + def test_defaults_the_api_base(self): + assert _guardrail().api_base == "https://api.alice.io/v2/evaluate/litellm" + + def test_trailing_slash_does_not_double_up(self): + assert _guardrail(api_base="https://api.alice.io/").api_base == ("https://api.alice.io/v2/evaluate/litellm") + + +class TestAliceForwarding: + """The hook's arguments cross the wire as they were received — nothing selected, nothing + renamed — except the caller's raw credentials, which are stripped before request_data is + serialized (see TestAliceCredentialStripping).""" + + @pytest.mark.asyncio + async def test_forwards_the_hook_arguments_verbatim(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + inputs = {"texts": ["hello"], "structured_messages": [{"role": "user", "content": "hello"}]} + request_data = {"model": "gpt-4o", "metadata": {"user_api_key_alias": "payments-bot"}} + # Snapshot before the call: @log_guardrail_information writes its own entry into + # request_data["metadata"] afterwards, so the original is no longer what was sent. + sent_inputs = deepcopy(inputs) + sent_request_data = deepcopy(request_data) + + await guardrail.apply_guardrail(inputs=inputs, request_data=request_data, input_type="request") + + body = guardrail.async_handler.post.call_args.kwargs["json"] + assert body["input_type"] == "request" + assert body["inputs"] == sent_inputs + assert body["request_data"] == sent_request_data + + @pytest.mark.asyncio + async def test_sends_the_credential(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + + await guardrail.apply_guardrail(inputs={"texts": ["hi"]}, request_data={}, input_type="request") + + assert guardrail.async_handler.post.call_args.kwargs["headers"]["af-api-key"] == "test-key" + + @pytest.mark.asyncio + async def test_marks_a_completion_as_a_response(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + + await guardrail.apply_guardrail(inputs={"texts": ["answer"]}, request_data={}, input_type="response") + + assert guardrail.async_handler.post.call_args.kwargs["json"]["input_type"] == "response" + + @pytest.mark.asyncio + async def test_nothing_selectable_reaches_no_evaluation(self): + """No texts, images, tools, tool_calls, or structured_messages: genuinely nothing to send.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock() + + result = await guardrail.apply_guardrail(inputs={"texts": []}, request_data={}, input_type="request") + + assert result == {"texts": []} + guardrail.async_handler.post.assert_not_called() + + @pytest.mark.asyncio + async def test_tool_calls_only_still_reaches_alice(self): + """A batch with empty texts but populated tool_calls is still a selection decision Alice + should make, not the plugin — see the class docstring.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + inputs = {"texts": [], "tool_calls": [{"id": "call_1", "function": {"name": "get_weather"}}]} + + await guardrail.apply_guardrail(inputs=inputs, request_data={}, input_type="request") + + guardrail.async_handler.post.assert_called_once() + assert guardrail.async_handler.post.call_args.kwargs["json"]["inputs"]["tool_calls"] == inputs["tool_calls"] + + @pytest.mark.asyncio + async def test_images_only_still_reaches_alice(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + + await guardrail.apply_guardrail( + inputs={"texts": [], "images": ["data:image/png;base64,abc"]}, request_data={}, input_type="request" + ) + + guardrail.async_handler.post.assert_called_once() + + @pytest.mark.asyncio + async def test_structured_messages_only_still_reaches_alice(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + + await guardrail.apply_guardrail( + inputs={"texts": [], "structured_messages": [{"role": "user", "content": []}]}, + request_data={}, + input_type="request", + ) + + guardrail.async_handler.post.assert_called_once() + + @pytest.mark.asyncio + async def test_makes_exactly_one_attempt(self): + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock(side_effect=httpx.ConnectError("refused")) + + await guardrail.apply_guardrail(inputs={"texts": ["hi"]}, request_data={}, input_type="request") + + assert guardrail.async_handler.post.call_count == 1 + + +class TestAliceCredentialStripping: + """request_data's raw-credential keys never leave the process.""" + + @pytest.mark.asyncio + async def test_secret_fields_and_api_key_are_stripped(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + request_data = { + "model": "gpt-4o", + "api_key": "sk-forwarded-provider-secret", + "secret_fields": {"raw_headers": {"authorization": "Bearer caller-virtual-key"}}, + "metadata": {"user_api_key_alias": "payments-bot"}, + } + + await guardrail.apply_guardrail(inputs={"texts": ["hi"]}, request_data=request_data, input_type="request") + + sent_request_data = guardrail.async_handler.post.call_args.kwargs["json"]["request_data"] + assert "secret_fields" not in sent_request_data + assert "api_key" not in sent_request_data + assert sent_request_data == {"model": "gpt-4o", "metadata": {"user_api_key_alias": "payments-bot"}} + + @pytest.mark.asyncio + async def test_nested_credentials_are_stripped_at_every_depth(self): + """Shaped after a real captured Claude Code payload: the caller's Authorization/x-api-key + lives under several independent nesting paths, none of which are the root.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + request_data = { + "model": "claude-3-5-sonnet", + "secret_fields": {"raw_headers": {"authorization": "Bearer caller-virtual-key"}}, + "provider_specific_header": {"extra_headers": {"authorization": "sk-ant-oat01-nested-oauth"}}, + "proxy_server_request": { + "url": "/v1/messages", + "headers": {"authorization": "Bearer inbound-caller-secret", "x-request-id": "req-1"}, + "body": { + "model": "claude-3-5-sonnet", + "metadata": {"headers": {"authorization": "Bearer body-metadata-secret"}}, + }, + }, + "metadata": { + "user_api_key_alias": "payments-bot", + "headers": {"authorization": "Bearer metadata-secret"}, + "requester_metadata": {"headers": {"authorization": "Bearer requester-metadata-secret"}}, + }, + "litellm_metadata": {"headers": {"authorization": "Bearer litellm-metadata-secret"}}, + } + + await guardrail.apply_guardrail(inputs={"texts": ["hi"]}, request_data=request_data, input_type="request") + + posted_body = guardrail.async_handler.post.call_args.kwargs["json"] + serialized = json.dumps(posted_body) + assert "authorization" not in serialized.lower() + assert "caller-virtual-key" not in serialized + assert "nested-oauth" not in serialized + assert "inbound-caller-secret" not in serialized + assert "body-metadata-secret" not in serialized + assert "metadata-secret" not in serialized + assert "requester-metadata-secret" not in serialized + assert "litellm-metadata-secret" not in serialized + + sent_request_data = posted_body["request_data"] + assert sent_request_data["model"] == "claude-3-5-sonnet" + assert sent_request_data["proxy_server_request"]["url"] == "/v1/messages" + assert "headers" not in sent_request_data["proxy_server_request"] + assert sent_request_data["proxy_server_request"]["body"]["model"] == "claude-3-5-sonnet" + assert "headers" not in sent_request_data["proxy_server_request"]["body"]["metadata"] + assert sent_request_data["metadata"]["user_api_key_alias"] == "payments-bot" + assert "headers" not in sent_request_data["metadata"] + assert "requester_metadata" in sent_request_data["metadata"] + assert "headers" not in sent_request_data["metadata"]["requester_metadata"] + assert "headers" not in sent_request_data["litellm_metadata"] + assert "secret_fields" not in sent_request_data + assert "provider_specific_header" not in sent_request_data + + @pytest.mark.asyncio + async def test_the_original_request_data_is_not_mutated(self): + """Stripping must only affect the outbound copy — api_key still has to reach the + provider, and secret_fields still has to reach the rest of the request pipeline.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + request_data = {"api_key": "sk-forwarded-provider-secret", "secret_fields": {"raw_headers": {}}} + + await guardrail.apply_guardrail(inputs={"texts": ["hi"]}, request_data=request_data, input_type="request") + + assert request_data["api_key"] == "sk-forwarded-provider-secret" + assert request_data["secret_fields"] == {"raw_headers": {}} + + +class TestAliceVerdicts: + @pytest.mark.asyncio + async def test_allow_leaves_the_inputs_untouched(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "ALLOW", "categories": []})) + + result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + assert result["texts"] == ["hello"] + + @pytest.mark.asyncio + async def test_block_surfaces_the_policy_message(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=_verdict( + { + "verdict": "BLOCK", + "categories": ["self_harm"], + "correlation_id": "c1", + "message": "Blocked by your organization's policy", + } + ) + ) + + with pytest.raises(GuardrailRaisedException) as error: + await guardrail.apply_guardrail(inputs={"texts": ["bad"]}, request_data={}, input_type="request") + + assert "Blocked by your organization's policy" in str(error.value) + + @pytest.mark.asyncio + async def test_block_without_a_message_still_blocks(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "BLOCK", "categories": []})) + + with pytest.raises(GuardrailRaisedException): + await guardrail.apply_guardrail(inputs={"texts": ["bad"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_mask_substitutes_by_position(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=_verdict( + { + "verdict": "MASK", + "categories": ["pii"], + "replacements": [{"index": 1, "text": "my ssn is ***"}], + } + ) + ) + + result = await guardrail.apply_guardrail( + inputs={"texts": ["untouched", "my ssn is 123-45-6789"]}, + request_data={}, + input_type="request", + ) + + assert result["texts"] == ["untouched", "my ssn is ***"] + + @pytest.mark.asyncio + async def test_mask_that_lands_nowhere_blocks(self): + """A mask that wrote nothing would let the text through under a verdict that said not to.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=_verdict({"verdict": "MASK", "categories": [], "replacements": [{"index": 9, "text": "***"}]}) + ) + + with pytest.raises(GuardrailRaisedException): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_mask_with_no_replacements_blocks(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(return_value=_verdict({"verdict": "MASK", "categories": []})) + + with pytest.raises(GuardrailRaisedException): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_mask_with_one_invalid_replacement_blocks_entirely(self): + """A mixed valid/invalid replacement list must not let the valid half through: that + would leave the content named by the invalid entry unmasked while looking like success.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=_verdict( + { + "verdict": "MASK", + "categories": ["pii"], + "replacements": [{"index": 0, "text": "***"}, {"index": 9, "text": "***"}], + } + ) + ) + + with pytest.raises(GuardrailRaisedException): + await guardrail.apply_guardrail( + inputs={"texts": ["my ssn is 123-45-6789"]}, request_data={}, input_type="request" + ) + + @pytest.mark.asyncio + async def test_mask_leaves_structured_messages_identical(self): + """A new structured_messages object makes the translation layer skip the texts write-back.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=_verdict({"verdict": "MASK", "categories": [], "replacements": [{"index": 0, "text": "***"}]}) + ) + messages = [{"role": "user", "content": "secret"}] + + result = await guardrail.apply_guardrail( + inputs={"texts": ["secret"], "structured_messages": messages}, + request_data={}, + input_type="request", + ) + + assert result["structured_messages"] is messages + + @pytest.mark.asyncio + async def test_detect_allows_and_leaves_the_text_alone(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=_verdict({"verdict": "DETECT", "categories": ["profanity"], "correlation_id": "c1"}) + ) + + result = await guardrail.apply_guardrail(inputs={"texts": ["mild"]}, request_data={}, input_type="request") + + assert result["texts"] == ["mild"] + + +class TestAliceUnreachable: + @pytest.mark.parametrize( + "failure", + [ + pytest.param({"side_effect": httpx.ConnectError("refused")}, id="connect-error"), + pytest.param({"return_value": _verdict({"verdict": "MAYBE"})}, id="unrecognized-verdict"), + pytest.param({"return_value": _verdict({})}, id="no-verdict"), + ], + ) + @pytest.mark.asyncio + async def test_fails_closed_by_default(self, failure: dict): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock(**failure) + + with pytest.raises(GuardrailRaisedException, match="unavailable"): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_fails_open_when_configured(self): + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock(side_effect=httpx.ConnectError("refused")) + + result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + assert result["texts"] == ["hello"] + + +class TestAliceTransportFailures: + """Every path out of the HTTP call, since each decides whether traffic flows unscreened.""" + + @pytest.mark.asyncio + async def test_a_timeout_is_unreachable(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + side_effect=litellm.exceptions.Timeout(message="slow", model="gpt-4o", llm_provider="openai") + ) + + with pytest.raises(GuardrailRaisedException, match="unavailable"): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.parametrize("status", [500, 502, 503, 504]) + @pytest.mark.asyncio + async def test_upstream_5xx_is_unreachable(self, status: int): + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock( + side_effect=httpx.HTTPStatusError( + "server error", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + response=_verdict({}, status_code=status), + ) + ) + + result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + assert result["texts"] == ["hello"] + + @pytest.mark.asyncio + async def test_a_500_fails_closed_by_default(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + side_effect=httpx.HTTPStatusError( + "server error", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + response=_verdict({}, status_code=500), + ) + ) + + with pytest.raises(GuardrailRaisedException, match="unavailable"): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_a_4xx_is_not_treated_as_unreachable(self): + """A rejected credential is our misconfiguration, not an outage — it must not fail open.""" + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock( + side_effect=httpx.HTTPStatusError( + "unauthorized", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + response=_verdict({}, status_code=401), + ) + ) + + with pytest.raises(httpx.HTTPStatusError): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_a_non_object_body_fails_closed_by_default(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=Response( + status_code=200, + json=["not", "an", "object"], + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + ) + + with pytest.raises(GuardrailRaisedException, match="unavailable"): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_a_non_object_body_fails_open_when_configured(self): + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock( + return_value=Response( + status_code=200, + json=["not", "an", "object"], + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + ) + + result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + assert result["texts"] == ["hello"] + + @pytest.mark.asyncio + async def test_malformed_json_fails_closed_by_default(self): + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=Response( + status_code=200, + content=b"not json", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + ) + + with pytest.raises(GuardrailRaisedException, match="unavailable"): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_malformed_json_fails_open_when_configured(self): + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock( + return_value=Response( + status_code=200, + content=b"not json", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + ) + + result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + assert result["texts"] == ["hello"] + + @pytest.mark.asyncio + async def test_an_undecodable_body_fails_closed_by_default(self): + """UnicodeDecodeError is a sibling of JSONDecodeError under ValueError, not a subclass.""" + guardrail = _guardrail() + guardrail.async_handler.post = AsyncMock( + return_value=Response( + status_code=200, + content=b"\xff\xfe not utf-8", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + ) + + with pytest.raises(GuardrailRaisedException, match="unavailable"): + await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + @pytest.mark.asyncio + async def test_an_undecodable_body_fails_open_when_configured(self): + guardrail = _guardrail(unreachable_fallback="fail_open") + guardrail.async_handler.post = AsyncMock( + return_value=Response( + status_code=200, + content=b"\xff\xfe not utf-8", + request=Request("POST", "https://api.alice.io/v2/evaluate/litellm"), + ) + ) + + result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request") + + assert result["texts"] == ["hello"] + + +class TestAliceSerialization: + """`request_data` carries live objects, so it cannot be posted as it stands.""" + + def test_drops_what_cannot_serialize_and_keeps_the_rest(self): + class Span: + pass + + result = _json_safe({"model": "x", "metadata": {"span": Span(), "user": "u1"}, "n": 1}) + + assert result == {"model": "x", "metadata": {"span": None, "user": "u1"}, "n": 1} + + def test_survives_a_cycle(self): + data: dict = {"a": 1} + data["self"] = data + + assert _json_safe(data) == {"a": 1, "self": None} + + def test_drops_a_model_that_will_not_dump(self): + class Stubborn: + def model_dump(self, mode: str = "python") -> dict: + raise RuntimeError("cannot serialise") + + assert _json_safe({"m": Stubborn()}) == {"m": None} + + def test_drops_a_bare_unserialisable_value(self): + class Span: + pass + + assert _json_safe(Span()) is None + + def test_dumps_pydantic_models(self): + from pydantic import BaseModel + + class Model(BaseModel): + name: str + + assert _json_safe({"m": Model(name="x")}) == {"m": {"name": "x"}} + + +def test_config_model_is_exposed_for_the_ui(): + config_model = AliceGuardrail.get_config_model() + + assert config_model is not None + assert config_model.ui_friendly_name() == "Alice" + + +def test_guardrail_name_constant(): + assert GUARDRAIL_NAME == "alice" diff --git a/tests/test_litellm/proxy/management_helpers/test_object_permission_utils.py b/tests/test_litellm/proxy/management_helpers/test_object_permission_utils.py index 5ef83344c1a..f2b6b799271 100644 --- a/tests/test_litellm/proxy/management_helpers/test_object_permission_utils.py +++ b/tests/test_litellm/proxy/management_helpers/test_object_permission_utils.py @@ -1,11 +1,9 @@ import json +from unittest.mock import AsyncMock, MagicMock, patch import pytest from fastapi import HTTPException - -from unittest.mock import AsyncMock, MagicMock, patch - from litellm.proxy._types import ( LiteLLM_ObjectPermissionBase, LiteLLM_ObjectPermissionTable, @@ -13,10 +11,10 @@ from litellm.proxy._types import ( SpecialMCPServerName, ) from litellm.proxy.management_helpers.object_permission_utils import ( + _drop_stale_object_permission_mcp_servers, _extract_requested_mcp_access_groups, _extract_requested_mcp_server_ids, _resolve_team_allowed_mcp_servers, - _rewrite_object_permission_mcp_servers, _set_object_permission, enforce_all_proxy_mcp_servers_grant_is_admin_only, validate_key_mcp_servers_against_team, @@ -153,10 +151,10 @@ def test_extract_requested_mcp_server_ids_excludes_no_mcp_servers_sentinel(): assert _extract_requested_mcp_server_ids(obj_perm) == {"server-1"} -def test_rewrite_object_permission_mcp_servers_preserves_sentinel(): - obj_perm = {"mcp_servers": ["no-mcp-servers", "alias-1"]} - _rewrite_object_permission_mcp_servers(obj_perm, {"alias-1": {"server-1"}}) - assert obj_perm["mcp_servers"] == ["no-mcp-servers", "server-1"] +def test_drop_stale_object_permission_mcp_servers_preserves_sentinel_and_alias(): + obj_perm = {"mcp_servers": ["no-mcp-servers", "alias-1", "gone-id"]} + _drop_stale_object_permission_mcp_servers(obj_perm, {"alias-1": {"server-1"}, "gone-id": set()}) + assert obj_perm["mcp_servers"] == ["no-mcp-servers", "alias-1"] @pytest.mark.asyncio @@ -692,9 +690,10 @@ async def test_validate_mcp_server_alias_outside_team_scope_raises( new_callable=AsyncMock, return_value=[], ) -async def test_validate_mcp_server_alias_is_normalized_before_save( - mock_access_groups, mock_allow_all -): +async def test_validate_mcp_server_alias_persists_verbatim(mock_access_groups, mock_allow_all): + """Regression for the multi-region shared-DB setup: an alias grant must be + stored as the alias, so every instance can expand it to its own local id. + Rewriting to this instance's server_id breaks access on the other region.""" team_obj = _make_team_obj(mcp_servers=["allowed-server-id"]) object_permission = { "mcp_servers": ["allowed-alias"], @@ -706,8 +705,27 @@ async def test_validate_mcp_server_alias_is_normalized_before_save( team_obj=team_obj, ) - assert object_permission["mcp_servers"] == ["allowed-server-id"] - assert object_permission["mcp_tool_permissions"] == {"allowed-server-id": ["tool1"]} + assert object_permission["mcp_servers"] == ["allowed-alias"] + assert object_permission["mcp_tool_permissions"] == {"Allowed Server": ["tool1"]} + + +def test_alias_grant_expands_on_other_region_after_save(): + """Cross-region flow: the west instance saves an alias grant (its resolver maps + the alias to west's hash-derived id), then the central instance, whose registry + maps the same alias to a different id, expands the persisted grant. Rewriting + to west's id at save time is exactly the regression this guards against.""" + west_mgr = _make_mock_mcp_manager(servers=[_make_mock_mcp_server("west-id", alias="github-mcp")]) + central_mgr = _make_mock_mcp_manager(servers=[_make_mock_mcp_server("central-id", alias="github-mcp")]) + + object_permission = {"mcp_servers": ["github-mcp"]} + _drop_stale_object_permission_mcp_servers(object_permission, {"github-mcp": {"west-id"}}) + assert object_permission["mcp_servers"] == ["github-mcp"] + + from litellm.proxy._experimental.mcp_server.mcp_server_manager import MCPServerManager + + expand = MCPServerManager.expand_permission_list + assert expand(west_mgr, object_permission["mcp_servers"]) == ["west-id"] + assert expand(central_mgr, object_permission["mcp_servers"]) == ["central-id"] @pytest.mark.asyncio diff --git a/tests/test_litellm/proxy/policy_engine/test_pipeline_executor.py b/tests/test_litellm/proxy/policy_engine/test_pipeline_executor.py index 054a5af4148..4fcb7d22588 100644 --- a/tests/test_litellm/proxy/policy_engine/test_pipeline_executor.py +++ b/tests/test_litellm/proxy/policy_engine/test_pipeline_executor.py @@ -749,6 +749,106 @@ async def test_single_step_pipeline_allow(monkeypatch): assert guard.calls == 1 +@pytest.mark.asyncio +async def test_allow_restores_independent_guardrails_list(monkeypatch): + """ + Request activates an independent guardrail; an unrelated pipeline runs and allows. + Expected: no modified_data escapes, so the request's guardrails list survives + and the independent guardrail still runs at later lifecycle stages (post_call). + Regression: LIT-6587 (pipeline clobbered the list with its last step's guardrail). + """ + pipeline_guard = AlwaysPassGuardrail(guardrail_name="input-scan") + + pipeline = GuardrailPipeline( + mode="pre_call", + steps=[PipelineStep(guardrail="input-scan", on_fail="block", on_pass="allow")], + ) + + monkeypatch.setattr(litellm, "callbacks", [pipeline_guard]) + + data = { + "messages": [{"role": "user", "content": "clean content"}], + "metadata": {"guardrails": ["independent-output-guard"]}, + } + result = await PipelineExecutor.execute_steps( + steps=pipeline.steps, + mode=pipeline.mode, + data=data, + user_api_key_dict=MagicMock(), + call_type="completion", + policy_name="input-pipeline-policy", + ) + + assert pipeline_guard.calls == 1 + assert result.terminal_action == "allow" + propagated = result.modified_data or data + assert propagated["metadata"]["guardrails"] == ["independent-output-guard"] + assert data["metadata"]["guardrails"] == ["independent-output-guard"] + + +@pytest.mark.asyncio +async def test_allow_does_not_leak_guardrails_into_bare_request(monkeypatch): + """A request without metadata must not gain a metadata.guardrails list from the pipeline.""" + pipeline_guard = AlwaysPassGuardrail(guardrail_name="input-scan") + + pipeline = GuardrailPipeline( + mode="pre_call", + steps=[PipelineStep(guardrail="input-scan", on_fail="block", on_pass="allow")], + ) + + monkeypatch.setattr(litellm, "callbacks", [pipeline_guard]) + + data = {"messages": [{"role": "user", "content": "clean content"}]} + result = await PipelineExecutor.execute_steps( + steps=pipeline.steps, + mode=pipeline.mode, + data=data, + user_api_key_dict=MagicMock(), + call_type="completion", + policy_name="input-pipeline-policy", + ) + + assert result.terminal_action == "allow" + propagated = result.modified_data or data + assert "guardrails" not in propagated.get("metadata", {}) + assert "metadata" not in data + + +@pytest.mark.asyncio +async def test_data_forwarding_keeps_changes_and_restores_guardrails_list(monkeypatch): + """A pass_data pipeline's modifications propagate while the request's guardrails list is restored.""" + pii_guard = PiiMaskingGuardrail(guardrail_name="pii-masker") + content_guard = ContentCheckGuardrail(guardrail_name="content-check") + + pipeline = GuardrailPipeline( + mode="pre_call", + steps=[ + PipelineStep(guardrail="pii-masker", on_fail="block", on_pass="next", pass_data=True), + PipelineStep(guardrail="content-check", on_fail="block", on_pass="allow"), + ], + ) + + monkeypatch.setattr(litellm, "callbacks", [pii_guard, content_guard]) + + data = { + "messages": [{"role": "user", "content": "Hello John Smith"}], + "metadata": {"guardrails": ["independent-output-guard"]}, + } + result = await PipelineExecutor.execute_steps( + steps=pipeline.steps, + mode=pipeline.mode, + data=data, + user_api_key_dict=MagicMock(), + call_type="completion", + policy_name="pii-then-safety", + ) + + assert result.terminal_action == "allow" + assert result.modified_data is not None + assert result.modified_data["messages"][0]["content"] == "Hello [REDACTED]" + assert result.modified_data["metadata"]["guardrails"] == ["independent-output-guard"] + + @pytest.mark.asyncio async def test_step_results_include_duration(monkeypatch): """Step results should include timing information.""" diff --git a/tests/test_litellm/proxy/spend_tracking/test_savings.py b/tests/test_litellm/proxy/spend_tracking/test_savings.py index c3297ee6ae9..7dd18587df3 100644 --- a/tests/test_litellm/proxy/spend_tracking/test_savings.py +++ b/tests/test_litellm/proxy/spend_tracking/test_savings.py @@ -14,6 +14,8 @@ from litellm.proxy.spend_tracking.savings import ( from litellm.router import Router from litellm.types.utils import Usage +pytestmark = pytest.mark.usefixtures("local_model_cost_map") + def _anthropic_costs(model: str) -> tuple[float, float]: info = litellm.get_model_info(model=model, custom_llm_provider="anthropic") diff --git a/tests/test_litellm/proxy/test_proxy_cli.py b/tests/test_litellm/proxy/test_proxy_cli.py index 6ea6f208bb5..5e2dd358d75 100644 --- a/tests/test_litellm/proxy/test_proxy_cli.py +++ b/tests/test_litellm/proxy/test_proxy_cli.py @@ -1737,7 +1737,8 @@ class TestRunServerDbSetup: mock_atexit_register, mock_subprocess_run, ): - """Test that use_prisma_db_push flag correctly controls PrismaManager.setup_database use_migrate parameter""" + """Which resolver and which migration mode run_server hands setup_database, + across the db push flag, the v2/legacy flag pair and USE_V2_MIGRATION_RESOLVER.""" from litellm.proxy.proxy_cli import run_server # Mock subprocess.run to simulate prisma being available @@ -1787,7 +1788,7 @@ class TestRunServerDbSetup: # use_prisma_db_push should be False (default), so use_migrate should be True run_server.main(["--local", "--skip_server_startup"], standalone_mode=False) mock_setup_database.assert_called_with( - use_migrate=True, use_v2_resolver=False + use_migrate=True, use_v2_resolver=True ) # Reset mocks @@ -1802,9 +1803,38 @@ class TestRunServerDbSetup: standalone_mode=False, ) mock_setup_database.assert_called_with( - use_migrate=False, use_v2_resolver=False + use_migrate=False, use_v2_resolver=True ) + for argv, env_value, expected_v2 in ( + ([], None, True), + (["--use_v2_migration_resolver"], None, True), + (["--use_legacy_migration_resolver"], None, False), + ([], "false", False), + ([], "true", True), + (["--use_v2_migration_resolver"], "false", True), + (["--use_legacy_migration_resolver"], "true", False), + ): + mock_setup_database.reset_mock() + mock_should_update_schema.reset_mock() + mock_should_update_schema.return_value = True + + resolver_env = ( + {"USE_V2_MIGRATION_RESOLVER": env_value} + if env_value is not None + else {} + ) + os.environ.pop("USE_V2_MIGRATION_RESOLVER", None) + with patch.dict(os.environ, resolver_env): + run_server.main( + ["--local", "--skip_server_startup", *argv], + standalone_mode=False, + ) + assert mock_setup_database.call_args.kwargs == { + "use_migrate": True, + "use_v2_resolver": expected_v2, + }, f"argv={argv} env={env_value}" + @patch("subprocess.run") @patch("atexit.register") @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database") @@ -1869,7 +1899,7 @@ class TestRunServerDbSetup: ) assert exc_info.value.code == 1 mock_setup_database.assert_called_once_with( - use_migrate=True, use_v2_resolver=False + use_migrate=True, use_v2_resolver=True ) @patch("subprocess.run") @@ -1981,7 +2011,6 @@ class TestRunServerDbSetup: use_migrate=True, use_v2_resolver=True ) - # --- Module-level helpers for worker startup hook tests --- _dummy_hook_called = False diff --git a/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py b/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py index bb59e576568..b2b8eb5da80 100644 --- a/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py +++ b/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py @@ -648,6 +648,72 @@ class TestLiteLLMCompletionResponsesConfig: assert responses_api_response.status == "incomplete" + def test_tool_call_only_response_emits_no_null_text_message_item(self): + """A tool-calls-only turn (message content None, e.g. from Anthropic) + must not emit a message output item whose output_text has text null. + OpenAI rejects such an item on replay with + "Invalid type for 'input[..].content[..].text': expected a string, but + got null instead." Native OpenAI tool-only turns carry no message item.""" + chat_completion_response = ModelResponse( + id="test-response-id", + created=1234567890, + model="claude-sonnet-4-5", + object="chat.completion", + choices=[ + Choices( + finish_reason="tool_calls", + index=0, + message=Message( + content=None, + role="assistant", + tool_calls=[ + ChatCompletionMessageToolCall( + id="toolu_01OnlyToolCall", + type="function", + function=Function(name="get_weather", arguments='{"city": "SF"}'), + ) + ], + ), + ) + ], + ) + + responses_api_response = LiteLLMCompletionResponsesConfig.transform_chat_completion_response_to_responses_api_response( + request_input="what's the weather in SF?", + responses_api_request={}, + chat_completion_response=chat_completion_response, + ) + + output_types = [item.type for item in responses_api_response.output] + assert "message" not in output_types + assert "function_call" in output_types + + def test_content_bearing_response_still_emits_message_item(self): + """Turns with real text content must keep their message output item.""" + chat_completion_response = ModelResponse( + id="test-response-id", + created=1234567890, + model="claude-sonnet-4-5", + object="chat.completion", + choices=[ + Choices( + finish_reason="stop", + index=0, + message=Message(content="It is sunny.", role="assistant"), + ) + ], + ) + + responses_api_response = LiteLLMCompletionResponsesConfig.transform_chat_completion_response_to_responses_api_response( + request_input="what's the weather in SF?", + responses_api_request={}, + chat_completion_response=chat_completion_response, + ) + + message_items = [item for item in responses_api_response.output if item.type == "message"] + assert len(message_items) == 1 + assert message_items[0].content[0].text == "It is sunny." + def test_transform_chat_completion_response_preserves_hidden_params(self): """Test that _hidden_params from chat completion response are preserved in responses API response""" # Setup @@ -985,6 +1051,55 @@ class TestFunctionCallTransformation: assert result[0]["tool_calls"][0]["function"]["arguments"] == "{}" + def test_function_call_transformation_json_encodes_object_arguments(self): + """A decoded arguments object must be JSON-encoded, not str()'d. + + Clients and providers sometimes send `arguments` as an object rather + than a JSON string; `str()` on a dict produces a Python repr with + single quotes, which downstream JSON parsers reject with errors like + "Expecting ',' delimiter". + """ + function_call_item = { + "type": "function_call", + "name": "shell", + "arguments": {"command": "ls", "timeout": 30, "flags": ["-l", "-a"]}, + "call_id": "call_123", + "id": "call_123", + "status": "completed", + } + + result = LiteLLMCompletionResponsesConfig._transform_responses_api_function_call_to_chat_completion_message( + function_call=function_call_item + ) + + arguments = result[0].get("tool_calls", [])[0].get("function", {}).get("arguments") + assert json.loads(arguments) == {"command": "ls", "timeout": 30, "flags": ["-l", "-a"]} + assert "'" not in arguments + + def test_create_tool_call_chunk_json_encodes_object_arguments(self): + """Cached tool_call definitions with object arguments stay valid JSON.""" + chunk = LiteLLMCompletionResponsesConfig._create_tool_call_chunk( + tool_use_definition={ + "id": "call_456", + "type": "function", + "function": {"name": "shell", "arguments": {"command": "ls"}}, + }, + tool_call_id="call_456", + index=0, + ) + + assert json.loads(chunk["function"]["arguments"]) == {"command": "ls"} + + def test_create_tool_call_chunk_keeps_empty_arguments_default(self): + """Missing arguments still fall back to an empty JSON object.""" + chunk = LiteLLMCompletionResponsesConfig._create_tool_call_chunk( + tool_use_definition={"id": "call_789", "type": "function", "function": {"name": "shell"}}, + tool_call_id="call_789", + index=0, + ) + + assert chunk["function"]["arguments"] == "{}" + def test_complete_input_transformation_with_function_calls(self): """Test the complete transformation with the exact input from the issue""" test_input = [ @@ -3294,6 +3409,7 @@ class TestEnsureOutputItemContentPartAdded: iterator._pending_tool_events = [] iterator._tool_output_index_by_call_id = {} iterator._tool_args_by_call_id = {} + iterator._tool_item_id_by_call_id = {} iterator._tool_call_id_by_index = {} iterator._ambiguous_tool_call_indexes = set() iterator._next_tool_output_index = 1 diff --git a/tests/test_litellm/responses/litellm_completion_transformation/test_streaming_iterator_transformation.py b/tests/test_litellm/responses/litellm_completion_transformation/test_streaming_iterator_transformation.py index 823f656ddc5..4a03913f55a 100644 --- a/tests/test_litellm/responses/litellm_completion_transformation/test_streaming_iterator_transformation.py +++ b/tests/test_litellm/responses/litellm_completion_transformation/test_streaming_iterator_transformation.py @@ -10,6 +10,7 @@ before response.completed, and that every event of a bridged stream carries the spend tracking stores, so a follow-up previous_response_id still finds the conversation. """ +import json from unittest.mock import AsyncMock, MagicMock import pytest @@ -131,7 +132,7 @@ def test_tool_call_delta_is_emitted_as_responses_events(): evt2 = iterator._transform_chat_completion_chunk_to_response_api_chunk(chunk) assert evt2 is not None assert evt2.type == ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DELTA - assert evt2.item_id == "call_1" + assert evt2.item_id == "fc_call_1" assert evt2.output_index == 1 # The delta will be a chunk of the arguments, not the full arguments assert len(evt2.delta) <= 10 # Chunks are max 10 characters @@ -196,7 +197,7 @@ def test_tool_calls_present_only_in_final_response_are_emitted_before_completed( # The last event should be FUNCTION_CALL_ARGUMENTS_DONE assert evt.type == ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DONE - assert evt.item_id == "call_2" + assert evt.item_id == "fc_call_2" assert evt.output_index == 1 assert evt.arguments == '{"y":2}' @@ -290,7 +291,7 @@ def test_tool_call_arguments_are_chunked_to_match_openai_behavior(): # Verify each delta is at most 10 characters for evt in delta_events: assert len(evt.delta) <= 10 - assert evt.item_id == "call_test" + assert evt.item_id == "fc_call_test" assert evt.output_index == 1 assert hasattr(evt, "__dict__") and "sequence_number" in evt.__dict__ @@ -348,7 +349,8 @@ def test_tool_call_delta_without_id_uses_index_mapping(): if evt.type == ResponsesAPIStreamEvents.OUTPUT_ITEM_ADDED ] assert len(output_item_added_events) == 1 - assert output_item_added_events[0].item.id == "call_abc123" + assert output_item_added_events[0].item.id == "fc_call_abc123" + assert output_item_added_events[0].item.call_id == "call_abc123" def test_parallel_tool_calls_without_ids_use_index_mapping(): @@ -403,8 +405,8 @@ def test_parallel_tool_calls_without_ids_use_index_mapping(): arguments_by_call_id.setdefault(evt.item_id, "") arguments_by_call_id[evt.item_id] += evt.delta - assert arguments_by_call_id["call_a"] == '{"x":1}' - assert arguments_by_call_id["call_b"] == '{"y":2}' + assert arguments_by_call_id["fc_call_a"] == '{"x":1}' + assert arguments_by_call_id["fc_call_b"] == '{"y":2}' def test_reused_index_with_new_call_id_marks_fallback_ambiguous(): @@ -460,10 +462,10 @@ def test_reused_index_with_new_call_id_marks_fallback_ambiguous(): arguments_by_call_id.setdefault(evt.item_id, "") arguments_by_call_id[evt.item_id] += evt.delta - assert arguments_by_call_id["call_a"] == '{"a":' - assert arguments_by_call_id["call_b"] == '{"b":' - assert arguments_by_call_id["call_a"] != '{"a":1}' - assert arguments_by_call_id["call_b"] != '{"b":1}' + assert arguments_by_call_id["fc_call_a"] == '{"a":' + assert arguments_by_call_id["fc_call_b"] == '{"b":' + assert arguments_by_call_id["fc_call_a"] != '{"a":1}' + assert arguments_by_call_id["fc_call_b"] != '{"b":1}' @pytest.mark.asyncio @@ -523,3 +525,91 @@ async def test_streaming_response_id_falls_back_when_upstream_yields_nothing(): assert response_ids assert len(set(response_ids)) == 1 assert response_ids[0].startswith("resp_") + + +def test_object_tool_call_arguments_stream_as_valid_json(): + """A provider that sends decoded object arguments must still stream valid JSON. + + `str()` on a dict yields a Python repr with single quotes, which clients + parsing function_call_arguments reject with errors like + "Expecting ',' delimiter". + """ + iterator = LiteLLMCompletionStreamingIterator( + model="test-model", + litellm_custom_stream_wrapper=AsyncMock(), + request_input="Test input", + responses_api_request={}, + ) + iterator._queue_tool_call_delta_events( + [ + { + "index": 0, + "id": "call_obj", + "type": "function", + "function": {"name": "shell", "arguments": {"command": "ls", "flags": ["-l"]}}, + } + ] + ) + + streamed_arguments = "".join( + evt.delta + for evt in iterator._pending_tool_events + if evt.type == ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DELTA + ) + + assert json.loads(streamed_arguments) == {"command": "ls", "flags": ["-l"]} + + +def test_streamed_anthropic_tool_call_events_correlate_on_normalized_item_id(): + iterator = LiteLLMCompletionStreamingIterator( + model="test-model", + litellm_custom_stream_wrapper=AsyncMock(), + request_input="Test input", + responses_api_request={}, + ) + + response = ModelResponse( + id="resp-anthropic", + created=123, + model="test-model", + object="chat.completion", + choices=[ + { + "index": 0, + "finish_reason": "tool_calls", + "message": { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "toolu_01AbCdEf", + "type": "function", + "function": {"name": "get_weather", "arguments": '{"city":"Paris"}'}, + "index": 0, + } + ], + }, + } + ], + ) + iterator.litellm_model_response = response + + events = [] + while True: + evt = iterator.common_done_event_logic(sync_mode=True) + events.append(evt) + if evt.type == ResponsesAPIStreamEvents.OUTPUT_ITEM_DONE: + break + + added = [e for e in events if e.type == ResponsesAPIStreamEvents.OUTPUT_ITEM_ADDED] + deltas = [e for e in events if e.type == ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DELTA] + dones = [e for e in events if e.type == ResponsesAPIStreamEvents.FUNCTION_CALL_ARGUMENTS_DONE] + item_dones = [e for e in events if e.type == ResponsesAPIStreamEvents.OUTPUT_ITEM_DONE] + + assert len(added) == 1 and len(dones) == 1 and len(item_dones) == 1 and deltas + assert added[0].item.id == "fc_toolu_01AbCdEf" + assert added[0].item.call_id == "toolu_01AbCdEf" + assert item_dones[0].item.id == "fc_toolu_01AbCdEf" + assert item_dones[0].item.call_id == "toolu_01AbCdEf" + for evt in deltas + dones: + assert evt.item_id == added[0].item.id diff --git a/tests/test_litellm/responses/test_custom_tool_call.py b/tests/test_litellm/responses/test_custom_tool_call.py index c605ef24934..5122c1c1d67 100644 --- a/tests/test_litellm/responses/test_custom_tool_call.py +++ b/tests/test_litellm/responses/test_custom_tool_call.py @@ -20,6 +20,7 @@ from litellm.responses.litellm_completion_transformation.transformation import ( from litellm.responses.litellm_completion_transformation.custom_tools import ( extract_custom_tool_names, is_custom_tool_call, + openai_shaped_tool_call_item_id, unwrap_custom_tool_arguments, build_tool_call_item_kwargs, convert_custom_tool_to_function_tool, @@ -129,6 +130,41 @@ class TestCustomToolUtilities: assert kwargs["arguments"] == raw assert "input" not in kwargs + def test_openai_shaped_tool_call_item_id_prefixes_foreign_ids(self): + """Anthropic-style tool ids must be normalized to OpenAI's item id + shapes (fc/ctc prefixes) so replaying the item to OpenAI does not 400 + with "Expected an ID that begins with 'fc'".""" + assert openai_shaped_tool_call_item_id("function_call", "toolu_01Abc") == "fc_toolu_01Abc" + assert openai_shaped_tool_call_item_id("function_call", "srvtoolu_01Xyz") == "fc_srvtoolu_01Xyz" + assert openai_shaped_tool_call_item_id("custom_tool_call", "toolu_01Abc") == "ctc_toolu_01Abc" + assert openai_shaped_tool_call_item_id("function_call", "fc_already") == "fc_already" + assert openai_shaped_tool_call_item_id("custom_tool_call", "ctc_already") == "ctc_already" + assert openai_shaped_tool_call_item_id("function_call", "") == "" + assert openai_shaped_tool_call_item_id("message", "toolu_01Abc") == "toolu_01Abc" + + def test_build_tool_call_item_kwargs_normalizes_item_id_keeps_call_id(self): + """The streaming item id gets the OpenAI shape while call_id stays raw + so tool_result pairing (which keys off call_id) keeps working.""" + function_kwargs = build_tool_call_item_kwargs( + call_id="toolu_01Abc", + name="get_weather", + arguments_or_input="{}", + status="completed", + custom_tool_names=set(), + ) + assert function_kwargs["id"] == "fc_toolu_01Abc" + assert function_kwargs["call_id"] == "toolu_01Abc" + + custom_kwargs = build_tool_call_item_kwargs( + call_id="toolu_01Def", + name="apply_patch", + arguments_or_input=json.dumps({"content": "patch"}), + status="completed", + custom_tool_names={"apply_patch"}, + ) + assert custom_kwargs["id"] == "ctc_toolu_01Def" + assert custom_kwargs["call_id"] == "toolu_01Def" + def test_unwrap_custom_tool_arguments_oversized_returns_raw(self): """Arguments larger than the safety cap are returned unchanged to avoid OOM on JSON parsing a pathologically large string.""" @@ -293,6 +329,52 @@ class TestTransformationCustomTools: assert item.name == "regular_tool" assert item.arguments == json.dumps({"param": "value"}) + def test_transform_anthropic_tool_call_ids_get_openai_item_id_shape(self): + """Anthropic tool ids (toolu_/srvtoolu_) surfacing through the bridge + must be emitted with fc/ctc-prefixed item ids so a Responses client can + replay them to OpenAI verbatim, while call_id stays raw for pairing.""" + from litellm.types.utils import ModelResponse, Choices, Message, ChatCompletionMessageToolCall, Function + + client_call = ChatCompletionMessageToolCall( + id="toolu_01ClientCall", + type="function", + function=Function(name="get_weather", arguments=json.dumps({"city": "SF"})), + ) + server_call = ChatCompletionMessageToolCall( + id="srvtoolu_01ServerCall", + type="function", + function=Function(name="web_search", arguments=json.dumps({"query": "zig"})), + ) + custom_call = ChatCompletionMessageToolCall( + id="toolu_01CustomCall", + type="function", + function=Function(name="apply_patch", arguments=json.dumps({"content": "patch content"})), + ) + + message = Message(role="assistant", content=None, tool_calls=[client_call, server_call, custom_call]) + choices = [Choices(index=0, message=message, finish_reason="tool_calls")] + response = ModelResponse( + id="test_response", choices=choices, created=1234567890, model="claude-sonnet-4-5", object="chat.completion" + ) + responses_api_request = { + "tools": [{"type": "custom", "name": "apply_patch"}, {"type": "function", "name": "get_weather"}] + } + + result = LiteLLMCompletionResponsesConfig.transform_chat_completion_tools_to_responses_tools( + response, responses_api_request=responses_api_request + ) + + assert [item.id for item in result] == [ + "fc_toolu_01ClientCall", + "fc_srvtoolu_01ServerCall", + "ctc_toolu_01CustomCall", + ] + assert [item.call_id for item in result] == [ + "toolu_01ClientCall", + "srvtoolu_01ServerCall", + "toolu_01CustomCall", + ] + def test_transform_mixed_tool_calls(self): """Test transformation with both custom and regular tool calls.""" from litellm.types.utils import ModelResponse, Choices, Message, ChatCompletionMessageToolCall, Function diff --git a/tests/test_litellm/test_claude_fable_5_config.py b/tests/test_litellm/test_claude_fable_5_config.py index 99c59ffa58e..3ecf94602d9 100644 --- a/tests/test_litellm/test_claude_fable_5_config.py +++ b/tests/test_litellm/test_claude_fable_5_config.py @@ -1,5 +1,5 @@ """ -Validate Claude Fable 5 model configuration entries. +Validate Claude Fable 5 and Claude Fable 5.1 model configuration entries. Fable 5 is a new tier above Opus ($10/$50 per MTok) with the same adaptive-only API surface as Opus 4.7/4.8. The cost-map entries below are what make the model @@ -210,6 +210,156 @@ def test_adaptive_thinking_detected_for_fable_5(local_model_cost_map, model): assert AnthropicModelInfo._is_adaptive_thinking_model(model, "anthropic") is True +FABLE_5_1_VARIANTS = ( + "claude-fable-5-1", + "anthropic.claude-fable-5-1", + "global.anthropic.claude-fable-5-1", + "us.anthropic.claude-fable-5-1", + "eu.anthropic.claude-fable-5-1", + "vertex_ai/claude-fable-5-1", + "vertex_ai/claude-fable-5-1@default", + "azure_ai/claude-fable-5-1", +) + + +def test_fable_5_1_model_pricing_and_capabilities(): + model_data = _load_root_cost_map() + + expected_models = [ + ("claude-fable-5-1", "anthropic"), + ("anthropic.claude-fable-5-1", "bedrock_converse"), + ("vertex_ai/claude-fable-5-1", "vertex_ai-anthropic_models"), + ("azure_ai/claude-fable-5-1", "azure_ai"), + ] + + for model_name, provider in expected_models: + assert model_name in model_data, f"Missing model entry: {model_name}" + info = model_data[model_name] + + assert info["litellm_provider"] == provider + assert info["mode"] == "chat" + assert info["max_input_tokens"] == 1000000 + assert info["max_output_tokens"] == 128000 + assert info["max_tokens"] == 128000 + + assert info["input_cost_per_token"] == 1e-05 + assert info["output_cost_per_token"] == 5e-05 + assert info["cache_creation_input_token_cost"] == 1.25e-05 + assert info["cache_creation_input_token_cost_above_1hr"] == 2e-05 + + assert "input_cost_per_token_above_200k_tokens" not in info + assert "output_cost_per_token_above_200k_tokens" not in info + + assert info["supports_assistant_prefill"] is False + assert info["supports_forced_tool_use"] is False + assert info["supports_function_calling"] is True + assert info["supports_prompt_caching"] is True + assert info["supports_reasoning"] is True + assert info["supports_tool_choice"] is True + assert info["supports_vision"] is True + assert info["supports_xhigh_reasoning_effort"] is True + assert info["supports_max_reasoning_effort"] is True + assert info["prompt_cache_min_tokens"] == 512 + + +@pytest.mark.parametrize( + "cost_map", + [_load_root_cost_map(), GetModelCostMap.load_local_model_cost_map()], + ids=["root", "bundled_backup"], +) +def test_fable_5_1_cache_reads_cost_a_quarter_of_fable_5(cost_map): + """Fable 5.1 prices cache hits at 0.025x base input instead of the usual + 0.1x, so copying Fable 5's cache-read price overcharges every cache hit 4x.""" + for model_name in FABLE_5_1_VARIANTS: + info = cost_map[model_name] + geo_premium = model_name.startswith(("us.", "eu.")) + expected = 2.75e-07 if geo_premium else 2.5e-07 + assert info["cache_read_input_token_cost"] == expected, model_name + assert info["cache_read_input_token_cost"] == pytest.approx( + info["input_cost_per_token"] * 0.025 + ), model_name + + +def test_fable_5_1_bedrock_regional_model_pricing(): + model_data = _load_root_cost_map() + + expected_models = { + "global.anthropic.claude-fable-5-1": { + "input_cost_per_token": 1e-05, + "output_cost_per_token": 5e-05, + "cache_creation_input_token_cost": 1.25e-05, + "cache_read_input_token_cost": 2.5e-07, + }, + "us.anthropic.claude-fable-5-1": { + "input_cost_per_token": 1.1e-05, + "output_cost_per_token": 5.5e-05, + "cache_creation_input_token_cost": 1.375e-05, + "cache_read_input_token_cost": 2.75e-07, + }, + "eu.anthropic.claude-fable-5-1": { + "input_cost_per_token": 1.1e-05, + "output_cost_per_token": 5.5e-05, + "cache_creation_input_token_cost": 1.375e-05, + "cache_read_input_token_cost": 2.75e-07, + }, + } + + for model_name, expected in expected_models.items(): + assert model_name in model_data, f"Missing model entry: {model_name}" + info = model_data[model_name] + assert info["litellm_provider"] == "bedrock_converse" + assert info["max_input_tokens"] == 1000000 + assert info["max_output_tokens"] == 128000 + assert info["bedrock_output_config_effort_ceiling"] == "xhigh" + for key, value in expected.items(): + assert info[key] == value + + +def test_fable_5_1_geo_multiplier_without_fast_mode(): + """Fable 5.1 has no fast mode, so a ``fast`` key here would misprice + ``speed='fast'`` requests.""" + model_data = _load_root_cost_map() + assert model_data["claude-fable-5-1"]["provider_specific_entry"] == {"us": 1.1} + + +def test_fable_5_1_present_in_bundled_backup(): + backup = GetModelCostMap.load_local_model_cost_map() + root = _load_root_cost_map() + for model_name in FABLE_5_1_VARIANTS: + assert model_name in backup, f"Missing from backup cost map: {model_name}" + assert backup[model_name] == root[model_name], model_name + + +def test_fable_5_1_registered_for_bedrock_converse(): + assert "anthropic.claude-fable-5-1" in BEDROCK_CONVERSE_MODELS + + +def test_fable_5_1_provider_resolves_via_model_info(local_model_cost_map): + info = litellm.get_model_info(model="claude-fable-5-1") + assert info["litellm_provider"] == "anthropic" + assert info["max_input_tokens"] == 1000000 + assert info["max_output_tokens"] == 128000 + + +@pytest.mark.parametrize( + "model", + [ + "claude-fable-5-1", + "anthropic/claude-fable-5-1", + "anthropic.claude-fable-5-1", + "bedrock/us.anthropic.claude-fable-5-1", + "bedrock/invoke/eu.anthropic.claude-fable-5-1", + "bedrock/global.anthropic.claude-fable-5-1", + "vertex_ai/claude-fable-5-1", + "azure_ai/claude-fable-5-1", + ], +) +def test_adaptive_thinking_detected_for_fable_5_1(local_model_cost_map, model): + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + assert AnthropicModelInfo._is_adaptive_thinking_model(model, "anthropic") is True + + @pytest.mark.parametrize( "cost_map", [_load_root_cost_map(), GetModelCostMap.load_local_model_cost_map()], diff --git a/tests/test_litellm/test_fireworks_serverless_model_costs.py b/tests/test_litellm/test_fireworks_serverless_model_costs.py index 0458af0da0e..a7a9e0fc37d 100644 --- a/tests/test_litellm/test_fireworks_serverless_model_costs.py +++ b/tests/test_litellm/test_fireworks_serverless_model_costs.py @@ -84,3 +84,41 @@ def test_bare_fireworks_ids_resolve_through_prefixed_entries(): assert info["output_cost_per_token"] == pytest.approx(expected["output_cost_per_token"]) assert info["max_input_tokens"] == expected["max_input_tokens"] assert info["max_output_tokens"] == expected["max_output_tokens"] + + +TWIN_PINNED_PRICES = { + "deepseek-v4-flash-0731": { + "input_cost_per_token": 2.2e-07, + "cache_read_input_token_cost": 7e-09, + "output_cost_per_token": 6.6e-07, + }, +} + + +def test_deepseek_v4_flash_0731_twins_pin_published_pricing(model_data): + """Both 0731 entries carry the price published at docs.fireworks.ai/serverless/pricing.""" + for bare_suffix, expected in TWIN_PINNED_PRICES.items(): + for key in ( + f"fireworks_ai/{bare_suffix}", + f"fireworks_ai/accounts/fireworks/models/{bare_suffix}", + ): + entry = model_data[key] + for field, value in expected.items(): + assert entry[field] == pytest.approx(value), f"{key}.{field}" + + +def test_fireworks_account_prefixed_twins_agree_on_price(model_data): + """Every accounts/fireworks/models/X entry prices identically to its bare fireworks_ai/X twin.""" + prefix = "fireworks_ai/accounts/fireworks/models/" + pairs_checked = 0 + for key, entry in model_data.items(): + if not key.startswith(prefix): + continue + bare_key = f"fireworks_ai/{key[len(prefix):]}" + bare_entry = model_data.get(bare_key) + if bare_entry is None: + continue + pairs_checked += 1 + for field in sorted({f for f in (*entry, *bare_entry) if "cost" in f}): + assert entry.get(field) == bare_entry.get(field), f"{key} vs {bare_key}: {field}" + assert pairs_checked >= 20 diff --git a/tests/test_litellm/test_openai_embedding_encoding_format_default.py b/tests/test_litellm/test_openai_embedding_encoding_format_default.py index 94e4e3c81e5..7a42eaf0f0a 100644 --- a/tests/test_litellm/test_openai_embedding_encoding_format_default.py +++ b/tests/test_litellm/test_openai_embedding_encoding_format_default.py @@ -1,124 +1,121 @@ -from unittest.mock import MagicMock, patch +import json +from typing import Final +import httpx import pytest +import respx -from litellm import embedding +import litellm -@pytest.mark.parametrize( - "set_env, env_value, expected", - [ - (False, None, "float"), - (True, "base64", "base64"), - ], -) -def test_openai_embedding_encoding_format_default( - monkeypatch, set_env, env_value, expected -): - monkeypatch.delenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", raising=False) - if set_env: - monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", env_value) - - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", - "object": "list", - "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } +def _mock_openai_embedding_route(respx_mock: respx.MockRouter) -> respx.Route: + return respx_mock.post("https://api.openai.com/v1/embeddings").mock( + return_value=httpx.Response( + 200, + json={ + "object": "list", + "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], + "model": "text-embedding-3-small", + "usage": {"prompt_tokens": 2, "total_tokens": 2}, + }, + ) ) - mock_response.headers = {} - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response - ) - embedding( - model="text-embedding-ada-002", - input="Hello world", - ) +@pytest.fixture(autouse=True) +def clear_default_encoding_format_env(monkeypatch: pytest.MonkeyPatch): + monkeypatch.delenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", raising=False) - call_kwargs = ( - mock_client_instance.embeddings.with_raw_response.create.call_args[1] - ) - assert call_kwargs["encoding_format"] == expected + +def test_embedding_openai_omits_encoding_format_when_client_omits_it(respx_mock: respx.MockRouter) -> None: + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + response: Final = litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert "encoding_format" not in request_body + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] + + +def test_embedding_openai_forwards_explicit_encoding_format(respx_mock: respx.MockRouter) -> None: + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + litellm.embedding( + model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test", encoding_format="base64" + ) + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert request_body["encoding_format"] == "base64" + + +def test_embedding_openai_explicit_encoding_format_wins_over_env_var( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch +) -> None: + monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", "float") + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + litellm.embedding( + model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test", encoding_format="base64" + ) + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert request_body["encoding_format"] == "base64" + + +@pytest.mark.parametrize("env_value", ["float", "base64"]) +def test_embedding_openai_env_var_sets_default_encoding_format( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch, env_value: str +) -> None: + monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", env_value) + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert request_body["encoding_format"] == env_value @pytest.mark.parametrize("env_none", ["none", "NONE", " none "]) -def test_openai_embedding_encoding_format_env_none_omits_param( - monkeypatch, env_none -): - """LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT=none omits encoding_format (provider default).""" +def test_embedding_openai_env_none_omits_encoding_format( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch, env_none: str +) -> None: monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", env_none) + mock_route: Final = _mock_openai_embedding_route(respx_mock) - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", - "object": "list", - "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } + litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert "encoding_format" not in request_body + + +@pytest.mark.asyncio +async def test_aembedding_openai_omits_encoding_format_when_client_omits_it( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch +) -> None: + monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + response: Final = await litellm.aembedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert "encoding_format" not in request_body + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] + + +def test_embedding_openai_omitted_encoding_format_maps_provider_errors( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch +) -> None: + respx_mock.post("https://api.openai.com/v1/embeddings").mock( + return_value=httpx.Response( + 429, + headers={"retry-after": "42", "x-should-retry": "false"}, + json={"error": {"message": "rate limited", "type": "rate_limit_error"}}, + ) ) - mock_response.headers = {} - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response + with pytest.raises(litellm.RateLimitError) as exc_info: + litellm.embedding( + model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test", max_retries=0 ) - embedding( - model="text-embedding-ada-002", - input="Hello world", - ) - - call_kwargs = ( - mock_client_instance.embeddings.with_raw_response.create.call_args[1] - ) - assert "encoding_format" not in call_kwargs - - -def test_openai_embedding_encoding_format_explicit_overrides_env(monkeypatch): - """Request `encoding_format` wins over LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT.""" - monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", "float") - - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", - "object": "list", - "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } - ) - mock_response.headers = {} - - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response - ) - - embedding( - model="text-embedding-ada-002", - input="Hello world", - encoding_format="base64", - ) - - call_kwargs = ( - mock_client_instance.embeddings.with_raw_response.create.call_args[1] - ) - assert call_kwargs["encoding_format"] == "base64" + assert int(exc_info.value.litellm_response_headers["retry-after"]) == 42 diff --git a/tests/test_litellm/test_register_model_custom_pricing.py b/tests/test_litellm/test_register_model_custom_pricing.py index 39f498b4e58..452a15334ef 100644 --- a/tests/test_litellm/test_register_model_custom_pricing.py +++ b/tests/test_litellm/test_register_model_custom_pricing.py @@ -793,3 +793,203 @@ def test_embedding_direct_sdk_custom_pricing_still_registers_shared_key(): finally: litellm.model_cost.pop(model_key, None) _invalidate_model_cost_lowercase_map() + + +def test_update_dictionary_merges_nested_dicts_without_aliasing(): + """A nested dict must be merged copy-on-write: the pre-existing nested dict + object stays untouched, and the caller's incoming nested dict is never + inserted by reference into the merged result. + """ + from litellm.utils import _update_dictionary + + existing_nested = {"hours_utc": "01:00-02:00"} + existing = {"off_peak_pricing": existing_nested} + incoming_nested = {"windows": [{"hours_utc": "16:00-19:00", "weekdays": [2]}]} + incoming = {"off_peak_pricing": incoming_nested} + + merged = _update_dictionary(existing, incoming) + + assert merged["off_peak_pricing"] == { + "hours_utc": "01:00-02:00", + "windows": [{"hours_utc": "16:00-19:00", "weekdays": [2]}], + } + assert existing_nested == {"hours_utc": "01:00-02:00"} + assert merged["off_peak_pricing"] is not incoming_nested + + fresh = _update_dictionary({}, incoming) + assert fresh["off_peak_pricing"] == incoming_nested + assert fresh["off_peak_pricing"] is not incoming_nested + + +def test_router_deployments_sharing_backend_keep_their_own_off_peak_pricing(): + """Two deployments of the same backend model with different + ``off_peak_pricing`` blocks must each keep their own schedule under their + unique model id, and neither block may leak onto the shared backend keys. + + Before the fix, ``register_model`` inserted the first deployment's block by + reference into the built-in ``gpt-4o-mini`` entry, and the second + deployment's registration merged its keys into that same object, corrupting + the first deployment's schedule and polluting the built-in entry. + """ + from litellm import Router + + active_block = { + "windows": [{"hours_utc": "16:00-19:00", "weekdays": [2]}], + "input_cost_per_token": 5e-07, + "output_cost_per_token": 1e-06, + } + inactive_block = { + "hours_utc": "05:00-06:00", + "input_cost_per_token": 5e-07, + "output_cost_per_token": 1e-06, + } + shared_keys = ["gpt-4o-mini", "openai/gpt-4o-mini"] + deployment_ids = ["offpeak-alias-dep-1", "offpeak-alias-dep-2"] + original_entries = _snapshot_model_cost_entries(shared_keys) + + router = Router( + model_list=[ + { + "model_name": "offpeak-active-weekday", + "litellm_params": { + "model": "openai/gpt-4o-mini", + "api_key": "fake-key-for-registration", + }, + "model_info": { + "id": deployment_ids[0], + "input_cost_per_token": 1e-06, + "output_cost_per_token": 2e-06, + "off_peak_pricing": dict(active_block), + }, + }, + { + "model_name": "offpeak-inactive-hours", + "litellm_params": { + "model": "openai/gpt-4o-mini", + "api_key": "fake-key-for-registration", + }, + "model_info": { + "id": deployment_ids[1], + "input_cost_per_token": 1e-06, + "output_cost_per_token": 2e-06, + "off_peak_pricing": dict(inactive_block), + }, + }, + ] + ) + + try: + registered_first = litellm.model_cost[deployment_ids[0]]["off_peak_pricing"] + registered_second = litellm.model_cost[deployment_ids[1]]["off_peak_pricing"] + assert registered_first == active_block + assert registered_second == inactive_block + for shared_key in shared_keys: + shared_entry = litellm.model_cost.get(shared_key) or {} + assert not shared_entry.get("off_peak_pricing") + finally: + for deployment_id in deployment_ids: + litellm.model_cost.pop(deployment_id, None) + _restore_model_cost_entries(original_entries) + del router + + +def test_router_off_peak_only_deployment_inherits_builtin_base_rates(): + """A deployment that sets only ``off_peak_pricing`` on its model_info must + still be costed from its deployment-scoped entry: the base token rates are + inherited from the backend model's built-in cost map entry, since the + shared backend key deliberately never carries the off-peak block. + """ + from litellm import Router + + block = { + "hours_utc": "00:00-00:00", + "input_cost_per_token": 5e-05, + "output_cost_per_token": 1e-04, + } + shared_keys = ["gpt-4o-mini", "openai/gpt-4o-mini"] + deployment_id = "offpeak-only-dep-1" + original_entries = _snapshot_model_cost_entries(shared_keys + [deployment_id]) + builtin_info = litellm.get_model_info(model="openai/gpt-4o-mini") + + router = Router( + model_list=[ + { + "model_name": "offpeak-only", + "litellm_params": { + "model": "openai/gpt-4o-mini", + "api_key": "fake-key-for-registration", + }, + "model_info": {"id": deployment_id, "off_peak_pricing": dict(block)}, + } + ] + ) + + try: + entry = litellm.model_cost[deployment_id] + assert entry["off_peak_pricing"] == block + assert entry["input_cost_per_token"] is not None + assert entry["input_cost_per_token"] == builtin_info["input_cost_per_token"] + assert entry["output_cost_per_token"] == builtin_info["output_cost_per_token"] + for shared_key in shared_keys: + shared_entry = litellm.model_cost.get(shared_key) or {} + assert not shared_entry.get("off_peak_pricing") + finally: + _restore_model_cost_entries(original_entries) + del router + + +def test_use_custom_pricing_for_model_sees_off_peak_only_model_info(): + from litellm.litellm_core_utils.litellm_logging import use_custom_pricing_for_model + + block = {"hours_utc": "00:00-00:00", "input_cost_per_token": 5e-05} + assert use_custom_pricing_for_model({"metadata": {"model_info": {"off_peak_pricing": block}}}) is True + assert use_custom_pricing_for_model({"metadata": {"model_info": {"off_peak_pricing": None}}}) is False + assert use_custom_pricing_for_model({"metadata": {"model_info": {"id": "some-id"}}}) is False + + +def test_completion_cost_applies_off_peak_only_deployment_pricing(): + """End to end through the cost calculator: with ``custom_pricing`` set and + a ``router_model_id`` whose entry carries only an always-on off-peak block, + the request bills at the block's rates rather than the shared backend rate. + """ + from litellm import Router + from litellm.types.utils import ModelResponse, Usage + + block = { + "hours_utc": "00:00-00:00", + "input_cost_per_token": 5e-05, + "output_cost_per_token": 1e-04, + } + shared_keys = ["gpt-4o-mini", "openai/gpt-4o-mini"] + deployment_id = "offpeak-only-dep-2" + original_entries = _snapshot_model_cost_entries(shared_keys + [deployment_id]) + + router = Router( + model_list=[ + { + "model_name": "offpeak-only", + "litellm_params": { + "model": "openai/gpt-4o-mini", + "api_key": "fake-key-for-registration", + }, + "model_info": {"id": deployment_id, "off_peak_pricing": dict(block)}, + } + ] + ) + + try: + response = ModelResponse( + model="gpt-4o-mini", + usage=Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150), + ) + cost = litellm.completion_cost( + completion_response=response, + model="openai/gpt-4o-mini", + custom_llm_provider="openai", + custom_pricing=True, + router_model_id=deployment_id, + ) + assert cost == pytest.approx(100 * 5e-05 + 50 * 1e-04) + finally: + _restore_model_cost_entries(original_entries) + del router diff --git a/tests/test_litellm/test_router_model_cost_isolation.py b/tests/test_litellm/test_router_model_cost_isolation.py index b580b03574e..7b7a962bf00 100644 --- a/tests/test_litellm/test_router_model_cost_isolation.py +++ b/tests/test_litellm/test_router_model_cost_isolation.py @@ -538,6 +538,157 @@ def test_inherit_builtin_cache_pricing_noop_for_unknown_backend(): assert model_info == {"input_cost_per_token": 0.000003} +def test_inherit_builtin_base_rates_for_off_peak_fills_missing_rates(): + """Direct unit test of the helper: an entry carrying only an + off_peak_pricing block inherits the backend model's built-in base token + rates, so cost lookup via the deployment id can bill standard rates + outside the windows. + """ + backend_model = "gpt-4o-mini" + builtin_info = litellm.get_model_info(model=backend_model, custom_llm_provider="openai") + off_peak_block = { + "hours_utc": "00:00-00:00", + "input_cost_per_token": 5e-07, + "output_cost_per_token": 1e-06, + } + model_info = {"off_peak_pricing": off_peak_block} + + Router._inherit_builtin_base_rates_for_off_peak( + model_info=model_info, + backend_model=backend_model, + custom_llm_provider="openai", + ) + + assert model_info["input_cost_per_token"] == builtin_info["input_cost_per_token"] + assert model_info["output_cost_per_token"] == builtin_info["output_cost_per_token"] + assert model_info["off_peak_pricing"] == off_peak_block + + +def test_inherit_builtin_base_rates_for_off_peak_carries_threshold_rates(): + """A backend with above-threshold pricing hands the whole rate structure to + the deployment entry, so peak-hour billing of large prompts through that + entry matches the shared backend entry instead of flattening to the base + rate. + """ + backend_model = "gemini/gemini-2.5-pro" + builtin_info = litellm.get_model_info(model=backend_model) + assert builtin_info["input_cost_per_token_above_200k_tokens"] is not None + + model_info = { + "off_peak_pricing": {"hours_utc": "00:00-00:00", "input_cost_per_token": 5e-07}, + } + + Router._inherit_builtin_base_rates_for_off_peak( + model_info=model_info, + backend_model=backend_model, + custom_llm_provider="gemini", + ) + + assert model_info["input_cost_per_token"] == builtin_info["input_cost_per_token"] + assert ( + model_info["input_cost_per_token_above_200k_tokens"] + == builtin_info["input_cost_per_token_above_200k_tokens"] + ) + assert ( + model_info["output_cost_per_token_above_200k_tokens"] + == builtin_info["output_cost_per_token_above_200k_tokens"] + ) + + +def test_inherit_builtin_base_rates_for_off_peak_carries_companion_billing_fields(): + """Billing rules that are not literal cost rates, like the web search + billing unit, must ride along, or grounding and regional uplifts would + bill differently through the deployment entry than through the shared + backend entry. + """ + backend_model = "gemini-3-pro-image" + raw_entry = litellm.model_cost[backend_model] + assert raw_entry.get("web_search_billing_unit") is not None + + model_info = { + "off_peak_pricing": {"hours_utc": "00:00-00:00", "input_cost_per_token": 5e-07}, + } + + Router._inherit_builtin_base_rates_for_off_peak( + model_info=model_info, + backend_model=backend_model, + custom_llm_provider=None, + ) + + assert model_info["web_search_billing_unit"] == raw_entry["web_search_billing_unit"] + assert model_info["input_cost_per_token"] == raw_entry["input_cost_per_token"] + + +def test_inherit_builtin_base_rates_for_off_peak_tiered_only_backend_stores_no_zero(): + """A tiered-only backend has no flat token rates; get_model_info synthesizes + zeros for them, and storing those would mark the deployment explicitly + priced free. The tier table itself must carry over as an isolated copy so + mutating the deployment entry never touches the shared cost map. + """ + backend_model = "dashscope/qwen-flash" + raw_tiers = litellm.model_cost[backend_model]["tiered_pricing"] + + model_info = { + "off_peak_pricing": {"hours_utc": "00:00-00:00", "input_cost_per_token": 5e-07}, + } + + Router._inherit_builtin_base_rates_for_off_peak( + model_info=model_info, + backend_model=backend_model, + custom_llm_provider="dashscope", + ) + + assert model_info.get("input_cost_per_token") != 0 + assert model_info.get("output_cost_per_token") != 0 + assert model_info["tiered_pricing"] == raw_tiers + assert model_info["tiered_pricing"] is not raw_tiers + assert model_info["tiered_pricing"][0] is not raw_tiers[0] + + original_first_tier = copy.deepcopy(raw_tiers[0]) + model_info["tiered_pricing"][0]["input_cost_per_token"] = 123.0 + assert raw_tiers[0] == original_first_tier + + +def test_inherit_builtin_base_rates_for_off_peak_leaves_explicit_rates_alone(): + """An entry that sets its own base rate beside the block already counts as + a full custom pricing entry; the helper must not mix builtin rates into it. + """ + model_info = { + "off_peak_pricing": {"hours_utc": "00:00-00:00", "input_cost_per_token": 5e-07}, + "input_cost_per_token": 3e-06, + } + + Router._inherit_builtin_base_rates_for_off_peak( + model_info=model_info, + backend_model="gpt-4o-mini", + custom_llm_provider="openai", + ) + + assert model_info["input_cost_per_token"] == 3e-06 + assert "output_cost_per_token" not in model_info + + +def test_inherit_builtin_base_rates_for_off_peak_noop_without_block_or_backend(): + """Nothing happens without an off_peak_pricing block, and an unmapped + backend model leaves the entry unchanged rather than raising. + """ + plain_info = {"id": "dep-1"} + Router._inherit_builtin_base_rates_for_off_peak( + model_info=plain_info, + backend_model="gpt-4o-mini", + custom_llm_provider="openai", + ) + assert plain_info == {"id": "dep-1"} + + off_peak_info = {"off_peak_pricing": {"hours_utc": "00:00-00:00", "input_cost_per_token": 5e-07}} + Router._inherit_builtin_base_rates_for_off_peak( + model_info=off_peak_info, + backend_model="this-backend-model-does-not-exist-x9y8z7", + custom_llm_provider=None, + ) + assert "input_cost_per_token" not in off_peak_info + + def test_custom_pricing_field_denylist_covers_all_builtin_pricing_fields(): """The shared-backend-key stripping in Router relies on CustomPricingLiteLLMParams enumerating every per-deployment pricing field. diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index 627f341fe2f..521e91daded 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -1005,6 +1005,7 @@ def test_aaamodel_prices_and_context_window_json_is_valid(): "gemini_native_audio": {"type": "boolean"}, "gemini_audio_only_live": {"type": "boolean"}, "supports_embedding_image_input": {"type": "boolean"}, + "supports_forced_tool_use": {"type": "boolean"}, "supports_function_calling": {"type": "boolean"}, "supports_image_input": {"type": "boolean"}, "supports_nova_canvas_image_edit": {"type": "boolean"}, diff --git a/tests/test_litellm/test_video_generation.py b/tests/test_litellm/test_video_generation.py index b166e902d6e..2a60ff9c4b5 100644 --- a/tests/test_litellm/test_video_generation.py +++ b/tests/test_litellm/test_video_generation.py @@ -532,6 +532,41 @@ class TestVideoGeneration: assert abs(cost_for("runwayml/seedance2_5", "480p", 8.0) - 1.6) < 0.001 assert abs(cost_for("runwayml/gen4.5", None, 8.0) - 0.96) < 0.001 + def test_completion_cost_veo_31_tiers_pin_published_rates(self, monkeypatch): + """The gemini and vertex_ai veo 3.1 entries bill Google's published per-second tier rates.""" + from litellm.cost_calculator import completion_cost + + local_map_path = os.path.join( + os.path.dirname(__file__), "..", "..", "model_prices_and_context_window.json" + ) + with open(local_map_path, "r") as f: + monkeypatch.setattr(litellm, "model_cost", json.load(f)) + + def cost_for(model: str, provider: str, resolution: str | None, duration: float) -> float: + mock_response = MagicMock() + mock_response.usage = { + "duration_seconds": duration, + **({"video_resolution": resolution} if resolution else {}), + } + type(mock_response)._hidden_params = {} + return completion_cost( + completion_response=mock_response, + model=model, + call_type="create_video", + custom_llm_provider=provider, + ) + + for provider in ("gemini", "vertex_ai"): + for suffix in ("generate-preview", "generate-001"): + standard = f"{provider}/veo-3.1-{suffix}" + fast = f"{provider}/veo-3.1-fast-{suffix}" + assert abs(cost_for(standard, provider, None, 8.0) - 3.2) < 1e-6 + assert abs(cost_for(standard, provider, "1080p", 8.0) - 3.2) < 1e-6 + assert abs(cost_for(standard, provider, "4k", 8.0) - 4.8) < 1e-6 + assert abs(cost_for(fast, provider, "720p", 8.0) - 0.8) < 1e-6 + assert abs(cost_for(fast, provider, "1080p", 8.0) - 0.96) < 1e-6 + assert abs(cost_for(fast, provider, "4k", 8.0) - 2.4) < 1e-6 + def test_video_generation_with_files(self): """Test video generation with file uploads.""" config = OpenAIVideoConfig() diff --git a/type-discipline-budget.json b/type-discipline-budget.json index 83c49afb538..3d2e97d55a5 100644 --- a/type-discipline-budget.json +++ b/type-discipline-budget.json @@ -1,9 +1,9 @@ { "LIT001": { - "limit": 22403 + "limit": 22367 }, "LIT002": { - "limit": 26780 + "limit": 26777 }, "LIT003": { "limit": 269 @@ -27,10 +27,10 @@ "limit": 0 }, "LIT010": { - "limit": 16512 + "limit": 16507 }, "LIT011": { - "limit": 5537 + "limit": 5535 }, "LIT012": { "limit": 4495 diff --git a/ui/litellm-dashboard/eslint-budgets.json b/ui/litellm-dashboard/eslint-budgets.json index bbf69c4a77a..e8207d179bd 100644 --- a/ui/litellm-dashboard/eslint-budgets.json +++ b/ui/litellm-dashboard/eslint-budgets.json @@ -4,5 +4,8 @@ "complexity": { "max": 140, "target": 80 }, "max-depth": { "max": 70, "target": 30 }, "local/no-large-inline-object-arg": { "max": 559, "target": 300 }, - "local/no-long-condition-chain": { "max": 265, "target": 120 } + "local/no-long-condition-chain": { "max": 265, "target": 120 }, + "testing-library/no-container": { "max": 133, "target": 50 }, + "testing-library/no-node-access": { "max": 716, "target": 500 }, + "testing-library/prefer-screen-queries": { "max": 18, "target": 18 } } diff --git a/ui/litellm-dashboard/eslint.config.mjs b/ui/litellm-dashboard/eslint.config.mjs index 23cc5096bb1..f5e3b23b3ec 100644 --- a/ui/litellm-dashboard/eslint.config.mjs +++ b/ui/litellm-dashboard/eslint.config.mjs @@ -104,10 +104,13 @@ const eslintConfig = [ plugins: { "testing-library": testingLibrary, "jest-dom": jestDom }, rules: { "testing-library/await-async-queries": "error", + "testing-library/no-container": "warn", + "testing-library/no-node-access": "warn", "testing-library/no-wait-for-multiple-assertions": "error", "testing-library/no-wait-for-side-effects": "error", "testing-library/prefer-find-by": "error", "testing-library/prefer-presence-queries": "error", + "testing-library/prefer-screen-queries": "warn", "jest-dom/prefer-checked": "error", "jest-dom/prefer-empty": "error", "jest-dom/prefer-enabled-disabled": "error", diff --git a/ui/litellm-dashboard/package-lock.json b/ui/litellm-dashboard/package-lock.json index d2a64b93384..4e5b0c1dda6 100644 --- a/ui/litellm-dashboard/package-lock.json +++ b/ui/litellm-dashboard/package-lock.json @@ -4891,9 +4891,9 @@ } }, "node_modules/baseline-browser-mapping": { - "version": "2.10.27", - "resolved": "https://registry.npmjs.org/baseline-browser-mapping/-/baseline-browser-mapping-2.10.27.tgz", - "integrity": "sha512-zEs/ufmZoUd7WftKpKyXaT6RFxpQ5Qm9xytKRHvJfxFV9DFJkZph9RvJ1LcOUi0Z1ZVijMte65JbILeV+8QQEA==", + "version": "2.11.20", + "resolved": "https://registry.npmjs.org/baseline-browser-mapping/-/baseline-browser-mapping-2.11.20.tgz", + "integrity": "sha512-H0ulySigv6icDJ1F7SjtdCD6PrhTpdYCmP0CactWy1+ekh0AFd0o1Wn5T8b+hnTmdBx19u9yhL6wvCylXMY7zw==", "license": "Apache-2.0", "bin": { "baseline-browser-mapping": "dist/cli.cjs" @@ -4939,9 +4939,9 @@ } }, "node_modules/browserslist": { - "version": "4.28.2", - "resolved": "https://registry.npmjs.org/browserslist/-/browserslist-4.28.2.tgz", - "integrity": "sha512-48xSriZYYg+8qXna9kwqjIVzuQxi+KYWp2+5nCYnYKPTr0LvD89Jqk2Or5ogxz0NUMfIjhh2lIUX/LyX9B4oIg==", + "version": "4.28.8", + "resolved": "https://registry.npmjs.org/browserslist/-/browserslist-4.28.8.tgz", + "integrity": "sha512-V2NpofLblG64mfOtSgDhOJESZEGogzDMBv/q+W6oc4LXWP/q75eOXoOaaOu1EOadB9U4Bwx/e0yzbvwKH8zalA==", "dev": true, "funding": [ { @@ -4959,11 +4959,11 @@ ], "license": "MIT", "dependencies": { - "baseline-browser-mapping": "^2.10.12", - "caniuse-lite": "^1.0.30001782", - "electron-to-chromium": "^1.5.328", - "node-releases": "^2.0.36", - "update-browserslist-db": "^1.2.3" + "baseline-browser-mapping": "^2.11.12", + "caniuse-lite": "^1.0.30001809", + "electron-to-chromium": "^1.5.402", + "node-releases": "^2.0.53", + "update-browserslist-db": "^1.3.0" }, "bin": { "browserslist": "cli.js" @@ -5042,9 +5042,9 @@ } }, "node_modules/caniuse-lite": { - "version": "1.0.30001791", - "resolved": "https://registry.npmjs.org/caniuse-lite/-/caniuse-lite-1.0.30001791.tgz", - "integrity": "sha512-yk0l/YSrOnFZk3UROpDLQD9+kC1l4meK/wed583AXrzoarMGJcbRi2Q4RaUYbKxYAsZ8sWmaSa/DsLmdBeI1vQ==", + "version": "1.0.30001810", + "resolved": "https://registry.npmjs.org/caniuse-lite/-/caniuse-lite-1.0.30001810.tgz", + "integrity": "sha512-TITQPUkaz+aVk5GL6NhOdwk1aEaNTSDPsGFWrTuhKGtjTF70jL/Oht2W4c6rXUe5fu7Ie19VIahAXHIIiWWNeg==", "funding": [ { "type": "opencollective", @@ -5706,9 +5706,9 @@ } }, "node_modules/electron-to-chromium": { - "version": "1.5.349", - "resolved": "https://registry.npmjs.org/electron-to-chromium/-/electron-to-chromium-1.5.349.tgz", - "integrity": "sha512-QsWVGyRuY07Aqb234QytTfwd5d9AJlfNIQ5wIOl1L+PZDzI9d9+Fn0FRale/QYlFxt/bUnB0/nLd1jFPGxGK1A==", + "version": "1.5.416", + "resolved": "https://registry.npmjs.org/electron-to-chromium/-/electron-to-chromium-1.5.416.tgz", + "integrity": "sha512-K6bvB2BjnNrugtIih6ewlbBI9DXa976jIdiIlRLHhBoEI9a4JaQjjHyF+A1IQI543aQYR4LnmOrT/K5fZj0aPA==", "dev": true, "license": "ISC" }, @@ -9901,11 +9901,14 @@ } }, "node_modules/node-releases": { - "version": "2.0.38", - "resolved": "https://registry.npmjs.org/node-releases/-/node-releases-2.0.38.tgz", - "integrity": "sha512-3qT/88Y3FbH/Kx4szpQQ4HzUbVrHPKTLVpVocKiLfoYvw9XSGOX2FmD2d6DrXbVYyAQTF2HeF6My8jmzx7/CRw==", + "version": "2.0.54", + "resolved": "https://registry.npmjs.org/node-releases/-/node-releases-2.0.54.tgz", + "integrity": "sha512-YHs7BmmcsdAI5Ozuf8JZo6PT0mv2GIWC9vMfvUC3dp65M8hn7Ux8CPL+2oBI7juNuj9d0ndhTcznq2ODBps9cQ==", "dev": true, - "license": "MIT" + "license": "MIT", + "engines": { + "node": ">=18" + } }, "node_modules/nuqs": { "version": "2.9.4", @@ -12350,9 +12353,9 @@ } }, "node_modules/update-browserslist-db": { - "version": "1.2.3", - "resolved": "https://registry.npmjs.org/update-browserslist-db/-/update-browserslist-db-1.2.3.tgz", - "integrity": "sha512-Js0m9cx+qOgDxo0eMiFGEueWztz+d4+M3rGlmKPT+T4IS/jP4ylw3Nwpu6cpTTP8R1MAC1kF4VbdLt3ARf209w==", + "version": "1.3.2", + "resolved": "https://registry.npmjs.org/update-browserslist-db/-/update-browserslist-db-1.3.2.tgz", + "integrity": "sha512-UQ+MSxlhRm1bzjhU+DcuXfjFO1FzNtqhK5+9Yvlp90ItDLk5vT932A0rFu619nf7RVS+Y/VeaUW1jaRDqZ8VJw==", "dev": true, "funding": [ { diff --git a/ui/litellm-dashboard/public/assets/logos/alice.svg b/ui/litellm-dashboard/public/assets/logos/alice.svg new file mode 100644 index 00000000000..f18f887b98c --- /dev/null +++ b/ui/litellm-dashboard/public/assets/logos/alice.svg @@ -0,0 +1,4 @@ + + + + diff --git a/ui/litellm-dashboard/src/app/(dashboard)/agents/_components/add_agent_form.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/agents/_components/add_agent_form.integration.test.tsx index bffc7b0fbc8..dad8599e967 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/agents/_components/add_agent_form.integration.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/agents/_components/add_agent_form.integration.test.tsx @@ -5,6 +5,7 @@ import { describe, it, expect, vi, beforeEach } from "vitest"; import AddAgentForm from "./add_agent_form"; import * as networking from "@/components/networking"; import type { AgentCreateInfo } from "@/components/networking"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; vi.mock("@/components/networking", () => ({ createAgentCall: vi.fn(), @@ -309,8 +310,7 @@ describe("AddAgentForm submit payload", () => { await user.type(await screen.findByLabelText("Allowed Models"), "gpt-4o,"); await user.keyboard("{Escape}"); - await user.click(screen.getByLabelText("Allowed Agents (Sub-Agents)")); - await user.click(await screen.findByTitle("Sub Agent One")); + await chooseSelectOption(user, screen.getByLabelText("Allowed Agents (Sub-Agents)"), "Sub Agent One"); await user.keyboard("{Escape}"); await user.click(screen.getByText(/Configure which models, agents, and MCP tools/)); await user.click(screen.getByRole("button", { name: /^Next/ })); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/api-reference/_components/APIReferenceView.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/api-reference/_components/APIReferenceView.test.tsx index b2e0a42eecb..dae19032e20 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/api-reference/_components/APIReferenceView.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/api-reference/_components/APIReferenceView.test.tsx @@ -13,17 +13,17 @@ describe("APIReferenceView", () => { it("uses the API doc base url when provided", () => { const apiDocUrl = "https://docs.litellm.test"; - const { getAllByTestId } = render(); + render(); - const codeBlocks = getAllByTestId(codeBlockTestId); + const codeBlocks = screen.getAllByTestId(codeBlockTestId); expect(codeBlocks[0]).toHaveTextContent(new RegExp(apiDocUrl)); }); it("falls back to the proxy base url when the docs url is missing", () => { const proxyUrl = "https://proxy.litellm.test"; - const { getAllByTestId } = render(); + render(); - const codeBlocks = getAllByTestId(codeBlockTestId); + const codeBlocks = screen.getAllByTestId(codeBlockTestId); expect(codeBlocks[0]).toHaveTextContent(new RegExp(proxyUrl)); }); @@ -31,7 +31,7 @@ describe("APIReferenceView", () => { const apiDocUrl = "https://docs-preferred.litellm.test"; const proxyUrl = "https://proxy-backup.litellm.test"; - const { getAllByTestId } = render( + render( { />, ); - const codeBlocks = getAllByTestId(codeBlockTestId); + const codeBlocks = screen.getAllByTestId(codeBlockTestId); const renderedCode = codeBlocks[0].textContent ?? ""; expect(renderedCode).toContain(apiDocUrl); expect(renderedCode).not.toContain(proxyUrl); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/budget_modal.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/budget_modal.integration.test.tsx index bc920e55abd..17a297d203d 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/budget_modal.integration.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/budget_modal.integration.test.tsx @@ -4,6 +4,7 @@ import React from "react"; import { beforeEach, describe, expect, it, vi } from "vitest"; import BudgetModal from "./budget_modal"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; const { createMock } = vi.hoisted(() => ({ createMock: vi.fn() })); @@ -63,8 +64,7 @@ describe("BudgetModal", () => { await openOptionalSettings(user); fireEvent.change(screen.getByLabelText("Max Budget (USD)"), { target: { value: "42.567" } }); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("monthly")); + await chooseSelectOption(user, screen.getByRole("combobox"), "monthly"); await create(user); @@ -80,8 +80,7 @@ describe("BudgetModal", () => { await openOptionalSettings(user); fireEvent.change(screen.getByLabelText("Max Budget (USD)"), { target: { value: "42.567" } }); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("monthly")); + await chooseSelectOption(user, screen.getByRole("combobox"), "monthly"); await user.click(screen.getByText("Optional Settings")); await waitFor(() => expect(screen.queryByLabelText("Max Budget (USD)")).not.toBeInTheDocument()); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/edit_budget_modal.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/edit_budget_modal.integration.test.tsx index 3fa96b54f1d..fe0ecfc10dd 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/edit_budget_modal.integration.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/budgets/_components/edit_budget_modal.integration.test.tsx @@ -6,6 +6,7 @@ import { beforeEach, describe, expect, it, vi } from "vitest"; import type { components } from "@/lib/http/schema"; import EditBudgetModal from "./edit_budget_modal"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; const { updateMock } = vi.hoisted(() => ({ updateMock: vi.fn() })); @@ -73,8 +74,7 @@ describe("EditBudgetModal", () => { await user.clear(screen.getByLabelText("Max Budget (USD)")); fireEvent.change(screen.getByLabelText("Max Budget (USD)"), { target: { value: "42.567" } }); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("monthly")); + await chooseSelectOption(user, screen.getByRole("combobox"), "monthly"); await save(user); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/cache_settings/RedisTypeSelector.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/cache_settings/RedisTypeSelector.test.tsx index 9d4d5a6d425..372f27e2be1 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/cache_settings/RedisTypeSelector.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/cache_settings/RedisTypeSelector.test.tsx @@ -1,12 +1,10 @@ import { describe, expect, it } from "vitest"; import RedisTypeSelector from "./RedisTypeSelector"; -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; describe("RedisTypeSelector", () => { it("should render the component", () => { - const { getAllByText } = render( - {}} />, - ); - expect(getAllByText(/Redis/i).length).toBeGreaterThan(0); + render( {}} />); + expect(screen.getAllByText(/Redis/i).length).toBeGreaterThan(0); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/coordination_redis_settings/CoordinationRedisTypeSelector.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/coordination_redis_settings/CoordinationRedisTypeSelector.test.tsx index 76287e6e724..563c684237a 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/coordination_redis_settings/CoordinationRedisTypeSelector.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/caching/_components/coordination_redis_settings/CoordinationRedisTypeSelector.test.tsx @@ -5,6 +5,7 @@ import userEvent from "@testing-library/user-event"; import { renderWithProviders } from "@/../tests/test-utils"; import CoordinationRedisTypeSelector from "./CoordinationRedisTypeSelector"; import { COORDINATION_REDIS_TYPE_DESCRIPTIONS } from "./coordinationRedisFields"; +import { chooseSelectOption } from "../../../../../../tests/test-utils"; describe("CoordinationRedisTypeSelector", () => { it("labels the control and shows the current selection", () => { @@ -36,8 +37,7 @@ describe("CoordinationRedisTypeSelector", () => { const user = userEvent.setup(); renderWithProviders(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Cluster")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Cluster"); expect(onTypeChange).toHaveBeenCalledTimes(1); expect(onTypeChange.mock.calls[0][0]).toBe("cluster"); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx index 8c36b934789..f320d8e0f97 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx @@ -1,4 +1,4 @@ -import { fireEvent, render } from "@testing-library/react"; +import { fireEvent, render, screen } from "@testing-library/react"; import { describe, expect, it, vi } from "vitest"; import type { DailyData, KeyMetricWithMetadata, SpendMetrics } from "@/components/UsagePage/types"; @@ -79,25 +79,25 @@ const renderWith = (results: DailyData[], overrides: Partial describe("CacheLeakageCard", () => { it("ranks leaking keys by uncached prompt tokens and shows cache hit ratio", () => { - const { getByText, getByLabelText } = renderWith([ + renderWith([ dayWithKeys("2026-07-12", { "hash-caching": key("caching-key", { prompt_tokens: 1000, cache_read_input_tokens: 900 }), "hash-leaky": key("leaky-key", { prompt_tokens: 10000, cache_read_input_tokens: 0 }), }), ]); - expect(getByText("leaky-key")).toBeInTheDocument(); - expect(getByText("0.0%")).toBeInTheDocument(); - expect(getByText("90.0%")).toBeInTheDocument(); + expect(screen.getByText("leaky-key")).toBeInTheDocument(); + expect(screen.getByText("0.0%")).toBeInTheDocument(); + expect(screen.getByText("90.0%")).toBeInTheDocument(); [ "Input tokens you sent in this range that weren't served from or written to the cache", "Share of your input tokens that were served from the cache", "About how much you'd save if this uncached input used prompt caching. Estimated as uncached input tokens times what your cached traffic already nets per cached token (realized cache savings, after write premiums, ÷ cache read and write tokens). Blank when caching is not currently saving anything overall.", - ].forEach((info) => expect(getByLabelText(info)).toBeInTheDocument()); + ].forEach((info) => expect(screen.getByLabelText(info)).toBeInTheDocument()); }); it("sorts by the clicked column, worst cache hit rate first", () => { - const { getAllByRole, getByText } = renderWith([ + renderWith([ dayWithKeys("2026-07-12", { "hash-a": key("alpha", { prompt_tokens: 10000, @@ -111,48 +111,48 @@ describe("CacheLeakageCard", () => { }), }), ]); - const firstDataRow = () => getAllByRole("row")[1]; + const firstDataRow = () => screen.getAllByRole("row")[1]; expect(firstDataRow()).toHaveTextContent("alpha"); - fireEvent.click(getByText("Cache hit rate")); + fireEvent.click(screen.getByText("Cache hit rate")); expect(firstDataRow()).toHaveTextContent("bravo"); - fireEvent.click(getByText("Cache hit rate")); + fireEvent.click(screen.getByText("Cache hit rate")); expect(firstDataRow()).toHaveTextContent("alpha"); }); it("switches to the model view and lists only Anthropic models", () => { - const { getByText, queryByText } = renderWith([ + renderWith([ dayWithModels("2026-07-12", { "claude-sonnet-5": { prompt_tokens: 5000, cache_read_input_tokens: 0 }, "gpt-4o": { prompt_tokens: 8000, cache_read_input_tokens: 0 }, }), ]); - fireEvent.click(getByText("By model")); + fireEvent.click(screen.getByText("By model")); - expect(getByText("Cache leakage by model")).toBeInTheDocument(); - expect(getByText("claude-sonnet-5")).toBeInTheDocument(); - expect(queryByText("gpt-4o")).not.toBeInTheDocument(); + expect(screen.getByText("Cache leakage by model")).toBeInTheDocument(); + expect(screen.getByText("claude-sonnet-5")).toBeInTheDocument(); + expect(screen.queryByText("gpt-4o")).not.toBeInTheDocument(); }); it("shows an empty state when no key used tokens in the range", () => { - const { getByText, queryByRole } = renderWith([dayWithKeys("2026-07-12", {})]); + renderWith([dayWithKeys("2026-07-12", {})]); - expect(getByText("No key usage in this range.")).toBeInTheDocument(); - expect(queryByRole("table")).not.toBeInTheDocument(); + expect(screen.getByText("No key usage in this range.")).toBeInTheDocument(); + expect(screen.queryByRole("table")).not.toBeInTheDocument(); }); it("tells the user the table is still filling in while fallback pages stream", () => { const day = dayWithKeys("2026-07-12", { "hash-leaky": key("leaky-key", { prompt_tokens: 10000, cache_read_input_tokens: 0 }), }); - const { getByText, getByRole } = renderWith([day], { isFetchingMore: true }); + renderWith([day], { isFetchingMore: true }); - expect(getByRole("table")).toBeInTheDocument(); + expect(screen.getByRole("table")).toBeInTheDocument(); expect( - getByText("Data is still loading; rows and totals will update as the rest of the range arrives."), + screen.getByText("Data is still loading; rows and totals will update as the rest of the range arrives."), ).toBeInTheDocument(); }); @@ -160,10 +160,10 @@ describe("CacheLeakageCard", () => { const day = dayWithKeys("2026-07-12", { "hash-leaky": key("leaky-key", { prompt_tokens: 10000, cache_read_input_tokens: 0 }), }); - const { queryByText } = renderWith([day], { loading: true }); + renderWith([day], { loading: true }); expect( - queryByText("Data is still loading; rows and totals will update as the rest of the range arrives."), + screen.queryByText("Data is still loading; rows and totals will update as the rest of the range arrives."), ).not.toBeInTheDocument(); }); @@ -171,10 +171,10 @@ describe("CacheLeakageCard", () => { const day = dayWithKeys("2026-07-12", { "hash-leaky": key("leaky-key", { prompt_tokens: 10000, cache_read_input_tokens: 0 }), }); - const { queryByText } = renderWith([day]); + renderWith([day]); expect( - queryByText("Data is still loading; rows and totals will update as the rest of the range arrives."), + screen.queryByText("Data is still loading; rows and totals will update as the rest of the range arrives."), ).not.toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.activity.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.activity.test.tsx index 1cc7bec13d1..03250e3e53b 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.activity.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.activity.test.tsx @@ -1,5 +1,5 @@ import React from "react"; -import { fireEvent, render, waitFor } from "@testing-library/react"; +import { fireEvent, render, waitFor, screen } from "@testing-library/react"; import { describe, expect, it, vi } from "vitest"; import { QueryClient, QueryClientProvider } from "@tanstack/react-query"; @@ -56,7 +56,7 @@ describe("CostOptimizationView daily activity", () => { useAuthorizedMock.mockReturnValue({ accessToken: "test-token", userId: "u1", userRole: "proxy_admin" }); const queryClient = new QueryClient({ defaultOptions: { queries: { retry: false } } }); - const { getByRole, getByTestId, findByTestId, queryByText } = render( + render( , @@ -64,12 +64,12 @@ describe("CostOptimizationView daily activity", () => { await waitFor(() => expect(mockUserDailyActivityAggregatedCall).toHaveBeenCalledTimes(1)); - fireEvent.click(getByRole("tab", { name: "Prompt Caching" })); - await findByTestId("caching-settings"); + fireEvent.click(screen.getByRole("tab", { name: "Prompt Caching" })); + await screen.findByTestId("caching-settings"); expect(mockUserDailyActivityAggregatedCall).toHaveBeenCalledTimes(1); expect(mockUserDailyActivityCall).not.toHaveBeenCalled(); - expect(queryByText(/Currently fetching spend data/)).not.toBeInTheDocument(); + expect(screen.queryByText(/Currently fetching spend data/)).not.toBeInTheDocument(); }); it("shows the fetch-progress banner while the paginated fallback streams pages in", async () => { @@ -84,13 +84,13 @@ describe("CostOptimizationView daily activity", () => { useAuthorizedMock.mockReturnValue({ accessToken: "test-token", userId: "u1", userRole: "proxy_admin" }); const queryClient = new QueryClient({ defaultOptions: { queries: { retry: false } } }); - const { findByText, getByRole } = render( + render( , ); - expect(await findByText(/Currently fetching spend data: fetched 1 \/ 3 pages/)).toBeInTheDocument(); - expect(getByRole("button", { name: "Stop" })).toBeInTheDocument(); + expect(await screen.findByText(/Currently fetching spend data: fetched 1 \/ 3 pages/)).toBeInTheDocument(); + expect(screen.getByRole("button", { name: "Stop" })).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.test.tsx index d5df5aa75da..028367555a1 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CostOptimizationView.test.tsx @@ -1,5 +1,5 @@ import React from "react"; -import { fireEvent, render } from "@testing-library/react"; +import { fireEvent, render, screen } from "@testing-library/react"; import { beforeEach, describe, expect, it, vi } from "vitest"; import { QueryClient, QueryClientProvider } from "@tanstack/react-query"; @@ -45,32 +45,32 @@ describe("CostOptimizationView", () => { }); it("renders the standard page header with the sidebar's Cost Optimization icon", () => { - const { container, getByRole, getByText } = renderView(); + const { container } = renderView(); - expect(getByRole("heading", { level: 1, name: "Cost Optimization" })).toBeInTheDocument(); - expect(getByText(/Track and configure the mechanisms that save you money/)).toBeInTheDocument(); + expect(screen.getByRole("heading", { level: 1, name: "Cost Optimization" })).toBeInTheDocument(); + expect(screen.getByText(/Track and configure the mechanisms that save you money/)).toBeInTheDocument(); expect(container.querySelector(".lucide-piggy-bank")).not.toBeNull(); }); it("renders the four cost-optimization tabs", () => { - const { getByText } = renderView(); + renderView(); - expect(getByText("Overall")).toBeInTheDocument(); - expect(getByText("Prompt Compression")).toBeInTheDocument(); - expect(getByText("Prompt Caching")).toBeInTheDocument(); - expect(getByText("Auto-Router")).toBeInTheDocument(); + expect(screen.getByText("Overall")).toBeInTheDocument(); + expect(screen.getByText("Prompt Compression")).toBeInTheDocument(); + expect(screen.getByText("Prompt Caching")).toBeInTheDocument(); + expect(screen.getByText("Auto-Router")).toBeInTheDocument(); }); it("defaults to the Overall tab and switches the active tab on click", () => { - const { getByRole } = renderView(); + renderView(); - expect(getByRole("tab", { name: "Overall" })).toHaveAttribute("aria-selected", "true"); - expect(getByRole("tab", { name: "Prompt Compression" })).toHaveAttribute("aria-selected", "false"); + expect(screen.getByRole("tab", { name: "Overall" })).toHaveAttribute("aria-selected", "true"); + expect(screen.getByRole("tab", { name: "Prompt Compression" })).toHaveAttribute("aria-selected", "false"); - fireEvent.click(getByRole("tab", { name: "Prompt Compression" })); + fireEvent.click(screen.getByRole("tab", { name: "Prompt Compression" })); - expect(getByRole("tab", { name: "Overall" })).toHaveAttribute("aria-selected", "false"); - expect(getByRole("tab", { name: "Prompt Compression" })).toHaveAttribute("aria-selected", "true"); + expect(screen.getByRole("tab", { name: "Overall" })).toHaveAttribute("aria-selected", "false"); + expect(screen.getByRole("tab", { name: "Prompt Compression" })).toHaveAttribute("aria-selected", "true"); }); // Unlike the other three pages in this cleanup, Cost Optimization keeps its @@ -80,21 +80,21 @@ describe("CostOptimizationView", () => { // are proxy-admin-only, so those are what disappear. describe("proxy-admin-only tabs", () => { it.each(["Internal User", "Internal Viewer", "Org Admin"])("shows %s the Overall tab only", (userRole) => { - const { getByRole, queryByRole } = renderView(userRole); + renderView(userRole); - expect(getByRole("tab", { name: "Overall" })).toBeInTheDocument(); - expect(queryByRole("tab", { name: "Prompt Compression" })).not.toBeInTheDocument(); - expect(queryByRole("tab", { name: "Prompt Caching" })).not.toBeInTheDocument(); - expect(queryByRole("tab", { name: "Auto-Router" })).not.toBeInTheDocument(); + expect(screen.getByRole("tab", { name: "Overall" })).toBeInTheDocument(); + expect(screen.queryByRole("tab", { name: "Prompt Compression" })).not.toBeInTheDocument(); + expect(screen.queryByRole("tab", { name: "Prompt Caching" })).not.toBeInTheDocument(); + expect(screen.queryByRole("tab", { name: "Auto-Router" })).not.toBeInTheDocument(); }); it("never mounts the panels behind the admin-only endpoints for an internal user", () => { - const { getByTestId, queryByTestId } = renderView("Internal User"); + renderView("Internal User"); - expect(getByTestId("usage-tab")).toBeInTheDocument(); - expect(queryByTestId("compression-tab")).not.toBeInTheDocument(); - expect(queryByTestId("caching-tab")).not.toBeInTheDocument(); - expect(queryByTestId("autorouter-benchmarks-tab")).not.toBeInTheDocument(); + expect(screen.getByTestId("usage-tab")).toBeInTheDocument(); + expect(screen.queryByTestId("compression-tab")).not.toBeInTheDocument(); + expect(screen.queryByTestId("caching-tab")).not.toBeInTheDocument(); + expect(screen.queryByTestId("autorouter-benchmarks-tab")).not.toBeInTheDocument(); }); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/PromptCachingTab.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/PromptCachingTab.test.tsx index 38517dab0ab..2c602033171 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/PromptCachingTab.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/PromptCachingTab.test.tsx @@ -1,4 +1,4 @@ -import { render, waitFor } from "@testing-library/react"; +import { render, waitFor, screen } from "@testing-library/react"; import { describe, expect, it, vi } from "vitest"; const mockGetGeneralSettingsCall = vi.fn(); @@ -37,10 +37,10 @@ describe("PromptCachingTab", () => { cancelled: false, cancel: vi.fn(), }; - const { getByTestId } = render(); + render(); - expect(getByTestId("caching-settings")).toBeInTheDocument(); - expect(getByTestId("cache-leakage-card")).toBeInTheDocument(); + expect(screen.getByTestId("caching-settings")).toBeInTheDocument(); + expect(screen.getByTestId("cache-leakage-card")).toBeInTheDocument(); await waitFor(() => expect(mockCacheLeakageCard).toHaveBeenCalledWith(expect.objectContaining({ activity }))); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/ShadowEvalSection.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/ShadowEvalSection.test.tsx index 64e03985f57..a1de608d0bb 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/ShadowEvalSection.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/ShadowEvalSection.test.tsx @@ -97,6 +97,7 @@ import { useStopShadowEval, type ShadowEvalJob, } from "./useShadowEval"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; const job = (overrides: Partial = {}): ShadowEvalJob => ({ job_id: "job-1", @@ -437,8 +438,7 @@ describe("ShadowEvalSection", () => { await user.click(within(keyList).getByText("prod-alpha")); await user.click(keyInput); await user.click(within(keyList).getByText("staging-beta")); - await user.click(screen.getByPlaceholderText("Select up to 4 auto-routers")); - await user.click(await screen.findByText("gpt-auto")); + await chooseSelectOption(user, screen.getByPlaceholderText("Select up to 4 auto-routers"), "gpt-auto"); expect(screen.getByText("Start shadow eval")).toBeDisabled(); @@ -470,8 +470,7 @@ describe("ShadowEvalSection", () => { await user.click(screen.getByPlaceholderText("Search teams by alias")); const teamList = await screen.findByTestId("paginated-multi-select-list"); await user.click(within(teamList).getByText("engineering")); - await user.click(screen.getByPlaceholderText("Select up to 4 auto-routers")); - await user.click(await screen.findByText("gpt-auto")); + await chooseSelectOption(user, screen.getByPlaceholderText("Select up to 4 auto-routers"), "gpt-auto"); await user.click(screen.getByPlaceholderText("Select a judge model")); await user.click(await screen.findByRole("option", { name: /anthropic\/claude-sonnet-5/ })); await user.click(screen.getByText("Start shadow eval")); @@ -502,8 +501,7 @@ describe("ShadowEvalSection", () => { await user.click(screen.getByPlaceholderText("Search keys by alias")); const keyList = await screen.findByTestId("paginated-multi-select-list"); await user.click(within(keyList).getByText("prod-alpha")); - await user.click(screen.getByPlaceholderText("Select up to 4 auto-routers")); - await user.click(await screen.findByText("gpt-auto")); + await chooseSelectOption(user, screen.getByPlaceholderText("Select up to 4 auto-routers"), "gpt-auto"); await user.click(screen.getByPlaceholderText("Select a judge model")); await user.click(await screen.findByRole("option", { name: /anthropic\/claude-sonnet-5/ })); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/UsageTab.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/UsageTab.test.tsx index df23e5509bf..f85a667a074 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/UsageTab.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/UsageTab.test.tsx @@ -1,4 +1,4 @@ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { beforeEach, describe, expect, it, vi } from "vitest"; import type { ToolSpendResponse } from "@/components/networking"; @@ -152,13 +152,13 @@ describe("UsageTab", () => { gateway_injected_caching_savings_spend: 0.006, compression_saved_tokens: 100000, }; - const { getByText } = renderWith([day("2026-07-12", firstDay), day("2026-07-13", secondDay)]); + renderWith([day("2026-07-12", firstDay), day("2026-07-13", secondDay)]); - expect(getByText("$0.1500")).toBeInTheDocument(); - expect(getByText("$0.1400")).toBeInTheDocument(); - expect(getByText("$0.0100")).toBeInTheDocument(); - expect(getByText("$0.0160")).toBeInTheDocument(); - expect(getByText("140,000 tokens compressed")).toBeInTheDocument(); + expect(screen.getByText("$0.1500")).toBeInTheDocument(); + expect(screen.getByText("$0.1400")).toBeInTheDocument(); + expect(screen.getByText("$0.0100")).toBeInTheDocument(); + expect(screen.getByText("$0.0160")).toBeInTheDocument(); + expect(screen.getByText("140,000 tokens compressed")).toBeInTheDocument(); }); const twoDays = () => [ @@ -167,11 +167,11 @@ describe("UsageTab", () => { ]; it("opens on a running total anchored at $0 at the start of the range", () => { - const { getByTestId } = renderWith(twoDays()); + renderWith(twoDays()); // Cumulative prepends a synthetic $0 point at the range start (Jul 1) so the // line rises from zero rather than floating; the daily running totals follow. - const series = readSeries(getByTestId("area-chart")); + const series = readSeries(screen.getByTestId("area-chart")); expect(series).toHaveLength(3); expect(series[0]).toMatchObject({ date: "Jul 1", Compression: 0, "Prompt caching": 0 }); expect(series[1]).toMatchObject({ Compression: 0.04, "Prompt caching": 0.006 }); @@ -183,12 +183,12 @@ describe("UsageTab", () => { // The original complaint: a one-day range plotted a single floating dot. The // synthetic start anchor gives the line a zero origin to climb from. const oneDay = new Date(2026, 6, 24); - const { getByTestId } = renderWith( - [day("2026-07-24", { compression_savings_spend: 0.2, gateway_injected_caching_savings_spend: 0.05 })], - { from: oneDay, to: oneDay }, - ); + renderWith([day("2026-07-24", { compression_savings_spend: 0.2, gateway_injected_caching_savings_spend: 0.05 })], { + from: oneDay, + to: oneDay, + }); - const series = readSeries(getByTestId("area-chart")); + const series = readSeries(screen.getByTestId("area-chart")); expect(series).toHaveLength(2); expect(series[0]).toMatchObject({ date: "Jul 24", Compression: 0, "Prompt caching": 0 }); expect(series[1]).toMatchObject({ date: "Jul 24", Compression: 0.2, "Prompt caching": 0.05 }); @@ -202,49 +202,49 @@ describe("UsageTab", () => { day("2026-07-13", { gateway_injected_caching_savings_spend: 0.1 }), day("2026-07-12", { gateway_injected_caching_savings_spend: 0.04 }), ]; - const { getByTestId, getByRole } = renderWith(newestFirst); + renderWith(newestFirst); // The $0 anchor leads, then the days climb oldest to newest. - const cumulative = readSeries(getByTestId("area-chart")); + const cumulative = readSeries(screen.getByTestId("area-chart")); expect(cumulative.map((p: { date: string }) => p.date)).toEqual(["Jul 1", "Jul 12", "Jul 13"]); expect(cumulative[1]["Prompt caching"]).toBeCloseTo(0.04, 5); expect(cumulative[2]["Prompt caching"]).toBeCloseTo(0.14, 5); expect(cumulative[2]["Prompt caching"]).toBeGreaterThan(cumulative[1]["Prompt caching"]); - await userEvent.click(getByRole("tab", { name: "Per day" })); - const perDay = readSeries(getByTestId("bar-chart")); + await userEvent.click(screen.getByRole("tab", { name: "Per day" })); + const perDay = readSeries(screen.getByTestId("bar-chart")); expect(perDay.map((p: { date: string }) => p.date)).toEqual(["Jul 12", "Jul 13"]); }); it("draws bars of the raw per-interval readings on the other tab", async () => { - const { getByRole, getByTestId, queryByTestId } = renderWith(twoDays()); + renderWith(twoDays()); // Cumulative opens on the area line. - expect(getByTestId("area-chart")).toBeInTheDocument(); + expect(screen.getByTestId("area-chart")).toBeInTheDocument(); - await userEvent.click(getByRole("tab", { name: "Per day" })); + await userEvent.click(screen.getByRole("tab", { name: "Per day" })); // Per day switches to a bar chart of the unaccumulated daily savings, with no // synthetic anchor prepended. - expect(queryByTestId("area-chart")).not.toBeInTheDocument(); - const series = readSeries(getByTestId("bar-chart")); + expect(screen.queryByTestId("area-chart")).not.toBeInTheDocument(); + const series = readSeries(screen.getByTestId("bar-chart")); expect(series).toHaveLength(2); expect(series[0]).toMatchObject({ Compression: 0.04, "Prompt caching": 0.006 }); expect(series[1]).toMatchObject({ Compression: 0.1, "Prompt caching": 0.01 }); }); it("says what the line means and over what range", async () => { - const { getByText, getByRole } = renderWith(twoDays()); + renderWith(twoDays()); - expect(getByText("Running total saved · Jul 1 – Jul 14 (UTC)")).toBeInTheDocument(); - await userEvent.click(getByRole("tab", { name: "Per day" })); - expect(getByText("Saved per day · Jul 1 – Jul 14 (UTC)")).toBeInTheDocument(); + expect(screen.getByText("Running total saved · Jul 1 – Jul 14 (UTC)")).toBeInTheDocument(); + await userEvent.click(screen.getByRole("tab", { name: "Per day" })); + expect(screen.getByText("Saved per day · Jul 1 – Jul 14 (UTC)")).toBeInTheDocument(); }); it("builds the per-driver donut from the range totals, not the running total", () => { - const { getByTestId } = renderWith(twoDays()); + renderWith(twoDays()); - const slices = JSON.parse(getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); + const slices = JSON.parse(screen.getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); expect(slices).toEqual([ { driver: "Compression", color: "emerald", usd: expect.closeTo(0.14, 5) }, { driver: "Prompt caching", color: "blue", usd: expect.closeTo(0.016, 5) }, @@ -252,9 +252,9 @@ describe("UsageTab", () => { }); it("omits a driver slice when that driver has no savings", () => { - const { getByTestId } = renderWith([day("2026-07-12", { compression_savings_spend: 0.04 })]); + renderWith([day("2026-07-12", { compression_savings_spend: 0.04 })]); - const slices = JSON.parse(getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); + const slices = JSON.parse(screen.getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); expect(slices).toEqual([{ driver: "Compression", color: "emerald", usd: expect.closeTo(0.04, 5) }]); }); @@ -262,7 +262,7 @@ describe("UsageTab", () => { // Stacking sums the series into one bar. Auto-router savings go negative when a // model switch pays for a cold cache, and that segment would be drawn below the // axis while the rest of the bar still read as the day's total. - const { getByRole, getByTestId } = renderWith([ + renderWith([ day("2026-07-12", { compression_savings_spend: 0.1, gateway_injected_caching_savings_spend: 0.02, @@ -270,8 +270,8 @@ describe("UsageTab", () => { }), ]); - await userEvent.click(getByRole("tab", { name: "Per day" })); - const bars = getByTestId("bar-chart"); + await userEvent.click(screen.getByRole("tab", { name: "Per day" })); + const bars = screen.getByTestId("bar-chart"); expect(bars).toHaveAttribute("data-stack", "false"); expect(readSeries(bars)[0]).toMatchObject({ "Auto-router": -0.05 }); }); @@ -281,10 +281,10 @@ describe("UsageTab", () => { // per day"). Hand-rolled rows made it compete with the legend and the toggle for // width, so the header grew a line on one tab and the chart moved with it. CardHeader // sizes the action column to its content and gives the rest to the title column. - const { getByRole, getByTestId, container } = renderWith(twoDays()); + const { container } = renderWith(twoDays()); const header = () => { - const legend = getByTestId("chart-legend"); + const legend = screen.getByTestId("chart-legend"); const action = legend.closest('[data-slot="card-action"]') as HTMLElement; const cardHeader = action.parentElement as HTMLElement; const description = cardHeader.querySelector('[data-slot="card-description"]') as HTMLElement; @@ -295,12 +295,12 @@ describe("UsageTab", () => { expect(before.action).toBeTruthy(); expect(before.description).toBeTruthy(); // the toggle rides in the same action slot as the legend, so neither moves alone - expect(before.action.contains(getByRole("tablist"))).toBe(true); + expect(before.action.contains(screen.getByRole("tablist"))).toBe(true); // the subtitle lives outside that slot, so its length cannot reposition the controls expect(before.action.contains(before.description)).toBe(false); expect(before.description).toHaveTextContent(/Running total saved/); - await userEvent.click(getByRole("tab", { name: "Per day" })); + await userEvent.click(screen.getByRole("tab", { name: "Per day" })); const after = header(); expect(after.action).toBe(before.action); @@ -314,7 +314,7 @@ describe("UsageTab", () => { // Switching models leaves the new one with a cold cache, so a route can cost more // than the baseline would have. A negative slice is meaningless in a donut, but the // total has to keep the loss or the page can only ever report good news. - const { getByText, getByTestId } = renderWith([ + renderWith([ day("2026-07-12", { compression_savings_spend: 0.1, gateway_injected_caching_savings_spend: 0.02, @@ -322,16 +322,16 @@ describe("UsageTab", () => { }), ]); - expect(getByText("$0.0700")).toBeInTheDocument(); - expect(getByText("-$0.0500")).toBeInTheDocument(); + expect(screen.getByText("$0.0700")).toBeInTheDocument(); + expect(screen.getByText("-$0.0500")).toBeInTheDocument(); - const slices = JSON.parse(getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); + const slices = JSON.parse(screen.getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); expect(slices.map((d: { driver: string }) => d.driver)).toEqual(["Compression", "Prompt caching"]); - expect(getByTestId("donut-chart")).toHaveAttribute("data-label", "$0.1200"); + expect(screen.getByTestId("donut-chart")).toHaveAttribute("data-label", "$0.1200"); }); it("carries auto-router savings into the summary card, donut slice, and cumulative series", () => { - const { getByText, getByTestId } = renderWith([ + renderWith([ day("2026-07-12", { compression_savings_spend: 0.04, gateway_injected_caching_savings_spend: 0.006, @@ -345,11 +345,11 @@ describe("UsageTab", () => { ]); // Total saved now sums three drivers, and the auto-router card carries its own total. - expect(getByText("$0.2260")).toBeInTheDocument(); - expect(getByText("$0.0700")).toBeInTheDocument(); + expect(screen.getByText("$0.2260")).toBeInTheDocument(); + expect(screen.getByText("$0.0700")).toBeInTheDocument(); // The driver donut gains a third slice priced from the range totals. - const slices = JSON.parse(getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); + const slices = JSON.parse(screen.getByTestId("donut-chart").getAttribute("data-slices") ?? "[]"); expect(slices).toEqual([ { driver: "Compression", color: "emerald", usd: expect.closeTo(0.14, 5) }, { driver: "Prompt caching", color: "blue", usd: expect.closeTo(0.016, 5) }, @@ -357,7 +357,7 @@ describe("UsageTab", () => { ]); // And the cumulative line accumulates the auto-router series alongside the others. - const series = readSeries(getByTestId("area-chart")); + const series = readSeries(screen.getByTestId("area-chart")); expect(series[2]["Auto-router"]).toBeCloseTo(0.07, 5); }); @@ -371,9 +371,9 @@ describe("UsageTab", () => { start_date: "2026-07-12", end_date: "2026-07-12", }; - const { findAllByTestId } = renderWith([day("2026-07-12", {})], { toolSpend }); + renderWith([day("2026-07-12", {})], { toolSpend }); - const bars = await findAllByTestId("bar-chart"); + const bars = await screen.findAllByTestId("bar-chart"); const series = JSON.parse(bars[0].getAttribute("data-series") ?? "[]"); expect(series[0]).toMatchObject({ tool_name: "search", spend: 4.0 }); // The 64px bar cap is this card's opt-in; the shared BarChart must not cap @@ -391,14 +391,16 @@ describe("UsageTab", () => { start_date: "2026-07-12", end_date: "2026-07-12", }; - const { findAllByTestId, getAllByTestId } = renderWith([day("2026-07-12", {})], { toolSpend }); + renderWith([day("2026-07-12", {})], { toolSpend }); - const bars = await findAllByTestId("bar-chart"); + const bars = await screen.findAllByTestId("bar-chart"); const [totalByTool, dailyByTool] = bars.slice(-2); expect(dailyByTool).toHaveAttribute("data-show-legend", "false"); expect(totalByTool).toHaveAttribute("data-colors", dailyByTool.getAttribute("data-colors")); - const toolLegends = getAllByTestId("chart-legend").filter((legend) => legend.textContent === "search,read_file"); + const toolLegends = screen + .getAllByTestId("chart-legend") + .filter((legend) => legend.textContent === "search,read_file"); expect(toolLegends).toHaveLength(1); }); @@ -415,23 +417,23 @@ describe("UsageTab", () => { it.each(["Internal User", "Internal Viewer", "Org Admin"])( "hides the card and never calls the endpoint for %s", async (userRole) => { - const { queryByText, getByTestId } = renderWith([day("2026-07-12", { compression_savings_spend: 0.04 })], { + renderWith([day("2026-07-12", { compression_savings_spend: 0.04 })], { toolSpend, userRole, }); // Liveness gate: the daily-activity charts still render for this role, // so the absence below is the gate, not an empty tab. - expect(getByTestId("donut-chart")).toBeInTheDocument(); - expect(queryByText("Spend by tool")).not.toBeInTheDocument(); + expect(screen.getByTestId("donut-chart")).toBeInTheDocument(); + expect(screen.queryByText("Spend by tool")).not.toBeInTheDocument(); await vi.waitFor(() => expect(mockGetToolSpend).not.toHaveBeenCalled()); }, ); it("keeps the card and the endpoint call for an admin", async () => { - const { findByText } = renderWith([day("2026-07-12", { compression_savings_spend: 0.04 })], { toolSpend }); + renderWith([day("2026-07-12", { compression_savings_spend: 0.04 })], { toolSpend }); - expect(await findByText("Spend by tool")).toBeInTheDocument(); + expect(await screen.findByText("Spend by tool")).toBeInTheDocument(); expect(mockGetToolSpend).toHaveBeenCalled(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_configs.ts b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_configs.ts index 03cfeed42ff..7785a8e44ab 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_configs.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_configs.ts @@ -312,4 +312,10 @@ export const GUARDRAIL_PRESETS: Record = { mode: "pre_call", defaultOn: false, }, + alice: { + provider: "Alice", + guardrailNameSuggestion: "Alice", + mode: "pre_call", + defaultOn: false, + }, }; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.test.ts b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.test.ts index 13909e48185..1e486639840 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.test.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.test.ts @@ -27,6 +27,7 @@ const EXPECTED_PARTNER_LOGO_FILES: Record = { deepkeep: "deepkeep.svg", repelloai: "repelloai.png", straiker: "straiker.svg", + alice: "alice.svg", }; describe("guardrail_garden_data logos", () => { diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.ts b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.ts index 744af89a357..931b3a111d8 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_data.ts @@ -464,6 +464,16 @@ export const PARTNER_GUARDRAIL_CARDS: GuardrailCardInfo[] = [ tags: ["Agentic", "Prompt Injection", "Tool Misuse", "MCP", "Skills"], providerKey: "Straiker", }, + { + id: "alice", + name: "Alice", + description: + "Policy-based guardrails for prompts and model responses, evaluated per application so one proxy can enforce a different policy set per team or product.", + category: "partner", + logo: guardrailLogoMap["Alice"], + tags: ["Content Moderation", "Prompt Injection", "PII", "Policy"], + providerKey: "Alice", + }, ]; export const ALL_CARDS = [...LITELLM_CONTENT_FILTER_CARDS, ...PARTNER_GUARDRAIL_CARDS]; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_detail.tsx b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_detail.tsx index 0a0f70b6bc2..dda56a06ab1 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_detail.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_garden_detail.tsx @@ -1,6 +1,7 @@ import React, { useState } from "react"; import { ArrowLeft } from "lucide-react"; import { Button } from "@/components/ui/button"; +import { cn } from "@/lib/cva.config"; import AddGuardrailForm from "./add_guardrail_form"; import { Logo } from "@/components/molecules/logo/Logo"; import { GUARDRAIL_PRESETS } from "./guardrail_garden_configs"; @@ -40,7 +41,7 @@ const GuardrailDetailView: React.FC = ({ card, onBack, const tabs = [{ key: "overview", label: "Overview" }, ...(card.eval ? [{ key: "eval", label: "Eval Results" }] : [])]; return ( -
+
{/* Back link */}
= ({ card, onBack,
{/* ── Header block (Vertex-style) ── */} -
+
-

{card.name}

+

{card.name}

-

{card.description}

+

{card.description}

{/* Action buttons — outlined style like Vertex */}
@@ -66,21 +67,18 @@ const GuardrailDetailView: React.FC = ({ card, onBack,
{/* ── Tab bar ──────────────────────────────────── */} -
-
+
+
{tabs.map((tab) => (
setActiveTab(tab.key)} - style={{ - padding: "12px 20px", - fontSize: 14, - color: activeTab === tab.key ? "#1a73e8" : "#5f6368", - borderBottom: activeTab === tab.key ? "3px solid #1a73e8" : "3px solid transparent", - cursor: "pointer", - fontWeight: activeTab === tab.key ? 500 : 400, - marginBottom: -1, - }} + className={cn( + "-mb-px cursor-pointer border-b-[3px] px-5 py-3 text-sm", + activeTab === tab.key + ? "border-info font-medium text-info" + : "border-transparent font-normal text-muted-foreground", + )} > {tab.label}
@@ -90,31 +88,27 @@ const GuardrailDetailView: React.FC = ({ card, onBack, {/* ── Tab content ──────────────────────────────── */} {activeTab === "overview" && ( -
+
{/* Left column — overview + details table */} -
-

Overview

-

{card.description}

+
+

Overview

+

{card.description}

-

Guardrail Details

-

Details are as follows

+

Guardrail Details

+

Details are as follows

- +
- - - + + + {detailRows.map((row, i) => ( - - - + + + ))} @@ -122,37 +116,30 @@ const GuardrailDetailView: React.FC = ({ card, onBack, {/* Right column — metadata sidebar like Vertex */} -
+
{/* Guardrail ID */} -
-
Guardrail ID
-
litellm/{card.id}
+
+
Guardrail ID
+
litellm/{card.id}
{/* Type */} -
-
Type
-
+
+
Type
+
{card.category === "litellm" ? "Content Filter" : "Partner"}
{/* Tags — pill style like Vertex */} {card.tags.length > 0 && ( -
-
Tags
-
+
+
Tags
+
{card.tags.map((tag) => ( {tag} @@ -166,19 +153,19 @@ const GuardrailDetailView: React.FC = ({ card, onBack, {activeTab === "eval" && (
-

Eval Results

-
- Property - - {card.name} -
Property{card.name}
{row.property}{row.value}
{row.property}{row.value}
+

Eval Results

+
- - - + + + {evalRows.map((row, i) => ( - - - + + + ))} diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info.test.tsx index 2b90a1d8cbc..fcffc2122e7 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info.test.tsx @@ -1,5 +1,5 @@ import * as networking from "@/components/networking"; -import { fireEvent, render, waitFor, within } from "@testing-library/react"; +import { fireEvent, render, waitFor, within, screen } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { afterEach, describe, expect, it, vi } from "vitest"; import GuardrailInfoView from "./guardrail_info"; @@ -65,21 +65,19 @@ describe("Guardrail Info", () => { vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); - const { getAllByText, getByText } = render( - {}} accessToken="123" isAdmin={true} />, - ); + render( {}} accessToken="123" isAdmin={true} />); // Wait for the loading to complete and data to be rendered await waitFor(() => { // The guardrail name appears in multiple places (title and settings tab) - const elements = getAllByText("Test Guardrail"); + const elements = screen.getAllByText("Test Guardrail"); expect(elements.length).toBeGreaterThan(0); }); // Verify other key elements are present - expect(getByText("Back to Guardrails")).toBeInTheDocument(); - expect(getByText("Overview")).toBeInTheDocument(); - expect(getByText("Settings")).toBeInTheDocument(); + expect(screen.getByText("Back to Guardrails")).toBeInTheDocument(); + expect(screen.getByText("Overview")).toBeInTheDocument(); + expect(screen.getByText("Settings")).toBeInTheDocument(); }); it("should render a tag-based mode object rather than crashing the detail view", async () => { @@ -105,11 +103,9 @@ describe("Guardrail Info", () => { vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); - const { findAllByText } = render( - {}} accessToken="123" isAdmin={true} />, - ); + render( {}} accessToken="123" isAdmin={true} />); - expect(await findAllByText("pre_call, post_call (tag-based)")).not.toHaveLength(0); + expect(await screen.findAllByText("pre_call, post_call (tag-based)")).not.toHaveLength(0); }); it("should render the provider logo from the bundled guardrail logo map", async () => { @@ -135,11 +131,9 @@ describe("Guardrail Info", () => { vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); - const { findByAltText } = render( - {}} accessToken="123" isAdmin={true} />, - ); + render( {}} accessToken="123" isAdmin={true} />); - const logo = await findByAltText("Presidio PII logo"); + const logo = await screen.findByAltText("Presidio PII logo"); expect(logo).toHaveAttribute("src", expect.stringContaining("microsoft_azure.svg")); }); @@ -167,25 +161,27 @@ describe("Guardrail Info", () => { vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); - const { getByText, findByText, container } = render( + const { container } = render( {}} accessToken="123" isAdmin={true} />, ); await waitFor(() => { - expect(getByText("Settings")).toBeInTheDocument(); + expect(screen.getByText("Settings")).toBeInTheDocument(); }); // Click the Settings tab - fireEvent.click(getByText("Settings")); + fireEvent.click(screen.getByText("Settings")); // Wait for the Settings panel to render await waitFor(() => { - expect(getByText("Guardrail Settings")).toBeInTheDocument(); + expect(screen.getByText("Guardrail Settings")).toBeInTheDocument(); }); await userEvent.hover(within(container).getByRole("img", { name: "Config guardrail details" })); - expect(await findByText("Guardrail is defined in the config file and cannot be edited.")).toBeInTheDocument(); + expect( + await screen.findByText("Guardrail is defined in the config file and cannot be edited."), + ).toBeInTheDocument(); }); it("should render the guardrail info", async () => { @@ -216,12 +212,10 @@ describe("Guardrail Info", () => { vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); - const { getByText } = render( - {}} accessToken="123" isAdmin={true} />, - ); + render( {}} accessToken="123" isAdmin={true} />); await waitFor(() => { - expect(getByText("PII Entity Configuration")).toBeInTheDocument(); + expect(screen.getByText("PII Entity Configuration")).toBeInTheDocument(); }); }); it("should handle content filter updates correctly", async () => { @@ -251,30 +245,28 @@ describe("Guardrail Info", () => { vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); vi.mocked(networking.updateGuardrailCall).mockResolvedValue({ status: "success" }); - const { getByText, getByLabelText } = render( - {}} accessToken="123" isAdmin={true} />, - ); + render( {}} accessToken="123" isAdmin={true} />); await waitFor(() => { - expect(getByText("Settings")).toBeInTheDocument(); + expect(screen.getByText("Settings")).toBeInTheDocument(); }); // Go to Settings tab - fireEvent.click(getByText("Settings")); + fireEvent.click(screen.getByText("Settings")); await waitFor(() => { - expect(getByText("Guardrail Settings")).toBeInTheDocument(); + expect(screen.getByText("Guardrail Settings")).toBeInTheDocument(); }); // Enter Edit Mode - fireEvent.click(getByText("Edit Settings")); + fireEvent.click(screen.getByText("Edit Settings")); // Modify Guardrail Name to force an update - const nameInput = getByLabelText("Guardrail Name"); + const nameInput = screen.getByLabelText("Guardrail Name"); fireEvent.change(nameInput, { target: { value: "Updated Name" } }); // Save with only name change - const saveButton = getByText("Save Changes"); + const saveButton = screen.getByText("Save Changes"); fireEvent.click(saveButton); await waitFor(() => { @@ -300,16 +292,16 @@ describe("Guardrail Info", () => { // Enter Edit Mode again to make changes await waitFor(() => { - expect(getByText("Edit Settings")).toBeInTheDocument(); + expect(screen.getByText("Edit Settings")).toBeInTheDocument(); }); - fireEvent.click(getByText("Edit Settings")); + fireEvent.click(screen.getByText("Edit Settings")); // Now modify the values using the mock button - const simulateChangeButton = getByText("Simulate Change"); + const simulateChangeButton = screen.getByText("Simulate Change"); fireEvent.click(simulateChangeButton); // Save again - fireEvent.click(getByText("Save Changes")); + fireEvent.click(screen.getByText("Save Changes")); await waitFor(() => { expect(networking.updateGuardrailCall).toHaveBeenCalled(); @@ -339,12 +331,10 @@ describe("Guardrail Info", () => { }); vi.mocked(networking.getGuardrailProviderSpecificParams).mockResolvedValue({}); - const { findByRole, getByRole, getByText } = render( - {}} accessToken="123" isAdmin={true} />, - ); + render( {}} accessToken="123" isAdmin={true} />); - expect(await findByRole("tab", { name: "Overview" })).toHaveAttribute("aria-selected", "true"); - expect(getByRole("tab", { name: "Settings" })).toHaveAttribute("aria-selected", "false"); - expect(getByText("Guardrail Settings")).toBeInTheDocument(); + expect(await screen.findByRole("tab", { name: "Overview" })).toHaveAttribute("aria-selected", "true"); + expect(screen.getByRole("tab", { name: "Settings" })).toHaveAttribute("aria-selected", "false"); + expect(screen.getByText("Guardrail Settings")).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info_helpers.tsx b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info_helpers.tsx index 83038b8e0e7..c1f2ddcf51c 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info_helpers.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/guardrail_info_helpers.tsx @@ -1,5 +1,6 @@ import aimSecurityLogo from "../../../../../public/assets/logos/aim_security.jpeg"; import aktoLogo from "../../../../../public/assets/logos/akto.svg"; +import aliceLogo from "../../../../../public/assets/logos/alice.svg"; import aporiaLogo from "../../../../../public/assets/logos/aporia.png"; import bedrockLogo from "../../../../../public/assets/logos/bedrock.svg"; import catoNetworksLogo from "../../../../../public/assets/logos/cato_networks.svg"; @@ -83,6 +84,7 @@ export const guardrail_provider_map: Record = { Deepkeep: "deepkeep", QostodianNexus: "qostodian_nexus", Repelloai: "repelloai", + Alice: "alice", }; // Function to populate provider map from API response - updates the original map @@ -204,6 +206,7 @@ export const guardrailLogoMap = { "Qostodian Nexus": qohashLogo.src, "RepelloAI Argus": repelloAiLogo.src, Straiker: straikerLogo.src, + Alice: aliceLogo.src, } satisfies Record; export const getGuardrailLogo = (displayName: string): string | undefined => diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_components.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_components.test.tsx index 2f839487111..30ff2cf7c5f 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_components.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_components.test.tsx @@ -1,4 +1,4 @@ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import { describe, it, expect } from "vitest"; import { CategoryFilter, QuickActions, PiiEntityList } from "./pii_components"; import type { PiiEntityCategory } from "@/components/guardrails/types"; @@ -6,25 +6,21 @@ import type { PiiEntityCategory } from "@/components/guardrails/types"; describe("CategoryFilter", () => { it("should render", () => { const emptyCategories: PiiEntityCategory[] = []; - const { getByText } = render( - {}} />, - ); - expect(getByText("Filter by category")).toBeInTheDocument(); + render( {}} />); + expect(screen.getByText("Filter by category")).toBeInTheDocument(); }); }); describe("QuickActions", () => { it("should render", () => { - const { getByText } = render( - {}} onUnselectAll={() => {}} hasSelectedEntities={false} />, - ); - expect(getByText("Quick Actions")).toBeInTheDocument(); + render( {}} onUnselectAll={() => {}} hasSelectedEntities={false} />); + expect(screen.getByText("Quick Actions")).toBeInTheDocument(); }); }); describe("PiiEntityList", () => { it("should render", () => { - const { getByText } = render( + render( { entityToCategoryMap={new Map()} />, ); - expect(getByText("No PII types match your filter criteria")).toBeInTheDocument(); + expect(screen.getByText("No PII types match your filter criteria")).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_configuration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_configuration.test.tsx index 00c568ef35b..4f822578fe8 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_configuration.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/guardrails/_components/pii_configuration.test.tsx @@ -1,10 +1,10 @@ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import { describe, it, expect } from "vitest"; import PiiConfiguration from "./pii_configuration"; describe("PiiConfiguration", () => { it("should render", () => { - const { getByText } = render( + render( { entityCategories={[]} />, ); - expect(getByText("Configure PII Protection")).toBeInTheDocument(); + expect(screen.getByText("Configure PII Protection")).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/ToolTestPanel.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/ToolTestPanel.test.tsx index 26e66c8338c..81c55e7982a 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/ToolTestPanel.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/ToolTestPanel.test.tsx @@ -6,6 +6,7 @@ import { describe, expect, it, vi } from "vitest"; import { ToolTestPanel } from "./ToolTestPanel"; import { InputSchema, MCPTool } from "@/components/mcp_tools/types"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; const buildTool = (schema: InputSchema | string): MCPTool => ({ name: "demo-tool", @@ -229,8 +230,7 @@ describe("ToolTestPanel argument payload", () => { const onSubmit = await submitPanel( { type: "object", properties: { active: { type: "boolean", default: false } } }, async (user) => { - await user.click(screen.getByLabelText("active")); - await user.click(await screen.findByText("True")); + await chooseSelectOption(user, screen.getByLabelText("active"), "True"); }, ); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx index 8dca87e24ef..8abb8855e3d 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/mcp-servers/_components/mcp_servers.test.tsx @@ -45,7 +45,7 @@ describe("MCPServers", () => { vi.mocked(networking.fetchMCPServers).mockResolvedValue([]); const queryClient = createQueryClient(); - const { getByText } = render( + render( , @@ -53,11 +53,11 @@ describe("MCPServers", () => { // Wait for the component to load and check if title renders await waitFor(() => { - expect(getByText("MCP Servers")).toBeInTheDocument(); + expect(screen.getByText("MCP Servers")).toBeInTheDocument(); }); // Verify the title is rendered - expect(getByText("MCP Servers")).toBeInTheDocument(); + expect(screen.getByText("MCP Servers")).toBeInTheDocument(); }); it("should render mocked MCP servers data in the table", async () => { @@ -96,7 +96,7 @@ describe("MCPServers", () => { vi.mocked(networking.fetchMCPServers).mockResolvedValue(mockServers); const queryClient = createQueryClient(); - const { getByText, getAllByText } = render( + render( , @@ -104,19 +104,19 @@ describe("MCPServers", () => { // Wait for the component to load await waitFor(() => { - expect(getByText("MCP Servers")).toBeInTheDocument(); + expect(screen.getByText("MCP Servers")).toBeInTheDocument(); }); // Wait for the mocked data to render in the table await waitFor(() => { - expect(getByText("Test Server 1")).toBeInTheDocument(); + expect(screen.getByText("Test Server 1")).toBeInTheDocument(); }); // Verify the mocked server data is rendered in the table - expect(getByText("Test Server 1")).toBeInTheDocument(); - expect(getByText("Test Server 2")).toBeInTheDocument(); - expect(getAllByText("test-server-1").length).toBeGreaterThan(0); - expect(getAllByText("test-server-2").length).toBeGreaterThan(0); + expect(screen.getByText("Test Server 1")).toBeInTheDocument(); + expect(screen.getByText("Test Server 2")).toBeInTheDocument(); + expect(screen.getAllByText("test-server-1").length).toBeGreaterThan(0); + expect(screen.getAllByText("test-server-2").length).toBeGreaterThan(0); // Verify the API was called // Note: useMCPServers uses useAuthorized() internally, which returns "123" from global mock @@ -168,7 +168,7 @@ describe("MCPServers", () => { vi.mocked(networking.fetchMCPServerHealth).mockResolvedValue(mockHealthStatuses); const queryClient = createQueryClient(); - const { getByText } = render( + render( , @@ -176,7 +176,7 @@ describe("MCPServers", () => { // Wait for the component to load await waitFor(() => { - expect(getByText("MCP Servers")).toBeInTheDocument(); + expect(screen.getByText("MCP Servers")).toBeInTheDocument(); }); // Verify the health check API was called (without a server ID filter — the hook always @@ -211,7 +211,7 @@ describe("MCPServers", () => { ); const queryClient = createQueryClient(); - const { getByText } = render( + render( , @@ -219,7 +219,7 @@ describe("MCPServers", () => { // Wait for the component to load await waitFor(() => { - expect(getByText("MCP Servers")).toBeInTheDocument(); + expect(screen.getByText("MCP Servers")).toBeInTheDocument(); }); // Verify that health check was initiated diff --git a/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/components/PriceDataManagementTab.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/components/PriceDataManagementTab.test.tsx index 8b34d61ebad..7cd418bc176 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/components/PriceDataManagementTab.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/components/PriceDataManagementTab.test.tsx @@ -1,5 +1,5 @@ /* @vitest-environment jsdom */ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import { describe, expect, it, vi } from "vitest"; import PriceDataManagementTab from "./PriceDataManagementTab"; @@ -11,7 +11,7 @@ vi.mock("@/app/(dashboard)/hooks/models/useModelCostMap", () => ({ describe("PriceDataManagementTab", () => { it("renders its content standalone, without a tab-panel ancestor", () => { - const { getByText } = render(); - expect(getByText("Price Data Management")).toBeInTheDocument(); + render(); + expect(screen.getByText("Price Data Management")).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.test.tsx index 521f89a39f2..a504d75bb63 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/page.test.tsx @@ -1,6 +1,6 @@ /* @vitest-environment jsdom */ import { QueryClient, QueryClientProvider } from "@tanstack/react-query"; -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { beforeEach, describe, expect, it, vi } from "vitest"; import ModelsAndEndpointsPage from "./page"; @@ -64,48 +64,48 @@ describe("ModelsAndEndpointsPage", () => { }); it("renders the admin tab bar and the All Models panel by default", () => { - const { getByRole, getByTestId } = renderPage(); - expect(getByRole("tab", { name: "All Models" })).toBeInTheDocument(); - expect(getByRole("tab", { name: "LLM Credentials" })).toBeInTheDocument(); - expect(getByRole("tab", { name: "Health Status" })).toBeInTheDocument(); - expect(getByTestId("panel-all-models")).toBeInTheDocument(); + renderPage(); + expect(screen.getByRole("tab", { name: "All Models" })).toBeInTheDocument(); + expect(screen.getByRole("tab", { name: "LLM Credentials" })).toBeInTheDocument(); + expect(screen.getByRole("tab", { name: "Health Status" })).toBeInTheDocument(); + expect(screen.getByTestId("panel-all-models")).toBeInTheDocument(); }); it("switches tabs in-memory, mounting only the active panel", async () => { const user = userEvent.setup(); - const { getByRole, getByTestId, queryByTestId } = renderPage(); - await user.click(getByRole("tab", { name: "Health Status" })); - expect(getByTestId("panel-health")).toBeInTheDocument(); - expect(queryByTestId("panel-all-models")).not.toBeInTheDocument(); + renderPage(); + await user.click(screen.getByRole("tab", { name: "Health Status" })); + expect(screen.getByTestId("panel-health")).toBeInTheDocument(); + expect(screen.queryByTestId("panel-all-models")).not.toBeInTheDocument(); }); it("renders the model detail overlay from the ?model drill-in and hides the tabs", () => { detailState.modelId = "abc-123"; - const { getByTestId, queryByRole } = renderPage(); - expect(getByTestId("model-info")).toHaveTextContent("model:abc-123"); - expect(queryByRole("tab", { name: "All Models" })).not.toBeInTheDocument(); + renderPage(); + expect(screen.getByTestId("model-info")).toHaveTextContent("model:abc-123"); + expect(screen.queryByRole("tab", { name: "All Models" })).not.toBeInTheDocument(); }); it("renders the team detail overlay from the ?team drill-in", () => { detailState.teamId = "team-9"; - const { getByTestId } = renderPage(); - expect(getByTestId("team-info")).toHaveTextContent("team:team-9"); + renderPage(); + expect(screen.getByTestId("team-info")).toHaveTextContent("team:team-9"); }); it("hides admin-only tabs for a non-admin user", () => { mockUseAuthorized.mockReturnValue(NON_ADMIN); - const { queryByRole } = renderPage(); - expect(queryByRole("tab", { name: "LLM Credentials" })).not.toBeInTheDocument(); - expect(queryByRole("tab", { name: "Health Status" })).not.toBeInTheDocument(); + renderPage(); + expect(screen.queryByRole("tab", { name: "LLM Credentials" })).not.toBeInTheDocument(); + expect(screen.queryByRole("tab", { name: "Health Status" })).not.toBeInTheDocument(); }); // Auto-routers are excluded from the All Models table, so this tab is their home: the only // place in the product to list, create, edit or delete one. describe("Auto-Routers tab", () => { it("sits third, after All Models and Add Model", () => { - const { getAllByRole } = renderPage(); + renderPage(); - const tabs = getAllByRole("tab").map((tab) => tab.textContent); + const tabs = screen.getAllByRole("tab").map((tab) => tab.textContent); expect(tabs[0]).toContain("All Models"); expect(tabs[1]).toBe("Add Model"); expect(tabs[2]).toContain("Auto-Routers"); @@ -115,17 +115,17 @@ describe("ModelsAndEndpointsPage", () => { it("renders its panel when selected", async () => { const user = userEvent.setup(); - const { getByRole, getByTestId } = renderPage(); + renderPage(); - await user.click(getByRole("tab", { name: /Auto-Routers/ })); - expect(getByTestId("panel-auto-routers")).toBeInTheDocument(); + await user.click(screen.getByRole("tab", { name: /Auto-Routers/ })); + expect(screen.getByTestId("panel-auto-routers")).toBeInTheDocument(); }); it("is hidden from non-admins, who cannot write models", () => { mockUseAuthorized.mockReturnValue(NON_ADMIN); - const { queryByRole } = renderPage(); + renderPage(); - expect(queryByRole("tab", { name: /Auto-Routers/ })).not.toBeInTheDocument(); + expect(screen.queryByRole("tab", { name: /Auto-Routers/ })).not.toBeInTheDocument(); }); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.test.tsx index a83c11d1444..ed02e7e16cc 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.test.tsx @@ -93,13 +93,13 @@ describe("ChatMessageBubble", () => { ])("should paint the $role surface from theme tokens, not fixed colours", ({ role, bubble, avatar }) => { render(); - const header = screen.getByText(role).closest("div") as HTMLElement; - const surface = header.parentElement as HTMLElement; + const surface = screen.getByTestId("message-surface"); + const avatarEl = screen.getByTestId("message-avatar"); expect(surface).toHaveClass(...bubble); expect(surface).not.toHaveAttribute("style"); - expect(header.firstElementChild).toHaveClass(avatar); - expect(header.firstElementChild).not.toHaveAttribute("style"); + expect(avatarEl).toHaveClass(avatar); + expect(avatarEl).not.toHaveAttribute("style"); }); it("should show model badge for assistant messages when model is provided", () => { diff --git a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.tsx b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.tsx index 8c54d9e89fa..bd6a4bc49a4 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatMessageBubble.tsx @@ -46,6 +46,7 @@ function ChatMessageBubble({ return (
{ describe("CompareUI", () => { it("should render", () => { - const { getByTestId } = render(); - expect(getByTestId("comparison-panel-1")).toBeInTheDocument(); - expect(getByTestId("comparison-panel-2")).toBeInTheDocument(); - expect(getByTestId("message-input")).toBeInTheDocument(); + render(); + expect(screen.getByTestId("comparison-panel-1")).toBeInTheDocument(); + expect(screen.getByTestId("comparison-panel-2")).toBeInTheDocument(); + expect(screen.getByTestId("message-input")).toBeInTheDocument(); }); it("adds a comparison when Add Comparison button is clicked", async () => { const user = userEvent.setup(); - const { container, getByTestId } = render( - , - ); + const { container } = render(); // Verify initial state: 2 comparison panels - expect(getByTestId("comparison-panel-1")).toBeInTheDocument(); - expect(getByTestId("comparison-panel-2")).toBeInTheDocument(); + expect(screen.getByTestId("comparison-panel-1")).toBeInTheDocument(); + expect(screen.getByTestId("comparison-panel-2")).toBeInTheDocument(); let comparisonPanels = container.querySelectorAll('[data-testid^="comparison-panel-"]'); expect(comparisonPanels).toHaveLength(2); @@ -117,15 +115,13 @@ describe("CompareUI", () => { }); // Verify the original 2 panels are still there - expect(getByTestId("comparison-panel-1")).toBeInTheDocument(); - expect(getByTestId("comparison-panel-2")).toBeInTheDocument(); + expect(screen.getByTestId("comparison-panel-1")).toBeInTheDocument(); + expect(screen.getByTestId("comparison-panel-2")).toBeInTheDocument(); }); it("should handle image upload and send message with attachment", async () => { const user = userEvent.setup(); - const { getByTestId, queryByTestId } = render( - , - ); + render(); const file = new File(["test content"], "test-image.png", { type: "image/png" }); @@ -138,13 +134,13 @@ describe("CompareUI", () => { } await waitFor(() => { - expect(getByTestId("has-attachment")).toBeInTheDocument(); + expect(screen.getByTestId("has-attachment")).toBeInTheDocument(); }); - const textarea = getByTestId("message-textarea"); + const textarea = screen.getByTestId("message-textarea"); fireEvent.change(textarea, { target: { value: "Describe this image" } }); - const sendButton = getByTestId("send-button"); + const sendButton = screen.getByTestId("send-button"); expect(sendButton).toBeEnabled(); await user.click(sendButton); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/compareUI/components/MessageDisplay.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/compareUI/components/MessageDisplay.test.tsx index 72a1d41f9fe..69b50d4088f 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/compareUI/components/MessageDisplay.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/compareUI/components/MessageDisplay.test.tsx @@ -1,4 +1,4 @@ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import { describe, expect, it, vi } from "vitest"; import type { MessageType } from "@/components/chat_ui/types"; import { MessageDisplay } from "./MessageDisplay"; @@ -39,9 +39,9 @@ describe("MessageDisplay", () => { model: "gpt-4", }, ]; - const { getByText } = render(); - expect(getByText("Hello")).toBeInTheDocument(); - expect(getByText("Hi there!")).toBeInTheDocument(); + render(); + expect(screen.getByText("Hello")).toBeInTheDocument(); + expect(screen.getByText("Hi there!")).toBeInTheDocument(); }); it("displays user and assistant messages with proper grouping and shows loading state", () => { @@ -64,13 +64,13 @@ describe("MessageDisplay", () => { }, }, ]; - const { getByText, getByTestId } = render(); - expect(getByText("You")).toBeInTheDocument(); - expect(getByText("What is 2+2?")).toBeInTheDocument(); - expect(getByText("gpt-4")).toBeInTheDocument(); - expect(getByText("calculator")).toBeInTheDocument(); - expect(getByText("2+2 equals 4")).toBeInTheDocument(); - expect(getByTestId("response-metrics")).toBeInTheDocument(); + render(); + expect(screen.getByText("You")).toBeInTheDocument(); + expect(screen.getByText("What is 2+2?")).toBeInTheDocument(); + expect(screen.getByText("gpt-4")).toBeInTheDocument(); + expect(screen.getByText("calculator")).toBeInTheDocument(); + expect(screen.getByText("2+2 equals 4")).toBeInTheDocument(); + expect(screen.getByTestId("response-metrics")).toBeInTheDocument(); }); it("should display image attachment in user message", () => { @@ -86,10 +86,10 @@ describe("MessageDisplay", () => { model: "gpt-4", }, ]; - const { getByTestId, getByText } = render(); - expect(getByText("What is in this image? [Image attached]")).toBeInTheDocument(); - expect(getByTestId("chat-image-renderer")).toBeInTheDocument(); - const image = getByTestId("chat-image-renderer").querySelector("img"); + render(); + expect(screen.getByText("What is in this image? [Image attached]")).toBeInTheDocument(); + expect(screen.getByTestId("chat-image-renderer")).toBeInTheDocument(); + const image = screen.getByTestId("chat-image-renderer").querySelector("img"); expect(image).toHaveAttribute("src", "blob:test-image-url"); }); }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/policies/_components/policy_test_panel.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/policies/_components/policy_test_panel.test.tsx index c8779be06c4..f23145a917a 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/policies/_components/policy_test_panel.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/policies/_components/policy_test_panel.test.tsx @@ -5,6 +5,7 @@ import userEvent, { PointerEventsCheckLevel } from "@testing-library/user-event" import { renderWithProviders } from "@/../tests/test-utils"; import * as networking from "@/components/networking"; import PolicyTestPanel from "./policy_test_panel"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; vi.mock("@/components/networking"); @@ -24,8 +25,7 @@ const setup = () => { }; const pickOption = async (user: ReturnType, label: string, option: string) => { - await user.click(screen.getByLabelText(label)); - await user.click(await screen.findByTitle(option)); + await chooseSelectOption(user, screen.getByLabelText(label), option); }; const simulate = async (user: ReturnType) => { diff --git a/ui/litellm-dashboard/src/app/(dashboard)/prompts/_components/index.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/prompts/_components/index.test.tsx index d6a4aaea2e1..3b0ee3a3ce2 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/prompts/_components/index.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/prompts/_components/index.test.tsx @@ -5,6 +5,7 @@ import { beforeEach, describe, expect, it, vi } from "vitest"; import { deletePromptCall, getPromptsList } from "@/components/networking"; import PromptsPanel from "./index"; +import { chooseSelectOption } from "../../../../../tests/test-utils"; vi.mock("@/components/networking", () => ({ getPromptsList: vi.fn(), @@ -118,8 +119,7 @@ describe("PromptsPanel toolbar", () => { expect(screen.getByText("All Environments")).toBeInTheDocument(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Production")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Production"); await waitFor(() => expect(mockGetPromptsList).toHaveBeenLastCalledWith("sk-test", "production")); }); @@ -131,12 +131,10 @@ describe("PromptsPanel toolbar", () => { renderPanel("Admin"); await screen.findByText("table-loaded"); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Production")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Production"); await waitFor(() => expect(screen.getByRole("combobox")).toHaveTextContent("Production")); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("All Environments")); + await chooseSelectOption(user, screen.getByRole("combobox"), "All Environments"); await waitFor(() => expect(screen.getByRole("combobox")).toHaveTextContent("All Environments")); await waitFor(() => expect(mockGetPromptsList).toHaveBeenLastCalledWith("sk-test", undefined)); diff --git a/ui/litellm-dashboard/src/components/CreateUserButton.test.tsx b/ui/litellm-dashboard/src/components/CreateUserButton.test.tsx index 3777c46973f..68bfd94bd63 100644 --- a/ui/litellm-dashboard/src/components/CreateUserButton.test.tsx +++ b/ui/litellm-dashboard/src/components/CreateUserButton.test.tsx @@ -5,6 +5,7 @@ import { describe, it, expect, vi, beforeEach } from "vitest"; import { CreateUserButton } from "./CreateUserButton"; import * as networking from "./networking"; import { toast } from "@/lib/toast"; +import { expectControlBesideLabel } from "../../tests/fieldOrientation"; vi.mock("./networking", () => ({ userCreateCall: vi.fn(), @@ -294,6 +295,20 @@ describe("CreateUserButton", () => { }); }); + it("lays the send invitation email checkbox out beside its label", async () => { + const user = userEvent.setup({ pointerEventsCheck: PointerEventsCheckLevel.Never }); + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByRole("button", { name: /\+ invite user/i })).toBeInTheDocument(); + }); + await user.click(screen.getByRole("button", { name: /\+ invite user/i })); + + const dialog = screen.getByRole("dialog", { name: /invite user/i }); + + expectControlBesideLabel(within(dialog).getByRole("checkbox")); + }); + describe("organizations", () => { it("should send organizations list in POST body when organizations are selected", async () => { const { useOrganizations } = await import("@/app/(dashboard)/hooks/organizations/useOrganizations"); diff --git a/ui/litellm-dashboard/src/components/CreateUserButton.tsx b/ui/litellm-dashboard/src/components/CreateUserButton.tsx index e052a9e0818..0f7c356b8cc 100644 --- a/ui/litellm-dashboard/src/components/CreateUserButton.tsx +++ b/ui/litellm-dashboard/src/components/CreateUserButton.tsx @@ -270,7 +270,7 @@ export const CreateUserButton: React.FC = ({ ); const sendInviteEmailField = ( - + {({ id, value, onChange, onBlur }) => ( )} diff --git a/ui/litellm-dashboard/src/components/EntityUsageExport/EntityUsageExportModal.test.tsx b/ui/litellm-dashboard/src/components/EntityUsageExport/EntityUsageExportModal.test.tsx index 8edf2174eee..cb04323e4c9 100644 --- a/ui/litellm-dashboard/src/components/EntityUsageExport/EntityUsageExportModal.test.tsx +++ b/ui/litellm-dashboard/src/components/EntityUsageExport/EntityUsageExportModal.test.tsx @@ -10,6 +10,7 @@ */ import { describe, it, expect, vi, beforeEach } from "vitest"; +import { screen } from "@testing-library/react"; import { renderWithProviders } from "../../../tests/test-utils"; import userEvent from "@testing-library/user-event"; import EntityUsageExportModal from "./EntityUsageExportModal"; @@ -73,13 +74,13 @@ describe("EntityUsageExportModal", () => { const user = userEvent.setup(); const { handleExportCSV } = await import("./utils"); - const { getByRole } = renderWithProviders(); + renderWithProviders(); // Default primary action reflects CSV export - expect(getByRole("button", { name: /Export CSV/i })).toBeInTheDocument(); + expect(screen.getByRole("button", { name: /Export CSV/i })).toBeInTheDocument(); // Click export - await user.click(getByRole("button", { name: /Export CSV/i })); + await user.click(screen.getByRole("button", { name: /Export CSV/i })); // Verifies export function was invoked with correct parameters expect(handleExportCSV).toHaveBeenCalledWith(baseProps.spendData, "daily", "Tag", "tag", {}); @@ -97,14 +98,14 @@ describe("EntityUsageExportModal", () => { const user = userEvent.setup(); const { handleExportCSV } = await import("./utils"); - const { getByText, getByRole } = renderWithProviders(); + renderWithProviders(); // Choose the alternate export type - click the label to trigger radio - const dailyModelLabel = getByText(/Day-by-day by tag and model/i); + const dailyModelLabel = screen.getByText(/Day-by-day by tag and model/i); await user.click(dailyModelLabel); // Export with default CSV format - const exportBtn = getByRole("button", { name: /Export CSV/i }); + const exportBtn = screen.getByRole("button", { name: /Export CSV/i }); await user.click(exportBtn); // Ensure the selected scope flowed through diff --git a/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.test.tsx b/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.test.tsx index 54566e1d75a..eddc5716305 100644 --- a/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.test.tsx +++ b/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.test.tsx @@ -9,6 +9,7 @@ import BaseSSOSettingsForm, { submitMountedSSOValues, useSSOSettingsForm, } from "./BaseSSOSettingsForm"; +import { expectControlBesideLabel } from "../../../../../../tests/fieldOrientation"; const user = () => userEvent.setup({ pointerEventsCheck: 0 }); @@ -233,6 +234,38 @@ describe("BaseSSOSettingsForm", () => { expect(screen.queryByText("Use Team Mappings")).not.toBeInTheDocument(); }); + + it("lays a provider checkbox field out beside its label", async () => { + const TestWrapper = () => { + const form = useSSOSettingsForm("sso-settings"); + + return ; + }; + + renderWithProviders(); + + await openProviderDropdown(); + await user().click(await screen.findByText(/saml sso/i)); + + expectControlBesideLabel( + await screen.findByRole("checkbox", { name: "Allow IdP-initiated (unsolicited) responses" }), + ); + }); + + it.each(["Use Role Mappings", "Use Team Mappings"])("lays the %s toggle out beside its label", async (label) => { + const TestWrapper = () => { + const form = useSSOSettingsForm("sso-settings"); + + return ; + }; + + renderWithProviders(); + + await openProviderDropdown(); + await user().click(await screen.findByText(/okta/i)); + + expectControlBesideLabel(await screen.findByRole("checkbox", { name: label })); + }); }); describe("renderProviderFields", () => { diff --git a/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.tsx b/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.tsx index 5216da382d0..cb97304f77a 100644 --- a/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.tsx +++ b/ui/litellm-dashboard/src/components/Settings/AdminSettings/SSOSettings/Modals/BaseSSOSettingsForm.tsx @@ -303,7 +303,7 @@ const SSOProviderField = ({ field }: { field: SSOProviderConfig["fields"][number if (field.type === "checkbox") { return ( - + {({ value, onChange, onBlur, id, ...rest }) => ( (); return ( - + {({ value, onChange, onBlur, id, ...rest }) => ( { await user.click(await screen.findByText("Edit team")); expect(onEditTeam).toHaveBeenCalledWith(expect.objectContaining({ team_id: "team-1" })); - await user.click(screen.getByTestId("team-actions-team-1")); - await user.click(await screen.findByText("Delete team")); + await chooseSelectOption(user, screen.getByTestId("team-actions-team-1"), "Delete team", "menuitem"); expect(onDeleteTeam).toHaveBeenCalledWith(expect.objectContaining({ team_id: "team-1" })); }); diff --git a/ui/litellm-dashboard/src/components/VirtualKeysPage/VirtualKeysTable.test.tsx b/ui/litellm-dashboard/src/components/VirtualKeysPage/VirtualKeysTable.test.tsx index d09ccbf8826..b06448912d0 100644 --- a/ui/litellm-dashboard/src/components/VirtualKeysPage/VirtualKeysTable.test.tsx +++ b/ui/litellm-dashboard/src/components/VirtualKeysPage/VirtualKeysTable.test.tsx @@ -2,7 +2,7 @@ import { screen, waitFor, within, fireEvent } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import type { OnUrlUpdateFunction } from "nuqs/adapters/testing"; import { vi, it, expect, beforeEach, describe, Mock, MockedFunction } from "vitest"; -import { renderWithProviders } from "../../../tests/test-utils"; +import { chooseSelectOption, renderWithProviders } from "../../../tests/test-utils"; import { VirtualKeysTable } from "./VirtualKeysTable"; import { KeyResponse, Team } from "../key_team_helpers/key_list"; import { useKeyInfo } from "@/app/(dashboard)/hooks/keys/useKeyInfo"; @@ -327,8 +327,7 @@ it("sorts by the backend max_budget field when 'Budget descending' is chosen fro const user = userEvent.setup(); renderWithProviders(); - await user.click(screen.getByTestId("sort-trigger-spend")); - await user.click(await screen.findByText("Budget descending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-spend"), "Budget descending", "menuitem"); await waitFor(() => { expect(mockUseKeys).toHaveBeenLastCalledWith( @@ -343,8 +342,7 @@ it("emphasizes the active field in the Spend / Budget header so the sorted colum const user = userEvent.setup(); renderWithProviders(); - await user.click(screen.getByTestId("sort-trigger-spend")); - await user.click(await screen.findByText("Budget descending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-spend"), "Budget descending", "menuitem"); await waitFor(() => { expect(screen.getByText("Budget", { selector: "[data-sort-field='max_budget']" })).toHaveClass("font-semibold"); @@ -356,8 +354,7 @@ it("sorts by spend ascending when 'Spend ascending' is chosen from the Spend / B const user = userEvent.setup(); renderWithProviders(); - await user.click(screen.getByTestId("sort-trigger-spend")); - await user.click(await screen.findByText("Spend ascending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-spend"), "Spend ascending", "menuitem"); await waitFor(() => { expect(mockUseKeys).toHaveBeenLastCalledWith(1, 50, expect.objectContaining({ sortBy: "spend", sortOrder: "asc" })); diff --git a/ui/litellm-dashboard/src/components/add_model/advanced_settings.test.tsx b/ui/litellm-dashboard/src/components/add_model/advanced_settings.test.tsx index 01e00d903aa..ceb9d1ca713 100644 --- a/ui/litellm-dashboard/src/components/add_model/advanced_settings.test.tsx +++ b/ui/litellm-dashboard/src/components/add_model/advanced_settings.test.tsx @@ -1,4 +1,4 @@ -import { act, fireEvent, render, waitFor } from "@testing-library/react"; +import { act, fireEvent, render, waitFor, screen } from "@testing-library/react"; import { beforeEach, describe, expect, it, vi } from "vitest"; import { MountedFormHost } from "../../../tests/mounted-form-host"; import AdvancedSettings from "./advanced_settings"; @@ -35,51 +35,51 @@ describe("AdvancedSettings", () => { }); it("should render tags list", async () => { - const { getByText } = renderAdvancedSettings(); - fireEvent.click(getByText("Advanced Settings")); + renderAdvancedSettings(); + fireEvent.click(screen.getByText("Advanced Settings")); await waitFor(() => { - expect(getByText("Tags")).toBeInTheDocument(); + expect(screen.getByText("Tags")).toBeInTheDocument(); }); }); it("should render the litellm params", async () => { - const { getByText } = renderAdvancedSettings(); + renderAdvancedSettings(); act(() => { - fireEvent.click(getByText("Advanced Settings")); + fireEvent.click(screen.getByText("Advanced Settings")); }); await waitFor(() => { - expect(getByText("LiteLLM Params")).toBeInTheDocument(); + expect(screen.getByText("LiteLLM Params")).toBeInTheDocument(); }); }); it("hides every PTU field when PTU cost attribution is disabled", async () => { - const { getByText, queryByText } = renderAdvancedSettings(); + renderAdvancedSettings(); act(() => { - fireEvent.click(getByText("Advanced Settings")); + fireEvent.click(screen.getByText("Advanced Settings")); }); await waitFor(() => { - expect(getByText("Tags")).toBeInTheDocument(); + expect(screen.getByText("Tags")).toBeInTheDocument(); }); for (const label of PTU_LABELS) { - expect(queryByText(label)).not.toBeInTheDocument(); + expect(screen.queryByText(label)).not.toBeInTheDocument(); } - expect(queryByText("PTU Effective To (UTC)")).not.toBeInTheDocument(); + expect(screen.queryByText("PTU Effective To (UTC)")).not.toBeInTheDocument(); }); it("shows every PTU field when PTU cost attribution is enabled", async () => { mockUsePtuCostAttributionEnabled.mockReturnValue(true); - const { getByText } = renderAdvancedSettings(); + renderAdvancedSettings(); act(() => { - fireEvent.click(getByText("Advanced Settings")); + fireEvent.click(screen.getByText("Advanced Settings")); }); await waitFor(() => { - expect(getByText("PTU Count")).toBeInTheDocument(); + expect(screen.getByText("PTU Count")).toBeInTheDocument(); }); for (const label of PTU_LABELS) { - expect(getByText(label)).toBeInTheDocument(); + expect(screen.getByText(label)).toBeInTheDocument(); } - expect(getByText("PTU Effective To (UTC)")).toBeInTheDocument(); + expect(screen.getByText("PTU Effective To (UTC)")).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/components/add_model/litellm_model_name.test.tsx b/ui/litellm-dashboard/src/components/add_model/litellm_model_name.test.tsx index 64074481603..d6476089cd7 100644 --- a/ui/litellm-dashboard/src/components/add_model/litellm_model_name.test.tsx +++ b/ui/litellm-dashboard/src/components/add_model/litellm_model_name.test.tsx @@ -1,4 +1,4 @@ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import { describe, expect, it } from "vitest"; import { getPlaceholder, Providers } from "../provider_info_helpers"; import { MountedFormHost } from "../../../tests/mounted-form-host"; @@ -6,7 +6,7 @@ import LiteLLMModelNameField from "./litellm_model_name"; describe("LitellmModelNameField", () => { it("should render", () => { - const { getByText } = render( + render( { /> , ); - expect(getByText("LiteLLM Model Name(s)")).toBeInTheDocument(); + expect(screen.getByText("LiteLLM Model Name(s)")).toBeInTheDocument(); }); it("should show Azure placeholder as 'my-deployment'", () => { - const { getByPlaceholderText, queryByPlaceholderText } = render( + render( , ); - expect(getByPlaceholderText("my-deployment")).toBeInTheDocument(); - expect(queryByPlaceholderText("gpt-3.5-turbo")).not.toBeInTheDocument(); + expect(screen.getByPlaceholderText("my-deployment")).toBeInTheDocument(); + expect(screen.queryByPlaceholderText("gpt-3.5-turbo")).not.toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/components/add_pass_through.integration.test.tsx b/ui/litellm-dashboard/src/components/add_pass_through.integration.test.tsx index df4dbff4111..daaacd7f0e2 100644 --- a/ui/litellm-dashboard/src/components/add_pass_through.integration.test.tsx +++ b/ui/litellm-dashboard/src/components/add_pass_through.integration.test.tsx @@ -1,6 +1,6 @@ import { describe, it, expect, vi, beforeEach } from "vitest"; import userEvent, { PointerEventsCheckLevel } from "@testing-library/user-event"; -import { fireEvent, renderWithProviders, screen, waitFor } from "../../tests/test-utils"; +import { chooseSelectOption, fireEvent, renderWithProviders, screen, waitFor } from "../../tests/test-utils"; import AddPassThroughEndpoint from "./add_pass_through"; const createPassThroughEndpoint = vi.fn(); @@ -153,8 +153,7 @@ describe("add_pass_through submit payload", () => { await openModal(user); await fillRequiredFields(user); - await user.click(screen.getByLabelText(/HTTP Methods/)); - await user.click(await screen.findByTitle("POST")); + await chooseSelectOption(user, screen.getByLabelText(/HTTP Methods/), "POST"); await submit(user); diff --git a/ui/litellm-dashboard/src/components/bulk_create_users_button.test.tsx b/ui/litellm-dashboard/src/components/bulk_create_users_button.test.tsx index 24685f9c129..e1faf8be1df 100644 --- a/ui/litellm-dashboard/src/components/bulk_create_users_button.test.tsx +++ b/ui/litellm-dashboard/src/components/bulk_create_users_button.test.tsx @@ -26,8 +26,8 @@ const openUploadStep = async () => { describe("BulkCreateUsersButton", () => { it("should render", () => { - const { getByText } = render(); - expect(getByText("+ Bulk Invite Users")).toBeInTheDocument(); + render(); + expect(screen.getByText("+ Bulk Invite Users")).toBeInTheDocument(); }); it("parses a CSV chosen through the file input", async () => { diff --git a/ui/litellm-dashboard/src/components/claude_code_plugins/skill_detail.tsx b/ui/litellm-dashboard/src/components/claude_code_plugins/skill_detail.tsx index 25b35c34861..7cbbb08b623 100644 --- a/ui/litellm-dashboard/src/components/claude_code_plugins/skill_detail.tsx +++ b/ui/litellm-dashboard/src/components/claude_code_plugins/skill_detail.tsx @@ -1,5 +1,6 @@ import React, { useState } from "react"; import { ArrowLeft, Check, Copy, Link2 } from "lucide-react"; +import { cn } from "@/lib/cva.config"; import { buildMarketplaceSettingsSnippet, formatInstallCommand } from "./helpers"; import { Plugin } from "./types"; @@ -50,48 +51,37 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { ]; return ( -
+
{/* Back link */}
Skills
{/* Header */} -
-

{skill.name}

+
+

{skill.name}

{skill.description && ( -

{skill.description}

+

{skill.description}

)}
{/* Tab bar */} -
-
+
+
{tabs.map((tab) => (
setActiveTab(tab.key)} - style={{ - padding: "12px 20px", - fontSize: 14, - color: activeTab === tab.key ? "#1a73e8" : "#5f6368", - borderBottom: activeTab === tab.key ? "3px solid #1a73e8" : "3px solid transparent", - cursor: "pointer", - fontWeight: activeTab === tab.key ? 500 : 400, - marginBottom: -1, - }} + className={cn( + "-mb-px cursor-pointer border-b-[3px] px-5 py-3 text-sm", + activeTab === tab.key + ? "border-info font-medium text-info" + : "border-transparent font-normal text-muted-foreground", + )} > {tab.label}
@@ -101,27 +91,23 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { {/* Overview tab */} {activeTab === "overview" && ( -
+
{/* Left column */} -
-

Skill Details

-

Metadata registered with this skill

-
MetricValue
MetricValue
{row.metric}{row.value}
{row.metric}{row.value}
+
+

Skill Details

+

Metadata registered with this skill

+
- - - + + + {detailRows.map((row, i) => ( - - - + + + ))} @@ -129,38 +115,27 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { {/* Right sidebar */} -
-
-
Status
+
+
+
Status
{skill.enabled ? "Public" : "Draft"}
{sourceUrl && ( -
-
Source
+
+
Source
{sourceUrl.replace("https://", "")} @@ -169,20 +144,13 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { )} {skill.keywords && skill.keywords.length > 0 && ( -
-
Tags
-
+
+
Tags
+
{skill.keywords.map((kw) => ( {kw} @@ -192,10 +160,8 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { )}
-
Skill ID
-
- {skill.id} -
+
Skill ID
+
{skill.id}
@@ -203,93 +169,43 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { {/* How to Use tab */} {activeTab === "usage" && ( -
-

Using this skill

-

+

+

Using this skill

+

Once your proxy is set as a marketplace, enable this skill in Claude Code with one command:

{/* Install command */} -
-
- Run in Claude Code +
+
+ Run in Claude Code
- 
-              {installCommand}
-
+ 
{installCommand}
{/* Shown when the marketplace catalog is stale and the plugin isn't found yet */} -
-

+

+

If you see "Plugin {skill.name} not found in marketplace", update the catalog first:

- 
+            
               /plugin marketplace update litellm
-

+

Don't have the marketplace configured yet?{" "} - setActiveTab("setup")} style={{ color: "#1a73e8", cursor: "pointer" }}> + setActiveTab("setup")} className="cursor-pointer text-info"> See one-time setup →

@@ -298,126 +214,56 @@ const SkillDetail: React.FC = ({ skill, onBack }) => { {/* Setup tab (linked from usage) */} {activeTab === "setup" && ( -
-

- One-time marketplace setup -

+
+

One-time marketplace setup

{/* Option 1: single command — fastest path for most users */} -

+

Run this command in Claude Code to register the marketplace:

-
-
- Run in Claude Code +
+
+ Run in Claude Code
- 
+            
               {`/plugin marketplace add ${typeof window !== "undefined" ? window.location.origin : ""}/claude-code/marketplace.json`}
{/* Option 2: settings.json — for persistent config or managed deployments. extraKnownMarketplaces requires source to be a nested object, not a flat string. */} -

- Or add this to{" "} - - ~/.claude/settings.json - {" "} +

+ Or add this to ~/.claude/settings.json{" "} for a persistent configuration:

-
-
- ~/.claude/settings.json +
+
+ ~/.claude/settings.json
- 
-              {settingsSnippet}
-
+ 
{settingsSnippet}
)} diff --git a/ui/litellm-dashboard/src/components/common_components/KeyLifecycleSettings.test.tsx b/ui/litellm-dashboard/src/components/common_components/KeyLifecycleSettings.test.tsx index 6e627e3b45c..df21444cbad 100644 --- a/ui/litellm-dashboard/src/components/common_components/KeyLifecycleSettings.test.tsx +++ b/ui/litellm-dashboard/src/components/common_components/KeyLifecycleSettings.test.tsx @@ -2,7 +2,7 @@ import React, { useState } from "react"; import { Controller, useForm } from "react-hook-form"; import userEvent, { PointerEventsCheckLevel } from "@testing-library/user-event"; import { describe, expect, it, vi, beforeEach } from "vitest"; -import { fireEvent, renderWithProviders, screen, waitFor } from "../../../tests/test-utils"; +import { chooseSelectOption, fireEvent, renderWithProviders, screen, waitFor } from "../../../tests/test-utils"; import KeyLifecycleSettings from "./KeyLifecycleSettings"; const CREATE_PLACEHOLDER = "e.g., 30d or leave empty to never expire"; @@ -167,8 +167,7 @@ describe("KeyLifecycleSettings", () => { await user.click(screen.getByRole("switch")); expect(await screen.findByText("Rotation Interval")).toBeInTheDocument(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("90 days")); + await chooseSelectOption(user, screen.getByRole("combobox"), "90 days"); await waitFor(() => expect(screen.getAllByTitle("90 days").some(isRenderedSelection)).toBe(true)); expect(screen.getByTestId("rotation-interval-value")).toHaveTextContent("90d"); @@ -181,8 +180,7 @@ describe("KeyLifecycleSettings", () => { await user.click(screen.getByRole("switch")); expect(await screen.findByText("Rotation Interval")).toBeInTheDocument(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Custom interval")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Custom interval"); expect(await screen.findByPlaceholderText("e.g., 1s, 5m, 2h, 14d")).toBeInTheDocument(); expect(screen.getByText("Supported formats: seconds (s), minutes (m), hours (h), days (d)")).toBeInTheDocument(); @@ -196,8 +194,7 @@ describe("KeyLifecycleSettings", () => { await user.click(screen.getByRole("switch")); expect(await screen.findByText("Rotation Interval")).toBeInTheDocument(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Custom interval")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Custom interval"); const customInput = await screen.findByPlaceholderText("e.g., 1s, 5m, 2h, 14d"); fireEvent.change(customInput, { target: { value: "14d" } }); @@ -213,14 +210,12 @@ describe("KeyLifecycleSettings", () => { await user.click(screen.getByRole("switch")); expect(await screen.findByText("Rotation Interval")).toBeInTheDocument(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Custom interval")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Custom interval"); const customInput = await screen.findByPlaceholderText("e.g., 1s, 5m, 2h, 14d"); fireEvent.change(customInput, { target: { value: "14d" } }); await waitFor(() => expect(screen.getByTestId("rotation-interval-value")).toHaveTextContent("14d")); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("7 days")); + await chooseSelectOption(user, screen.getByRole("combobox"), "7 days"); await waitFor(() => expect(screen.getByTestId("rotation-interval-value")).toHaveTextContent("7d")); expect(screen.queryByPlaceholderText("e.g., 1s, 5m, 2h, 14d")).not.toBeInTheDocument(); diff --git a/ui/litellm-dashboard/src/components/common_components/ModelSelector.test.tsx b/ui/litellm-dashboard/src/components/common_components/ModelSelector.test.tsx index bd7b56f1817..7a5db4667e9 100644 --- a/ui/litellm-dashboard/src/components/common_components/ModelSelector.test.tsx +++ b/ui/litellm-dashboard/src/components/common_components/ModelSelector.test.tsx @@ -2,6 +2,7 @@ import { act, fireEvent, render, screen } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { afterEach, describe, expect, it, vi } from "vitest"; import ModelSelector from "./ModelSelector"; +import { chooseSelectOption } from "../../../tests/test-utils"; vi.mock("@/components/llm_calls/fetch_models", () => ({ fetchAvailableModels: vi.fn().mockResolvedValue([]), @@ -9,8 +10,7 @@ vi.mock("@/components/llm_calls/fetch_models", () => ({ const openCustomModelInput = async () => { const user = userEvent.setup(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Enter custom model")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Enter custom model"); return screen.getByPlaceholderText("Enter custom model name"); }; diff --git a/ui/litellm-dashboard/src/components/molecules/cost_optimization_feedback_banner.test.tsx b/ui/litellm-dashboard/src/components/molecules/cost_optimization_feedback_banner.test.tsx index bad92555bd5..3fc1d41dfdd 100644 --- a/ui/litellm-dashboard/src/components/molecules/cost_optimization_feedback_banner.test.tsx +++ b/ui/litellm-dashboard/src/components/molecules/cost_optimization_feedback_banner.test.tsx @@ -1,4 +1,4 @@ -import { fireEvent, render } from "@testing-library/react"; +import { fireEvent, render, screen } from "@testing-library/react"; import { beforeEach, describe, expect, it } from "vitest"; import CostOptimizationFeedbackBanner from "./cost_optimization_feedback_banner"; @@ -10,24 +10,24 @@ describe("CostOptimizationFeedbackBanner", () => { }); it("renders with a link to the feedback discussion", () => { - const { getByText } = render(); - const link = getByText("Share Feedback").closest("a"); + render(); + const link = screen.getByText("Share Feedback").closest("a"); expect(link).toHaveAttribute("href", "https://github.com/BerriAI/litellm/discussions/32172"); }); it("hides itself and persists the dismissal when the dismiss button is clicked", () => { - const { getByText, queryByText, getByLabelText } = render(); - expect(getByText("Help shape cost optimization")).toBeInTheDocument(); + render(); + expect(screen.getByText("Help shape cost optimization")).toBeInTheDocument(); - fireEvent.click(getByLabelText("Dismiss banner")); + fireEvent.click(screen.getByLabelText("Dismiss banner")); - expect(queryByText("Help shape cost optimization")).not.toBeInTheDocument(); + expect(screen.queryByText("Help shape cost optimization")).not.toBeInTheDocument(); expect(localStorage.getItem(STORAGE_KEY)).toBe("true"); }); it("stays dismissed on remount once persisted", () => { localStorage.setItem(STORAGE_KEY, "true"); - const { queryByText } = render(); - expect(queryByText("Help shape cost optimization")).not.toBeInTheDocument(); + render(); + expect(screen.queryByText("Help shape cost optimization")).not.toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/components/organization/organization_view.test.tsx b/ui/litellm-dashboard/src/components/organization/organization_view.test.tsx index b8d8e3ba9c8..700d19eb13c 100644 --- a/ui/litellm-dashboard/src/components/organization/organization_view.test.tsx +++ b/ui/litellm-dashboard/src/components/organization/organization_view.test.tsx @@ -108,7 +108,7 @@ beforeEach(() => { test("renders organization view after loading data", async () => { mockUseOrganization.mockReturnValue({ data: mockOrg, isLoading: false } as any); - const { findAllByText } = renderWithProviders( + renderWithProviders( {}} @@ -120,7 +120,7 @@ test("renders organization view after loading data", async () => { />, ); - const [orgName] = await findAllByText("Acme Corp"); + const [orgName] = await screen.findAllByText("Acme Corp"); expect(orgName).toBeInTheDocument(); }); diff --git a/ui/litellm-dashboard/src/components/provider_info_helpers.tsx b/ui/litellm-dashboard/src/components/provider_info_helpers.tsx index 5baa8138960..d01a6a34cbe 100644 --- a/ui/litellm-dashboard/src/components/provider_info_helpers.tsx +++ b/ui/litellm-dashboard/src/components/provider_info_helpers.tsx @@ -150,6 +150,8 @@ export enum Providers { PETALS = "Petals", PG_VECTOR = "Pg Vector", PREDIBASE = "Predibase", + Qwen_AI_Platform = "Qwen AI Platform", + QwenCloud = "QwenCloud", RECRAFT = "Recraft", REPLICATE = "Replicate", RunwayML = "RunwayML", @@ -262,6 +264,8 @@ export const provider_map: Record = { PETALS: "petals", PG_VECTOR: "pg_vector", PREDIBASE: "predibase", + Qwen_AI_Platform: "qwen_ai_platform", + QwenCloud: "qwencloud", RECRAFT: "recraft", REPLICATE: "replicate", RunwayML: "runwayml", @@ -357,6 +361,8 @@ export const providerLogoMap: Partial> = { [Providers.Openrouter]: openrouterLogo.src, [Providers.Oracle]: oracleLogo.src, [Providers.Perplexity]: perplexityAiLogo.src, + [Providers.Qwen_AI_Platform]: qwenLogo.src, + [Providers.QwenCloud]: qwenLogo.src, [Providers.RECRAFT]: recraftLogo.src, [Providers.REPLICATE]: replicateLogo.src, [Providers.RunwayML]: runwayLogo.src, diff --git a/ui/litellm-dashboard/src/components/settings.test.tsx b/ui/litellm-dashboard/src/components/settings.test.tsx index b97fef32402..762b23f413e 100644 --- a/ui/litellm-dashboard/src/components/settings.test.tsx +++ b/ui/litellm-dashboard/src/components/settings.test.tsx @@ -77,21 +77,21 @@ describe("Settings", () => { }); it("should render the logging callbacks tab when access token is provided", async () => { - const { getByText } = render(); + render(); await waitFor(() => { - expect(getByText("Active Logging Callbacks")).toBeInTheDocument(); + expect(screen.getByText("Active Logging Callbacks")).toBeInTheDocument(); }); }); it("should display additional settings tabs", async () => { - const { getByText } = render(); + render(); await waitFor(() => { - expect(getByText("CloudZero Cost Tracking")).toBeInTheDocument(); - expect(getByText("Alerting Types")).toBeInTheDocument(); - expect(getByText("Alerting Settings")).toBeInTheDocument(); - expect(getByText("Email Alerts")).toBeInTheDocument(); + expect(screen.getByText("CloudZero Cost Tracking")).toBeInTheDocument(); + expect(screen.getByText("Alerting Types")).toBeInTheDocument(); + expect(screen.getByText("Alerting Settings")).toBeInTheDocument(); + expect(screen.getByText("Email Alerts")).toBeInTheDocument(); }); }); @@ -279,13 +279,13 @@ describe("Settings", () => { }); it("should display CloudZero Cost Tracking tab", async () => { - const { getByText } = render(); + render(); await waitFor(() => { - expect(getByText("Active Logging Callbacks")).toBeInTheDocument(); + expect(screen.getByText("Active Logging Callbacks")).toBeInTheDocument(); }); - expect(getByText("CloudZero Cost Tracking")).toBeInTheDocument(); + expect(screen.getByText("CloudZero Cost Tracking")).toBeInTheDocument(); }); }); diff --git a/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.test.tsx b/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.test.tsx index 58a0cd94997..f502ea77127 100644 --- a/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.test.tsx @@ -1,5 +1,5 @@ import type { ColumnDef, ExpandedState } from "@tanstack/react-table"; -import { render, screen, waitFor } from "@testing-library/react"; +import { render, screen, waitFor, within } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { useState } from "react"; import { describe, expect, it, vi } from "vitest"; @@ -7,6 +7,7 @@ import { describe, expect, it, vi } from "vitest"; import { DataTable } from "./DataTable"; import { DataTableMultiSortHeader, DataTableSortHeader } from "./DataTableSortHeader"; import { DataTableViewOptions } from "./DataTableViewOptions"; +import { chooseSelectOption } from "../../../../tests/test-utils"; interface Person { id: string; @@ -21,6 +22,12 @@ function person(id: string, name: string, flagged = false): Person { const names = (): (string | null)[] => screen.getAllByTestId("name-cell").map((el) => el.textContent); +const heightClassesOf = (el: HTMLElement | undefined): string[] => + (el?.className ?? "") + .split(/\s+/) + .filter((cls) => cls.startsWith("h-")) + .sort(); + const nameCellColumns: ColumnDef[] = [ { accessorKey: "name", @@ -170,16 +177,13 @@ describe("DataTable sorting", () => { const user = userEvent.setup(); render(); - await user.click(screen.getByTestId("sort-trigger-name")); - await user.click(await screen.findByText("Ascending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-name"), "Ascending", "menuitem"); expect(names()).toEqual(["Alice", "Bob", "Charlie"]); - await user.click(screen.getByTestId("sort-trigger-name")); - await user.click(await screen.findByText("Descending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-name"), "Descending", "menuitem"); expect(names()).toEqual(["Charlie", "Bob", "Alice"]); - await user.click(screen.getByTestId("sort-trigger-name")); - await user.click(await screen.findByText("Reset")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-name"), "Reset", "menuitem"); expect(names()).toEqual(["Charlie", "Alice", "Bob"]); }); @@ -196,12 +200,10 @@ describe("DataTable sorting", () => { />, ); - await user.click(screen.getByTestId("sort-trigger-spend")); - await user.click(await screen.findByText("Budget descending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-spend"), "Budget descending", "menuitem"); expect(onSortingChange).toHaveBeenLastCalledWith([{ id: "max_budget", desc: true }]); - await user.click(screen.getByTestId("sort-trigger-spend")); - await user.click(await screen.findByText("Spend ascending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-spend"), "Spend ascending", "menuitem"); expect(onSortingChange).toHaveBeenLastCalledWith([{ id: "spend", desc: false }]); }); @@ -229,12 +231,10 @@ describe("DataTable sorting", () => { describe("DataTable layout", () => { it("stretches the table to fill the container when resizing is on, so hidden columns leave no right-side gap", () => { - const { container } = render(); + render(); - const table = container.querySelector("table"); - expect(table).not.toBeNull(); // width pins the natural column total (horizontal scroll on overflow); minWidth:100% fills the gap on underflow. - expect(table?.style.minWidth).toBe("100%"); + expect(screen.getByRole("table")).toHaveStyle({ minWidth: "100%" }); }); }); @@ -354,17 +354,21 @@ describe("DataTable loading", () => { const { rerender } = render( , ); - const skeletonRow = screen.getAllByTestId("skeleton-row").at(0); - const loadedRowHeight = "h-8"; - expect(skeletonRow?.className).toContain(loadedRowHeight); + const skeletonHeight = heightClassesOf(screen.getAllByRole("row").at(-1)); rerender(); - expect(document.querySelector("[data-row-id]")?.className).toContain(loadedRowHeight); + const loadedHeight = heightClassesOf(screen.getByRole("row", { name: /Charlie/ })); + + expect(loadedHeight).not.toEqual([]); + expect(skeletonHeight).toEqual(loadedHeight); }); it("does not force the compact height on default-size skeleton rows", () => { - render(); - expect(screen.getAllByTestId("skeleton-row").at(0)?.className).not.toContain("h-8"); + const { rerender } = render(); + const skeletonHeight = heightClassesOf(screen.getAllByRole("row").at(-1)); + + rerender(); + expect(heightClassesOf(screen.getAllByRole("row").at(-1))).not.toEqual(skeletonHeight); }); it("varies skeleton shape and width per column instead of one fixed bar", () => { @@ -420,7 +424,7 @@ describe("DataTable loading", () => { describe("DataTable column visibility", () => { it("hides a column when toggled off in the view-options menu", async () => { const user = userEvent.setup(); - const { container } = render( + render( { />, ); - expect(container.querySelector('th[data-header-id="email"]')).not.toBeNull(); + expect(screen.getByRole("columnheader", { name: "Email" })).toBeInTheDocument(); await user.click(screen.getByTestId("view-options-trigger")); await user.click(await screen.findByTestId("view-option-email")); - await waitFor(() => expect(container.querySelector('th[data-header-id="email"]')).toBeNull()); + await waitFor(() => expect(screen.queryByRole("columnheader", { name: "Email" })).not.toBeInTheDocument()); await user.click(screen.getByTestId("view-option-email")); - await waitFor(() => expect(container.querySelector('th[data-header-id="email"]')).not.toBeNull()); + expect(await screen.findByRole("columnheader", { name: "Email" })).toBeInTheDocument(); }); it("omits columns that opt out of hiding from the menu", async () => { @@ -468,14 +472,10 @@ describe("DataTable column visibility", () => { describe("DataTable pinned columns", () => { it("applies sticky positioning to a pinned column only", () => { - const { container } = render(); + render(); - const pinnedHead = container.querySelector('th[data-header-id="name"]'); - const normalHead = container.querySelector('th[data-header-id="email"]'); - - expect(pinnedHead?.style.position).toBe("sticky"); - expect(pinnedHead?.style.left).toBe("0px"); - expect(normalHead?.style.position).toBe(""); + expect(screen.getByRole("columnheader", { name: "Name" })).toHaveStyle({ position: "sticky", left: "0px" }); + expect(screen.getByRole("columnheader", { name: "Email" })).not.toHaveStyle({ position: "sticky" }); }); }); @@ -570,7 +570,7 @@ describe("DataTable expansion", () => { describe("DataTable row styling and footer", () => { it("applies rowClassName to the matching row only", () => { const data = [person("a", "Alice", true), person("b", "Bob", false)]; - const { container } = render( + render( { />, ); - expect(container.querySelector('tr[data-row-id="a"]')?.className).toContain("flagged-row"); - expect(container.querySelector('tr[data-row-id="b"]')?.className).not.toContain("flagged-row"); + expect(screen.getByRole("row", { name: /Alice/ })).toHaveClass("flagged-row"); + expect(screen.getByRole("row", { name: /Bob/ })).not.toHaveClass("flagged-row"); }); it("renders the footer slot inside a tfoot element", () => { @@ -596,63 +596,57 @@ describe("DataTable row styling and footer", () => { />, ); - expect(screen.getByTestId("footer-row").closest("tfoot")).not.toBeNull(); + const rowGroups = screen.getAllByRole("rowgroup"); + expect(within(rowGroups.at(-1) as HTMLElement).getByText("Total: 3")).toBeInTheDocument(); }); }); describe("DataTable layout", () => { it("exposes resize handles with stable selectors only when resizing is enabled", () => { - const { container, rerender } = render( - , - ); - expect(container.querySelectorAll("[data-resizer][data-header-id]").length).toBe(2); + const { rerender } = render(); + expect(screen.getByTestId("column-resizer-name")).toBeInTheDocument(); + expect(screen.getByTestId("column-resizer-email")).toBeInTheDocument(); rerender(); - expect(container.querySelectorAll("[data-resizer]").length).toBe(0); + expect(screen.queryByTestId("column-resizer-name")).not.toBeInTheDocument(); }); it("makes the header sticky and constrains body height when maxBodyHeight is set", () => { - const { container } = render(); - expect(container.querySelector("thead")?.className).toContain("sticky"); - const scroller = container.querySelector('[data-slot="table-container"]')?.parentElement as HTMLElement; - expect(scroller).toHaveStyle({ maxHeight: "240px" }); + render(); + expect(screen.getByTestId("data-table-head")).toHaveClass("sticky"); + expect(screen.getByTestId("data-table-scroller")).toHaveStyle({ maxHeight: "240px" }); }); it("caps fillHeight at the parent's height instead of stretching to it, so a short table stays short", () => { - const { container } = render(); - const scroller = container.querySelector('[data-slot="table-container"]')?.parentElement as HTMLElement; - const frame = scroller.parentElement as HTMLElement; - const outer = frame.parentElement as HTMLElement; + render(); + const outer = screen.getByTestId("data-table-root"); + const frame = screen.getByTestId("data-table-frame"); + const scroller = screen.getByTestId("data-table-scroller"); // A ceiling, not a stretch: flex-1 here would hold the footer at the bottom on a two-row table. - expect(outer.className).toContain("max-h-full"); - expect(outer.className).not.toContain("flex-1"); - expect(frame.className).not.toContain("flex-1"); - expect(scroller.className).not.toContain("flex-1"); + expect(outer).toHaveClass("max-h-full", "flex-col"); + expect(outer).not.toHaveClass("flex-1"); + expect(frame).toHaveClass("flex-col"); + expect(frame).not.toHaveClass("flex-1"); + expect(scroller).not.toHaveClass("flex-1"); - expect(outer.className).toContain("flex-col"); - expect(frame.className).toContain("flex-col"); - expect(scroller.className).toContain("min-h-0"); - expect(scroller.className).toContain("overflow-auto"); + expect(scroller).toHaveClass("min-h-0", "overflow-auto"); expect(scroller).toHaveStyle({ maxHeight: "" }); // Without this the Table primitive's own overflow container captures the sticky header. - expect(scroller.className).toContain("[&_[data-slot=table-container]]:overflow-visible"); + expect(scroller).toHaveClass("[&_[data-slot=table-container]]:overflow-visible"); - const thead = container.querySelector("thead") as HTMLElement; - expect(thead.className).toContain("sticky"); // Rows pass under the header, so the semi-transparent row tint alone would let them show through. - expect(thead.className).toContain("bg-background"); + expect(screen.getByTestId("data-table-head")).toHaveClass("sticky", "bg-background"); }); it("leaves the default layout untouched when neither height mode is set", () => { - const { container } = render(); - const scroller = container.querySelector('[data-slot="table-container"]')?.parentElement as HTMLElement; + render(); + const scroller = screen.getByTestId("data-table-scroller"); - expect(scroller.className).toContain("overflow-x-auto"); - expect(scroller.className).not.toContain("min-h-0"); + expect(scroller).toHaveClass("overflow-x-auto"); + expect(scroller).not.toHaveClass("min-h-0"); expect(scroller).toHaveStyle({ maxHeight: "" }); - expect((scroller.parentElement as HTMLElement).className).not.toContain("flex-col"); - expect(container.querySelector("thead")?.className).not.toContain("sticky"); - expect(container.querySelector("thead")?.className).not.toContain("bg-background"); + expect(screen.getByTestId("data-table-frame")).not.toHaveClass("flex-col"); + expect(screen.getByTestId("data-table-head")).not.toHaveClass("sticky", "bg-background"); }); }); diff --git a/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.tsx b/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.tsx index e8357b37715..60267606951 100644 --- a/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.tsx +++ b/ui/litellm-dashboard/src/components/shared/DataTable/DataTable.tsx @@ -195,8 +195,7 @@ function DataTableHeadCell({ header, size, stickyHeader, enableColumnResi )} {canResize && (
column.resetSize()} @@ -589,15 +588,19 @@ export function DataTable(props: DataTableProps -
+
+
{toolbar !== undefined &&
{toolbar(table)}
}
- + {table.getHeaderGroups().map((headerGroup) => ( {headerGroup.headers.map((header) => ( diff --git a/ui/litellm-dashboard/src/components/shared/DataTable/DataTableSortHeader.test.tsx b/ui/litellm-dashboard/src/components/shared/DataTable/DataTableSortHeader.test.tsx index 70403ba2efe..aeb6fa44e91 100644 --- a/ui/litellm-dashboard/src/components/shared/DataTable/DataTableSortHeader.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/DataTable/DataTableSortHeader.test.tsx @@ -13,6 +13,7 @@ import { useState } from "react"; import { describe, expect, it, vi } from "vitest"; import { DataTableSortHeader, type DataTableSortVariant } from "./DataTableSortHeader"; +import { chooseSelectOption } from "../../../../tests/test-utils"; interface Item { name: string; @@ -84,16 +85,13 @@ describe("DataTableSortHeader", () => { const user = userEvent.setup(); render(); - await user.click(screen.getByTestId("sort-trigger-name")); - await user.click(await screen.findByText("Descending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-name"), "Descending", "menuitem"); expect(screen.getByTestId("sort-trigger-name").querySelector('[data-sort-indicator="desc"]')).not.toBeNull(); - await user.click(screen.getByTestId("sort-trigger-name")); - await user.click(await screen.findByText("Ascending")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-name"), "Ascending", "menuitem"); expect(screen.getByTestId("sort-trigger-name").querySelector('[data-sort-indicator="asc"]')).not.toBeNull(); - await user.click(screen.getByTestId("sort-trigger-name")); - await user.click(await screen.findByText("Reset")); + await chooseSelectOption(user, screen.getByTestId("sort-trigger-name"), "Reset", "menuitem"); expect(screen.getByTestId("sort-trigger-name").querySelector('[data-sort-indicator="none"]')).not.toBeNull(); }); diff --git a/ui/litellm-dashboard/src/components/shared/PaginatedSearchSelect.test.tsx b/ui/litellm-dashboard/src/components/shared/PaginatedSearchSelect.test.tsx index d248cf311cc..04c8d3e9012 100644 --- a/ui/litellm-dashboard/src/components/shared/PaginatedSearchSelect.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/PaginatedSearchSelect.test.tsx @@ -5,6 +5,7 @@ import { describe, expect, it, vi } from "vitest"; import { PaginatedSearchSelect } from "./PaginatedSearchSelect"; import type { SearchSelectOption } from "./SearchSelect"; +import { chooseSelectOption } from "../../../tests/test-utils"; const OPTIONS: SearchSelectOption[] = [ { label: "alias-alpha", value: "alias-alpha" }, @@ -63,8 +64,7 @@ describe("PaginatedSearchSelect", () => { } render(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("alias-beta")); + await chooseSelectOption(user, screen.getByRole("combobox"), "alias-beta"); expect(screen.getByRole("combobox")).toHaveValue("alias-beta"); await new Promise((resolve) => setTimeout(resolve, 400)); @@ -136,8 +136,7 @@ describe("PaginatedSearchSelect", () => { const onValueChange = vi.fn(); renderSelect({ onValueChange }); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("alias-beta")); + await chooseSelectOption(user, screen.getByRole("combobox"), "alias-beta"); expect(onValueChange).toHaveBeenCalledWith("alias-beta"); }); @@ -255,8 +254,7 @@ describe("PaginatedSearchSelect", () => { } render(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Beta Team")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Beta Team"); await user.click(screen.getByRole("button", { name: "refetch" })); expect(screen.getByRole("combobox")).toHaveValue("Beta Team"); diff --git a/ui/litellm-dashboard/src/components/shared/PaginationStatusAlerts.test.tsx b/ui/litellm-dashboard/src/components/shared/PaginationStatusAlerts.test.tsx index 3698b68155e..5bd48b1aa4c 100644 --- a/ui/litellm-dashboard/src/components/shared/PaginationStatusAlerts.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/PaginationStatusAlerts.test.tsx @@ -1,4 +1,4 @@ -import { fireEvent, render } from "@testing-library/react"; +import { fireEvent, render, screen } from "@testing-library/react"; import { describe, expect, it, vi } from "vitest"; import PaginationStatusAlerts from "./PaginationStatusAlerts"; @@ -6,7 +6,7 @@ import PaginationStatusAlerts from "./PaginationStatusAlerts"; describe("PaginationStatusAlerts", () => { it("shows page progress and wires the Stop button while fetching", () => { const cancel = vi.fn(); - const { getByRole, getByText } = render( + render( { />, ); - expect(getByText(/Currently fetching spend data: fetched 7 \/ 42 pages/)).toBeInTheDocument(); - fireEvent.click(getByRole("button", { name: "Stop" })); + expect(screen.getByText(/Currently fetching spend data: fetched 7 \/ 42 pages/)).toBeInTheDocument(); + fireEvent.click(screen.getByRole("button", { name: "Stop" })); expect(cancel).toHaveBeenCalledTimes(1); }); it("shows the partial-data notice after a cancel, frozen at the last fetched page", () => { - const { getByText } = render( + render( { />, ); - expect(getByText("Showing partial spend data (7/42 pages loaded)")).toBeInTheDocument(); + expect(screen.getByText("Showing partial spend data (7/42 pages loaded)")).toBeInTheDocument(); }); it("names the subject it is fetching", () => { - const { getByText } = render( + render( { />, ); - expect(getByText(/Currently fetching agent data: fetched 1 \/ 3 pages/)).toBeInTheDocument(); + expect(screen.getByText(/Currently fetching agent data: fetched 1 \/ 3 pages/)).toBeInTheDocument(); }); it("renders nothing when idle", () => { diff --git a/ui/litellm-dashboard/src/components/shared/SearchSelect.test.tsx b/ui/litellm-dashboard/src/components/shared/SearchSelect.test.tsx index c0b9bc7d8ae..62a510268dd 100644 --- a/ui/litellm-dashboard/src/components/shared/SearchSelect.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/SearchSelect.test.tsx @@ -3,6 +3,7 @@ import userEvent from "@testing-library/user-event"; import { describe, expect, it, vi } from "vitest"; import { SearchSelect } from "./SearchSelect"; +import { chooseSelectOption } from "../../../tests/test-utils"; const OPTIONS = [ { label: "Acme Prod", value: "team-1" }, @@ -71,8 +72,7 @@ describe("SearchSelect", () => { const onValueChange = vi.fn(); const user = userEvent.setup(); render(); - await user.click(screen.getByRole("combobox")); - await user.click(await screen.findByText("Growth")); + await chooseSelectOption(user, screen.getByRole("combobox"), "Growth"); expect(onValueChange).toHaveBeenCalledWith("team-2"); }); diff --git a/ui/litellm-dashboard/src/components/shared/charts/area_chart.test.tsx b/ui/litellm-dashboard/src/components/shared/charts/area_chart.test.tsx index 6e4ab14be33..c81ade526dd 100644 --- a/ui/litellm-dashboard/src/components/shared/charts/area_chart.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/charts/area_chart.test.tsx @@ -1,4 +1,4 @@ -import { render } from "@testing-library/react"; +import { render, screen } from "@testing-library/react"; import React from "react"; import { describe, expect, it } from "vitest"; import { AreaChart } from "./area_chart"; @@ -21,9 +21,9 @@ describe("AreaChart", () => { }); it("renders the No data placeholder instead of a chart when data is empty", () => { - const { container, getByText } = render(); + const { container } = render(); - expect(getByText("No data")).toBeInTheDocument(); + expect(screen.getByText("No data")).toBeInTheDocument(); expect(container.querySelector('[data-slot="chart"]')).toBeNull(); }); diff --git a/ui/litellm-dashboard/src/components/shared/charts/bar_chart.test.tsx b/ui/litellm-dashboard/src/components/shared/charts/bar_chart.test.tsx index a322eeb3ad0..3cd7bd2fd08 100644 --- a/ui/litellm-dashboard/src/components/shared/charts/bar_chart.test.tsx +++ b/ui/litellm-dashboard/src/components/shared/charts/bar_chart.test.tsx @@ -21,9 +21,9 @@ describe("BarChart", () => { }); it("renders the No data placeholder instead of a chart when data is empty", () => { - const { container, getByText } = render(); + const { container } = render(); - expect(getByText("No data")).toBeInTheDocument(); + expect(screen.getByText("No data")).toBeInTheDocument(); expect(container.querySelector('[data-slot="chart"]')).toBeNull(); }); diff --git a/ui/litellm-dashboard/src/components/templates/KeyInfoView.handleKeyUpdate.test.tsx b/ui/litellm-dashboard/src/components/templates/KeyInfoView.handleKeyUpdate.test.tsx index 6592ff8c357..705679fe2e6 100644 --- a/ui/litellm-dashboard/src/components/templates/KeyInfoView.handleKeyUpdate.test.tsx +++ b/ui/litellm-dashboard/src/components/templates/KeyInfoView.handleKeyUpdate.test.tsx @@ -445,7 +445,7 @@ describe("KeyInfoView handleKeyUpdate budget_duration", () => { ); fireEvent.click(screen.getByText("Settings")); - expect(screen.getByText("Budget Reset").parentElement?.textContent).toContain("Every 30d"); + expect(screen.getByTestId("budget-reset-value")).toHaveTextContent("Every 30d"); fireEvent.click(screen.getByText("Edit Settings")); (globalThis as any).__TEST_FORM_VALUES = { @@ -456,7 +456,7 @@ describe("KeyInfoView handleKeyUpdate budget_duration", () => { fireEvent.click(screen.getByText("Mock Submit")); await waitFor(() => { - expect(screen.getByText("Budget Reset").parentElement?.textContent).toBe("Budget ResetNever"); + expect(screen.getByTestId("budget-reset-value")).toHaveTextContent("Never"); }); }); }); diff --git a/ui/litellm-dashboard/src/components/templates/key_edit_view.test.tsx b/ui/litellm-dashboard/src/components/templates/key_edit_view.test.tsx index bf8f43b8b5f..97b09e00808 100644 --- a/ui/litellm-dashboard/src/components/templates/key_edit_view.test.tsx +++ b/ui/litellm-dashboard/src/components/templates/key_edit_view.test.tsx @@ -1,7 +1,7 @@ import { fireEvent, screen, waitFor } from "@testing-library/react"; import userEvent from "@testing-library/user-event"; import { beforeEach, describe, expect, it, vi } from "vitest"; -import { renderWithProviders } from "../../../tests/test-utils"; +import { chooseSelectOption, renderWithProviders } from "../../../tests/test-utils"; import { KeyResponse } from "../key_team_helpers/key_list"; import { MODEL_MAX_BUDGET_PREMIUM_HINT } from "../key_team_helpers/ModelMaxBudgetEditor"; import { @@ -300,7 +300,7 @@ describe("KeyEditView", () => { }); it("should render", async () => { - const { getByText } = renderWithProviders( + renderWithProviders( {}} @@ -313,12 +313,12 @@ describe("KeyEditView", () => { ); await waitFor(() => { - expect(getByText("Save Changes")).toBeInTheDocument(); + expect(screen.getByText("Save Changes")).toBeInTheDocument(); }); }); it("should render tags", async () => { - const { getByText } = renderWithProviders( + renderWithProviders( {}} @@ -331,12 +331,12 @@ describe("KeyEditView", () => { ); await waitFor(() => { - expect(getByText("test-tag")).toBeInTheDocument(); + expect(screen.getByText("test-tag")).toBeInTheDocument(); }); }); it("should not render tags in metadata textarea", async () => { - const { getByLabelText } = renderWithProviders( + renderWithProviders( {}} @@ -348,7 +348,7 @@ describe("KeyEditView", () => { />, ); - const metadataTextarea = getByLabelText("Metadata") as HTMLTextAreaElement; + const metadataTextarea = screen.getByLabelText("Metadata") as HTMLTextAreaElement; await waitFor(() => { expect(metadataTextarea).toHaveValue("{}"); }); @@ -963,10 +963,7 @@ describe("KeyEditView", () => { />, ); - await userEvent.click(await screen.findByLabelText("Reset Budget")); - - const weeklyOption = await screen.findByText("weekly"); - await userEvent.click(weeklyOption); + await chooseSelectOption(userEvent, await screen.findByLabelText("Reset Budget"), "weekly"); const submitButton = screen.getByRole("button", { name: /save changes/i }); await userEvent.click(submitButton); @@ -1042,8 +1039,7 @@ describe("KeyEditView", () => { ); const resetBudget = await screen.findByLabelText("Reset Budget"); - await userEvent.click(resetBudget); - await userEvent.click(await screen.findByText("Never resets")); + await chooseSelectOption(userEvent, resetBudget, "Never resets"); await waitFor(() => { expect(resetBudget).toHaveTextContent("Never resets"); @@ -1074,8 +1070,7 @@ describe("KeyEditView", () => { />, ); - await userEvent.click(await screen.findByLabelText("Reset Budget")); - await userEvent.click(await screen.findByText("Never resets")); + await chooseSelectOption(userEvent, await screen.findByLabelText("Reset Budget"), "Never resets"); await userEvent.click(screen.getByRole("button", { name: /save changes/i })); @@ -1946,8 +1941,7 @@ describe("KeyEditView", () => { await userEvent.clear(duration); await userEvent.type(duration, "45d"); - await userEvent.click(screen.getByLabelText(/TPM Rate Limit Type/)); - await userEvent.click(await screen.findByTitle("Guaranteed throughput")); + await chooseSelectOption(userEvent, screen.getByLabelText(/TPM Rate Limit Type/), /^Guaranteed throughput/); await userEvent.click(screen.getByRole("button", { name: /save changes/i })); @@ -2103,8 +2097,7 @@ describe("KeyEditView", () => { renderForPayload(onSubmitMock); await screen.findByRole("button", { name: /save changes/i }); - await userEvent.click(screen.getByLabelText(/RPM Rate Limit Type/)); - await userEvent.click(await screen.findByTitle("Guaranteed throughput")); + await chooseSelectOption(userEvent, screen.getByLabelText(/RPM Rate Limit Type/), /^Guaranteed throughput/); await userEvent.click(screen.getByRole("button", { name: /save changes/i })); diff --git a/ui/litellm-dashboard/src/components/templates/key_info_view.budget_display.test.tsx b/ui/litellm-dashboard/src/components/templates/key_info_view.budget_display.test.tsx index f506c0e51d7..a02f80f7800 100644 --- a/ui/litellm-dashboard/src/components/templates/key_info_view.budget_display.test.tsx +++ b/ui/litellm-dashboard/src/components/templates/key_info_view.budget_display.test.tsx @@ -381,6 +381,6 @@ describe("KeyInfoView budget reset visibility", () => { await waitFor(() => { expect(screen.getByText("Budget Reset")).toBeInTheDocument(); }); - expect(screen.getByText("Budget Reset").parentElement).toHaveTextContent("Never"); + expect(screen.getByTestId("budget-reset-value")).toHaveTextContent("Never"); }); }); diff --git a/ui/litellm-dashboard/src/components/templates/key_info_view.tsx b/ui/litellm-dashboard/src/components/templates/key_info_view.tsx index 0be80c3e173..0dd0dd6d6af 100644 --- a/ui/litellm-dashboard/src/components/templates/key_info_view.tsx +++ b/ui/litellm-dashboard/src/components/templates/key_info_view.tsx @@ -895,7 +895,7 @@ export default function KeyInfoView({

Budget Reset

-

+

{currentKeyData.budget_reset_at ? `${currentKeyData.budget_duration ? `Every ${currentKeyData.budget_duration}, next ` : ""}${formatTimestamp(currentKeyData.budget_reset_at)}` : "Never"} diff --git a/ui/litellm-dashboard/src/components/ui/field.test.tsx b/ui/litellm-dashboard/src/components/ui/field.test.tsx index 4bb3d0ec665..5765a637526 100644 --- a/ui/litellm-dashboard/src/components/ui/field.test.tsx +++ b/ui/litellm-dashboard/src/components/ui/field.test.tsx @@ -14,6 +14,7 @@ import { FieldSet, FieldTitle, } from "./field"; +import { ROW_LAYOUT_CLASSES, STRETCH_CHILDREN_CLASS } from "../../../tests/fieldOrientation"; describe("FieldError", () => { it("renders nothing when there are no errors and no children", () => { @@ -85,6 +86,20 @@ describe("Field", () => { expect(screen.getByRole("group")).toHaveAttribute("data-orientation", "horizontal"); }); + + it("stretches every child when vertical, which is what inputs, selects and textareas want", () => { + render(); + + expect(screen.getByRole("group")).toHaveClass("flex-col", STRETCH_CHILDREN_CLASS); + }); + + it("lays children in a row at their own width when horizontal, so a checkbox stays square", () => { + render(); + const field = screen.getByRole("group"); + + expect(field).toHaveClass(...ROW_LAYOUT_CLASSES); + expect(field).not.toHaveClass(STRETCH_CHILDREN_CLASS); + }); }); describe("field primitives forward refs to their DOM node", () => { diff --git a/ui/litellm-dashboard/src/components/view_logs/RequestLogsFilters.test.tsx b/ui/litellm-dashboard/src/components/view_logs/RequestLogsFilters.test.tsx index e0ec66e1ae2..5a35c7ae16b 100644 --- a/ui/litellm-dashboard/src/components/view_logs/RequestLogsFilters.test.tsx +++ b/ui/litellm-dashboard/src/components/view_logs/RequestLogsFilters.test.tsx @@ -3,7 +3,7 @@ import userEvent from "@testing-library/user-event"; import { useState } from "react"; import { beforeEach, describe, expect, it, vi } from "vitest"; -import { renderWithProviders, testQueryClient } from "../../../tests/test-utils"; +import { chooseSelectOption, renderWithProviders, testQueryClient } from "../../../tests/test-utils"; import { ERROR_CODE_OPTIONS } from "./constants"; import { LOG_FILTER_IDS } from "./log_filter_logic"; import { RequestLogsFilters } from "./RequestLogsFilters"; @@ -125,8 +125,7 @@ describe("RequestLogsFilters", () => { const user = userEvent.setup(); const { set } = renderFilters(); - await user.click(await screen.findByPlaceholderText("Search an internal user")); - await user.click(await screen.findByText("alice@example.com")); + await chooseSelectOption(user, await screen.findByPlaceholderText("Search an internal user"), "alice@example.com"); expect(set).toHaveBeenCalledWith(LOG_FILTER_IDS.USER_ID, "alice@example.com"); }); diff --git a/ui/litellm-dashboard/tests/fieldOrientation.ts b/ui/litellm-dashboard/tests/fieldOrientation.ts new file mode 100644 index 00000000000..49ba5056d66 --- /dev/null +++ b/ui/litellm-dashboard/tests/fieldOrientation.ts @@ -0,0 +1,18 @@ +import { expect } from "vitest"; + +export const ROW_LAYOUT_CLASSES = ["flex-row", "items-center"] as const; +export const STRETCH_CHILDREN_CLASS = "*:w-full"; + +/** + * Asserts a control sits beside its label at its own width instead of being stretched across the + * field. Reaches for the resolved classes because the defect is purely visual: nothing accessible + * distinguishes a square checkbox from a full-width bar. + */ +export const expectControlBesideLabel = (control: HTMLElement): void => { + const field = control.closest('[data-slot="field"]'); + if (field === null) throw new Error("control is not rendered inside a form field"); + + expect(field).toHaveAttribute("data-orientation", "horizontal"); + expect(field).toHaveClass(...ROW_LAYOUT_CLASSES); + expect(field).not.toHaveClass(STRETCH_CHILDREN_CLASS); +}; diff --git a/ui/litellm-dashboard/tests/test-utils.tsx b/ui/litellm-dashboard/tests/test-utils.tsx index 66966201a9c..553726faff0 100644 --- a/ui/litellm-dashboard/tests/test-utils.tsx +++ b/ui/litellm-dashboard/tests/test-utils.tsx @@ -44,20 +44,14 @@ const pointerBlocked = (element: HTMLElement): boolean => { return false; }; -/** - * Opens a Base UI Select and picks an option by its accessible name. - * - * The option is in the DOM one render before the popup finishes entering, and until then its - * positioner still carries `pointer-events: none`, which user-event refuses to click. Waiting on - * the option text alone is a race that React 19's flush timing loses. - */ export const chooseSelectOption = async ( - user: ReturnType, + user: Pick, "click">, trigger: HTMLElement, optionName: string | RegExp, + role: "option" | "menuitem" | "menuitemradio" = "option", ) => { await user.click(trigger); - const option = await screen.findByRole("option", { name: optionName }); + const option = await screen.findByRole(role, { name: optionName }); await waitFor(() => expect(pointerBlocked(option)).toBe(false)); await user.click(option); }; diff --git a/uv.lock b/uv.lock index 7df1a35b28a..27be919eea1 100644 --- a/uv.lock +++ b/uv.lock @@ -10,7 +10,7 @@ resolution-markers = [ ] [options] -exclude-newer = "2026-08-26T18:33:25.773031Z" +exclude-newer = "2026-08-29T17:58:57.633306Z" exclude-newer-span = "P3D" [manifest] @@ -4266,7 +4266,7 @@ wheels = [ [[package]] name = "litellm" -version = "1.100.0" +version = "1.101.0" source = { editable = "." } dependencies = [ { name = "aiohttp" }, @@ -4669,12 +4669,12 @@ proxy-dev = [ [[package]] name = "litellm-enterprise" -version = "0.1.62" +version = "0.1.63" source = { editable = "enterprise" } [[package]] name = "litellm-proxy-extras" -version = "0.4.91" +version = "0.4.92" source = { editable = "litellm-proxy-extras" } [[package]]

- Property - - {skill.name} -
Property{skill.name}
{row.property}{row.value}
{row.property}{row.value}