diff --git a/.circleci/config.yml b/.circleci/config.yml index 12e3cb1f6b6..790efc79862 100644 --- a/.circleci/config.yml +++ b/.circleci/config.yml @@ -42,7 +42,7 @@ commands: "pydantic==2.11.0" "mcp==1.25.0" "requests-mock>=1.12.1" \ "responses==0.25.7" "pytest-xdist==3.6.1" "pytest-timeout==2.2.0" \ "pytest-cov==5.0.0" "semantic_router==0.1.10" "fastapi-offline==1.7.3" \ - "a2a" + "a2a" "parameterized>=0.9.0" - setup_litellm_enterprise_pip - save_cache: paths: @@ -1115,7 +1115,7 @@ jobs: for dir in "${IGNORE_DIRS[@]}"; do IGNORE_ARGS="$IGNORE_ARGS --ignore=$dir" done - python -m pytest -v tests/llm_translation $IGNORE_ARGS --junitxml=test-results/junit.xml --durations=20 -n 8 --timeout=120 --timeout_method=thread + python -m pytest -v tests/llm_translation $IGNORE_ARGS --junitxml=test-results/junit.xml --durations=20 -n 8 --timeout=120 --timeout_method=thread --retries 2 --retry-delay 5 no_output_timeout: 15m # Store test results @@ -1331,7 +1331,7 @@ jobs: command: | pwd ls - python -m pytest -vv tests/unified_google_tests --cov=litellm --cov-report=xml -x -s -v --junitxml=test-results/junit.xml --durations=5 + python -m pytest -vv tests/unified_google_tests --cov=litellm --cov-report=xml -x -s -v --junitxml=test-results/junit.xml --durations=5 --retries 3 --retry-delay 5 no_output_timeout: 15m - run: name: Rename the coverage files diff --git a/.github/workflows/test-linting.yml b/.github/workflows/test-linting.yml index fc0f84a20d4..4cedb8b5bae 100644 --- a/.github/workflows/test-linting.yml +++ b/.github/workflows/test-linting.yml @@ -28,9 +28,12 @@ jobs: find . -type d -name "__pycache__" -exec rm -rf {} + || true find . -name "*.pyc" -delete || true + - name: Check poetry.lock is up to date + run: | + poetry check --lock || (echo "❌ poetry.lock is out of sync with pyproject.toml. Run 'poetry lock' locally and commit the result." && exit 1) + - name: Install dependencies run: | - poetry lock poetry install --with dev - name: Check Black formatting diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index 9396f323e45..2bc361bc48f 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -14,12 +14,12 @@ repos: types: [python] files: (litellm/|litellm_proxy_extras/|enterprise/).*\.py exclude: ^litellm/__init__.py$ - # - id: black - # name: black - # entry: poetry run black - # language: system - # types: [python] - # files: (litellm/|litellm_proxy_extras/|enterprise/).*\.py + - id: black + name: black + entry: poetry run black + language: system + types: [python] + files: (litellm/|litellm_proxy_extras/).*\.py - repo: https://github.com/pycqa/flake8 rev: 7.0.0 # The version of flake8 to use hooks: diff --git a/ci_cd/security_scans.sh b/ci_cd/security_scans.sh index e0f370e0035..801b700f64f 100755 --- a/ci_cd/security_scans.sh +++ b/ci_cd/security_scans.sh @@ -163,6 +163,9 @@ run_grype_scans() { "CVE-2026-25639" # axios - full fix requires 1.x major version bump; pinned to >=0.30.2 to clear other axios CVEs, upgrade to 1.x in follow-up "CVE-2026-2297" # Python 3.13 SourcelessFileLoader audit hook bypass - no fix available in base image "GHSA-qffp-2rhf-9h96" # tar hardlink path traversal - from nodejs_wheel bundled npm, not used in application runtime code + "CVE-2026-2673" # OpenSSL 3.6.1 TLS 1.3 key exchange group negotiation issue - no fix available yet + "CVE-2026-3644" # Python 3.13 vulnerability - no fix available in base image + "CVE-2026-4224" # Python 3.13 Expat parser stack overflow in ElementDeclHandler - no fix available in base image ) # Build JSON array of allowlisted CVE IDs for jq diff --git a/docs/my-website/blog/guardrail_logging_secret_exposure_incident/index.md b/docs/my-website/blog/guardrail_logging_secret_exposure_incident/index.md new file mode 100644 index 00000000000..71f9e3da011 --- /dev/null +++ b/docs/my-website/blog/guardrail_logging_secret_exposure_incident/index.md @@ -0,0 +1,78 @@ +--- +slug: guardrail-logging-secret-exposure-incident +title: "Incident Report: Guardrail logging exposed secret headers in spend logs and traces" +date: 2026-03-18T10:00:00 +authors: + - litellm +tags: [incident-report, security, guardrails] +hide_table_of_contents: false +--- + +**Date:** March 18, 2026 +**Duration:** Unknown +**Severity:** High +**Status:** Resolved + +## Summary + +When a custom guardrail returned the full LiteLLM request/data dictionary, the guardrail response logged by LiteLLM could include `secret_fields.raw_headers`, including plaintext `Authorization` headers containing API keys or other credentials. + +This information could then propagate to logging and observability surfaces that consume guardrail metadata, including: + +- **Spend logs in the LiteLLM UI:** visible to admins with access to spend-log data +- **OpenTelemetry traces:** visible to anyone with access to the relevant telemetry backend + +LLM calls, proxy routing, and provider execution were not blocked by this bug. The impact was exposure of sensitive request headers in observability and logging paths. + +{/* truncate */} + +--- + +## Background + +LiteLLM keeps internal request data (including request headers) for use during the call. That data is not meant to be written to logs or telemetry. + +When custom guardrails run, their outcomes are logged so they can appear in spend logs, OpenTelemetry traces, and other observability backends. If a guardrail returned the full request payload instead of a minimal result, that internal request data could be included in what was logged. Before the fix, the guardrail logging path did not strip that data before sending it to those systems. + +```mermaid +flowchart TD + inboundRequest["1. Incoming proxy request"] --> storeSecrets["2. Store internal request data"] + storeSecrets --> guardrailRuns["3. Custom guardrail runs"] + guardrailRuns --> fullDataReturn["4. Guardrail returns full request payload"] + fullDataReturn --> loggingBuild["5. Build guardrail log payload"] + loggingBuild --> spendLogs["6a. Persist to spend logs / UI"] + loggingBuild --> otelTraces["6b. Attach to OTEL guardrail spans"] +``` + +--- + +## Root Cause + +The root cause was incomplete sanitization in the guardrail logging path. When building the payload that gets sent to spend logs and traces, LiteLLM prepared guardrail responses for logging but did not strip internal request data (such as headers) from them. If a guardrail returned a response that included that data, it was passed through to the logging and observability systems unchanged. + +--- + +## Impact + +This issue required all of the following: + +1. A custom guardrail returned the full LiteLLM request/data dictionary, or another response object containing `secret_fields`. +2. LiteLLM logged that guardrail response through the standard guardrail logging path. +3. An operator, admin, or telemetry consumer had access to the resulting logs or traces. + +When those conditions were met, sensitive values could become visible through: + +- **Spend logs / UI responses:** guardrail metadata could be included in spend-log payloads rendered in the admin UI. +- **OpenTelemetry traces:** `guardrail_response` could be written as a span attribute on guardrail spans. +- **Other downstream observability backends:** any integration consuming the same guardrail metadata could receive the leaked values. + +This was a logging and telemetry exposure bug. It did not let callers bypass auth, access other tenants directly, or change model behavior, but it could expose plaintext credentials to people with access to those observability systems. + +--- + +## Guidance For Users + +- Upgrade to LiteLLM 1.82.3+. +- If you operated custom guardrails that return the full request/data dict, review whether spend logs or telemetry traces were retained during the affected period. +- Rotate any credentials that may have appeared in `Authorization` or other forwarded request headers in those systems. +- Apply least-privilege access controls to spend-log views and telemetry backends that may contain request-derived metadata. diff --git a/docs/my-website/docs/proxy/config_settings.md b/docs/my-website/docs/proxy/config_settings.md index f5b611a85a6..d7542fc2c3d 100644 --- a/docs/my-website/docs/proxy/config_settings.md +++ b/docs/my-website/docs/proxy/config_settings.md @@ -401,8 +401,10 @@ router_settings: | AUTH_STRATEGY | Strategy used for authentication (e.g., OAuth, API key) | AUTO_REDIRECT_UI_LOGIN_TO_SSO | Flag to enable automatic redirect of UI login page to SSO when SSO is configured. Default is **false** | AUDIO_SPEECH_CHUNK_SIZE | Chunk size for audio speech processing. Default is 1024 -| ANTHROPIC_API_KEY | API key for Anthropic service +| ANTHROPIC_API_KEY | API key for Anthropic service. Uses `x-api-key` header for authentication. +| ANTHROPIC_AUTH_TOKEN | Alternative auth token for Anthropic service. Uses `Authorization: Bearer` header instead of `x-api-key`. Used as fallback when `ANTHROPIC_API_KEY` is not set. | ANTHROPIC_API_BASE | Base URL for Anthropic API. Default is https://api.anthropic.com +| ANTHROPIC_BASE_URL | Alternative to `ANTHROPIC_API_BASE` for setting the Anthropic API base URL. Used as fallback when `ANTHROPIC_API_BASE` is not set. | ANTHROPIC_TOKEN_COUNTING_BETA_VERSION | Beta version header for Anthropic token counting API. Default is `token-counting-2024-11-01` | AWS_ACCESS_KEY_ID | Access Key ID for AWS services | AWS_BATCH_ROLE_ARN | ARN of the AWS IAM role for batch operations @@ -902,6 +904,7 @@ router_settings: | OTEL_SERVICE_NAME | Service name identifier for OpenTelemetry | OTEL_TRACER_NAME | Tracer name for OpenTelemetry tracing | OTEL_LOGS_EXPORTER | Exporter type for OpenTelemetry logs (e.g., console) +| OTEL_IGNORE_CONTEXT_PROPAGATION | When true, ignore parent span context propagation in OpenTelemetry callbacks | PAGERDUTY_API_KEY | API key for PagerDuty Alerting | PANW_PRISMA_AIRS_API_KEY | API key for PANW Prisma AIRS service | PANW_PRISMA_AIRS_API_BASE | Base URL for PANW Prisma AIRS service diff --git a/docs/my-website/docs/proxy/configs.md b/docs/my-website/docs/proxy/configs.md index 56a8b9566db..84a6fac1210 100644 --- a/docs/my-website/docs/proxy/configs.md +++ b/docs/my-website/docs/proxy/configs.md @@ -602,6 +602,22 @@ Since you shouldn't use 12.5, round down to **10** to leave a safety buffer. Thi - Total maximum connections: 8 workers × 10 connections = 80 connections - This stays safely under your database's 100 connection limit +## LiteLLM License Key (Enterprise) + +To enable [LiteLLM Enterprise features](https://docs.litellm.ai/docs/proxy/enterprise), set your license key as an environment variable: + +```bash +export LITELLM_LICENSE="eyJ..." +``` + +The license key is a JWT token provided when you purchase a LiteLLM Enterprise license. Once set, LiteLLM will automatically detect and activate enterprise features. + +You can also add it to your `.env` file: + +```env +LITELLM_LICENSE="eyJ..." +``` + ## Extras diff --git a/docs/my-website/release_notes/v1.82.3/index.md b/docs/my-website/release_notes/v1.82.3/index.md index 15df33fdb80..20be8826718 100644 --- a/docs/my-website/release_notes/v1.82.3/index.md +++ b/docs/my-website/release_notes/v1.82.3/index.md @@ -47,6 +47,10 @@ pip install litellm==1.82.3 - **FLUX Kontext image editing** — `flux-kontext-pro` and `flux-kontext-max` added to Black Forest Labs, alongside `flux-pro-1.0-fill` and `flux-pro-1.0-expand` for inpainting and outpainting - **116 new models, 132 deprecated models cleaned up** — Major model map refresh including Mistral Magistral, Dashscope Qwen3 VL, xAI Grok via Azure AI, ZAI GLM-5, Serper Search; removal of OpenAI GPT-3.5/GPT-4 legacy variants, Gemini 1.5, and Vertex AI PaLM2 - **SageMaker Nova provider** — [New `sagemaker_nova` provider for Amazon Nova models on SageMaker](../../docs/providers/aws_sagemaker) - [PR #21542](https://github.com/BerriAI/litellm/pull/21542) +- **Hashicorp Vault secret manager** — Config override backend powered by Hashicorp Vault, with full UI for managing vault-sourced credentials - [PR #22939](https://github.com/BerriAI/litellm/pull/22939), [PR #23036](https://github.com/BerriAI/litellm/pull/23036) +- **Responses API WebSocket streaming** — Real-time WebSocket streaming for the Responses API, including support across all providers - [PR #22559](https://github.com/BerriAI/litellm/pull/22559), [PR #22771](https://github.com/BerriAI/litellm/pull/22771) +- **Org Admin RBAC expansion** — Org Admins can now access team management endpoints, view and invite internal users, and manage team membership without requiring a global admin role - [PR #23085](https://github.com/BerriAI/litellm/pull/23085), [PR #23080](https://github.com/BerriAI/litellm/pull/23080) +- **Guardrail mode defaults and tag-based modes** — Set a default guardrail mode list globally, and specify a list of modes in tag-based guardrail configs - [PR #22676](https://github.com/BerriAI/litellm/pull/22676), [PR #23020](https://github.com/BerriAI/litellm/pull/23020) - **Secret redaction in logs** — API keys, tokens, and credentials automatically scrubbed from all proxy log output. Enabled by default; opt out with `LITELLM_DISABLE_REDACT_SECRETS=true` - [PR #23668](https://github.com/BerriAI/litellm/pull/23668) - **Streaming stability fix** — Critical fix for `RuntimeError: Cannot send a request, as the client has been closed.` crashes after ~1 hour in production - [PR #22926](https://github.com/BerriAI/litellm/pull/22926) @@ -54,7 +58,7 @@ pip install litellm==1.82.3 ## New Providers and Endpoints -### New Providers (5 new providers) +### New Providers (7 new providers) | Provider | Supported LiteLLM Endpoints | Description | | -------- | --------------------------- | ----------- | @@ -63,6 +67,8 @@ pip install litellm==1.82.3 | [Black Forest Labs](../../docs/providers/black_forest_labs) (`black_forest_labs/`) | `/images/generations`, `/images/edits` | FLUX image generation and editing — Kontext Pro/Max, Pro 1.0 Fill/Expand | | [Serper](../../docs/providers/serper) (`serper/`) | `/search` | Web search via Serper API | | [SageMaker Nova](../../docs/providers/aws_sagemaker) (`sagemaker_nova/`) | `/chat/completions` | Amazon Nova models via SageMaker endpoint | +| [Google Search API](../../docs/providers/google_search) (`google_search/`) | `/search` | Google Search API integration - [PR #22752](https://github.com/BerriAI/litellm/pull/22752) | +| [Bedrock Mantle](../../docs/providers/bedrock) (`bedrock_mantle/`) | `/chat/completions` | Amazon Bedrock via Mantle — alternative auth and routing path for Bedrock models - [PR #22866](https://github.com/BerriAI/litellm/pull/22866) | --- @@ -238,22 +244,92 @@ pip install litellm==1.82.3 - **[Responses API](../../docs/response_api)** - Handle `response.failed`, `response.incomplete`, and `response.cancelled` terminal event types in background streaming — previously only `response.completed` was handled - [PR #23492](https://github.com/BerriAI/litellm/pull/23492) + - WebSocket streaming support for Responses API — real-time streaming via WebSocket for all providers - [PR #22559](https://github.com/BerriAI/litellm/pull/22559), [PR #22771](https://github.com/BerriAI/litellm/pull/22771) + - WebRTC support for real-time audio/video communication - [PR #23446](https://github.com/BerriAI/litellm/pull/23446) + - Responses API support for OpenAI-compatible JSON providers (`openai_like`) - [PR #21398](https://github.com/BerriAI/litellm/pull/21398) + - Route `gpt-5.4+` calls using both tools and reasoning to the Responses API automatically - [PR #23577](https://github.com/BerriAI/litellm/pull/23577) -#### Bug Fixes +- **[Anthropic Files API](../../docs/providers/anthropic)** + - Full Anthropic Files API support — upload, retrieve, list, and delete files; use file references in messages - [PR #16594](https://github.com/BerriAI/litellm/pull/16594) + +- **[Mistral](../../docs/providers/mistral)** + - Voxtral audio transcription support — `mistral/voxtral-mini-*` and `mistral/voxtral-*` for audio transcription via Mistral - [PR #22801](https://github.com/BerriAI/litellm/pull/22801) + +- **[OpenAI](../../docs/providers/openai)** + - `litellm.acount_tokens()` public API — async token counting with full OpenAI provider support - [PR #22809](https://github.com/BerriAI/litellm/pull/22809) + - Normalize `reasoning_effort` dict to string for chat completion API - [PR #22981](https://github.com/BerriAI/litellm/pull/22981) + +- **[OpenRouter](../../docs/providers/openrouter)** + - Image edit support for OpenRouter models - [PR #22403](https://github.com/BerriAI/litellm/pull/22403) + +- **[Google Vertex AI](../../docs/providers/vertex)** + - VIDEO modality token usage tracking in `completion_tokens_details` - [PR #22550](https://github.com/BerriAI/litellm/pull/22550) + +- **Images API** + - `input_fidelity` parameter for image edit API - [PR #23201](https://github.com/BerriAI/litellm/pull/23201) + +- **General** + - Per-request `enable_json_schema_validation` flag for thread-safe JSON schema validation - [PR #21233](https://github.com/BerriAI/litellm/pull/21233) + - Model cost aliases expansion — define aliases in the cost map that inherit pricing from a parent model - [PR #23314](https://github.com/BerriAI/litellm/pull/23314), [PR #23457](https://github.com/BerriAI/litellm/pull/23457) + - Wildcards model support for the Files API - [PR #22740](https://github.com/BerriAI/litellm/pull/22740) + +#### Bugs - **[Anthropic](../../docs/providers/anthropic)** - Preserve native tool format (web_search, bash, tool_search, etc.) when guardrails convert tools for the Anthropic Messages API - [PR #23526](https://github.com/BerriAI/litellm/pull/23526) + - Enforce `type: "object"` on tool input schemas in `_map_tool_helper` — fixes tool call failures for strict-schema providers - [PR #23103](https://github.com/BerriAI/litellm/pull/23103) + - Deduplicate `tool_result` messages by `tool_call_id` — prevents duplicate tool result errors in multi-turn conversations - [PR #23104](https://github.com/BerriAI/litellm/pull/23104) + - Map `reasoning_effort` to `output_config` for Claude 4.6 models - [PR #22220](https://github.com/BerriAI/litellm/pull/22220) + +- **[Google Gemini](../../docs/providers/gemini)** + - Correct streaming `finish_reason` for tool calls — was incorrectly returning `null` instead of `tool_calls` - [PR #21577](https://github.com/BerriAI/litellm/pull/21577) + - Preserve `$ref` in JSON Schema for Gemini 2.0+ — schema references were being stripped, breaking structured output - [PR #21597](https://github.com/BerriAI/litellm/pull/21597) + - Handle `minimal` `reasoning_effort` param for Gemini 3.1 models - [PR #22920](https://github.com/BerriAI/litellm/pull/22920) + +- **[Google Vertex AI](../../docs/providers/vertex)** + - Pass through native Gemini `imageConfig` params for image generation - [PR #21585](https://github.com/BerriAI/litellm/pull/21585) + - Prevent content truncation when `finish_reason` races ahead of content chunks in streaming - [PR #22692](https://github.com/BerriAI/litellm/pull/22692) + - Strip LiteLLM-internal keys from `extra_body` before merging to Gemini request body - [PR #23131](https://github.com/BerriAI/litellm/pull/23131) + - Drop unsupported `output_config` parameter from all Vertex AI requests - [PR #22884](https://github.com/BerriAI/litellm/pull/22884) + - Skip schema transforms for Gemini 2.0+ tool parameters — avoids breaking native Gemini schema handling - [PR #23265](https://github.com/BerriAI/litellm/pull/23265) + +- **[OpenRouter](../../docs/providers/openrouter)** + - Pattern-based fix for native model double-stripping when provider prefix matches model name - [PR #22320](https://github.com/BerriAI/litellm/pull/22320) + - Use provider-reported usage in streaming responses when `stream_options` is not set - [PR #21592](https://github.com/BerriAI/litellm/pull/21592) + +- **[AWS Bedrock](../../docs/providers/bedrock)** + - Extract region and model ID from `bedrock/{region}/{model}` path format - [PR #22546](https://github.com/BerriAI/litellm/pull/22546) + - Strip `scope` from `cache_control` for Anthropic messages on Bedrock and Azure AI - [PR #22867](https://github.com/BerriAI/litellm/pull/22867) + - Populate `completion_tokens_details` in Responses API responses - [PR #23243](https://github.com/BerriAI/litellm/pull/23243) + +- **[Azure AI](../../docs/providers/azure_ai)** + - Resolve `api_base` from environment variable in Document Intelligence OCR - [PR #21581](https://github.com/BerriAI/litellm/pull/21581) - **[Moonshot / Kimi](../../docs/providers/openai_compatible)** - Auto-fill `reasoning_content` for Moonshot Kimi reasoning models - [PR #23580](https://github.com/BerriAI/litellm/pull/23580) + - Preserve `image_url` blocks in multimodal messages for Moonshot - [PR #21595](https://github.com/BerriAI/litellm/pull/21595) - **[HuggingFace](../../docs/providers/huggingface)** - Forward `extra_headers` to HuggingFace embedding API - [PR #23525](https://github.com/BerriAI/litellm/pull/23525) +- **Token Counting / Cost** + - Fix `count_tokens` to include system prompts and tools in token counting API requests - [PR #22301](https://github.com/BerriAI/litellm/pull/22301) + - Pass all custom pricing fields to `register_model` in `completion()` and `embedding()` - [PR #22552](https://github.com/BerriAI/litellm/pull/22552) + +- **Tools / Function Calling** + - Gracefully repair truncated JSON in tool call arguments — prevents crashes on malformed tool responses - [PR #22503](https://github.com/BerriAI/litellm/pull/22503) + - Fix `output_item.done` for function calls not emitting `finish_reason` in streaming - [PR #22553](https://github.com/BerriAI/litellm/pull/22553) + - Preserve thinking block order with multiple web searches - [PR #23093](https://github.com/BerriAI/litellm/pull/23093) + - **General** - Normalize `content_filtered` finish reason across providers - [PR #23564](https://github.com/BerriAI/litellm/pull/23564) + - Unify `finish_reason` mapping to OpenAI-compatible values across all providers - [PR #22138](https://github.com/BerriAI/litellm/pull/22138) - Fix custom cost tracking on deployments for `/v1/messages` and `/v1/responses` - [PR #23647](https://github.com/BerriAI/litellm/pull/23647) - Fix per-request custom pricing when `router_model_id` has no pricing data — now falls back to model name + - Fix batch list showing stale `validating` status after completion - [PR #22982](https://github.com/BerriAI/litellm/pull/22982) + - Fix batch retrieve returning raw `output_file_id` when `model_id` is missing - [PR #23194](https://github.com/BerriAI/litellm/pull/23194) + - Encode batch IDs when `x-litellm-model` header is used - [PR #22653](https://github.com/BerriAI/litellm/pull/22653) + - Map `reasoning` to `reasoning_content` in streaming Delta for gpt-oss providers - [PR #22803](https://github.com/BerriAI/litellm/pull/22803) --- @@ -264,10 +340,31 @@ pip install litellm==1.82.3 - **Virtual Keys** - Add Organization dropdown to Create/Edit Key form — `organization_id` is now a first-class field in Key Ownership - [PR #23595](https://github.com/BerriAI/litellm/pull/23595) - Allow setting `organization_id` on `/key/update` — keys can be assigned or moved to a different organization after creation - [PR #23557](https://github.com/BerriAI/litellm/pull/23557) + - Manual Spend Reset for virtual keys from the UI — admins can reset key spend to zero on demand - [PR #22715](https://github.com/BerriAI/litellm/pull/22715) + - BYOK (Bring Your Own Key) — client-side provider API key takes precedence over proxy key for Anthropic `/v1/messages` - [PR #22964](https://github.com/BerriAI/litellm/pull/22964) + - UI login session duration configurable via `LITELLM_UI_SESSION_DURATION` environment variable - [PR #22182](https://github.com/BerriAI/litellm/pull/22182) + - Auto-redirect UI login to SSO via `auto_redirect_ui_login_to_sso: true` in config.yaml - [PR #23367](https://github.com/BerriAI/litellm/pull/23367) + +- **Access Control (RBAC)** + - Org Admins can now access team management endpoints — `/team/new`, `/team/update`, `/team/delete`, `/team/member_add`, `/team/member_delete` - [PR #23085](https://github.com/BerriAI/litellm/pull/23085), [PR #23095](https://github.com/BerriAI/litellm/pull/23095) + - Org Admins can view and invite internal users — full user management without requiring global admin role - [PR #23080](https://github.com/BerriAI/litellm/pull/23080) + - Allow Admin Viewers to access Audit Logs — view-only admin role now includes audit log access - [PR #23419](https://github.com/BerriAI/litellm/pull/23419) + - RBAC for Vector Stores and Agents — key/team-level access control for vector store and agent resources - [PR #22858](https://github.com/BerriAI/litellm/pull/22858) + - User filter scope (`scope_user_search_to_org`) is now opt-in — previously default-on, causing unintended restriction - [PR #23057](https://github.com/BerriAI/litellm/pull/23057) + +- **Vector Stores** + - Vector Store management endpoints — retrieve, list, update, and delete vector stores via `/v1/vector_stores/*` - [PR #23435](https://github.com/BerriAI/litellm/pull/23435) + +- **Teams** + - Batch expiry setting for teams — configure a default expiry duration for all team keys - [PR #22705](https://github.com/BerriAI/litellm/pull/22705) + - Team Admin can reset key spend - [PR #22725](https://github.com/BerriAI/litellm/pull/22725) - **Internal Users** - Add/Remove Team Membership directly from the Internal Users info page — includes searchable dropdown and role selector; no longer requires navigating to each team - [PR #23638](https://github.com/BerriAI/litellm/pull/23638) +- **Models** + - Attach knowledge base to model via UI - [PR #22656](https://github.com/BerriAI/litellm/pull/22656) + - **Default Team Settings** - Modernize page to antd (consistent with rest of app) - [PR #23614](https://github.com/BerriAI/litellm/pull/23614) - Fix: default team params (budget, duration, tpm, rpm, permissions) now correctly applied on `/team/new` - [PR #23614](https://github.com/BerriAI/litellm/pull/23614) @@ -290,6 +387,13 @@ pip install litellm==1.82.3 - Fix Public Model Hub not showing config-defined models after save - [PR #23501](https://github.com/BerriAI/litellm/pull/23501) - Fix fallback popup model dropdown z-index issue - [PR #23516](https://github.com/BerriAI/litellm/pull/23516) - Fix double-counting bug in org/team key limit checks on `/key/update` +- Fix invite link allowing multiple password resets for the same link - [PR #22462](https://github.com/BerriAI/litellm/pull/22462) +- Fix key expiry default duration not being applied when `duration` is not set - [PR #22956](https://github.com/BerriAI/litellm/pull/22956) +- Fix all proxy models not including model access groups in key creation - [PR #23236](https://github.com/BerriAI/litellm/pull/23236) +- Fix admin viewers unable to see all organizations - [PR #22940](https://github.com/BerriAI/litellm/pull/22940) +- Fix Audit Logs UI: added server-side pagination, filtering, and drawer view - [PR #22476](https://github.com/BerriAI/litellm/pull/22476) +- Fix virtual keys in teams view not applying the team filter correctly - [PR #23065](https://github.com/BerriAI/litellm/pull/23065) +- Fix team expiry enforcement validation - [PR #22728](https://github.com/BerriAI/litellm/pull/22728) --- @@ -297,6 +401,13 @@ pip install litellm==1.82.3 ### Logging +- **[Helicone](../../docs/observability/helicone_integration)** + - Add Gemini and Vertex AI support to HeliconeLogger — routes Gemini and Vertex AI requests through the correct Helicone provider URL - [PR #19288](https://github.com/BerriAI/litellm/pull/19288) + - Fix correct provider URL for Vertex AI Gemini models - [PR #22603](https://github.com/BerriAI/litellm/pull/22603) + +- **[Langfuse](../../docs/proxy/logging#langfuse)** + - Fix failure path kwargs inconsistency causing dropped traces on failed requests - [PR #22390](https://github.com/BerriAI/litellm/pull/22390) + - **[Vantage](https://vantage.sh)** - Add Vantage integration for FOCUS 1.2 CSV export — export LiteLLM proxy spend data as FinOps Open Cost & Usage Specification reports, with time-windowed filenames to prevent overwrites - [PR #23333](https://github.com/BerriAI/litellm/pull/23333) @@ -305,7 +416,10 @@ pip install litellm==1.82.3 ### Guardrails -No major guardrail changes in this release. +- **Guardrail mode default list** — Configure a default list of guardrail modes applied globally when no per-request mode is specified - [PR #22676](https://github.com/BerriAI/litellm/pull/22676) +- **Tag-based guardrail mode lists** — Specify a list of modes in tag-based guardrail configs instead of a single mode - [PR #23020](https://github.com/BerriAI/litellm/pull/23020) +- **Fix presidio PII token leak** — Edge case where Anthropic handle in Presidio caused PII data exposure in token response - [PR #22627](https://github.com/BerriAI/litellm/pull/22627) +- **Fix OTEL orphaned guardrail traces** — Span redundancy and missing response IDs in OpenTelemetry guardrail traces - [PR #23001](https://github.com/BerriAI/litellm/pull/23001) ### Prompt Management @@ -313,7 +427,32 @@ No major prompt management changes in this release. ### Secret Managers -No major secret manager changes in this release. +- **[Hashicorp Vault](../../docs/secret_managers)** — Full Hashicorp Vault integration as a config override backend — secrets defined in Vault are fetched at startup and override `config.yaml` values. UI support for managing vault-sourced credentials included - [PR #22939](https://github.com/BerriAI/litellm/pull/22939), [PR #23036](https://github.com/BerriAI/litellm/pull/23036) + +--- + +## MCP Gateway + +#### Features + +- **Token authentication for MCP servers** — configure `auth_type: "bearer"` per MCP server to require token-based auth on tool calls - [PR #23260](https://github.com/BerriAI/litellm/pull/23260) +- **Team-scoped MCP server filtering** — keys created under a team only see MCP servers available to that team - [PR #23323](https://github.com/BerriAI/litellm/pull/23323) +- **Per-server health recheck in UI** — trigger a health check for individual MCP servers without reloading all servers - [PR #23328](https://github.com/BerriAI/litellm/pull/23328) + +#### Bugs + +- Fix MCP server URL and tools management issues causing tool discovery to fail - [PR #22751](https://github.com/BerriAI/litellm/pull/22751) +- Fix MCP server health checks triggering on server deletion - [PR #23063](https://github.com/BerriAI/litellm/pull/23063) + +--- + +## Spend Tracking, Budgets and Rate Limiting + +- **Fix budget-linked keys never having spend reset** — Keys linked to budget objects were not having their spend reset on the configured reset interval - [PR #20688](https://github.com/BerriAI/litellm/pull/20688) +- **Flex pricing support** — Add `flex_pricing` field to cost map for providers that offer dynamic pricing tiers - [PR #22992](https://github.com/BerriAI/litellm/pull/22992) +- **Fix spend log cleanup** — Resolved lock tracking, integer retention, and skip-log-level issues in spend log cleanup job - [PR #22687](https://github.com/BerriAI/litellm/pull/22687) +- **Fix WebSearch spend log deduplication** — WebSearch interception was failing with thinking enabled; fixed along with spend log dedup - [PR #22679](https://github.com/BerriAI/litellm/pull/22679) +- **Fix TypeError when request has no API key** — Spend tracking was throwing unhandled exception when API key was absent from request - [PR #23363](https://github.com/BerriAI/litellm/pull/23363) --- @@ -323,6 +462,10 @@ No major secret manager changes in this release. - **Fix OOM / Prisma connection loss** on large installs — unbounded managed-object poll was exhausting Prisma connections after ~60–70 minutes on instances with 336K+ queued response rows - [PR #23472](https://github.com/BerriAI/litellm/pull/23472) - **Centralize logging kwarg updates** — root cause fix migrating all logging updates to a single function, eliminating kwarg inconsistencies across logging paths - [PR #23659](https://github.com/BerriAI/litellm/pull/23659) - **Fix tiktoken cache for non-root offline containers** — tiktoken cache now works correctly in offline environments running as non-root users - [PR #23498](https://github.com/BerriAI/litellm/pull/23498) +- **Block proxy startup when Redis transaction buffer has no Redis** — prevents silent data loss when `use_redis_transaction_buffer: true` is set without a Redis connection - [PR #23019](https://github.com/BerriAI/litellm/pull/23019) +- **Fix `InFlightRequestsMiddleware` crash** — undefined kwargs in middleware were causing request failures - [PR #22523](https://github.com/BerriAI/litellm/pull/22523) +- **Fix `BaseModelResponseIterator` crash on non-string stream chunks** — streaming was crashing when providers returned non-string chunk data - [PR #23497](https://github.com/BerriAI/litellm/pull/23497) +- **Fix `SERVER_ROOT_PATH` prefix handling** — strip prefix before checking mapped pass-through routes to prevent double-prefix issues - [PR #23414](https://github.com/BerriAI/litellm/pull/23414) - **Add CodSpeed continuous performance benchmarks** — automated performance regression tracking on CI - [PR #23676](https://github.com/BerriAI/litellm/pull/23676) --- @@ -342,6 +485,16 @@ No major secret manager changes in this release. --- +## Documentation Updates + +- Add Anthropic `/v1/messages` → `/responses` parameter mapping reference - [PR #22893](https://github.com/BerriAI/litellm/pull/22893) +- Update Okta SSO docs and custom SSO handler example - [PR #22786](https://github.com/BerriAI/litellm/pull/22786) +- Add `LITELLM_MAX_BUDGET_PER_SESSION_TTL` to environment variables reference - [PR #23186](https://github.com/BerriAI/litellm/pull/23186) +- Add DB query performance guidelines to `CLAUDE.md` - [PR #23196](https://github.com/BerriAI/litellm/pull/23196) +- Add Gemini Vertex AI PayGo/priority cost tracking docs - [PR #22948](https://github.com/BerriAI/litellm/pull/22948) + +--- + ## New Contributors * @ryanh-ai made their first contribution in [PR #21542](https://github.com/BerriAI/litellm/pull/21542) @@ -359,14 +512,17 @@ No major secret manager changes in this release. ## Diff Summary ## 03/16/2026 -* New Providers: 5 +* New Providers: 7 * New Models / Updated Models: 116 new, 132 removed -* LLM API Endpoints: 5 -* Management Endpoints / UI: 11 -* AI Integrations: 2 -* Performance / Reliability: 5 +* LLM API Endpoints: 37 +* Management Endpoints / UI: 31 +* AI Integrations: 8 +* MCP Gateway: 5 +* Spend Tracking, Budgets and Rate Limiting: 5 +* Performance / Loadbalancing / Reliability improvements: 9 * Security: 3 * Database / Proxy Operations: 2 +* Documentation Updates: 5 --- diff --git a/docs/my-website/sidebars.js b/docs/my-website/sidebars.js index 7db72da276a..4c0471fb8f4 100644 --- a/docs/my-website/sidebars.js +++ b/docs/my-website/sidebars.js @@ -48,6 +48,20 @@ const sidebars = { slug: "/guardrail_providers" }, items: [ + { + type: "category", + label: "Contributing to Guardrails", + items: [ + "adding_provider/generic_guardrail_api", + "adding_provider/simple_guardrail_tutorial", + "adding_provider/adding_guardrail_support", + ] + }, + { + type: "doc", + id: "proxy/guardrails/team_based_guardrails", + label: "Team Bring-Your-Own Guardrails", + }, ...[ "proxy/guardrails/qualifire", "proxy/guardrails/aim_security", diff --git a/litellm/batches/main.py b/litellm/batches/main.py index e176dc42921..17d73aae6ad 100644 --- a/litellm/batches/main.py +++ b/litellm/batches/main.py @@ -524,6 +524,7 @@ def _handle_retrieve_batch_providers_without_provider_config( optional_params.api_base or litellm.api_base or get_secret_str("ANTHROPIC_API_BASE") + or get_secret_str("ANTHROPIC_BASE_URL") ) api_key = ( optional_params.api_key diff --git a/litellm/constants.py b/litellm/constants.py index 89c59ee9326..c0dd115210c 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -1444,6 +1444,7 @@ SENTRY_DENYLIST = [ "credential", "OPENAI_API_KEY", "ANTHROPIC_API_KEY", + "ANTHROPIC_AUTH_TOKEN", "AZURE_API_KEY", "COHERE_API_KEY", "REPLICATE_API_KEY", diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py index ee3c344169b..29d28b8c896 100644 --- a/litellm/cost_calculator.py +++ b/litellm/cost_calculator.py @@ -757,7 +757,7 @@ def _map_traffic_type_to_service_tier(traffic_type: Optional[str]) -> Optional[s """ if traffic_type is None: return None - service_tier = _GEMINI_TRAFFIC_TYPE_TO_SERVICE_TIER.get(traffic_type.upper()) + service_tier = _GEMINI_TRAFFIC_TYPE_TO_SERVICE_TIER.get(str(traffic_type).upper()) return service_tier diff --git a/litellm/integrations/datadog/datadog.py b/litellm/integrations/datadog/datadog.py index 64e0b26a8e7..4de3644b581 100644 --- a/litellm/integrations/datadog/datadog.py +++ b/litellm/integrations/datadog/datadog.py @@ -291,7 +291,7 @@ class DataDogLogger( dd_payload = DatadogPayload( ddsource=get_datadog_source(), - ddtags=get_datadog_tags(), + ddtags=",".join(get_datadog_tags()), hostname=get_datadog_hostname(), message=safe_dumps(message_payload), service=get_datadog_service(), @@ -442,7 +442,9 @@ class DataDogLogger( verbose_logger.debug("Datadog: Logger - Logging payload = %s", json_payload) dd_payload = DatadogPayload( ddsource=get_datadog_source(), - ddtags=get_datadog_tags(standard_logging_object=standard_logging_object), + ddtags=",".join( + get_datadog_tags(standard_logging_object=standard_logging_object) + ), hostname=get_datadog_hostname(), message=json_payload, service=get_datadog_service(), @@ -545,7 +547,7 @@ class DataDogLogger( _dd_message_str = safe_dumps(_payload_dict) _dd_payload = DatadogPayload( ddsource=get_datadog_source(), - ddtags=get_datadog_tags(), + ddtags=",".join(get_datadog_tags()), hostname=get_datadog_hostname(), message=_dd_message_str, service=get_datadog_service(), @@ -587,7 +589,7 @@ class DataDogLogger( _dd_message_str = safe_dumps(_payload_dict) _dd_payload = DatadogPayload( ddsource=get_datadog_source(), - ddtags=get_datadog_tags(), + ddtags=",".join(get_datadog_tags()), hostname=get_datadog_hostname(), message=_dd_message_str, service=get_datadog_service(), @@ -678,7 +680,7 @@ class DataDogLogger( dd_payload = DatadogPayload( ddsource=get_datadog_source(), - ddtags=get_datadog_tags(), + ddtags=",".join(get_datadog_tags()), hostname=get_datadog_hostname(), message=json_payload, service=get_datadog_service(), diff --git a/litellm/integrations/datadog/datadog_handler.py b/litellm/integrations/datadog/datadog_handler.py index 0406f1e5d20..b6bb2b57037 100644 --- a/litellm/integrations/datadog/datadog_handler.py +++ b/litellm/integrations/datadog/datadog_handler.py @@ -38,8 +38,13 @@ def get_datadog_pod_name() -> str: def get_datadog_tags( standard_logging_object: Optional[StandardLoggingPayload] = None, -) -> str: - """Build Datadog tags string used by multiple integrations.""" +) -> List[str]: + """Build Datadog tags as a list of individual tag strings. + + Returns a list of "key:value" strings suitable for Datadog LLM Observability + (which expects tags as an array). For Datadog Logs API (ddtags), join with + comma: ",".join(get_datadog_tags(...)). + """ base_tags = { "env": get_datadog_env(), @@ -66,4 +71,4 @@ def get_datadog_tags( if team_tag: tags.append(f"team:{team_tag}") - return ",".join(tags) + return tags diff --git a/litellm/integrations/datadog/datadog_llm_obs.py b/litellm/integrations/datadog/datadog_llm_obs.py index de6cc02fa3d..ec6c00961b6 100644 --- a/litellm/integrations/datadog/datadog_llm_obs.py +++ b/litellm/integrations/datadog/datadog_llm_obs.py @@ -203,7 +203,7 @@ class DataDogLLMObsLogger(CustomBatchLogger): type="span", attributes=DDSpanAttributes( ml_app=get_datadog_service(), - tags=[get_datadog_tags()], + tags=get_datadog_tags(), spans=self.log_queue, ), ), @@ -315,7 +315,7 @@ class DataDogLLMObsLogger(CustomBatchLogger): duration=int((end_time - start_time).total_seconds() * 1e9), metrics=metrics, status="error" if error_info else "ok", - tags=[get_datadog_tags(standard_logging_object=standard_logging_payload)], + tags=get_datadog_tags(standard_logging_object=standard_logging_payload), ) apm_trace_id = self._get_apm_trace_id() diff --git a/litellm/integrations/langsmith.py b/litellm/integrations/langsmith.py index 03845af521d..9cc37359928 100644 --- a/litellm/integrations/langsmith.py +++ b/litellm/integrations/langsmith.py @@ -5,7 +5,6 @@ import os import random import traceback import types -from litellm._uuid import uuid from datetime import datetime, timezone from typing import Any, Dict, List, Optional @@ -14,10 +13,11 @@ from pydantic import BaseModel # type: ignore import litellm from litellm._logging import verbose_logger +from litellm._uuid import uuid from litellm.integrations.custom_batch_logger import CustomBatchLogger from litellm.integrations.langsmith_mock_client import ( - should_use_langsmith_mock, create_mock_langsmith_client, + should_use_langsmith_mock, ) from litellm.llms.custom_httpx.http_handler import ( get_async_httpx_client, @@ -110,6 +110,60 @@ class LangsmithLogger(CustomBatchLogger): LANGSMITH_TENANT_ID=_credentials_tenant_id, ) + def _extract_metadata_fields( + self, metadata: dict, credentials: LangsmithCredentialsObject + ): + return { + "project_name": metadata.get( + "project_name", credentials["LANGSMITH_PROJECT"] + ), + "run_name": metadata.get("run_name", self.langsmith_default_run_name), + "run_id": metadata.get("id", metadata.get("run_id", None)), + "parent_run_id": metadata.get("parent_run_id", None), + "trace_id": metadata.get("trace_id", None), + "session_id": metadata.get("session_id", None), + "dotted_order": metadata.get("dotted_order", None), + } + + def _build_extra_metadata(self, metadata: Dict): + extra_metadata = dict(metadata) + requester_metadata = extra_metadata.get("requester_metadata") + if requester_metadata and isinstance(requester_metadata, dict): + for key in ("session_id", "thread_id", "conversation_id"): + if key in requester_metadata and key not in extra_metadata: + extra_metadata[key] = requester_metadata[key] + return extra_metadata + + def _build_outputs_with_usage( + self, payload: StandardLoggingPayload + ) -> Dict[str, Any]: + response = payload["response"] + outputs: Dict[str, Any] + if isinstance(response, dict): + outputs = {**response} + else: + outputs = {"output": response} + outputs["usage_metadata"] = { + "input_tokens": payload.get("prompt_tokens", 0), + "output_tokens": payload.get("completion_tokens", 0), + "total_tokens": payload.get("total_tokens", 0), + "total_cost": payload.get("response_cost", 0), + } + return outputs + + def _ensure_required_ids(self, data: dict, run_id: Optional[str]): + if "id" not in data or data["id"] is None: + run_id = str(uuid.uuid4()) + data["id"] = run_id + + if "trace_id" not in data or data["trace_id"] is None: + if run_id is not None and isinstance(run_id, str): + data["trace_id"] = run_id + + if "dotted_order" not in data or data["dotted_order"] is None: + if run_id is not None and isinstance(run_id, str): + data["dotted_order"] = self.make_dot_order(run_id=run_id) + def _prepare_log_data( self, kwargs, @@ -121,44 +175,28 @@ class LangsmithLogger(CustomBatchLogger): try: _litellm_params = kwargs.get("litellm_params", {}) or {} metadata = _litellm_params.get("metadata", {}) or {} - project_name = metadata.get( - "project_name", credentials["LANGSMITH_PROJECT"] - ) - run_name = metadata.get("run_name", self.langsmith_default_run_name) - run_id = metadata.get("id", metadata.get("run_id", None)) - parent_run_id = metadata.get("parent_run_id", None) - trace_id = metadata.get("trace_id", None) - session_id = metadata.get("session_id", None) - dotted_order = metadata.get("dotted_order", None) + + fields = self._extract_metadata_fields(metadata, credentials) verbose_logger.debug( - f"Langsmith Logging - project_name: {project_name}, run_name {run_name}" + f"Langsmith Logging - project_name: {fields['project_name']}, run_name {fields['run_name']}" ) - # Ensure everything in the payload is converted to str payload: Optional[StandardLoggingPayload] = kwargs.get( "standard_logging_object", None ) - if payload is None: raise Exception("Error logging request payload. Payload=none.") - metadata = payload[ - "metadata" - ] # ensure logged metadata is json serializable - - extra_metadata = dict(metadata) - requester_metadata = extra_metadata.get("requester_metadata") - if requester_metadata and isinstance(requester_metadata, dict): - for key in ("session_id", "thread_id", "conversation_id"): - if key in requester_metadata and key not in extra_metadata: - extra_metadata[key] = requester_metadata[key] + metadata = payload["metadata"] + extra_metadata = self._build_extra_metadata(dict(metadata)) + outputs = self._build_outputs_with_usage(payload) data = { - "name": run_name, - "run_type": "llm", # this should always be llm, since litellm always logs llm calls. Langsmith allow us to log "chain" + "name": fields["run_name"], + "run_type": "llm", "inputs": payload, - "outputs": payload["response"], - "session_name": project_name, + "outputs": outputs, + "session_name": fields["project_name"], "start_time": payload["startTime"], "end_time": payload["endTime"], "tags": payload["request_tags"], @@ -168,46 +206,19 @@ class LangsmithLogger(CustomBatchLogger): if payload["error_str"] is not None and payload["status"] == "failure": data["error"] = payload["error_str"] - if run_id: - data["id"] = run_id - - if parent_run_id: - data["parent_run_id"] = parent_run_id - - if trace_id: - data["trace_id"] = trace_id - - if session_id: - data["session_id"] = session_id - - if dotted_order: - data["dotted_order"] = dotted_order - - run_id: Optional[str] = data.get("id") # type: ignore - if "id" not in data or data["id"] is None: - """ - for /batch langsmith requires id, trace_id and dotted_order passed as params - """ - run_id = str(uuid.uuid4()) - - data["id"] = run_id - - if ( - "trace_id" not in data - or data["trace_id"] is None - and (run_id is not None and isinstance(run_id, str)) + for key in ( + "id", + "parent_run_id", + "trace_id", + "session_id", + "dotted_order", ): - data["trace_id"] = run_id - - if ( - "dotted_order" not in data - or data["dotted_order"] is None - and (run_id is not None and isinstance(run_id, str)) - ): - data["dotted_order"] = self.make_dot_order(run_id=run_id) # type: ignore + field_key = "run_id" if key == "id" else key + if fields[field_key]: + data[key] = fields[field_key] + self._ensure_required_ids(data, fields["run_id"]) verbose_logger.debug("Langsmith Logging data on langsmith: %s", data) - return data except Exception: raise diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py index 826396a70d8..fea139a64b4 100644 --- a/litellm/litellm_core_utils/litellm_logging.py +++ b/litellm/litellm_core_utils/litellm_logging.py @@ -84,6 +84,8 @@ from litellm.types.llms.openai import ( OpenAIModerationResponse, ResponseAPIUsage, ResponseCompletedEvent, + ResponseFailedEvent, + ResponseIncompleteEvent, ResponsesAPIResponse, ) from litellm.types.mcp import MCPPostCallResponseObject @@ -516,6 +518,23 @@ class Logging(LiteLLMLoggingBaseClass): ), ) + def get_router_model_id(self) -> Optional[str]: + """Extract the router deployment model_id from litellm_params. + + Checks both litellm_metadata and metadata for model_info.id. + Used by cost calculators to look up custom pricing registered + under the deployment's model_info.id in litellm.model_cost. + """ + if not hasattr(self, "litellm_params"): + return None + for key in ("litellm_metadata", "metadata"): + meta = self.litellm_params.get(key, {}) or {} + info = meta.get("model_info", {}) or {} + model_id = info.get("id") + if model_id is not None: + return model_id + return None + def update_environment_variables( self, litellm_params: Dict, @@ -1455,6 +1474,12 @@ class Logging(LiteLLMLoggingBaseClass): ): # use model_id if not already set router_model_id = hidden_params["model_id"] + # Fallback: extract router_model_id from litellm_params when not available + # from the result object. ResponsesAPIResponse objects (used by /v1/responses + # streaming) don't carry _hidden_params["model_id"] like ModelResponse does. + if router_model_id is None: + router_model_id = self.get_router_model_id() + ## RESPONSE COST ## custom_pricing = use_custom_pricing_for_model( litellm_params=( @@ -2958,8 +2983,7 @@ class Logging(LiteLLMLoggingBaseClass): if ( isinstance(callback, CustomLogger) and is_sync_request - and self.call_type - != CallTypes.pass_through.value + and self.call_type != CallTypes.pass_through.value ): # custom logger class callback.log_failure_event( start_time=start_time, @@ -3307,7 +3331,10 @@ class Logging(LiteLLMLoggingBaseClass): return result elif isinstance(result, TextCompletionResponse): return result - elif isinstance(result, ResponseCompletedEvent): + elif isinstance( + result, + (ResponseCompletedEvent, ResponseIncompleteEvent, ResponseFailedEvent), + ): ## return unified Usage object if isinstance(result.response.usage, ResponseAPIUsage): transformed_usage = ( @@ -3328,7 +3355,6 @@ class Logging(LiteLLMLoggingBaseClass): return result.response else: return None - return None def _handle_anthropic_messages_response_logging(self, result: Any) -> ModelResponse: """ diff --git a/litellm/litellm_core_utils/streaming_handler.py b/litellm/litellm_core_utils/streaming_handler.py index 6e991e6911b..96e70845b28 100644 --- a/litellm/litellm_core_utils/streaming_handler.py +++ b/litellm/litellm_core_utils/streaming_handler.py @@ -160,6 +160,7 @@ class CustomStreamWrapper: self.chunks: List = ( [] ) # keep track of the returned chunks - used for calculating the input/output tokens for stream options + self._repeated_messages_count = 1 self.is_function_call = self.check_is_function_call(logging_obj=logging_obj) self.created: Optional[int] = None self._last_returned_hidden_params: Optional[dict] = None @@ -241,7 +242,7 @@ class CustomStreamWrapper: except Exception as e: raise e - def safety_checker(self) -> None: + def raise_on_model_repetition(self) -> None: """ Fixes - https://github.com/BerriAI/litellm/issues/5158 @@ -249,28 +250,35 @@ class CustomStreamWrapper: Raises - InternalServerError, if LLM enters infinite loop while streaming """ - if len(self.chunks) >= litellm.REPEATED_STREAMING_CHUNK_LIMIT: - # Get the last n chunks - last_chunks = self.chunks[-litellm.REPEATED_STREAMING_CHUNK_LIMIT :] + if len(self.chunks) < 2: + return - # Extract the relevant content from the chunks - last_contents = [chunk.choices[0].delta.content for chunk in last_chunks] + last_content = self.chunks[-1].choices[0].delta.content - # Check if all extracted contents are identical - if all(content == last_contents[0] for content in last_contents): - if ( - last_contents[0] is not None - and isinstance(last_contents[0], str) - and len(last_contents[0]) > 2 - ): # ignore empty content - https://github.com/BerriAI/litellm/issues/5158#issuecomment-2287156946 - # All last n chunks are identical - raise litellm.InternalServerError( - message="The model is repeating the same chunk = {}.".format( - last_contents[0] - ), - model="", - llm_provider="", - ) + if ( + last_content is None + or not isinstance(last_content, str) + or len(last_content) <= 2 + ): # ignore empty content - https://github.com/BerriAI/litellm/issues/5158#issuecomment-2287156946 + self._repeated_messages_count = 1 + return + + second_to_last_content = self.chunks[-2].choices[0].delta.content + + if last_content == second_to_last_content: + self._repeated_messages_count += 1 + else: + self._repeated_messages_count = 1 + + if self._repeated_messages_count >= litellm.REPEATED_STREAMING_CHUNK_LIMIT: + # All last n chunks are identical + raise litellm.InternalServerError( + message="The model is repeating the same chunk = {}.".format( + last_content + ), + model="", + llm_provider="", + ) def check_special_tokens(self, chunk: str, finish_reason: Optional[str]): """ @@ -924,7 +932,7 @@ class CustomStreamWrapper: if ( is_chunk_non_empty ): # cannot set content of an OpenAI Object to be an empty string - self.safety_checker() + self.raise_on_model_repetition() hold, model_response_str = self.check_special_tokens( chunk=completion_obj["content"], finish_reason=model_response.choices[0].finish_reason, @@ -1893,15 +1901,19 @@ class CustomStreamWrapper: "usage", getattr(complete_streaming_response, "usage"), ) + try: + _cache_copy = complete_streaming_response.model_copy(deep=True) + _log_copy = complete_streaming_response.model_copy(deep=True) + except RuntimeError: + _cache_copy = complete_streaming_response.model_copy() + _log_copy = complete_streaming_response.model_copy() self.cache_streaming_response( - processed_chunk=complete_streaming_response.model_copy( - deep=True - ), + processed_chunk=_cache_copy, cache_hit=cache_hit, ) executor.submit( self.logging_obj.success_handler, - complete_streaming_response.model_copy(deep=True), + _log_copy, None, None, cache_hit, @@ -2113,11 +2125,13 @@ class CustomStreamWrapper: "usage", getattr(complete_streaming_response, "usage"), ) + try: + _copy = complete_streaming_response.model_copy(deep=True) + except RuntimeError: + _copy = complete_streaming_response.model_copy() asyncio.create_task( self.async_cache_streaming_response( - processed_chunk=complete_streaming_response.model_copy( - deep=True - ), + processed_chunk=_copy, cache_hit=cache_hit, ) ) diff --git a/litellm/llms/anthropic/batches/transformation.py b/litellm/llms/anthropic/batches/transformation.py index 699f133f0f6..98c0588a091 100644 --- a/litellm/llms/anthropic/batches/transformation.py +++ b/litellm/llms/anthropic/batches/transformation.py @@ -42,9 +42,8 @@ class AnthropicBatchesConfig(BaseBatchesConfig): api_base: Optional[str] = None, ) -> dict: """Validate and prepare environment-specific headers and parameters.""" - # Resolve api_key from environment if not provided - api_key = api_key or self.anthropic_model_info.get_api_key() - if api_key is None: + auth_header = self.anthropic_model_info.get_auth_header(api_key) + if auth_header is None: raise ValueError( "Missing Anthropic API Key - A call is being made to anthropic but no key is set either in the environment variables or via params" ) @@ -52,8 +51,8 @@ class AnthropicBatchesConfig(BaseBatchesConfig): "accept": "application/json", "anthropic-version": "2023-06-01", "content-type": "application/json", - "x-api-key": api_key, } + _headers.update(auth_header) # Add beta header for message batches if "anthropic-beta" not in headers: headers["anthropic-beta"] = "message-batches-2024-09-24" diff --git a/litellm/llms/anthropic/chat/handler.py b/litellm/llms/anthropic/chat/handler.py index 5eebebc2e23..9f2ddcae2c7 100644 --- a/litellm/llms/anthropic/chat/handler.py +++ b/litellm/llms/anthropic/chat/handler.py @@ -48,6 +48,10 @@ from litellm.types.llms.openai import ( ChatCompletionToolCallChunk, ChatCompletionToolCallFunctionChunk, ) +from litellm.types.responses.main import ( + OutputCodeInterpreterCall, + build_code_interpreter_log_outputs, +) from litellm.types.utils import ( Delta, GenericStreamingChunk, @@ -538,6 +542,12 @@ class ModelResponseIterator: # Accumulate compaction blocks for multi-turn reconstruction self.compaction_blocks: List[Dict[str, Any]] = [] + # Track server tool use inputs and results for code_interpreter_results + self._server_tool_inputs: Dict[str, Any] = {} + self.tool_results: List[Dict[str, Any]] = [] + self._current_server_tool_id: Optional[str] = None + self._container_id: Optional[str] = None + def check_empty_tool_call_args(self) -> bool: """ Check if the tool call block so far has been an empty string @@ -682,6 +692,39 @@ class ModelResponseIterator: return content_block_start + def _build_code_interpreter_results(self) -> list: + """Convert accumulated tool_results to OutputCodeInterpreterCall objects. + + Called during streaming to produce provider-neutral code_interpreter_results + alongside the raw tool_results, so the Responses API layer doesn't need + Anthropic-specific knowledge. + + Returns the full cumulative list each time (not incremental), matching + how web_search_results works. stream_chunk_builder uses "last value + wins" for list-valued provider_specific_fields keys, so the last + emission must contain every result. + """ + results = [] + for tr in self.tool_results: + if tr.get("type") != "bash_code_execution_tool_result": + continue + call_id = tr.get("tool_use_id", "") + content = tr.get("content", {}) + log_outputs = build_code_interpreter_log_outputs(content) + tool_input = self._server_tool_inputs.get(call_id, {}) + code = tool_input.get("command", "") if isinstance(tool_input, dict) else "" + results.append( + OutputCodeInterpreterCall( + type="code_interpreter_call", + id=call_id, + code=code, + container_id=self._container_id, + status="completed", + outputs=log_outputs, + ) + ) + return results + def chunk_parser(self, chunk: dict) -> ModelResponseStream: # noqa: PLR0915 try: type_chunk = chunk.get("type", "") or "" @@ -748,6 +791,23 @@ class ModelResponseIterator: ), index=self.tool_index, ) + # Track server tool use inputs for code_interpreter_results. + # The initial input in content_block_start is typically {} + # for streaming; the full input arrives via input_json_delta + # and is assembled at content_block_stop. + if ( + content_block_start["content_block"]["type"] + == "server_tool_use" + ): + self._current_server_tool_id = content_block_start[ + "content_block" + ]["id"] + tool_input = content_block_start["content_block"].get( + "input", {} + ) + self._server_tool_inputs[ + self._current_server_tool_id + ] = tool_input # Include caller information if present (for programmatic tool calling) if "caller" in content_block_start["content_block"]: caller_data = content_block_start["content_block"]["caller"] @@ -808,10 +868,12 @@ class ModelResponseIterator: elif content_type != "tool_search_tool_result": # Handle other tool results (code execution, etc.) # Skip tool_search_tool_result as it's internal metadata - if not hasattr(self, "tool_results"): - self.tool_results = [] self.tool_results.append(content_block_start["content_block"]) provider_specific_fields["tool_results"] = self.tool_results + # Convert to provider-neutral code_interpreter_results + provider_specific_fields[ + "code_interpreter_results" + ] = self._build_code_interpreter_results() elif type_chunk == "content_block_stop": ContentBlockStop(**chunk) # type: ignore @@ -828,6 +890,26 @@ class ModelResponseIterator: ), index=self.tool_index, ) + # Update server_tool_inputs with fully assembled input + # from input_json_delta chunks (content_block_start has {}) + if ( + self.current_content_block_type == "server_tool_use" + and self._current_server_tool_id + ): + args = "" + for block in self.content_blocks: + if block["delta"]["type"] == "input_json_delta": + partial_json = block["delta"].get("partial_json") + if isinstance(partial_json, str): + args += partial_json + if args: + try: + self._server_tool_inputs[ + self._current_server_tool_id + ] = json.loads(args) + except (json.JSONDecodeError, TypeError): + pass + self._current_server_tool_id = None # Reset response_format tool tracking when block stops self.is_response_format_tool = False # Reset current content block type @@ -840,6 +922,17 @@ class ModelResponseIterator: finish_reason, usage, container = self._handle_message_delta(chunk) if container: provider_specific_fields["container"] = container + # Store container_id and re-emit code_interpreter_results + # so stream_chunk_builder's last-value-wins picks up the + # version with container_id populated. + container_id = ( + container.get("id") if isinstance(container, dict) else None + ) + if container_id and self.tool_results: + self._container_id = container_id + provider_specific_fields[ + "code_interpreter_results" + ] = self._build_code_interpreter_results() elif type_chunk == "message_start": """ Anthropic diff --git a/litellm/llms/anthropic/chat/transformation.py b/litellm/llms/anthropic/chat/transformation.py index 47cdd8287e0..73d1b02c76d 100644 --- a/litellm/llms/anthropic/chat/transformation.py +++ b/litellm/llms/anthropic/chat/transformation.py @@ -50,6 +50,10 @@ from litellm.types.llms.openai import ( OpenAIMcpServerTool, OpenAIWebSearchOptions, ) +from litellm.types.responses.main import ( + OutputCodeInterpreterCall, + build_code_interpreter_log_outputs, +) from litellm.types.utils import ( CacheCreationTokenDetails, CompletionTokensDetailsWrapper, @@ -1522,7 +1526,7 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): tool_results = [] tool_results.append(content) - elif content.get("thinking", None) is not None: + elif content.get("type") == "thinking": if thinking_blocks is None: thinking_blocks = [] thinking_blocks.append(cast(ChatCompletionThinkingBlock, content)) @@ -1682,6 +1686,96 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): ) return usage + def _build_code_by_id_map( + self, tool_calls: List[ChatCompletionToolCallChunk] + ) -> Dict[str, str]: + code_by_id: Dict[str, str] = {} + for tc in tool_calls: + try: + args = json.loads(tc.get("function", {}).get("arguments", "{}")) + call_id = tc.get("id") + command = args.get("command", "") + if isinstance(call_id, str): + code_by_id[call_id] = command if isinstance(command, str) else "" + except Exception: + pass + return code_by_id + + def _build_code_interpreter_results( + self, + tool_results: List[Any], + code_by_id: Dict[str, str], + container_id: Optional[str], + ) -> List[OutputCodeInterpreterCall]: + code_interpreter_results = [] + for tr in tool_results: + if tr.get("type") != "bash_code_execution_tool_result": + continue + call_id = tr.get("tool_use_id", "") + content = tr.get("content", {}) + log_outputs = build_code_interpreter_log_outputs(content) + code_interpreter_results.append( + OutputCodeInterpreterCall( + type="code_interpreter_call", + id=call_id, + code=code_by_id.get(call_id, ""), + container_id=container_id, + status="completed", + outputs=log_outputs, + ) + ) + return code_interpreter_results + + def _build_provider_specific_fields( + self, + completion_response: dict, + citations: Optional[List[Any]], + thinking_blocks: Optional[ + List[ + Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock] + ] + ], + web_search_results: Optional[List[Any]], + tool_results: Optional[List[Any]], + compaction_blocks: Optional[List[Any]], + tool_calls: List[ChatCompletionToolCallChunk], + ) -> Dict[str, Any]: + provider_specific_fields: Dict[str, Any] = { + "citations": citations, + "thinking_blocks": thinking_blocks, + } + + context_management = completion_response.get("context_management") + if context_management is not None: + provider_specific_fields["context_management"] = context_management + + if web_search_results is not None: + provider_specific_fields["web_search_results"] = web_search_results + + if tool_results is not None: + provider_specific_fields["tool_results"] = tool_results + container_id = ( + completion_response.get("container", {}).get("id") + if isinstance(completion_response.get("container"), dict) + else None + ) + code_by_id = self._build_code_by_id_map(tool_calls) + code_interpreter_results = self._build_code_interpreter_results( + tool_results, code_by_id, container_id + ) + provider_specific_fields[ + "code_interpreter_results" + ] = code_interpreter_results + + container = completion_response.get("container") + if container is not None: + provider_specific_fields["container"] = container + + if compaction_blocks is not None: + provider_specific_fields["compaction_blocks"] = compaction_blocks + + return provider_specific_fields + def transform_parsed_response( self, completion_response: dict, @@ -1702,98 +1796,73 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig): status_code=raw_response.status_code, headers=response_headers, ) - else: - text_content = "" - citations: Optional[List[Any]] = None - thinking_blocks: Optional[ - List[ - Union[ - ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock - ] - ] - ] = None - reasoning_content: Optional[str] = None - tool_calls: List[ChatCompletionToolCallChunk] = [] - ( - text_content, - citations, - thinking_blocks, - reasoning_content, - tool_calls, - web_search_results, - tool_results, - compaction_blocks, - ) = self.extract_response_content(completion_response=completion_response) + ( + text_content, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = self.extract_response_content(completion_response=completion_response) - if ( - prefix_prompt is not None - and not text_content.startswith(prefix_prompt) - and not litellm.disable_add_prefix_to_prompt - ): - text_content = prefix_prompt + text_content + if ( + prefix_prompt is not None + and not text_content.startswith(prefix_prompt) + and not litellm.disable_add_prefix_to_prompt + ): + text_content = prefix_prompt + text_content - context_management: Optional[Dict] = completion_response.get( - "context_management" - ) + provider_specific_fields = self._build_provider_specific_fields( + completion_response, + citations, + thinking_blocks, + web_search_results, + tool_results, + compaction_blocks, + tool_calls, + ) - container: Optional[Dict] = completion_response.get("container") + _message = litellm.Message( + tool_calls=tool_calls, + content=text_content or None, + provider_specific_fields=provider_specific_fields, + thinking_blocks=thinking_blocks, + reasoning_content=reasoning_content, + ) + _message.provider_specific_fields = provider_specific_fields - provider_specific_fields: Dict[str, Any] = { - "citations": citations, - "thinking_blocks": thinking_blocks, - } - if context_management is not None: - provider_specific_fields["context_management"] = context_management - if web_search_results is not None: - provider_specific_fields["web_search_results"] = web_search_results - if tool_results is not None: - provider_specific_fields["tool_results"] = tool_results - if container is not None: - provider_specific_fields["container"] = container - if compaction_blocks is not None: - provider_specific_fields["compaction_blocks"] = compaction_blocks + json_mode_message = self._transform_response_for_json_mode( + json_mode=json_mode, + tool_calls=tool_calls, + ) + if json_mode_message is not None: + completion_response["stop_reason"] = "stop" + _message = json_mode_message - _message = litellm.Message( - tool_calls=tool_calls, - content=text_content or None, - provider_specific_fields=provider_specific_fields, - thinking_blocks=thinking_blocks, - reasoning_content=reasoning_content, - ) - _message.provider_specific_fields = provider_specific_fields + model_response.choices[0].message = _message + model_response._hidden_params["original_response"] = completion_response[ + "content" + ] + model_response.choices[0].finish_reason = cast( + OpenAIChatCompletionFinishReason, + map_finish_reason(completion_response["stop_reason"]), + ) - ## HANDLE JSON MODE - anthropic returns single function call - json_mode_message = self._transform_response_for_json_mode( - json_mode=json_mode, - tool_calls=tool_calls, - ) - if json_mode_message is not None: - completion_response["stop_reason"] = "stop" - _message = json_mode_message - - model_response.choices[0].message = _message # type: ignore - model_response._hidden_params["original_response"] = completion_response[ - "content" - ] # allow user to access raw anthropic tool calling response - - model_response.choices[0].finish_reason = cast( - OpenAIChatCompletionFinishReason, - map_finish_reason(completion_response["stop_reason"]), - ) - - ## CALCULATING USAGE usage = self.calculate_usage( usage_object=completion_response["usage"], reasoning_content=reasoning_content, completion_response=completion_response, speed=speed, ) - setattr(model_response, "usage", usage) # type: ignore + setattr(model_response, "usage", usage) model_response.created = int(time.time()) model_response.model = completion_response["model"] + _hidden_params["provider_specific_fields"] = provider_specific_fields model_response._hidden_params = _hidden_params return model_response diff --git a/litellm/llms/anthropic/common_utils.py b/litellm/llms/anthropic/common_utils.py index ac352467878..7d2d0a74961 100644 --- a/litellm/llms/anthropic/common_utils.py +++ b/litellm/llms/anthropic/common_utils.py @@ -359,9 +359,7 @@ class AnthropicModelInfo(BaseLLMModelInfo): Returns: List of beta header strings """ - from litellm.types.llms.anthropic import ( - ANTHROPIC_EFFORT_BETA_HEADER, - ) + from litellm.types.llms.anthropic import ANTHROPIC_EFFORT_BETA_HEADER betas = [] @@ -390,7 +388,8 @@ class AnthropicModelInfo(BaseLLMModelInfo): def get_anthropic_headers( self, - api_key: str, + api_key: Optional[str] = None, + auth_token: Optional[str] = None, anthropic_version: Optional[str] = None, computer_tool_used: Optional[str] = None, prompt_caching_set: bool = False, @@ -451,7 +450,9 @@ class AnthropicModelInfo(BaseLLMModelInfo): headers["authorization"] = f"Bearer {api_key}" headers["anthropic-dangerous-direct-browser-access"] = "true" betas.add(ANTHROPIC_OAUTH_BETA_HEADER) - else: + elif auth_token and not api_key: + headers["authorization"] = f"Bearer {auth_token}" + elif api_key: headers["x-api-key"] = api_key if user_anthropic_beta_headers is not None: @@ -485,9 +486,14 @@ class AnthropicModelInfo(BaseLLMModelInfo): headers, api_key = optionally_handle_anthropic_oauth( headers=headers, api_key=api_key ) + api_key = AnthropicModelInfo.get_api_key(api_key) + # Resolve auth_token from ANTHROPIC_AUTH_TOKEN if api_key is not set + auth_token: Optional[str] = None if api_key is None: + auth_token = AnthropicModelInfo.get_auth_token() + if api_key is None and auth_token is None: raise litellm.AuthenticationError( - message="Missing Anthropic API Key - A call is being made to anthropic but no key is set either in the environment variables or via params. Please set `ANTHROPIC_API_KEY` in your environment vars", + message="Missing Anthropic API Key - A call is being made to anthropic but no key is set either in the environment variables or via params. Please set `ANTHROPIC_API_KEY` or `ANTHROPIC_AUTH_TOKEN` in your environment vars", llm_provider="anthropic", model=model, ) @@ -519,6 +525,7 @@ class AnthropicModelInfo(BaseLLMModelInfo): prompt_caching_set=prompt_caching_set, pdf_used=pdf_used, api_key=api_key, + auth_token=auth_token, file_id_used=file_id_used, web_search_tool_used=web_search_tool_used, is_vertex_request=optional_params.get("is_vertex_request", False), @@ -543,6 +550,7 @@ class AnthropicModelInfo(BaseLLMModelInfo): return ( api_base or get_secret_str("ANTHROPIC_API_BASE") + or get_secret_str("ANTHROPIC_BASE_URL") or "https://api.anthropic.com" ) @@ -552,6 +560,35 @@ class AnthropicModelInfo(BaseLLMModelInfo): return api_key or get_secret_str("ANTHROPIC_API_KEY") + @staticmethod + def get_auth_token(auth_token: Optional[str] = None) -> Optional[str]: + """Get auth token from ANTHROPIC_AUTH_TOKEN env var. + + Unlike api_key (which uses X-Api-Key header), auth_token uses + Authorization: Bearer header, matching the official Anthropic SDK behavior. + """ + from litellm.secret_managers.main import get_secret_str + + return auth_token or get_secret_str("ANTHROPIC_AUTH_TOKEN") + + @staticmethod + def get_auth_header(api_key: Optional[str] = None) -> Optional[dict]: + """Resolve Anthropic credentials and return the appropriate auth header dict. + + Checks ANTHROPIC_API_KEY first (-> x-api-key), then + ANTHROPIC_AUTH_TOKEN (-> Authorization: Bearer). + Returns None if neither is available. + """ + resolved_key = AnthropicModelInfo.get_api_key(api_key) + if resolved_key is not None: + if is_anthropic_oauth_key(resolved_key): + return {"authorization": f"Bearer {resolved_key}"} + return {"x-api-key": resolved_key} + auth_token = AnthropicModelInfo.get_auth_token() + if auth_token is not None: + return {"authorization": f"Bearer {auth_token}"} + return None + @staticmethod def get_base_model(model: Optional[str] = None) -> Optional[str]: return model.replace("anthropic/", "") if model else None @@ -560,14 +597,16 @@ class AnthropicModelInfo(BaseLLMModelInfo): self, api_key: Optional[str] = None, api_base: Optional[str] = None ) -> List[str]: api_base = AnthropicModelInfo.get_api_base(api_base) - api_key = AnthropicModelInfo.get_api_key(api_key) - if api_base is None or api_key is None: + auth_header = AnthropicModelInfo.get_auth_header(api_key) + if api_base is None or auth_header is None: raise ValueError( - "ANTHROPIC_API_BASE or ANTHROPIC_API_KEY is not set. Please set the environment variable, to query Anthropic's `/models` endpoint." + "ANTHROPIC_API_BASE/ANTHROPIC_BASE_URL or ANTHROPIC_API_KEY/ANTHROPIC_AUTH_TOKEN is not set. Please set the environment variable, to query Anthropic's `/models` endpoint." ) + headers = {"anthropic-version": "2023-06-01"} + headers.update(auth_header) response = litellm.module_level_client.get( url=f"{api_base}/v1/models", - headers={"x-api-key": api_key, "anthropic-version": "2023-06-01"}, + headers=headers, ) try: diff --git a/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py b/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py index e9ceea48220..9b60a58260b 100644 --- a/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py +++ b/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py @@ -23,7 +23,6 @@ from ...common_utils import ( optionally_handle_anthropic_oauth, ) -DEFAULT_ANTHROPIC_API_BASE = "https://api.anthropic.com" DEFAULT_ANTHROPIC_API_VERSION = "2023-06-01" @@ -127,7 +126,9 @@ class AnthropicMessagesConfig(BaseAnthropicMessagesConfig): litellm_params: dict, stream: Optional[bool] = None, ) -> str: - api_base = api_base or DEFAULT_ANTHROPIC_API_BASE + api_base = ( + AnthropicModelInfo.get_api_base(api_base) or "https://api.anthropic.com" + ) if not api_base.endswith("/v1/messages"): api_base = f"{api_base}/v1/messages" return api_base @@ -142,17 +143,15 @@ class AnthropicMessagesConfig(BaseAnthropicMessagesConfig): api_key: Optional[str] = None, api_base: Optional[str] = None, ) -> Tuple[dict, Optional[str]]: - import os - # Check for Anthropic OAuth token in Authorization header headers, api_key = optionally_handle_anthropic_oauth( headers=headers, api_key=api_key ) - if api_key is None: - api_key = os.getenv("ANTHROPIC_API_KEY") - if "x-api-key" not in headers and "authorization" not in headers and api_key: - headers["x-api-key"] = api_key + if "x-api-key" not in headers and "authorization" not in headers: + auth_header = AnthropicModelInfo.get_auth_header(api_key) + if auth_header is not None: + headers.update(auth_header) if "anthropic-version" not in headers: headers["anthropic-version"] = DEFAULT_ANTHROPIC_API_VERSION if "content-type" not in headers: diff --git a/litellm/llms/anthropic/files/handler.py b/litellm/llms/anthropic/files/handler.py index 77cc8c27316..c56799f30cf 100644 --- a/litellm/llms/anthropic/files/handler.py +++ b/litellm/llms/anthropic/files/handler.py @@ -8,10 +8,8 @@ import httpx import litellm from litellm._logging import verbose_logger from litellm._uuid import uuid -from litellm.llms.custom_httpx.http_handler import ( - get_async_httpx_client, -) from litellm.litellm_core_utils.litellm_logging import Logging +from litellm.llms.custom_httpx.http_handler import get_async_httpx_client from litellm.types.llms.openai import ( FileContentRequest, HttpxBinaryResponseContent, @@ -85,9 +83,9 @@ class AnthropicFilesHandler: # Get Anthropic API credentials api_base = self.anthropic_model_info.get_api_base(api_base) - api_key = api_key or self.anthropic_model_info.get_api_key() + auth_header = self.anthropic_model_info.get_auth_header(api_key) - if not api_key: + if auth_header is None: raise ValueError("Missing Anthropic API Key") # Construct the Anthropic batch results URL @@ -97,8 +95,8 @@ class AnthropicFilesHandler: headers = { "accept": "application/json", "anthropic-version": "2023-06-01", - "x-api-key": api_key, } + headers.update(auth_header) # Make the request to Anthropic async_client = get_async_httpx_client(llm_provider=LlmProviders.ANTHROPIC) diff --git a/litellm/llms/anthropic/files/transformation.py b/litellm/llms/anthropic/files/transformation.py index 98a548a1369..0545cefb071 100644 --- a/litellm/llms/anthropic/files/transformation.py +++ b/litellm/llms/anthropic/files/transformation.py @@ -94,14 +94,14 @@ class AnthropicFilesConfig(BaseFilesConfig): api_key: Optional[str] = None, api_base: Optional[str] = None, ) -> dict: - api_key = AnthropicModelInfo.get_api_key(api_key) - if not api_key: + auth_header = AnthropicModelInfo.get_auth_header(api_key) + if auth_header is None: raise ValueError( - "Anthropic API key is required. Set ANTHROPIC_API_KEY environment variable or pass api_key parameter." + "Anthropic API key is required. Set ANTHROPIC_API_KEY or ANTHROPIC_AUTH_TOKEN environment variable or pass api_key parameter." ) headers.update( { - "x-api-key": api_key, + **auth_header, "anthropic-version": "2023-06-01", "anthropic-beta": ANTHROPIC_FILES_BETA_HEADER, } diff --git a/litellm/llms/anthropic/skills/transformation.py b/litellm/llms/anthropic/skills/transformation.py index af9863534ed..a992d84d459 100644 --- a/litellm/llms/anthropic/skills/transformation.py +++ b/litellm/llms/anthropic/skills/transformation.py @@ -35,17 +35,18 @@ class AnthropicSkillsConfig(BaseSkillsAPIConfig): """Add Anthropic-specific headers""" from litellm.llms.anthropic.common_utils import AnthropicModelInfo - # Get API key + # Get API key from litellm_params if available api_key = None - if litellm_params: + if litellm_params is not None: api_key = litellm_params.api_key - api_key = AnthropicModelInfo.get_api_key(api_key) - if not api_key: - raise ValueError("ANTHROPIC_API_KEY is required for Skills API") + auth_header = AnthropicModelInfo.get_auth_header(api_key) + if auth_header is None: + raise ValueError( + "ANTHROPIC_API_KEY or ANTHROPIC_AUTH_TOKEN is required for Skills API" + ) - # Add required headers - headers["x-api-key"] = api_key + headers.update(auth_header) headers["anthropic-version"] = "2023-06-01" # Add beta header for skills API diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 181045809f8..879dd42be47 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -4462,6 +4462,78 @@ "supports_vision": true, "supports_web_search": true }, + "azure/gpt-5.4-mini": { + "cache_read_input_token_cost": 7.5e-08, + "input_cost_per_token": 7.5e-07, + "litellm_provider": "azure", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 4.5e-06, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/batch", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_service_tier": true, + "supports_vision": true, + "supports_web_search": true, + "supports_none_reasoning_effort": false, + "supports_xhigh_reasoning_effort": false + }, + "azure/gpt-5.4-nano": { + "cache_read_input_token_cost": 2e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "azure", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.25e-06, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/batch", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_service_tier": true, + "supports_vision": true, + "supports_web_search": true, + "supports_none_reasoning_effort": false, + "supports_xhigh_reasoning_effort": false + }, "azure/gpt-image-1": { "cache_read_input_image_token_cost": 2.5e-06, "cache_read_input_token_cost": 1.25e-06, @@ -37032,5 +37104,157 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true + }, + "volcengine/doubao-seed-2-0-pro-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4.6e-07, + "output_cost_per_token": 2.3e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 7e-07, + "output_cost_per_token": 3.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 7e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "volcengine/doubao-seed-2-0-lite-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 8.7e-08, + "output_cost_per_token": 5.2e-07, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 1.3e-07, + "output_cost_per_token": 7.8e-07, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 2.6e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "volcengine/doubao-seed-2-0-mini-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 2.9e-08, + "output_cost_per_token": 2.9e-07, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 5.8e-08, + "output_cost_per_token": 5.8e-07, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 1.2e-07, + "output_cost_per_token": 1.2e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "volcengine/doubao-seed-2-0-code-preview-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4.6e-07, + "output_cost_per_token": 2.3e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 7e-07, + "output_cost_per_token": 3.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 7e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] } } diff --git a/litellm/proxy/_experimental/mcp_server/discoverable_endpoints.py b/litellm/proxy/_experimental/mcp_server/discoverable_endpoints.py index 3385e7feef6..07309eb57f2 100644 --- a/litellm/proxy/_experimental/mcp_server/discoverable_endpoints.py +++ b/litellm/proxy/_experimental/mcp_server/discoverable_endpoints.py @@ -208,26 +208,52 @@ async def exchange_token_with_server( client_id: str, client_secret: Optional[str], code_verifier: Optional[str], + refresh_token: Optional[str] = None, + scope: Optional[str] = None, ): - if grant_type != "authorization_code": + if grant_type not in ("authorization_code", "refresh_token"): raise HTTPException(status_code=400, detail="Unsupported grant_type") if mcp_server.token_url is None: raise HTTPException(status_code=400, detail="MCP server token url is not set") - proxy_base_url = get_request_base_url(request) - token_data = { - "grant_type": "authorization_code", - "client_id": mcp_server.client_id if mcp_server.client_id else client_id, - "client_secret": mcp_server.client_secret - if mcp_server.client_secret - else client_secret, - "code": code, - "redirect_uri": f"{proxy_base_url}/callback", - } + resolved_client_id = mcp_server.client_id if mcp_server.client_id else client_id + resolved_client_secret = ( + mcp_server.client_secret if mcp_server.client_secret else client_secret + ) - if code_verifier: - token_data["code_verifier"] = code_verifier + if grant_type == "refresh_token": + if not refresh_token: + raise HTTPException( + status_code=400, + detail="refresh_token is required for refresh_token grant", + ) + token_data: dict = { + "grant_type": "refresh_token", + "refresh_token": refresh_token, + "client_id": resolved_client_id, + } + if resolved_client_secret is not None: + token_data["client_secret"] = resolved_client_secret + if scope: + token_data["scope"] = scope + else: + if not code: + raise HTTPException( + status_code=400, + detail="code is required for authorization_code grant", + ) + proxy_base_url = get_request_base_url(request) + token_data = { + "grant_type": "authorization_code", + "client_id": resolved_client_id, + "code": code, + "redirect_uri": f"{proxy_base_url}/callback", + } + if resolved_client_secret is not None: + token_data["client_secret"] = resolved_client_secret + if code_verifier: + token_data["code_verifier"] = code_verifier async_client = get_async_httpx_client(llm_provider=httpxSpecialProvider.Oauth2Check) response = await async_client.post( @@ -375,6 +401,8 @@ async def token_endpoint( client_id: str = Form(...), client_secret: Optional[str] = Form(None), code_verifier: str = Form(None), + refresh_token: Optional[str] = Form(None), + scope: Optional[str] = Form(None), mcp_server_name: Optional[str] = None, ): """ @@ -408,6 +436,8 @@ async def token_endpoint( client_id=client_id, client_secret=client_secret, code_verifier=code_verifier, + refresh_token=refresh_token, + scope=scope, ) diff --git a/litellm/proxy/_new_secret_config.yaml b/litellm/proxy/_new_secret_config.yaml index 508c1c94659..604e7d5f418 100644 --- a/litellm/proxy/_new_secret_config.yaml +++ b/litellm/proxy/_new_secret_config.yaml @@ -1,41 +1,32 @@ model_list: - - model_name: gpt-3.5-turbo - litellm_params: - model: openai/gpt-3.5-turbo - api_key: os.environ/OPENAI_API_KEY - - model_name: gpt-4o - litellm_params: - model: openai/gpt-4o - api_key: os.environ/OPENAI_API_KEY - - model_name: claude-sonnet-4-5-20250929 - litellm_params: - model: anthropic/claude-sonnet-4-5-20250929 - - model_name: gpt-4.1-mini + + # OpenAI model for /v1/chat/completions test — 200x custom pricing + - model_name: "gpt-4.1-mini" litellm_params: model: openai/gpt-4.1-mini - - model_name: gpt-5-mini + api_key: os.environ/OPENAI_API_KEY + model_info: + id: gpt-4.1-mini-custom-pricing + input_cost_per_token: 0.00004 # 100x standard ($0.40/1M = $0.0000004) + output_cost_per_token: 0.00016 # 100x standard ($1.60/1M = $0.0000016) + + # OpenAI model for /v1/responses test — 100x custom pricing + - model_name: "gpt-5" litellm_params: - model: openai/gpt-5-mini - - model_name: custom_litellm_model + model: openai/gpt-5 + api_key: os.environ/OPENAI_API_KEY + model_info: + id: gpt-5-custom-pricing + mode: "chat" + input_cost_per_token: 125 # 100x standard ($1.25/1M = $0.00000125) + output_cost_per_token: 10 # 100x standard ($10.00/1M = $0.00001) + + # Anthropic model for /v1/messages test — 100x custom pricing + - model_name: "claude-sonnet-4-20250514" litellm_params: - model: litellm_agent/claude-sonnet-4-5-20250929 - litellm_system_prompt: "Be a helpful assistant." - - -guardrails: - - guardrail_name: "tool_policy" - litellm_params: - guardrail: tool_policy - mode: [pre_call, post_call] - default_on: true - -mcp_servers: - my_http_server: - url: "http://0.0.0.0:8001/mcp" - transport: "http" - description: "My custom MCP server" - available_on_public_internet: true - -general_settings: - store_model_in_db: true - store_prompts_in_spend_logs: true + model: anthropic/claude-sonnet-4-20250514 + api_key: os.environ/ANTHROPIC_API_KEY + model_info: + id: claude-sonnet-4-custom-pricing + input_cost_per_token: 0.0003 # 100x standard ($0.000003) + output_cost_per_token: 0.0015 # 100x standard ($0.000015) \ No newline at end of file diff --git a/litellm/proxy/_types.py b/litellm/proxy/_types.py index 9e86680e355..91a953c217e 100644 --- a/litellm/proxy/_types.py +++ b/litellm/proxy/_types.py @@ -1672,6 +1672,7 @@ class NewTeamRequest(TeamBase): int ] = None # allow user to set TPM limit for all team members team_member_key_duration: Optional[str] = None # e.g. "1d", "1w", "1m" + team_member_budget_duration: Optional[str] = None # e.g. "30d", "1mo" allowed_vector_store_indexes: Optional[List[AllowedVectorStoreIndexItem]] = None enforced_batch_output_expires_after: Optional[dict] = None enforced_file_expires_after: Optional[dict] = None @@ -2955,7 +2956,9 @@ class LiteLLM_ErrorLogs(LiteLLMPydanticObjectBase): endTime: Union[str, datetime, None] -AUDIT_ACTIONS = Literal["created", "updated", "deleted", "blocked", "rotated"] +AUDIT_ACTIONS = Literal[ + "created", "updated", "deleted", "blocked", "unblocked", "rotated" +] class LiteLLM_AuditLogs(LiteLLMPydanticObjectBase): diff --git a/litellm/proxy/auth/auth_checks.py b/litellm/proxy/auth/auth_checks.py index d31a13e8bc6..1aa14fff574 100644 --- a/litellm/proxy/auth/auth_checks.py +++ b/litellm/proxy/auth/auth_checks.py @@ -29,6 +29,7 @@ from litellm.constants import ( DEFAULT_MAX_RECURSE_DEPTH, EMAIL_BUDGET_ALERT_MAX_SPEND_ALERT_PERCENTAGE, ) +from litellm.litellm_core_utils.dd_tracing import tracer from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider from litellm.proxy._types import ( RBAC_ROLES, @@ -407,18 +408,21 @@ async def common_checks( # noqa: PLR0915 # 2. If team can call model if _model and team_object: - if not await can_team_access_model( - model=_model, - team_object=team_object, - llm_router=llm_router, - team_model_aliases=valid_token.team_model_aliases if valid_token else None, - ): - raise ProxyException( - message=f"Team not allowed to access model. Team={team_object.team_id}, Model={_model}. Allowed team models = {team_object.models}", - type=ProxyErrorTypes.team_model_access_denied, - param="model", - code=status.HTTP_401_UNAUTHORIZED, - ) + with tracer.trace("litellm.proxy.auth.common_checks.can_team_access_model"): + if not await can_team_access_model( + model=_model, + team_object=team_object, + llm_router=llm_router, + team_model_aliases=valid_token.team_model_aliases + if valid_token + else None, + ): + raise ProxyException( + message=f"Team not allowed to access model. Team={team_object.team_id}, Model={_model}. Allowed team models = {team_object.models}", + type=ProxyErrorTypes.team_model_access_denied, + param="model", + code=status.HTTP_401_UNAUTHORIZED, + ) # Require trace id for agent keys when agent has require_trace_id_on_calls_by_agent if valid_token is not None and valid_token.agent_id: @@ -443,54 +447,62 @@ async def common_checks( # noqa: PLR0915 ## 2.1 If user can call model (if personal key) if _model and team_object is None and user_object is not None: - await can_user_call_model( - model=_model, - llm_router=llm_router, - user_object=user_object, - ) + with tracer.trace("litellm.proxy.auth.common_checks.can_user_call_model"): + await can_user_call_model( + model=_model, + llm_router=llm_router, + user_object=user_object, + ) # 1.1 - 2.2 - 3.0.2 - 3.0.3: Project checks (blocked, model access, budget) - await _run_project_checks( - project_object=project_object, - _model=_model, - llm_router=llm_router, - skip_budget_checks=skip_budget_checks, - valid_token=valid_token, - proxy_logging_obj=proxy_logging_obj, - ) + with tracer.trace("litellm.proxy.auth.common_checks.run_project_checks"): + await _run_project_checks( + project_object=project_object, + _model=_model, + llm_router=llm_router, + skip_budget_checks=skip_budget_checks, + valid_token=valid_token, + proxy_logging_obj=proxy_logging_obj, + ) # If this is a free model, skip all budget checks if not skip_budget_checks: # 3. If team is in budget - await _team_max_budget_check( - team_object=team_object, - proxy_logging_obj=proxy_logging_obj, - valid_token=valid_token, - ) + with tracer.trace("litellm.proxy.auth.common_checks.team_max_budget_check"): + await _team_max_budget_check( + team_object=team_object, + proxy_logging_obj=proxy_logging_obj, + valid_token=valid_token, + ) # 3.0.5. If team is over soft budget (alert only, doesn't block) - await _team_soft_budget_check( - team_object=team_object, - proxy_logging_obj=proxy_logging_obj, - valid_token=valid_token, - ) + with tracer.trace("litellm.proxy.auth.common_checks.team_soft_budget_check"): + await _team_soft_budget_check( + team_object=team_object, + proxy_logging_obj=proxy_logging_obj, + valid_token=valid_token, + ) # 3.1. If organization is in budget - await _organization_max_budget_check( - valid_token=valid_token, - team_object=team_object, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - proxy_logging_obj=proxy_logging_obj, - ) + with tracer.trace( + "litellm.proxy.auth.common_checks.organization_max_budget_check" + ): + await _organization_max_budget_check( + valid_token=valid_token, + team_object=team_object, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + proxy_logging_obj=proxy_logging_obj, + ) - await _tag_max_budget_check( - request_body=request_body, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - proxy_logging_obj=proxy_logging_obj, - valid_token=valid_token, - ) + with tracer.trace("litellm.proxy.auth.common_checks.tag_max_budget_check"): + await _tag_max_budget_check( + request_body=request_body, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + proxy_logging_obj=proxy_logging_obj, + valid_token=valid_token, + ) # 4. If user is in budget ## 4.1 check personal budget, if personal key @@ -508,14 +520,15 @@ async def common_checks( # noqa: PLR0915 ) ## 4.2 check team member budget, if team key - await _check_team_member_budget( - team_object=team_object, - user_object=user_object, - valid_token=valid_token, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - proxy_logging_obj=proxy_logging_obj, - ) + with tracer.trace("litellm.proxy.auth.common_checks.check_team_member_budget"): + await _check_team_member_budget( + team_object=team_object, + user_object=user_object, + valid_token=valid_token, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + proxy_logging_obj=proxy_logging_obj, + ) # 5. If end_user ('user' passed to /chat/completions, /embeddings endpoint) is in budget if ( @@ -554,19 +567,21 @@ async def common_checks( # noqa: PLR0915 ) # 11. [OPTIONAL] Vector store checks - is the object allowed to access the vector store - await vector_store_access_check( - request_body=request_body, - team_object=team_object, - valid_token=valid_token, - ) + with tracer.trace("litellm.proxy.auth.common_checks.vector_store_access_check"): + await vector_store_access_check( + request_body=request_body, + team_object=team_object, + valid_token=valid_token, + ) # 12. [OPTIONAL] Tool allowlist - key/team allowed_tools (no DB in hot path) - await check_tools_allowlist( - request_body=request_body, - valid_token=valid_token, - team_object=team_object, - route=route, - ) + with tracer.trace("litellm.proxy.auth.common_checks.check_tools_allowlist"): + await check_tools_allowlist( + request_body=request_body, + valid_token=valid_token, + team_object=team_object, + route=route, + ) return True diff --git a/litellm/proxy/auth/user_api_key_auth.py b/litellm/proxy/auth/user_api_key_auth.py index 044333ac134..eb6a5bdb994 100644 --- a/litellm/proxy/auth/user_api_key_auth.py +++ b/litellm/proxy/auth/user_api_key_auth.py @@ -548,13 +548,12 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 custom_auth_api_key: bool = False try: - # get the request body - - await pre_db_read_auth_checks( - request_data=request_data, - request=request, - route=route, - ) + with tracer.trace("litellm.proxy.auth.pre_db_read_auth_checks"): + await pre_db_read_auth_checks( + request_data=request_data, + request=request, + route=route, + ) pass_through_endpoints: Optional[List[dict]] = general_settings.get( "pass_through_endpoints", None ) @@ -588,9 +587,10 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 ### USER-DEFINED AUTH FUNCTION ### if enterprise_custom_auth is not None: - response = await enterprise_custom_auth( - request=request, api_key=api_key, user_custom_auth=user_custom_auth - ) + with tracer.trace("litellm.proxy.auth.enterprise_custom_auth"): + response = await enterprise_custom_auth( + request=request, api_key=api_key, user_custom_auth=user_custom_auth + ) if response is not None and isinstance(response, UserAPIKeyAuth): validated = UserAPIKeyAuth.model_validate(response) validated = await _run_post_custom_auth_checks( @@ -706,18 +706,19 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 # Fall through to virtual key checks if do_standard_jwt_auth: - result = await JWTAuthManager.auth_builder( - request_data=request_data, - general_settings=general_settings, - api_key=api_key, - jwt_handler=jwt_handler, - route=route, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - proxy_logging_obj=proxy_logging_obj, - parent_otel_span=parent_otel_span, - request_headers=_safe_get_request_headers(request), - ) + with tracer.trace("litellm.proxy.auth.jwt_auth_builder"): + result = await JWTAuthManager.auth_builder( + request_data=request_data, + general_settings=general_settings, + api_key=api_key, + jwt_handler=jwt_handler, + route=route, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + proxy_logging_obj=proxy_logging_obj, + parent_otel_span=parent_otel_span, + request_headers=_safe_get_request_headers(request), + ) is_proxy_admin = result["is_proxy_admin"] team_id = result["team_id"] @@ -909,15 +910,15 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 try: end_user_params["end_user_id"] = end_user_id - # get end-user object - _end_user_object = await get_end_user_object( - end_user_id=end_user_id, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - parent_otel_span=parent_otel_span, - proxy_logging_obj=proxy_logging_obj, - route=route, - ) + with tracer.trace("litellm.proxy.auth.get_end_user_object"): + _end_user_object = await get_end_user_object( + end_user_id=end_user_id, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + parent_otel_span=parent_otel_span, + proxy_logging_obj=proxy_logging_obj, + route=route, + ) if _end_user_object is not None: end_user_params[ "allowed_model_region" @@ -960,14 +961,15 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 if valid_token is None: ## Check CACHE try: - valid_token = await get_key_object( - hashed_token=hash_token(api_key), - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - parent_otel_span=parent_otel_span, - proxy_logging_obj=proxy_logging_obj, - check_cache_only=True, - ) + with tracer.trace("litellm.proxy.auth.get_key_object_check_cache"): + valid_token = await get_key_object( + hashed_token=hash_token(api_key), + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + parent_otel_span=parent_otel_span, + proxy_logging_obj=proxy_logging_obj, + check_cache_only=True, + ) except Exception: verbose_logger.debug("api key not found in cache.") valid_token = None @@ -1139,13 +1141,14 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 api_key = hash_token(token=api_key) try: - valid_token = await get_key_object( - hashed_token=api_key, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - parent_otel_span=parent_otel_span, - proxy_logging_obj=proxy_logging_obj, - ) + with tracer.trace("litellm.proxy.auth.get_key_object_from_db"): + valid_token = await get_key_object( + hashed_token=api_key, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + parent_otel_span=parent_otel_span, + proxy_logging_obj=proxy_logging_obj, + ) except ProxyException as e: if e.code == 401 or e.code == "401": e.message = "Authentication Error, Invalid proxy server token passed. Received API Key = {}, Key Hash (Token) ={}. Unable to find token in cache or `LiteLLM_VerificationTokenTable`".format( @@ -1233,14 +1236,15 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 # Check 2. If user_id for this token is in budget - done in common_checks() if valid_token.user_id is not None: try: - user_obj = await get_user_object( - user_id=valid_token.user_id, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - user_id_upsert=False, - parent_otel_span=parent_otel_span, - proxy_logging_obj=proxy_logging_obj, - ) + with tracer.trace("litellm.proxy.auth.get_user_object"): + user_obj = await get_user_object( + user_id=valid_token.user_id, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + user_id_upsert=False, + parent_otel_span=parent_otel_span, + proxy_logging_obj=proxy_logging_obj, + ) except Exception as e: verbose_logger.debug( "litellm.proxy.auth.user_api_key_auth.py::user_api_key_auth() - Unable to get user from db/cache. Setting user_obj to None. Exception received - {}".format( @@ -1329,71 +1333,73 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 ) if not skip_budget_checks: - # Check 4. Token Spend is under budget - if RouteChecks.is_llm_api_route(route=route): - await _virtual_key_max_budget_check( + with tracer.trace("litellm.proxy.auth.budget_checks"): + # Check 4. Token Spend is under budget + if RouteChecks.is_llm_api_route(route=route): + await _virtual_key_max_budget_check( + valid_token=valid_token, + proxy_logging_obj=proxy_logging_obj, + user_obj=user_obj, + ) + + # Check 5. Max Budget Alert Check + await _virtual_key_max_budget_alert_check( valid_token=valid_token, proxy_logging_obj=proxy_logging_obj, user_obj=user_obj, ) - # Check 5. Max Budget Alert Check - await _virtual_key_max_budget_alert_check( - valid_token=valid_token, - proxy_logging_obj=proxy_logging_obj, - user_obj=user_obj, - ) - - # Check 6. Soft Budget Check - await _virtual_key_soft_budget_check( - valid_token=valid_token, - proxy_logging_obj=proxy_logging_obj, - user_obj=user_obj, - ) - - # Check 5. Token Model Spend is under Model budget - max_budget_per_model = valid_token.model_max_budget - current_model = request_data.get("model", None) - - if ( - max_budget_per_model is not None - and isinstance(max_budget_per_model, dict) - and len(max_budget_per_model) > 0 - and prisma_client is not None - and current_model is not None - and valid_token.token is not None - ): - ## GET THE SPEND FOR THIS MODEL - await model_max_budget_limiter.is_key_within_model_budget( - user_api_key_dict=valid_token, - model=current_model, + # Check 6. Soft Budget Check + await _virtual_key_soft_budget_check( + valid_token=valid_token, + proxy_logging_obj=proxy_logging_obj, + user_obj=user_obj, ) - # Check 5b. End-user model max budget - end_user_mmb = valid_token.end_user_model_max_budget - if ( - end_user_mmb is not None - and isinstance(end_user_mmb, dict) - and len(end_user_mmb) > 0 - and current_model is not None - and valid_token.end_user_id is not None - ): - await model_max_budget_limiter.is_end_user_within_model_budget( - end_user_id=valid_token.end_user_id, - end_user_model_max_budget=end_user_mmb, - model=current_model, - ) + # Check 5. Token Model Spend is under Model budget + max_budget_per_model = valid_token.model_max_budget + current_model = request_data.get("model", None) + + if ( + max_budget_per_model is not None + and isinstance(max_budget_per_model, dict) + and len(max_budget_per_model) > 0 + and prisma_client is not None + and current_model is not None + and valid_token.token is not None + ): + ## GET THE SPEND FOR THIS MODEL + await model_max_budget_limiter.is_key_within_model_budget( + user_api_key_dict=valid_token, + model=current_model, + ) + + # Check 5b. End-user model max budget + end_user_mmb = valid_token.end_user_model_max_budget + if ( + end_user_mmb is not None + and isinstance(end_user_mmb, dict) + and len(end_user_mmb) > 0 + and current_model is not None + and valid_token.end_user_id is not None + ): + await model_max_budget_limiter.is_end_user_within_model_budget( + end_user_id=valid_token.end_user_id, + end_user_model_max_budget=end_user_mmb, + model=current_model, + ) # Check 6: Additional Common Checks across jwt + key auth if valid_token.team_id is not None: try: - _team_obj = await get_team_object( - team_id=valid_token.team_id, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - parent_otel_span=parent_otel_span, - proxy_logging_obj=proxy_logging_obj, - ) + with tracer.trace("litellm.proxy.auth.get_team_object"): + _team_obj = await get_team_object( + team_id=valid_token.team_id, + prisma_client=prisma_client, + user_api_key_cache=user_api_key_cache, + parent_otel_span=parent_otel_span, + proxy_logging_obj=proxy_logging_obj, + ) except HTTPException: _team_obj = LiteLLM_TeamTableCachedObj( team_id=valid_token.team_id, @@ -1431,11 +1437,14 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 litellm.max_budget > 0 and prisma_client is not None ): # user set proxy max budget cache_key = "{}:spend".format(litellm_proxy_admin_name) - global_proxy_spend = await _fetch_global_spend_with_event_coordination( - cache_key=cache_key, - user_api_key_cache=user_api_key_cache, - prisma_client=prisma_client, - ) + with tracer.trace("litellm.proxy.auth.get_global_proxy_spend"): + global_proxy_spend = ( + await _fetch_global_spend_with_event_coordination( + cache_key=cache_key, + user_api_key_cache=user_api_key_cache, + prisma_client=prisma_client, + ) + ) if global_proxy_spend is not None: call_info = CallInfo( @@ -1452,21 +1461,22 @@ async def _user_api_key_auth_builder( # noqa: PLR0915 user_info=call_info, ) ) - _ = await common_checks( - request=request, - request_body=request_data, - team_object=_team_obj, - user_object=user_obj, - end_user_object=_end_user_object, - general_settings=general_settings, - global_proxy_spend=global_proxy_spend, - route=route, - llm_router=llm_router, - proxy_logging_obj=proxy_logging_obj, - valid_token=valid_token, - skip_budget_checks=skip_budget_checks, - project_object=_project_obj, - ) + with tracer.trace("litellm.proxy.auth.common_checks"): + _ = await common_checks( + request=request, + request_body=request_data, + team_object=_team_obj, + user_object=user_obj, + end_user_object=_end_user_object, + general_settings=general_settings, + global_proxy_spend=global_proxy_spend, + route=route, + llm_router=llm_router, + proxy_logging_obj=proxy_logging_obj, + valid_token=valid_token, + skip_budget_checks=skip_budget_checks, + project_object=_project_obj, + ) # Token passed all checks if valid_token is None: raise HTTPException(401, detail="Invalid API key") diff --git a/litellm/proxy/common_request_processing.py b/litellm/proxy/common_request_processing.py index 72765aab7da..e5a31c36719 100644 --- a/litellm/proxy/common_request_processing.py +++ b/litellm/proxy/common_request_processing.py @@ -1260,7 +1260,9 @@ class ProxyBaseLLMRequestProcessing: custom_headers = ProxyBaseLLMRequestProcessing.get_custom_headers( user_api_key_dict=user_api_key_dict, call_id=( - _litellm_logging_obj.litellm_call_id if _litellm_logging_obj else None + _litellm_logging_obj.litellm_call_id + if _litellm_logging_obj + else self.data.get("litellm_call_id") ), model_id=model_id, version=version, diff --git a/litellm/proxy/discovery_endpoints/ui_discovery_endpoints.py b/litellm/proxy/discovery_endpoints/ui_discovery_endpoints.py index 2a38ceffba5..233df5c6c57 100644 --- a/litellm/proxy/discovery_endpoints/ui_discovery_endpoints.py +++ b/litellm/proxy/discovery_endpoints/ui_discovery_endpoints.py @@ -27,10 +27,17 @@ async def get_ui_config(): admin_ui_disabled = os.getenv("DISABLE_ADMIN_UI", "false").lower() == "true" sso_configured = _has_user_setup_sso() + + from litellm.proxy.proxy_server import proxy_config + + is_control_plane = len(proxy_config.worker_registry) > 0 + return UiDiscoveryEndpoints( server_root_path=get_server_root_path(), proxy_base_url=get_proxy_base_url(), auto_redirect_to_sso=sso_configured and auto_redirect_ui_login_to_sso, admin_ui_disabled=admin_ui_disabled, sso_configured=sso_configured, + is_control_plane=is_control_plane, + workers=proxy_config.worker_registry if is_control_plane else [], ) diff --git a/litellm/proxy/management_endpoints/key_management_endpoints.py b/litellm/proxy/management_endpoints/key_management_endpoints.py index 1016e6b149c..831922ec3f9 100644 --- a/litellm/proxy/management_endpoints/key_management_endpoints.py +++ b/litellm/proxy/management_endpoints/key_management_endpoints.py @@ -41,10 +41,8 @@ from litellm.proxy._experimental.mcp_server.db import ( from litellm.proxy._types import * from litellm.proxy._types import LiteLLM_VerificationToken from litellm.proxy.auth.auth_checks import ( - _cache_key_object, _delete_cache_key_object, can_team_access_model, - get_key_object, get_org_object, get_project_object, get_team_object, @@ -1656,7 +1654,7 @@ async def _get_and_validate_existing_key( LiteLLM_VerificationToken: The existing key row Raises: - HTTPException: If key is not found + ProxyException: 404 if key is not found """ if prisma_client is None: raise HTTPException( @@ -1664,16 +1662,18 @@ async def _get_and_validate_existing_key( detail={"error": "Database not connected"}, ) - existing_key_row = await prisma_client.get_data( - token=token, - table_name="key", - query_type="find_unique", + hashed_token = _hash_token_if_needed(token=token) + + existing_key_row = await prisma_client.db.litellm_verificationtoken.find_unique( + where={"token": hashed_token} ) if existing_key_row is None: - raise HTTPException( - status_code=404, - detail={"error": f"Key not found: {token}"}, + raise ProxyException( + message="Key not found.", + type=ProxyErrorTypes.not_found_error, + param="key", + code=status.HTTP_404_NOT_FOUND, ) return existing_key_row @@ -2111,19 +2111,11 @@ async def update_key_fn( key = data_json.pop("key") # get the row from db - if prisma_client is None: - raise Exception("Not connected to DB!") - - existing_key_row = await prisma_client.get_data( - token=data.key, table_name="key", query_type="find_unique" + existing_key_row = await _get_and_validate_existing_key( + token=data.key, + prisma_client=prisma_client, ) - if existing_key_row is None: - raise HTTPException( - status_code=404, - detail={"error": f"Team not found, passed team_id={data.team_id}"}, - ) - await _validate_update_key_data( data=data, existing_key_row=existing_key_row, @@ -2158,6 +2150,8 @@ async def update_key_fn( ) _data = {**non_default_values, "token": key} + if prisma_client is None: + raise Exception("Not connected to DB!") response = await prisma_client.update_data(token=key, data=_data) # Delete - key from cache, since it's been updated! @@ -2330,6 +2324,8 @@ async def bulk_update_keys( error_message = error_detail.get("error", str(e)) else: error_message = str(error_detail) + elif isinstance(e, ProxyException): + error_message = e.message else: error_message = str(e) @@ -4945,18 +4941,19 @@ async def block_key( route="/key/block", ) - if litellm.store_audit_logs is True: - # make an audit log for key update - record = await prisma_client.db.litellm_verificationtoken.find_unique( - where={"token": hashed_token} + # Check if the key exists before trying to block it + existing_record = await prisma_client.db.litellm_verificationtoken.find_unique( + where={"token": hashed_token} + ) + if existing_record is None: + raise ProxyException( + message="Key not found.", + type=ProxyErrorTypes.not_found_error, + param="key", + code=status.HTTP_404_NOT_FOUND, ) - if record is None: - raise ProxyException( - message=f"Key {data.key} not found", - type=ProxyErrorTypes.bad_request_error, - param="key", - code=status.HTTP_404_NOT_FOUND, - ) + + if litellm.store_audit_logs is True: asyncio.create_task( create_audit_log_for_update( request_data=LiteLLM_AuditLogs( @@ -4970,7 +4967,7 @@ async def block_key( object_id=hashed_token, action="blocked", updated_values="{}", - before_value=record.model_dump_json(), + before_value=existing_record.model_dump_json(), ) ) ) @@ -4979,24 +4976,9 @@ async def block_key( where={"token": hashed_token}, data={"blocked": True} # type: ignore ) - ## UPDATE KEY CACHE - - ### get cached object ### - key_object = await get_key_object( + ## UPDATE KEY CACHE - invalidate so next read re-fetches from DB + await _delete_cache_key_object( hashed_token=hashed_token, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - parent_otel_span=None, - proxy_logging_obj=proxy_logging_obj, - ) - - ### update cached object ### - key_object.blocked = True - - ### store cached object ### - await _cache_key_object( - hashed_token=hashed_token, - user_api_key_obj=key_object, user_api_key_cache=user_api_key_cache, proxy_logging_obj=proxy_logging_obj, ) @@ -5068,18 +5050,19 @@ async def unblock_key( route="/key/unblock", ) - if litellm.store_audit_logs is True: - # make an audit log for key update - record = await prisma_client.db.litellm_verificationtoken.find_unique( - where={"token": hashed_token} + # Check if the key exists before trying to unblock it + existing_record = await prisma_client.db.litellm_verificationtoken.find_unique( + where={"token": hashed_token} + ) + if existing_record is None: + raise ProxyException( + message="Key not found.", + type=ProxyErrorTypes.not_found_error, + param="key", + code=status.HTTP_404_NOT_FOUND, ) - if record is None: - raise ProxyException( - message=f"Key {data.key} not found", - type=ProxyErrorTypes.bad_request_error, - param="key", - code=status.HTTP_404_NOT_FOUND, - ) + + if litellm.store_audit_logs is True: asyncio.create_task( create_audit_log_for_update( request_data=LiteLLM_AuditLogs( @@ -5091,9 +5074,9 @@ async def unblock_key( changed_by_api_key=user_api_key_dict.api_key, table_name=LitellmTableNames.KEY_TABLE_NAME, object_id=hashed_token, - action="blocked", + action="unblocked", updated_values="{}", - before_value=record.model_dump_json(), + before_value=existing_record.model_dump_json(), ) ) ) @@ -5102,24 +5085,9 @@ async def unblock_key( where={"token": hashed_token}, data={"blocked": False} # type: ignore ) - ## UPDATE KEY CACHE - - ### get cached object ### - key_object = await get_key_object( + ## UPDATE KEY CACHE - invalidate so next read re-fetches from DB + await _delete_cache_key_object( hashed_token=hashed_token, - prisma_client=prisma_client, - user_api_key_cache=user_api_key_cache, - parent_otel_span=None, - proxy_logging_obj=proxy_logging_obj, - ) - - ### update cached object ### - key_object.blocked = False - - ### store cached object ### - await _cache_key_object( - hashed_token=hashed_token, - user_api_key_obj=key_object, user_api_key_cache=user_api_key_cache, proxy_logging_obj=proxy_logging_obj, ) diff --git a/litellm/proxy/management_endpoints/mcp_management_endpoints.py b/litellm/proxy/management_endpoints/mcp_management_endpoints.py index 3e5b729cea6..e4bb288cda9 100644 --- a/litellm/proxy/management_endpoints/mcp_management_endpoints.py +++ b/litellm/proxy/management_endpoints/mcp_management_endpoints.py @@ -1214,17 +1214,9 @@ if MCP_AVAILABLE: "error": "User does not have permission to create mcp servers. You can only create mcp servers if you are a PROXY_ADMIN." }, ) - elif payload.server_id is not None: - # fail if the mcp server with id already exists - mcp_server = await get_mcp_server(prisma_client, payload.server_id) - if mcp_server is not None: - raise HTTPException( - status_code=status.HTTP_400_BAD_REQUEST, - detail={ - "error": f"MCP Server with id {payload.server_id} already exists. Cannot create another." - }, - ) - elif ( + + # Block reserved special server IDs + if ( SpecialMCPServerName.all_team_servers == payload.server_id or SpecialMCPServerName.all_proxy_servers == payload.server_id ): @@ -1235,6 +1227,17 @@ if MCP_AVAILABLE: }, ) + if payload.server_id is not None: + # fail if the mcp server with id already exists + mcp_server = await get_mcp_server(prisma_client, payload.server_id) + if mcp_server is not None: + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail={ + "error": f"MCP Server with id {payload.server_id} already exists. Cannot create another." + }, + ) + # TODO: audit log for create # Admin-created servers are always active — clear any submission lifecycle @@ -1399,6 +1402,8 @@ if MCP_AVAILABLE: client_id: Optional[str] = Form(None), client_secret: Optional[str] = Form(None), code_verifier: Optional[str] = Form(None), + refresh_token: Optional[str] = Form(None), + scope: Optional[str] = Form(None), ): mcp_server = _get_cached_temporary_mcp_server_or_404(server_id) resolved_client_id = mcp_server.client_id or client_id or "" @@ -1422,6 +1427,8 @@ if MCP_AVAILABLE: client_id=resolved_client_id, client_secret=client_secret, code_verifier=code_verifier, + refresh_token=refresh_token, + scope=scope, ) @router.post( diff --git a/litellm/proxy/management_endpoints/team_endpoints.py b/litellm/proxy/management_endpoints/team_endpoints.py index b033d3fce0d..3643373be65 100644 --- a/litellm/proxy/management_endpoints/team_endpoints.py +++ b/litellm/proxy/management_endpoints/team_endpoints.py @@ -724,6 +724,7 @@ async def new_team( # noqa: PLR0915 - disable_global_guardrails: Optional[bool] - Whether to disable global guardrails for the key. - object_permission: Optional[LiteLLM_ObjectPermissionBase] - team-specific object permission. Example - {"vector_stores": ["vector_store_1", "vector_store_2"], "agents": ["agent_1", "agent_2"], "agent_access_groups": ["dev_group"]}. IF null or {} then no object permission. - team_member_budget: Optional[float] - The maximum budget allocated to an individual team member. + - team_member_budget_duration: Optional[str] - The duration of the budget for the team member. Doc [here](https://docs.litellm.ai/docs/proxy/team_budgets) - team_member_rpm_limit: Optional[int] - The RPM (Requests Per Minute) limit for individual team members. - team_member_tpm_limit: Optional[int] - The TPM (Tokens Per Minute) limit for individual team members. - team_member_key_duration: Optional[str] - The duration for a team member's key. e.g. "1d", "1w", "1mo" @@ -934,6 +935,7 @@ async def new_team( # noqa: PLR0915 team_member_budget=data.team_member_budget, team_member_rpm_limit=data.team_member_rpm_limit, team_member_tpm_limit=data.team_member_tpm_limit, + team_member_budget_duration=data.team_member_budget_duration, ): data_json = await TeamMemberBudgetHandler.create_team_member_budget_table( data=data, @@ -942,6 +944,7 @@ async def new_team( # noqa: PLR0915 team_member_budget=data.team_member_budget, team_member_rpm_limit=data.team_member_rpm_limit, team_member_tpm_limit=data.team_member_tpm_limit, + team_member_budget_duration=data.team_member_budget_duration, ) ## ADD TO TEAM TABLE @@ -3334,7 +3337,9 @@ def _convert_teams_to_response_models( use_deleted_table: bool, ) -> List[Union[TeamListItem, LiteLLM_TeamTable, LiteLLM_DeletedTeamTable]]: """Convert raw Prisma team rows to response models.""" - team_list: List[Union[TeamListItem, LiteLLM_TeamTable, LiteLLM_DeletedTeamTable]] = [] + team_list: List[ + Union[TeamListItem, LiteLLM_TeamTable, LiteLLM_DeletedTeamTable] + ] = [] for team in teams: try: team_dict = team.model_dump() diff --git a/litellm/proxy/management_endpoints/ui_sso.py b/litellm/proxy/management_endpoints/ui_sso.py index daf1d6f1316..d06ce56f816 100644 --- a/litellm/proxy/management_endpoints/ui_sso.py +++ b/litellm/proxy/management_endpoints/ui_sso.py @@ -16,6 +16,7 @@ import os import secrets from copy import deepcopy from typing import TYPE_CHECKING, Any, Dict, List, Literal, Optional, Tuple, Union, cast +from urllib.parse import urlencode, urlparse if TYPE_CHECKING: import httpx @@ -301,6 +302,7 @@ async def google_login( source: Optional[str] = None, key: Optional[str] = None, existing_key: Optional[str] = None, + return_to: Optional[str] = None, ): # noqa: PLR0915 """ Create Proxy API Keys using Google Workspace SSO. Requires setting PROXY_BASE_URL in .env @@ -394,13 +396,23 @@ async def google_login( is True ): verbose_proxy_logger.info(f"Redirecting to SSO login for {redirect_url}") - return await SSOAuthenticationHandler.get_sso_login_redirect( + sso_redirect = await SSOAuthenticationHandler.get_sso_login_redirect( redirect_url=redirect_url, microsoft_client_id=microsoft_client_id, google_client_id=google_client_id, generic_client_id=generic_client_id, state=cli_state, ) + if return_to is not None and sso_redirect is not None: + SSOAuthenticationHandler._validate_return_to(return_to) + sso_redirect.set_cookie( + key="litellm_cp_return_to", + value=return_to, + max_age=600, + httponly=True, + samesite="lax", + ) + return sso_redirect elif ui_username is not None: # No Google, Microsoft SSO # Use UI Credentials set in .env @@ -1312,12 +1324,17 @@ async def auth_callback(request: Request, state: Optional[str] = None): # noqa: request=request, key=key_id, existing_key=existing_key, result=result ) + # Control-plane cross-origin: read return_to from cookie. + # Starlette's cookie_parser already handles RFC 2109 unquoting. + cp_return_to: Optional[str] = request.cookies.get("litellm_cp_return_to") + return await SSOAuthenticationHandler.get_redirect_response_from_openid( result=result, request=request, received_response=received_response, generic_client_id=generic_client_id, ui_access_mode=ui_access_mode, + return_to=cp_return_to, ) @@ -1760,6 +1777,38 @@ class SSOAuthenticationHandler: Handler for SSO Authentication across all SSO providers """ + @staticmethod + def _validate_return_to(return_to: str) -> None: + """ + Validate that return_to matches the configured control_plane_url origin. + + Raises HTTPException(400) if: + - control_plane_url is not configured in general_settings + - return_to origin does not match control_plane_url origin + """ + from litellm.proxy.proxy_server import general_settings + + control_plane_url = general_settings.get("control_plane_url") + if control_plane_url is None: + raise HTTPException( + status_code=400, + detail="return_to is not allowed: control_plane_url is not configured", + ) + + def _origin(url: str) -> tuple: + parsed = urlparse(url) + scheme = (parsed.scheme or "").lower() + hostname = (parsed.hostname or "").lower() + default_port = 443 if scheme == "https" else 80 + port = parsed.port if parsed.port is not None else default_port + return (scheme, hostname, port) + + if _origin(return_to) != _origin(control_plane_url): + raise HTTPException( + status_code=400, + detail="return_to does not match the configured control_plane_url", + ) + @staticmethod async def get_sso_login_redirect( redirect_url: str, @@ -2358,6 +2407,7 @@ class SSOAuthenticationHandler: received_response: Optional[dict] = None, generic_client_id: Optional[str] = None, ui_access_mode: Optional[Dict] = None, + return_to: Optional[str] = None, ) -> RedirectResponse: import jwt @@ -2367,6 +2417,7 @@ class SSOAuthenticationHandler: master_key, premium_user, proxy_logging_obj, + redis_usage_cache, user_api_key_cache, user_custom_sso, ) @@ -2534,6 +2585,36 @@ class SSOAuthenticationHandler: master_key or "", algorithm="HS256", ) + + # Control-plane cross-origin: store JWT behind a single-use opaque + # code (60s TTL) so the token never appears in browser history / logs. + # The control plane redeems it via POST /v3/login/exchange. + if return_to is not None: + SSOAuthenticationHandler._validate_return_to(return_to) + + code = secrets.token_urlsafe(32) + cache_key = f"login_code:{code}" + cache_value = {"token": jwt_token, "redirect_url": return_to} + if redis_usage_cache is not None: + await redis_usage_cache.async_set_cache( + key=cache_key, value=cache_value, ttl=60 + ) + else: + await user_api_key_cache.async_set_cache( + key=cache_key, value=cache_value, ttl=60 + ) + + separator = "&" if "?" in return_to else "?" + redirect_url = ( + return_to + separator + urlencode({"login": "success", "code": code}) + ) + verbose_proxy_logger.info( + "Cross-origin SSO: redirecting to control plane with login code" + ) + redirect_response = RedirectResponse(url=redirect_url, status_code=303) + redirect_response.delete_cookie("litellm_cp_return_to") + return redirect_response + if user_id is not None and isinstance(user_id, str): litellm_dashboard_ui += "?login=success" verbose_proxy_logger.info(f"Redirecting to {litellm_dashboard_ui}") diff --git a/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py b/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py index 4e3e04a8474..534022cc133 100644 --- a/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py +++ b/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py @@ -22,6 +22,7 @@ from litellm.constants import ( ALLOWED_VERTEX_AI_PASSTHROUGH_HEADERS, BEDROCK_AGENT_RUNTIME_PASS_THROUGH_ROUTES, ) +from litellm.llms.anthropic.common_utils import AnthropicModelInfo from litellm.llms.vertex_ai.vertex_llm_base import VertexBase from litellm.proxy._types import * from litellm.proxy.auth.route_checks import RouteChecks @@ -585,7 +586,11 @@ async def anthropic_proxy_route( """ [Docs](https://docs.litellm.ai/docs/pass_through/anthropic_completion) """ - base_target_url = os.getenv("ANTHROPIC_API_BASE") or "https://api.anthropic.com" + base_target_url = ( + os.getenv("ANTHROPIC_API_BASE") + or os.getenv("ANTHROPIC_BASE_URL") + or "https://api.anthropic.com" + ) encoded_endpoint = httpx.URL(endpoint).path # Ensure endpoint starts with '/' for proper URL construction @@ -606,10 +611,11 @@ async def anthropic_proxy_route( is_streaming_request = await is_streaming_request_fn(request) ## CREATE PASS-THROUGH + auth_header = AnthropicModelInfo.get_auth_header(anthropic_api_key or None) endpoint_func = create_pass_through_route( endpoint=endpoint, target=str(updated_url), - custom_headers={"x-api-key": "{}".format(anthropic_api_key)}, + custom_headers=auth_header if auth_header is not None else {}, _forward_headers=True, is_streaming_request=is_streaming_request, ) # dynamically construct pass-through endpoint based on incoming path diff --git a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py index 20d06b7d531..fcb1e0b2e49 100644 --- a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py +++ b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py @@ -7,6 +7,7 @@ import httpx import litellm from litellm._logging import verbose_proxy_logger from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj +from litellm.litellm_core_utils.litellm_logging import use_custom_pricing_for_model from litellm.llms.anthropic import get_anthropic_config from litellm.llms.anthropic.chat.handler import ( ModelResponseIterator as AnthropicModelResponseIterator, @@ -124,10 +125,21 @@ class AnthropicPassthroughLoggingHandler: if custom_llm_provider and not model.startswith(f"{custom_llm_provider}/"): model_for_cost = f"{custom_llm_provider}/{model}" + router_model_id = logging_obj.get_router_model_id() + custom_pricing = use_custom_pricing_for_model( + litellm_params=( + logging_obj.litellm_params + if hasattr(logging_obj, "litellm_params") + else None + ) + ) + response_cost = litellm.completion_cost( completion_response=litellm_model_response, model=model_for_cost, custom_llm_provider=custom_llm_provider, + custom_pricing=custom_pricing, + router_model_id=router_model_id, ) kwargs["response_cost"] = response_cost @@ -319,9 +331,7 @@ class AnthropicPassthroughLoggingHandler: import base64 from litellm._uuid import uuid - from litellm.llms.anthropic.batches.transformation import ( - AnthropicBatchesConfig, - ) + from litellm.llms.anthropic.batches.transformation import AnthropicBatchesConfig from litellm.types.utils import Choices, SpecialEnums try: diff --git a/litellm/proxy/proxy_cli.py b/litellm/proxy/proxy_cli.py index 97d5de0d53d..c638e294268 100644 --- a/litellm/proxy/proxy_cli.py +++ b/litellm/proxy/proxy_cli.py @@ -12,6 +12,7 @@ import click import httpx from dotenv import load_dotenv +import litellm from litellm.constants import DEFAULT_NUM_WORKERS_LITELLM_PROXY from litellm.secret_managers.main import get_secret_bool @@ -387,7 +388,7 @@ class ProxyInitializationHelpers: @click.option("--api_base", default=None, help="API base URL.") @click.option( "--api_version", - default="2024-07-01-preview", + default=litellm.AZURE_DEFAULT_API_VERSION, help="For azure - pass in the api version.", ) @click.option( diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 9c29927c5cb..e982c934aa6 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -541,6 +541,7 @@ from litellm.types.llms.anthropic import ( AnthropicResponseUsageBlock, ) from litellm.types.llms.openai import HttpxBinaryResponseContent +from litellm.types.proxy.control_plane_endpoints import WorkerRegistryEntry from litellm.types.proxy.management_endpoints.model_management_endpoints import ( ModelGroupInfoProxy, ) @@ -1546,6 +1547,7 @@ user_custom_key_generate = None # Sentinel: prevents PKCE-no-Redis advisory from re-logging on config hot-reload. # Tests that need to reset it can patch 'litellm.proxy.proxy_server._pkce_no_redis_warning_emitted'. _pkce_no_redis_warning_emitted: bool = False +_cp_no_redis_warning_emitted: bool = False user_custom_sso = None user_custom_ui_sso_sign_in_handler = None use_background_health_checks = None @@ -2295,6 +2297,7 @@ class ProxyConfig: self.config: Dict[str, Any] = {} self._last_semantic_filter_config: Optional[Dict[str, Any]] = None self._last_hashicorp_vault_config: Optional[Dict[str, Any]] = None + self.worker_registry: List["WorkerRegistryEntry"] = [] def is_yaml(self, config_file_path: str) -> bool: if not os.path.isfile(config_file_path): @@ -3095,6 +3098,21 @@ class ProxyConfig: "Set PKCE_STRICT_CACHE_MISS=true to fail fast with a 401 on cache misses " "instead of continuing without a code_verifier." ) + + ### CONTROL PLANE CODE-EXCHANGE PREREQUISITE CHECK ### + cp_url = general_settings.get("control_plane_url") + if cp_url and redis_usage_cache is None: + global _cp_no_redis_warning_emitted + if not _cp_no_redis_warning_emitted: + _cp_no_redis_warning_emitted = True + verbose_proxy_logger.warning( + "control_plane_url is configured but Redis is not configured for LiteLLM caching. " + "Login codes (SSO and /v3/login) will not be shared across instances — " + "the /v3/login/exchange call may land on a different pod and fail with 401. " + "Configure Redis via the 'cache' section in your proxy config, " + "or ensure sticky sessions for single-instance deployments." + ) + ### STORE MODEL IN DB ### feature flag for `/model/new` store_model_in_db = general_settings.get("store_model_in_db", False) if store_model_in_db is None: @@ -3385,7 +3403,15 @@ class ProxyConfig: litellm.vector_store_registry.load_vector_stores_from_config( vector_store_registry_config ) - pass + + ## WORKER REGISTRY (Control Plane) + worker_registry_config = config.get("worker_registry", None) + if worker_registry_config: + self.worker_registry = [ + WorkerRegistryEntry(**e) for e in worker_registry_config + ] + else: + self.worker_registry = [] async def _init_policy_engine( self, @@ -11095,6 +11121,165 @@ async def login_v2(request: Request): # noqa: PLR0915 ) +@router.post( + "/v3/login", include_in_schema=False +) # control-plane login — always returns token in body for cross-origin use +async def login_v3(request: Request): # noqa: PLR0915 + global premium_user, general_settings, master_key + from litellm.proxy.auth.login_utils import authenticate_user, create_ui_token_object + from litellm.proxy.utils import get_custom_url + + try: + if not general_settings.get("control_plane_url"): + raise ProxyException( + message="/v3/login is only available on workers with control_plane_url configured", + type=ProxyErrorTypes.not_found_error, + param="control_plane_url", + code=status.HTTP_404_NOT_FOUND, + ) + + body = await request.json() + username = str(body.get("username")) + password = str(body.get("password")) + + login_result = await authenticate_user( + username=username, + password=password, + master_key=master_key, + prisma_client=prisma_client, + ) + + returned_ui_token_object = create_ui_token_object( + login_result=login_result, + general_settings=general_settings, + premium_user=premium_user, + ) + + import jwt + + jwt_token = jwt.encode( + cast(dict, returned_ui_token_object), + cast(str, master_key), + algorithm="HS256", + ) + + litellm_dashboard_ui = get_custom_url(str(request.base_url)) + if litellm_dashboard_ui.endswith("/"): + litellm_dashboard_ui += "ui/" + else: + litellm_dashboard_ui += "/ui/" + litellm_dashboard_ui += "?login=success" + + # Store JWT behind a single-use opaque code (60s TTL) + code = secrets.token_urlsafe(32) + cache_key = f"login_code:{code}" + cache_value = {"token": jwt_token, "redirect_url": litellm_dashboard_ui} + if redis_usage_cache is not None: + await redis_usage_cache.async_set_cache( + key=cache_key, value=cache_value, ttl=60 + ) + else: + await user_api_key_cache.async_set_cache( + key=cache_key, value=cache_value, ttl=60 + ) + + return JSONResponse( + content={"code": code, "expires_in": 60}, + status_code=status.HTTP_200_OK, + ) + except Exception as e: + verbose_proxy_logger.exception( + "litellm.proxy.proxy_server.login_v3(): Exception occurred - {}".format( + str(e) + ) + ) + if isinstance(e, ProxyException): + raise e + elif isinstance(e, HTTPException): + raise ProxyException( + message=getattr(e, "detail", str(e)), + type=ProxyErrorTypes.auth_error, + param=getattr(e, "param", "None"), + code=getattr(e, "status_code", status.HTTP_500_INTERNAL_SERVER_ERROR), + ) + else: + error_msg = f"{str(e)}" + raise ProxyException( + message=error_msg, + type=ProxyErrorTypes.auth_error, + param="None", + code=status.HTTP_500_INTERNAL_SERVER_ERROR, + ) + + +@router.post( + "/v3/login/exchange", include_in_schema=False +) # exchange single-use opaque code for JWT +async def login_v3_exchange(request: Request): + try: + if not general_settings.get("control_plane_url"): + raise ProxyException( + message="/v3/login/exchange is only available on workers with control_plane_url configured", + type=ProxyErrorTypes.not_found_error, + param="control_plane_url", + code=status.HTTP_404_NOT_FOUND, + ) + + body = await request.json() + code = body.get("code") + if not code: + raise ProxyException( + message="Missing 'code' parameter", + type=ProxyErrorTypes.auth_error, + param="code", + code=status.HTTP_400_BAD_REQUEST, + ) + + cache_key = f"login_code:{code}" + if redis_usage_cache is not None: + cached_data = await redis_usage_cache.async_get_cache(key=cache_key) + else: + cached_data = await user_api_key_cache.async_get_cache(key=cache_key) + + if not cached_data or not isinstance(cached_data, dict): + raise ProxyException( + message="Invalid or expired login code", + type=ProxyErrorTypes.auth_error, + param="code", + code=status.HTTP_401_UNAUTHORIZED, + ) + + # Single-use: delete immediately + if redis_usage_cache is not None: + await redis_usage_cache.async_delete_cache(key=cache_key) + else: + await user_api_key_cache.async_delete_cache(key=cache_key) + + json_response = JSONResponse( + content={ + "token": cached_data["token"], + "redirect_url": cached_data["redirect_url"], + }, + status_code=status.HTTP_200_OK, + ) + json_response.set_cookie(key="token", value=cached_data["token"]) + return json_response + except ProxyException: + raise + except Exception as e: + verbose_proxy_logger.exception( + "litellm.proxy.proxy_server.login_v3_exchange(): Exception occurred - {}".format( + str(e) + ) + ) + raise ProxyException( + message=str(e), + type=ProxyErrorTypes.auth_error, + param="None", + code=status.HTTP_500_INTERNAL_SERVER_ERROR, + ) + + @app.get("/onboarding/get_token", include_in_schema=False) async def onboarding(invite_link: str, request: Request): """ diff --git a/litellm/proxy/utils.py b/litellm/proxy/utils.py index df527d08af8..351bc23915a 100644 --- a/litellm/proxy/utils.py +++ b/litellm/proxy/utils.py @@ -1878,28 +1878,33 @@ class ProxyLogging: ) input: Union[list, str, dict] = "" + normalized_call_type: Optional[str] = None if "messages" in request_data and isinstance( request_data["messages"], list ): input = request_data["messages"] litellm_logging_obj.model_call_details["messages"] = input if litellm_logging_obj.call_type != CallTypes.pass_through.value: - litellm_logging_obj.call_type = CallTypes.acompletion.value + normalized_call_type = CallTypes.acompletion.value elif "prompt" in request_data and isinstance(request_data["prompt"], str): input = request_data["prompt"] litellm_logging_obj.model_call_details["prompt"] = input if litellm_logging_obj.call_type != CallTypes.pass_through.value: - litellm_logging_obj.call_type = CallTypes.atext_completion.value + normalized_call_type = CallTypes.atext_completion.value elif "input" in request_data and isinstance(request_data["input"], list): input = request_data["input"] litellm_logging_obj.model_call_details["input"] = input if litellm_logging_obj.call_type != CallTypes.pass_through.value: - litellm_logging_obj.call_type = CallTypes.aembedding.value + normalized_call_type = CallTypes.aembedding.value + if normalized_call_type is not None: + litellm_logging_obj.call_type = normalized_call_type + litellm_logging_obj.model_call_details[ + "call_type" + ] = normalized_call_type # Pass-through endpoints are logged via the callback loop's # async_post_call_failure_hook — skip pre_call and failure handlers. if litellm_logging_obj.call_type == CallTypes.pass_through.value: return - litellm_logging_obj.pre_call( input=input, api_key="", diff --git a/litellm/responses/litellm_completion_transformation/streaming_iterator.py b/litellm/responses/litellm_completion_transformation/streaming_iterator.py index ce037850b86..0672b03bcd7 100644 --- a/litellm/responses/litellm_completion_transformation/streaming_iterator.py +++ b/litellm/responses/litellm_completion_transformation/streaming_iterator.py @@ -107,6 +107,7 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): self._reasoning_done_emitted = False self._reasoning_item_id: Optional[str] = None self._accumulated_reasoning_content_parts: List[str] = [] + self._accumulated_provider_specific_fields: Dict[str, Any] = {} def _get_or_assign_tool_output_index(self, call_id: str) -> int: existing = self._tool_output_index_by_call_id.get(call_id) @@ -479,16 +480,36 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): event.__dict__["sequence_number"] = self._sequence_number return event - def create_litellm_model_response( - self, - ) -> Optional[ModelResponse]: - return cast( + def _merge_provider_specific_fields(self, src: dict) -> None: + """Merge provider_specific_fields using last-value-wins for lists. + + List-valued keys (web_search_results, tool_results, + code_interpreter_results, etc.) are emitted cumulatively — each + emission contains the full list so far. Using "last value wins" + matches stream_chunk_builder's semantics and avoids quadratic + growth from repeated extend calls. + """ + for key, val in src.items(): + self._accumulated_provider_specific_fields[key] = val + + def create_litellm_model_response(self) -> Optional[ModelResponse]: + response = cast( Optional[ModelResponse], stream_chunk_builder( chunks=self.collected_chat_completion_chunks, logging_obj=self.litellm_logging_obj, ), ) + if response is not None and self._accumulated_provider_specific_fields: + if ( + not hasattr(response, "_hidden_params") + or response._hidden_params is None + ): + response._hidden_params = {} + response._hidden_params.setdefault("provider_specific_fields", {}).update( + self._accumulated_provider_specific_fields + ) + return response @staticmethod def _snapshot_chunk_for_stream_chunk_builder( @@ -853,6 +874,17 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): if chunk is not None: chunk = cast(ModelResponseStream, chunk) self._ensure_output_item_for_chunk(chunk) + # Accumulate provider_specific_fields from chunk and delta + for src in ( + getattr(chunk, "provider_specific_fields", None), + getattr( + chunk.choices[0].delta if chunk.choices else None, + "provider_specific_fields", + None, + ), + ): + if src and isinstance(src, dict): + self._merge_provider_specific_fields(src) # Proceed to transformation self.collected_chat_completion_chunks.append( self._snapshot_chunk_for_stream_chunk_builder(chunk) @@ -964,6 +996,17 @@ class LiteLLMCompletionStreamingIterator(ResponsesAPIStreamingIterator): try: chunk = self.litellm_custom_stream_wrapper.__next__() self._ensure_output_item_for_chunk(chunk) + # Accumulate provider_specific_fields from chunk and delta + for src in ( + getattr(chunk, "provider_specific_fields", None), + getattr( + chunk.choices[0].delta if chunk.choices else None, + "provider_specific_fields", + None, + ), + ): + if src and isinstance(src, dict): + self._merge_provider_specific_fields(src) # Emit any just-queued output_item event if self._pending_response_events: return self._pending_response_events.pop(0) diff --git a/litellm/responses/litellm_completion_transformation/transformation.py b/litellm/responses/litellm_completion_transformation/transformation.py index 71fa88fb751..b6479a36998 100644 --- a/litellm/responses/litellm_completion_transformation/transformation.py +++ b/litellm/responses/litellm_completion_transformation/transformation.py @@ -42,6 +42,7 @@ from litellm.types.llms.openai import ( from litellm.types.responses.main import ( GenericResponseOutputItem, GenericResponseOutputItemContentAnnotation, + OutputCodeInterpreterCall, OutputFunctionToolCall, OutputImageGenerationCall, OutputText, @@ -1696,6 +1697,7 @@ class LiteLLMCompletionResponsesConfig: ) -> List[ Union[ GenericResponseOutputItem, + OutputCodeInterpreterCall, OutputFunctionToolCall, OutputImageGenerationCall, ResponseFunctionToolCall, @@ -1704,6 +1706,7 @@ class LiteLLMCompletionResponsesConfig: responses_output: List[ Union[ GenericResponseOutputItem, + OutputCodeInterpreterCall, OutputFunctionToolCall, OutputImageGenerationCall, ResponseFunctionToolCall, @@ -1725,8 +1728,63 @@ class LiteLLMCompletionResponsesConfig: chat_completion_response=chat_completion_response ) ) + + # Convert server-side tool results (e.g. Anthropic code execution) + # into code_interpreter_call output items, replacing the corresponding + # function_call items so the output matches OpenAI's native shape. + tool_result_items = ( + LiteLLMCompletionResponsesConfig._extract_tool_result_output_items( + chat_completion_response + ) + ) + if tool_result_items: + result_by_id = {item.id: item for item in tool_result_items} + replaced_ids = set(result_by_id.keys()) + responses_output = [ + ( + result_by_id[getattr(item, "call_id", None)] + if ( + getattr(item, "type", None) == "function_call" + and getattr(item, "call_id", None) in replaced_ids + ) + else item + ) + for item in responses_output + ] + return responses_output + @staticmethod + def _extract_tool_result_output_items( + chat_completion_response: ModelResponse, + ) -> list: + """Extract pre-built code_interpreter_call output items from provider_specific_fields. + + Provider transformers (e.g. Anthropic) convert their native tool results + into OutputCodeInterpreterCall objects and store them in + provider_specific_fields["code_interpreter_results"]. This method + simply retrieves them — no provider-specific parsing here. + """ + output_items: list = [] + for choice in chat_completion_response.choices or []: + message = getattr(choice, "message", None) + if not message: + continue + psf = getattr(message, "provider_specific_fields", None) + if not psf or not isinstance(psf, dict): + continue + results = psf.get("code_interpreter_results") + if results and isinstance(results, list): + for item in results: + # In the streaming path, items are plain dicts after + # model_dump() in stream_chunk_builder. Reconstruct + # Pydantic objects so responses_output has a uniform type. + if isinstance(item, dict): + output_items.append(OutputCodeInterpreterCall(**item)) + else: + output_items.append(item) + return output_items + @staticmethod def _extract_reasoning_output_items( chat_completion_response: ModelResponse, diff --git a/litellm/responses/streaming_iterator.py b/litellm/responses/streaming_iterator.py index 073ee926063..10a74a5b3c6 100644 --- a/litellm/responses/streaming_iterator.py +++ b/litellm/responses/streaming_iterator.py @@ -166,11 +166,12 @@ class BaseResponsesAPIStreamingIterator: ) setattr(item, "encrypted_content", wrapped_content) - # Store the completed response - if ( - openai_responses_api_chunk - and getattr(openai_responses_api_chunk, "type", None) - == ResponsesAPIStreamEvents.RESPONSE_COMPLETED + # Store the completed response (also for incomplete/failed so logging still fires) + _chunk_type = getattr(openai_responses_api_chunk, "type", None) + if openai_responses_api_chunk and _chunk_type in ( + ResponsesAPIStreamEvents.RESPONSE_COMPLETED, + ResponsesAPIStreamEvents.RESPONSE_INCOMPLETE, + ResponsesAPIStreamEvents.RESPONSE_FAILED, ): self.completed_response = openai_responses_api_chunk # Add cost to usage object if include_cost_in_streaming_usage is True @@ -195,10 +196,12 @@ class BaseResponsesAPIStreamingIterator: if cost is not None: setattr(usage_obj, "cost", cost) except Exception: - # If cost calculation fails, continue without cost pass - self._handle_logging_completed_response() + if _chunk_type == ResponsesAPIStreamEvents.RESPONSE_FAILED: + self._handle_logging_failed_response() + else: + self._handle_logging_completed_response() return openai_responses_api_chunk @@ -216,6 +219,32 @@ class BaseResponsesAPIStreamingIterator: """Base implementation - should be overridden by subclasses""" pass + def _handle_logging_failed_response(self): + """ + Handle logging for RESPONSE_FAILED events by routing to failure handlers. + + Unlike _handle_logging_completed_response (which calls success handlers), + this constructs an exception from the response error and routes to + async_failure_handler / failure_handler so logging integrations correctly + record the call as failed. + """ + response_obj = ( + getattr(self.completed_response, "response", None) + if self.completed_response + else None + ) + error_info = getattr(response_obj, "error", None) if response_obj else None + error_message = "Response failed" + if isinstance(error_info, dict): + error_message = error_info.get("message", str(error_info)) + exception = litellm.APIError( + status_code=500, + message=error_message, + llm_provider=self.custom_llm_provider or "", + model=self.model or "", + ) + self._handle_failure(exception) + async def _call_post_streaming_deployment_hook(self, chunk): """ Allow callbacks to modify streaming chunks before returning (parity with chat). diff --git a/litellm/router.py b/litellm/router.py index 46998abb160..36046ebf302 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -3874,14 +3874,23 @@ class Router: The response from the handler function """ handler_name = original_function.__name__ + metadata_variable_name = _get_router_metadata_variable_name( + function_name="generic_api_call" + ) try: verbose_router_logger.debug( f"Inside _generic_api_call() - handler: {handler_name}, model: {model}; kwargs: {kwargs}" ) + self._update_kwargs_before_fallbacks( + model=model, + kwargs=kwargs, + metadata_variable_name=metadata_variable_name, + ) deployment = self.get_available_deployment( model=model, messages=kwargs.get("messages", None), specific_deployment=kwargs.pop("specific_deployment", None), + request_kwargs=kwargs, ) self._update_kwargs_with_deployment( deployment=deployment, kwargs=kwargs, function_name="generic_api_call" diff --git a/litellm/types/llms/openai.py b/litellm/types/llms/openai.py index a2df3f2e0d6..5a80b40d61f 100644 --- a/litellm/types/llms/openai.py +++ b/litellm/types/llms/openai.py @@ -84,6 +84,7 @@ from typing_extensions import Annotated, Dict, Required, TypedDict, override from litellm.types.llms.base import BaseLiteLLMOpenAIResponseObject from litellm.types.responses.main import ( GenericResponseOutputItem, + OutputCodeInterpreterCall, OutputFunctionToolCall, OutputImageGenerationCall, ) @@ -1242,6 +1243,7 @@ class ResponsesAPIResponse(BaseLiteLLMOpenAIResponseObject): List[ Union[ GenericResponseOutputItem, + OutputCodeInterpreterCall, OutputFunctionToolCall, OutputImageGenerationCall, ResponseFunctionToolCall, @@ -1308,13 +1310,16 @@ class ResponsesAPIResponse(BaseLiteLLMOpenAIResponseObject): if not isinstance(serialized, list): return serialized return [ - { - k: v - for k, v in item.items() - if v is not None or k not in ("status", "content", "encrypted_content") - } - if isinstance(item, dict) and item.get("type") == "reasoning" - else item + ( + { + k: v + for k, v in item.items() + if v is not None + or k not in ("status", "content", "encrypted_content") + } + if isinstance(item, dict) and item.get("type") == "reasoning" + else item + ) for item in serialized ] diff --git a/litellm/types/proxy/control_plane_endpoints.py b/litellm/types/proxy/control_plane_endpoints.py new file mode 100644 index 00000000000..8bf4c44b205 --- /dev/null +++ b/litellm/types/proxy/control_plane_endpoints.py @@ -0,0 +1,14 @@ +from pydantic import BaseModel, field_validator + + +class WorkerRegistryEntry(BaseModel): + worker_id: str + name: str + url: str + + @field_validator("url") + @classmethod + def url_must_be_http(cls, v: str) -> str: + if not v.startswith(("http://", "https://")): + raise ValueError("Worker URL must start with http:// or https://") + return v diff --git a/litellm/types/proxy/discovery_endpoints/ui_discovery_endpoints.py b/litellm/types/proxy/discovery_endpoints/ui_discovery_endpoints.py index 4a4cdaa2bae..46cd3f49f1a 100644 --- a/litellm/types/proxy/discovery_endpoints/ui_discovery_endpoints.py +++ b/litellm/types/proxy/discovery_endpoints/ui_discovery_endpoints.py @@ -1,7 +1,9 @@ -from typing import Optional +from typing import List, Optional from pydantic import BaseModel +from litellm.types.proxy.control_plane_endpoints import WorkerRegistryEntry + class UiDiscoveryEndpoints(BaseModel): server_root_path: str @@ -9,3 +11,5 @@ class UiDiscoveryEndpoints(BaseModel): auto_redirect_to_sso: bool admin_ui_disabled: bool sso_configured: bool + is_control_plane: bool = False + workers: List[WorkerRegistryEntry] = [] diff --git a/litellm/types/responses/main.py b/litellm/types/responses/main.py index 7a666d5e65f..ebd2ad5b5a8 100644 --- a/litellm/types/responses/main.py +++ b/litellm/types/responses/main.py @@ -49,6 +49,42 @@ class OutputImageGenerationCall(BaseLiteLLMOpenAIResponseObject): result: Optional[str] # Base64 encoded image data (without data:image prefix) +class OutputCodeInterpreterCallLog(BaseLiteLLMOpenAIResponseObject): + """Log output from a code interpreter call""" + + type: Literal["logs"] + logs: str + + +class OutputCodeInterpreterCall(BaseLiteLLMOpenAIResponseObject): + """A code interpreter / code execution call output""" + + type: Literal["code_interpreter_call"] + id: str + code: Optional[str] + container_id: Optional[str] + status: Literal["in_progress", "completed", "incomplete", "failed"] + outputs: Optional[List[OutputCodeInterpreterCallLog]] + + +def build_code_interpreter_log_outputs( + content: Any, +) -> Optional[List[OutputCodeInterpreterCallLog]]: + """Convert Anthropic bash_code_execution stdout/stderr to log outputs. + + Shared by streaming (handler.py) and non-streaming (transformation.py) paths. + """ + if not isinstance(content, dict): + return None + parts = [] + if content.get("stdout"): + parts.append(content["stdout"]) + if content.get("stderr"): + parts.append(f"STDERR: {content['stderr']}") + logs = "".join(parts) + return [OutputCodeInterpreterCallLog(type="logs", logs=logs)] if logs else None + + class GenericResponseOutputItem(BaseLiteLLMOpenAIResponseObject): """ Generic response API output item diff --git a/litellm/utils.py b/litellm/utils.py index 81d749ab821..c8272586dad 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -6160,7 +6160,10 @@ def validate_environment( # noqa: PLR0915 ["AZURE_API_BASE", "AZURE_API_VERSION", "AZURE_API_KEY"] ) elif custom_llm_provider == "anthropic": - if "ANTHROPIC_API_KEY" in os.environ: + if ( + "ANTHROPIC_API_KEY" in os.environ + or "ANTHROPIC_AUTH_TOKEN" in os.environ + ): keys_in_environment = True else: missing_keys.append("ANTHROPIC_API_KEY") @@ -6399,7 +6402,10 @@ def validate_environment( # noqa: PLR0915 missing_keys.append("OPENAI_API_KEY") ## anthropic elif model in litellm.anthropic_models: - if "ANTHROPIC_API_KEY" in os.environ: + if ( + "ANTHROPIC_API_KEY" in os.environ + or "ANTHROPIC_AUTH_TOKEN" in os.environ + ): keys_in_environment = True else: missing_keys.append("ANTHROPIC_API_KEY") @@ -8593,9 +8599,7 @@ class ProviderConfigManager: return ManusFilesConfig() elif LlmProviders.ANTHROPIC == provider: - from litellm.llms.anthropic.files.transformation import ( - AnthropicFilesConfig, - ) + from litellm.llms.anthropic.files.transformation import AnthropicFilesConfig return AnthropicFilesConfig() return None diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 181045809f8..879dd42be47 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -4462,6 +4462,78 @@ "supports_vision": true, "supports_web_search": true }, + "azure/gpt-5.4-mini": { + "cache_read_input_token_cost": 7.5e-08, + "input_cost_per_token": 7.5e-07, + "litellm_provider": "azure", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 4.5e-06, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/batch", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_service_tier": true, + "supports_vision": true, + "supports_web_search": true, + "supports_none_reasoning_effort": false, + "supports_xhigh_reasoning_effort": false + }, + "azure/gpt-5.4-nano": { + "cache_read_input_token_cost": 2e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "azure", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.25e-06, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/batch", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_service_tier": true, + "supports_vision": true, + "supports_web_search": true, + "supports_none_reasoning_effort": false, + "supports_xhigh_reasoning_effort": false + }, "azure/gpt-image-1": { "cache_read_input_image_token_cost": 2.5e-06, "cache_read_input_token_cost": 1.25e-06, @@ -37032,5 +37104,157 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true + }, + "volcengine/doubao-seed-2-0-pro-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4.6e-07, + "output_cost_per_token": 2.3e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 7e-07, + "output_cost_per_token": 3.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 7e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "volcengine/doubao-seed-2-0-lite-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 8.7e-08, + "output_cost_per_token": 5.2e-07, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 1.3e-07, + "output_cost_per_token": 7.8e-07, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 2.6e-07, + "output_cost_per_token": 1.6e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "volcengine/doubao-seed-2-0-mini-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 2.9e-08, + "output_cost_per_token": 2.9e-07, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 5.8e-08, + "output_cost_per_token": 5.8e-07, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 1.2e-07, + "output_cost_per_token": 1.2e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] + }, + "volcengine/doubao-seed-2-0-code-preview-260215": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_tool_choice": false, + "supports_vision": true, + "tiered_pricing": [ + { + "input_cost_per_token": 4.6e-07, + "output_cost_per_token": 2.3e-06, + "range": [ + 0, + 32000.0 + ] + }, + { + "input_cost_per_token": 7e-07, + "output_cost_per_token": 3.5e-06, + "range": [ + 32000.0, + 128000.0 + ] + }, + { + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 7e-06, + "range": [ + 128000.0, + 256000.0 + ] + } + ] } } diff --git a/poetry.lock b/poetry.lock index e3b083d778a..d43cdea726c 100644 --- a/poetry.lock +++ b/poetry.lock @@ -3473,15 +3473,15 @@ files = [ [[package]] name = "mcp" -version = "1.25.0" +version = "1.26.0" description = "Model Context Protocol SDK" optional = true python-versions = ">=3.10" groups = ["main"] markers = "python_version >= \"3.10\" and extra == \"proxy\"" files = [ - {file = "mcp-1.25.0-py3-none-any.whl", hash = "sha256:b37c38144a666add0862614cc79ec276e97d72aa8ca26d622818d4e278b9721a"}, - {file = "mcp-1.25.0.tar.gz", hash = "sha256:56310361ebf0364e2d438e5b45f7668cbb124e158bb358333cd06e49e83a6802"}, + {file = "mcp-1.26.0-py3-none-any.whl", hash = "sha256:904a21c33c25aa98ddbeb47273033c435e595bbacfdb177f4bd87f6dceebe1ca"}, + {file = "mcp-1.26.0.tar.gz", hash = "sha256:db6e2ef491eecc1a0d93711a76f28dec2e05999f93afd48795da1c1137142c66"}, ] [package.dependencies] @@ -8018,4 +8018,4 @@ utils = ["numpydoc"] [metadata] lock-version = "2.1" python-versions = ">=3.9,<4.0" -content-hash = "eda34dfd8b35474beffee18893d6782c7b3d0d3d2c610f66237eb97176f43527" +content-hash = "2cf958f1a04fd5f1ab0e5cfc33bdbf441b518ed6c82d0f2546bf64cd3d2f89be" diff --git a/pyproject.toml b/pyproject.toml index 3fb05cc8b34..73f495203bb 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [tool.poetry] name = "litellm" -version = "1.82.4" +version = "1.82.5" description = "Library to easily interface with LLM API providers" authors = ["BerriAI"] license = "MIT" @@ -184,7 +184,7 @@ requires = ["poetry-core", "wheel"] build-backend = "poetry.core.masonry.api" [tool.commitizen] -version = "1.82.4" +version = "1.82.5" version_files = [ "pyproject.toml:^version" ] diff --git a/tests/llm_responses_api_testing/test_base_responses_api_streaming_iterator.py b/tests/llm_responses_api_testing/test_base_responses_api_streaming_iterator.py index 860445d875d..e9181d810e1 100644 --- a/tests/llm_responses_api_testing/test_base_responses_api_streaming_iterator.py +++ b/tests/llm_responses_api_testing/test_base_responses_api_streaming_iterator.py @@ -30,9 +30,11 @@ from litellm.responses.streaming_iterator import BaseResponsesAPIStreamingIterat from litellm.responses.utils import ResponsesAPIRequestUtils from litellm.types.llms.openai import ( ResponseCompletedEvent, + ResponseFailedEvent, + ResponseIncompleteEvent, ResponsesAPIResponse, ResponsesAPIStreamEvents, - OutputTextDeltaEvent + OutputTextDeltaEvent, ) @@ -429,3 +431,155 @@ class TestBaseResponsesAPIStreamingIterator: mock_logging_obj.async_failure_handler.assert_not_called() mock_logging_obj.failure_handler.assert_not_called() + def test_process_chunk_response_failed_calls_failure_handler(self): + """ + Test that a RESPONSE_FAILED event routes to failure handlers, + not success handlers. Failed responses represent genuine LLM-level + errors and should be logged as failures. + """ + from litellm.responses.streaming_iterator import ResponsesAPIStreamingIterator + + mock_response = Mock() + mock_response.headers = {} + mock_response.aiter_lines = Mock() + mock_logging_obj = Mock(spec=LiteLLMLoggingObj) + mock_logging_obj.model_call_details = {"litellm_params": {}} + mock_logging_obj.async_failure_handler = Mock() + mock_logging_obj.failure_handler = Mock() + mock_logging_obj.async_success_handler = Mock() + mock_logging_obj.success_handler = Mock() + mock_config = Mock(spec=BaseResponsesAPIConfig) + + mock_responses_api_response = Mock(spec=ResponsesAPIResponse) + mock_responses_api_response.id = "resp_failed_123" + mock_responses_api_response.error = { + "type": "server_error", + "message": "The model encountered an error", + } + mock_responses_api_response.usage = None + + mock_failed_event = Mock(spec=ResponseFailedEvent) + mock_failed_event.type = ResponsesAPIStreamEvents.RESPONSE_FAILED + mock_failed_event.response = mock_responses_api_response + + mock_config.transform_streaming_response.return_value = mock_failed_event + + iterator = ResponsesAPIStreamingIterator( + response=mock_response, + model="gpt-4", + responses_api_provider_config=mock_config, + logging_obj=mock_logging_obj, + litellm_metadata={"model_info": {"id": "model_123"}}, + custom_llm_provider="openai", + ) + + test_chunk_data = { + "type": "response.failed", + "response": { + "id": "resp_failed_123", + "error": { + "type": "server_error", + "message": "The model encountered an error", + }, + }, + } + + with patch.object( + ResponsesAPIRequestUtils, + "_update_responses_api_response_id_with_model_id", + return_value=mock_responses_api_response, + ), patch( + "litellm.responses.streaming_iterator.run_async_function" + ) as mock_run_async, patch( + "litellm.responses.streaming_iterator.executor" + ) as mock_executor: + result = iterator._process_chunk(json.dumps(test_chunk_data)) + + assert result is not None + assert result.type == ResponsesAPIStreamEvents.RESPONSE_FAILED + assert iterator.completed_response == result + + # Failure handler should have been called via _handle_failure + mock_run_async.assert_called_once() + call_kwargs = mock_run_async.call_args + assert ( + call_kwargs[1]["async_function"] + == mock_logging_obj.async_failure_handler + ) + + mock_executor.submit.assert_called_once() + submit_args = mock_executor.submit.call_args + assert submit_args[0][0] == mock_logging_obj.failure_handler + + def test_process_chunk_response_incomplete_calls_success_handler(self): + """ + Test that a RESPONSE_INCOMPLETE event routes to success handlers. + Incomplete responses (e.g. max_output_tokens reached) are still valid + responses with usage data — analogous to finish_reason='length' in chat. + """ + from litellm.responses.streaming_iterator import ResponsesAPIStreamingIterator + + mock_response = Mock() + mock_response.headers = {} + mock_response.aiter_lines = Mock() + mock_logging_obj = Mock(spec=LiteLLMLoggingObj) + mock_logging_obj.model_call_details = {"litellm_params": {}} + mock_logging_obj.async_failure_handler = Mock() + mock_logging_obj.failure_handler = Mock() + mock_logging_obj.async_success_handler = Mock() + mock_logging_obj.success_handler = Mock() + mock_config = Mock(spec=BaseResponsesAPIConfig) + + mock_responses_api_response = Mock(spec=ResponsesAPIResponse) + mock_responses_api_response.id = "resp_incomplete_123" + mock_responses_api_response.incomplete_details = { + "reason": "max_output_tokens" + } + mock_responses_api_response.usage = None + + mock_incomplete_event = Mock(spec=ResponseIncompleteEvent) + mock_incomplete_event.type = ResponsesAPIStreamEvents.RESPONSE_INCOMPLETE + mock_incomplete_event.response = mock_responses_api_response + + mock_config.transform_streaming_response.return_value = mock_incomplete_event + + iterator = ResponsesAPIStreamingIterator( + response=mock_response, + model="gpt-4", + responses_api_provider_config=mock_config, + logging_obj=mock_logging_obj, + litellm_metadata={"model_info": {"id": "model_123"}}, + custom_llm_provider="openai", + ) + + test_chunk_data = { + "type": "response.incomplete", + "response": { + "id": "resp_incomplete_123", + "incomplete_details": {"reason": "max_output_tokens"}, + }, + } + + with patch.object( + ResponsesAPIRequestUtils, + "_update_responses_api_response_id_with_model_id", + return_value=mock_responses_api_response, + ), patch( + "asyncio.create_task" + ) as mock_create_task, patch( + "litellm.responses.streaming_iterator.executor" + ) as mock_executor: + result = iterator._process_chunk(json.dumps(test_chunk_data)) + + assert result is not None + assert result.type == ResponsesAPIStreamEvents.RESPONSE_INCOMPLETE + assert iterator.completed_response == result + + # Success handler should have been called (via _handle_logging_completed_response) + mock_create_task.assert_called_once() + mock_executor.submit.assert_called_once() + + # Failure handlers should NOT have been called + mock_logging_obj.async_failure_handler.assert_not_called() + mock_logging_obj.failure_handler.assert_not_called() + diff --git a/tests/logging_callback_tests/test_datadog.py b/tests/logging_callback_tests/test_datadog.py index fc4b3ff3cf7..4cfd4a6cc9c 100644 --- a/tests/logging_callback_tests/test_datadog.py +++ b/tests/logging_callback_tests/test_datadog.py @@ -593,7 +593,7 @@ def test_datadog_static_methods(): # Test tags format with default values assert ( "env:unknown,service:litellm-server,version:unknown,HOSTNAME:" - in get_datadog_tags() + in ",".join(get_datadog_tags()) ) # Test with custom environment variables @@ -631,7 +631,7 @@ def test_datadog_static_methods(): # Test tags format with custom values expected_custom_tags = "env:production,service:custom-service,version:1.0.0,HOSTNAME:test-host,POD_NAME:pod-123" print("DataDogLogger._get_datadog_tags()", get_datadog_tags()) - assert get_datadog_tags() == expected_custom_tags + assert ",".join(get_datadog_tags()) == expected_custom_tags @pytest.mark.asyncio @@ -672,11 +672,11 @@ def test_get_datadog_tags(): """Test the _get_datadog_tags static method with various inputs""" # Test with no standard_logging_object and default env vars base_tags = get_datadog_tags() - assert "env:" in base_tags - assert "service:" in base_tags - assert "version:" in base_tags - assert "POD_NAME:" in base_tags - assert "HOSTNAME:" in base_tags + assert any("env:" in t for t in base_tags) + assert any("service:" in t for t in base_tags) + assert any("version:" in t for t in base_tags) + assert any("POD_NAME:" in t for t in base_tags) + assert any("HOSTNAME:" in t for t in base_tags) # Test with custom env vars test_env = { @@ -705,12 +705,12 @@ def test_get_datadog_tags(): # Test with empty request_tags standard_logging_obj["request_tags"] = [] tags_empty_request = get_datadog_tags(standard_logging_obj) - assert "request_tag:" not in tags_empty_request + assert not any(t.startswith("request_tag:") for t in tags_empty_request) # Test with None request_tags standard_logging_obj["request_tags"] = None tags_none_request = get_datadog_tags(standard_logging_obj) - assert "request_tag:" not in tags_none_request + assert not any(t.startswith("request_tag:") for t in tags_none_request) @pytest.mark.asyncio diff --git a/tests/proxy_unit_tests/test_proxy_utils.py b/tests/proxy_unit_tests/test_proxy_utils.py index d77a7465c10..00d4cd24e4b 100644 --- a/tests/proxy_unit_tests/test_proxy_utils.py +++ b/tests/proxy_unit_tests/test_proxy_utils.py @@ -2278,6 +2278,75 @@ async def test_post_call_failure_hook_auth_error_llm_api_route(): mock_handle_logging.assert_called_once() +@pytest.mark.asyncio +@pytest.mark.parametrize( + "request_data, route, expected_call_type", + [ + ( + {"model": "bad-model", "messages": [{"role": "user", "content": "hello"}]}, + "/v1/chat/completions", + "acompletion", + ), + ( + {"model": "bad-model", "prompt": "hello"}, + "/v1/completions", + "atext_completion", + ), + ( + {"model": "bad-model", "input": ["hello"]}, + "/v1/embeddings", + "aembedding", + ), + ], +) +async def test_handle_logging_proxy_only_error_syncs_normalized_call_type( + request_data, route, expected_call_type +): + from fastapi import HTTPException + + from litellm.caching.caching import DualCache + from litellm.litellm_core_utils.litellm_logging import Logging + from litellm.proxy.utils import ProxyLogging + + cache = DualCache() + proxy_logging = ProxyLogging(user_api_key_cache=cache) + captured_logging_obj = {} + original_function_setup = litellm.utils.function_setup + + def _capture_function_setup(*args, **kwargs): + logging_obj, data = original_function_setup(*args, **kwargs) + captured_logging_obj["logging_obj"] = logging_obj + return logging_obj, data + + with patch( + "litellm.proxy.utils.litellm.utils.function_setup", + side_effect=_capture_function_setup, + ), patch.object( + Logging, "async_failure_handler", new=AsyncMock(return_value=None) + ), patch.object( + Logging, "failure_handler", return_value=None + ), patch( + "litellm.proxy.utils.threading.Thread" + ) as mock_thread: + mock_thread.return_value.start = Mock() + + await proxy_logging._handle_logging_proxy_only_error( + request_data=request_data, + user_api_key_dict=UserAPIKeyAuth( + api_key="test_key", + user_id="test_user", + token="test_token", + request_route=route, + ), + route=route, + original_exception=HTTPException(status_code=400, detail="bad request"), + ) + + logging_obj = captured_logging_obj["logging_obj"] + assert logging_obj.call_type == expected_call_type + assert logging_obj.model_call_details["call_type"] == expected_call_type + + @pytest.mark.asyncio async def test_during_call_hook_parallel_execution(): """ diff --git a/tests/router_unit_tests/test_router_helper_utils.py b/tests/router_unit_tests/test_router_helper_utils.py index 3001aec8b86..34a19f5ce79 100644 --- a/tests/router_unit_tests/test_router_helper_utils.py +++ b/tests/router_unit_tests/test_router_helper_utils.py @@ -1568,6 +1568,91 @@ def test_handle_clientside_credential_with_deployment_model_name(model_list): print("✓ _handle_clientside_credential test passed!") +def test_sync_generic_api_call_preserves_requested_model_group_in_logs(): + router = Router( + model_list=[ + { + "model_name": "claude-sonnet-4-6", + "litellm_params": { + "model": "bedrock/global.anthropic.claude-sonnet-4-6", + "aws_access_key_id": "test-access-key", + "aws_secret_access_key": "test-secret-key", + "aws_region_name": "us-west-2", + }, + } + ] + ) + + try: + captured_kwargs = {} + + def mock_original_function(**kwargs): + captured_kwargs.update(kwargs) + return {"status": "ok"} + + response = router._generic_api_call_with_fallbacks( + model="claude-sonnet-4-6", + original_function=mock_original_function, + ) + + assert response == {"status": "ok"} + assert ( + captured_kwargs["model"] == "bedrock/global.anthropic.claude-sonnet-4-6" + ) + assert ( + captured_kwargs["litellm_metadata"]["model_group"] == "claude-sonnet-4-6" + ) + assert ( + captured_kwargs["litellm_metadata"]["deployment"] + == "bedrock/global.anthropic.claude-sonnet-4-6" + ) + finally: + router.discard() + + +def test_sync_generic_api_call_uses_request_kwargs_for_deployment_selection(): + router = Router( + model_list=[ + { + "model_name": "regional-model", + "litellm_params": { + "model": "anthropic/us-model", + "api_key": "test-api-key", + "region_name": "us", + }, + }, + { + "model_name": "regional-model", + "litellm_params": { + "model": "anthropic/eu-model", + "api_key": "test-api-key", + "region_name": "eu", + }, + }, + ], + enable_pre_call_checks=True, + ) + + try: + captured_kwargs = {} + + def mock_original_function(**kwargs): + captured_kwargs.update(kwargs) + return {"status": "ok"} + + response = router._generic_api_call_with_fallbacks( + model="regional-model", + original_function=mock_original_function, + messages=[{"role": "user", "content": "Hello from Europe"}], + allowed_model_region="eu", + ) + + assert response == {"status": "ok"} + assert captured_kwargs["model"] == "anthropic/eu-model" + finally: + router.discard() + + @pytest.mark.parametrize( "function_name, expected_metadata_key", [ diff --git a/tests/test_litellm/integrations/datadog/test_datadog_tags_regression.py b/tests/test_litellm/integrations/datadog/test_datadog_tags_regression.py index 3f1d2be4137..cc9eae7a371 100644 --- a/tests/test_litellm/integrations/datadog/test_datadog_tags_regression.py +++ b/tests/test_litellm/integrations/datadog/test_datadog_tags_regression.py @@ -44,7 +44,7 @@ class TestDatadogTagsRegression: assert "env:test-env" in tags_legacy assert "service:test-service" in tags_legacy # Verify NO team tag (should not invent one) - assert "team:" not in tags_legacy + assert not any(t.startswith("team:") for t in tags_legacy) # Case 2: New feature (team info provided) payload_with_team = StandardLoggingPayload( diff --git a/tests/test_litellm/integrations/test_langsmith_init.py b/tests/test_litellm/integrations/test_langsmith_init.py index 9f7db4095bc..779e7b4c94b 100644 --- a/tests/test_litellm/integrations/test_langsmith_init.py +++ b/tests/test_litellm/integrations/test_langsmith_init.py @@ -132,3 +132,57 @@ class TestLangsmithLoggerInit: assert ( logger.sampling_rate >= 0.0 ), f"sampling_rate should be non-negative, got {logger.sampling_rate}" + + +class TestLangsmithPrepareLogData: + """Regression test for #24001: _prepare_log_data must inject + usage_metadata into outputs so LangSmith's Cost column is populated.""" + + @patch("asyncio.create_task") + @patch.dict(os.environ, {"LANGSMITH_SAMPLING_RATE": "1"}, clear=False) + def test_outputs_contain_usage_metadata(self, mock_create_task): + logger = LangsmithLogger( + langsmith_api_key="test-key", + langsmith_project="test-project", + ) + + payload = { + "id": "test-id", + "response": {"choices": [{"message": {"content": "hi"}}]}, + "metadata": {}, + "startTime": 1.0, + "endTime": 2.0, + "request_tags": [], + "error_str": None, + "status": "success", + "response_cost": 0.0042, + "prompt_tokens": 100, + "completion_tokens": 50, + "total_tokens": 150, + } + + kwargs = { + "litellm_params": {"metadata": {}}, + "standard_logging_object": payload, + } + + credentials = { + "LANGSMITH_API_KEY": "test-key", + "LANGSMITH_PROJECT": "test-project", + "LANGSMITH_BASE_URL": "https://api.smith.langchain.com", + } + + data = logger._prepare_log_data( + kwargs=kwargs, + response_obj=None, + start_time=1.0, + end_time=2.0, + credentials=credentials, + ) + + assert "usage_metadata" in data["outputs"] + um = data["outputs"]["usage_metadata"] + assert um["total_cost"] == 0.0042 + assert um["input_tokens"] == 100 + assert um["output_tokens"] == 50 + assert um["total_tokens"] == 150 diff --git a/tests/test_litellm/litellm_core_utils/test_litellm_logging.py b/tests/test_litellm/litellm_core_utils/test_litellm_logging.py index 6e9b72e96cb..0f950f6da77 100644 --- a/tests/test_litellm/litellm_core_utils/test_litellm_logging.py +++ b/tests/test_litellm/litellm_core_utils/test_litellm_logging.py @@ -11,7 +11,8 @@ sys.path.insert( import time from litellm.constants import SENTRY_DENYLIST, SENTRY_PII_DENYLIST -from litellm.litellm_core_utils.litellm_logging import Logging as LitellmLogging +from litellm.litellm_core_utils.litellm_logging import \ + Logging as LitellmLogging from litellm.litellm_core_utils.litellm_logging import set_callbacks from litellm.types.utils import ModelResponse, TextCompletionResponse @@ -139,7 +140,8 @@ def test_sentry_environment(): def test_use_custom_pricing_for_model(): - from litellm.litellm_core_utils.litellm_logging import use_custom_pricing_for_model + from litellm.litellm_core_utils.litellm_logging import \ + use_custom_pricing_for_model litellm_params = { "custom_llm_provider": "azure", @@ -154,7 +156,8 @@ def test_use_custom_pricing_for_model_via_litellm_metadata(): Generic API call routes (/messages, /responses) store model_info under litellm_metadata, not metadata. Regression test for #23185. """ - from litellm.litellm_core_utils.litellm_logging import use_custom_pricing_for_model + from litellm.litellm_core_utils.litellm_logging import \ + use_custom_pricing_for_model litellm_params = { "litellm_metadata": { @@ -170,7 +173,8 @@ def test_use_custom_pricing_for_model_via_litellm_metadata(): def test_use_custom_pricing_not_detected_litellm_metadata_no_pricing(): """Should return False when litellm_metadata.model_info has no pricing keys.""" - from litellm.litellm_core_utils.litellm_logging import use_custom_pricing_for_model + from litellm.litellm_core_utils.litellm_logging import \ + use_custom_pricing_for_model litellm_params = { "litellm_metadata": { @@ -180,6 +184,198 @@ def test_use_custom_pricing_not_detected_litellm_metadata_no_pricing(): assert use_custom_pricing_for_model(litellm_params) is False +def test_response_cost_calculator_uses_router_model_id_from_litellm_metadata(): + """_response_cost_calculator should extract router_model_id from + litellm_params.litellm_metadata.model_info.id when the result object + does not carry _hidden_params (e.g. ResponsesAPIResponse from /v1/responses + streaming). Regression test for custom pricing on streaming responses.""" + import litellm + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj + from litellm.types.llms.openai import ResponsesAPIResponse + + custom_model_id = "gpt-5-custom-pricing" + custom_input_cost = 125.0 + custom_output_cost = 10.0 + + litellm.register_model( + model_cost={ + custom_model_id: { + "input_cost_per_token": custom_input_cost, + "output_cost_per_token": custom_output_cost, + "max_tokens": 128000, + "max_input_tokens": 128000, + "max_output_tokens": 16384, + "litellm_provider": "openai", + } + } + ) + + try: + logging_obj = LiteLLMLoggingObj( + model="gpt-5", + messages=[{"role": "user", "content": "Hi"}], + stream=True, + call_type="aresponses", + start_time=time.time(), + litellm_call_id="test-123", + function_id="test-fn", + ) + + logging_obj.update_environment_variables( + model="gpt-5", + user="", + optional_params={}, + litellm_params={ + "api_base": "", + "litellm_metadata": { + "model_info": { + "id": custom_model_id, + "input_cost_per_token": custom_input_cost, + "output_cost_per_token": custom_output_cost, + }, + }, + }, + ) + + response_obj = ResponsesAPIResponse( + id="resp_abc", + created_at=1234567890, + model="gpt-5", + output=[], + usage={ + "input_tokens": 10, + "output_tokens": 5, + "total_tokens": 15, + }, + ) + + cost = logging_obj._response_cost_calculator(result=response_obj) + + assert cost is not None, "Cost should not be None" + expected_cost = (10 * custom_input_cost) + (5 * custom_output_cost) + assert cost == pytest.approx( + expected_cost + ), f"Expected {expected_cost}, got {cost}" + finally: + litellm.model_cost.pop(custom_model_id, None) + + +class TestGetRouterModelId: + """Tests for the get_router_model_id helper method.""" + + def test_returns_id_from_litellm_metadata(self, logging_obj): + """Should extract model_info.id from litellm_metadata.""" + logging_obj.litellm_params = { + "litellm_metadata": { + "model_info": {"id": "custom-deploy-1"}, + }, + } + assert logging_obj.get_router_model_id() == "custom-deploy-1" + + def test_returns_id_from_metadata(self, logging_obj): + """Should fall back to metadata when litellm_metadata has no model_info.""" + logging_obj.litellm_params = { + "metadata": { + "model_info": {"id": "custom-deploy-2"}, + }, + } + assert logging_obj.get_router_model_id() == "custom-deploy-2" + + def test_prefers_litellm_metadata_over_metadata(self, logging_obj): + """litellm_metadata should take priority over metadata.""" + logging_obj.litellm_params = { + "litellm_metadata": { + "model_info": {"id": "from-litellm-meta"}, + }, + "metadata": { + "model_info": {"id": "from-meta"}, + }, + } + assert logging_obj.get_router_model_id() == "from-litellm-meta" + + def test_returns_none_when_no_model_info(self, logging_obj): + """Should return None when no model_info is present.""" + logging_obj.litellm_params = {"api_base": ""} + assert logging_obj.get_router_model_id() is None + + def test_returns_none_when_no_litellm_params(self): + """Should return None when litellm_params is not set.""" + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj + + obj = LiteLLMLoggingObj( + model="test", + messages=[], + stream=False, + call_type="completion", + start_time=time.time(), + litellm_call_id="x", + function_id="x", + ) + # litellm_params exists but is empty by default + assert obj.get_router_model_id() is None + + +class TestAnthropicPassthroughCustomPricing: + """Verify the Anthropic pass-through handler forwards custom pricing.""" + + def test_completion_cost_receives_custom_pricing_args(self): + """_create_anthropic_response_logging_payload should pass + custom_pricing and router_model_id to litellm.completion_cost + when the logging object carries custom pricing in model_info.""" + from unittest.mock import patch + + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj + from litellm.proxy.pass_through_endpoints.llm_provider_handlers.anthropic_passthrough_logging_handler import \ + AnthropicPassthroughLoggingHandler + + logging_obj = LiteLLMLoggingObj( + model="claude-sonnet-4-20250514", + messages=[{"role": "user", "content": "Hi"}], + stream=False, + call_type="anthropic_messages", + start_time=time.time(), + litellm_call_id="test-456", + function_id="test-fn", + ) + logging_obj.update_environment_variables( + model="claude-sonnet-4-20250514", + user="", + optional_params={}, + litellm_params={ + "api_base": "", + "litellm_metadata": { + "model_info": { + "id": "claude-custom-pricing", + "input_cost_per_token": 0.5, + "output_cost_per_token": 1.5, + }, + }, + }, + ) + logging_obj.model_call_details["custom_llm_provider"] = "anthropic" + + mock_response = ModelResponse() + mock_response.usage = {"prompt_tokens": 10, "completion_tokens": 5} # type: ignore + + with patch("litellm.completion_cost", return_value=42.0) as mock_cost: + AnthropicPassthroughLoggingHandler._create_anthropic_response_logging_payload( + litellm_model_response=mock_response, + model="claude-sonnet-4-20250514", + kwargs={}, + start_time=time.time(), + end_time=time.time(), + logging_obj=logging_obj, + ) + + mock_cost.assert_called_once() + call_kwargs = mock_cost.call_args + assert call_kwargs.kwargs.get("custom_pricing") is True + assert call_kwargs.kwargs.get("router_model_id") == "claude-custom-pricing" + + class TestUpdateFromKwargs: """Tests for the update_from_kwargs convenience wrapper.""" @@ -245,9 +441,8 @@ class TestUpdateFromKwargs: def test_custom_pricing_detected_via_litellm_metadata(self, logging_obj): """Custom pricing in litellm_metadata.model_info should set custom_pricing flag.""" - from litellm.litellm_core_utils.litellm_logging import ( - use_custom_pricing_for_model, - ) + from litellm.litellm_core_utils.litellm_logging import \ + use_custom_pricing_for_model lm_meta = { "model_info": { @@ -306,7 +501,8 @@ async def test_datadog_logger_not_shadowed_by_llm_obs(monkeypatch): monkeypatch.setenv("DD_SITE", "us5.datadoghq.com") from litellm.integrations.datadog.datadog import DataDogLogger - from litellm.integrations.datadog.datadog_llm_obs import DataDogLLMObsLogger + from litellm.integrations.datadog.datadog_llm_obs import \ + DataDogLLMObsLogger from litellm.litellm_core_utils import litellm_logging as logging_module logging_module._in_memory_loggers.clear() @@ -347,7 +543,8 @@ async def test_logfire_logger_accepts_env_vars_for_base_url(monkeypatch): ) # no trailing slash on purpose # Import after env vars are set (important if module-level caching exists) - from litellm.integrations.opentelemetry import OpenTelemetry # logger class + from litellm.integrations.opentelemetry import \ + OpenTelemetry # logger class from litellm.litellm_core_utils import litellm_logging as logging_module logging_module._in_memory_loggers.clear() @@ -676,7 +873,8 @@ def test_success_handler_runs_guardrail_logging_hook_when_enabled(logging_obj): def test_get_user_agent_tags(): - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup tags = StandardLoggingPayloadSetup._get_user_agent_tags( proxy_server_request={ @@ -691,7 +889,8 @@ def test_get_user_agent_tags(): def test_get_request_tags(): - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup tags = StandardLoggingPayloadSetup._get_request_tags( litellm_params={"metadata": {"tags": ["test-tag"]}}, @@ -718,7 +917,8 @@ def test_get_request_tags_from_metadata_and_litellm_metadata(): 4. No tags in either 5. None values for metadata/litellm_metadata """ - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Test case 1: Tags in metadata only tags = StandardLoggingPayloadSetup._get_request_tags( @@ -799,7 +999,8 @@ def test_get_request_tags_does_not_mutate_original_tags(): would cause User-Agent tags to be duplicated because the function was mutating the original tags list instead of creating a copy. """ - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Create metadata with original tags original_tags = ["custom-tag-1", "custom-tag-2"] @@ -859,7 +1060,8 @@ def test_get_request_tags_does_not_mutate_original_tags(): def test_get_extra_header_tags(): """Test the _get_extra_header_tags method with various scenarios.""" import litellm - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Store original value to restore later original_extra_headers = getattr(litellm, "extra_spend_tag_headers", None) @@ -1080,7 +1282,8 @@ async def test_e2e_generate_cold_storage_object_key_successful(): from datetime import datetime, timezone from unittest.mock import patch - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Create test data start_time = datetime(2025, 1, 15, 10, 30, 45, 123456, timezone.utc) @@ -1122,7 +1325,8 @@ async def test_e2e_generate_cold_storage_object_key_with_custom_logger_s3_path() from datetime import datetime, timezone from unittest.mock import MagicMock, patch - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Create test data start_time = datetime(2025, 1, 15, 10, 30, 45, 123456, timezone.utc) @@ -1173,7 +1377,8 @@ async def test_e2e_generate_cold_storage_object_key_with_logger_no_s3_path(): from datetime import datetime, timezone from unittest.mock import MagicMock, patch - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Create test data start_time = datetime(2025, 1, 15, 10, 30, 45, 123456, timezone.utc) @@ -1220,7 +1425,8 @@ async def test_e2e_generate_cold_storage_object_key_not_configured(): from unittest.mock import patch import litellm - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Create test data start_time = datetime(2025, 1, 15, 10, 30, 45, 123456, timezone.utc) @@ -1244,7 +1450,8 @@ def test_get_final_response_obj_with_empty_response_obj_and_list_init(): When response_obj is empty (falsy), the method should return init_response_obj if it's a list. """ - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Create test objects class TestObject1: @@ -1280,7 +1487,8 @@ def test_get_usage_as_dict(): """ Test get_usage_as_dict returns usage as plain dict from response_obj or combined_usage_object. """ - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup from litellm.types.utils import Usage # Test case 1: None response_obj returns empty usage dict @@ -1318,7 +1526,8 @@ def test_append_system_prompt_messages(): """ Test append_system_prompt_messages prepends system message from kwargs to messages list. """ - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Test case 1: system in kwargs with existing messages kwargs = {"system": "You are a helpful assistant"} @@ -1389,7 +1598,8 @@ async def test_async_success_handler_sets_standard_logging_object_for_pass_throu from datetime import datetime from unittest.mock import patch - from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj from litellm.types.utils import StandardPassThroughResponseObject # Create a logging object for a pass-through endpoint @@ -1470,7 +1680,8 @@ async def test_async_success_handler_prevents_reprocessing_for_pass_through_endp from datetime import datetime from unittest.mock import patch - from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj from litellm.types.utils import StandardPassThroughResponseObject # Create a logging object for a pass-through endpoint @@ -1546,7 +1757,8 @@ async def test_async_success_handler_sets_standard_logging_object_for_streaming_ from datetime import datetime from unittest.mock import patch - from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj from litellm.types.utils import StandardPassThroughResponseObject # Create a logging object for a streaming pass-through endpoint @@ -1602,7 +1814,8 @@ def test_get_error_information_error_code_priority(): Test get_error_information prioritizes 'code' attribute over 'status_code' attribute and handles edge cases like empty strings and "None" string values. """ - from litellm.litellm_core_utils.litellm_logging import StandardLoggingPayloadSetup + from litellm.litellm_core_utils.litellm_logging import \ + StandardLoggingPayloadSetup # Test case 1: Exception with 'code' attribute (ProxyException style) class ProxyException(Exception): @@ -1795,7 +2008,8 @@ async def test_async_success_handler_preserves_response_cost_for_pass_through_en by pass-through handlers (Gemini/Vertex).""" from datetime import datetime - from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj + from litellm.litellm_core_utils.litellm_logging import \ + Logging as LiteLLMLoggingObj from litellm.types.utils import ModelResponse, Usage logging_obj = LiteLLMLoggingObj( 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 5d7b291e7b3..e0862629947 100644 --- a/tests/test_litellm/litellm_core_utils/test_streaming_handler.py +++ b/tests/test_litellm/litellm_core_utils/test_streaming_handler.py @@ -1340,6 +1340,121 @@ def test_is_chunk_non_empty_with_valid_tool_calls( ) +def _make_chunk(content: Optional[str]) -> ModelResponseStream: + return ModelResponseStream( + id="test", + created=1741037890, + model="test-model", + choices=[StreamingChoices(index=0, delta=Delta(content=content))], + ) + + +def _build_chunks(pattern: list[str], N: int) -> list[ModelResponseStream]: + """ + Build a list of chunks based on a pattern specification. + """ + chunks = [] + for i, p in enumerate(pattern): + if p == "same": + chunks.append(_make_chunk("same_chunk")) + elif p == "diff": + chunks.append(_make_chunk(f"chunk_{i}")) + else: + chunks.append(_make_chunk(p)) + return chunks + +_REPETITION_TEST_CASES = [ + # Basic cases + pytest.param( + ["same"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + True, + id="all_identical_raises", + ), + pytest.param( + ["same"] * (litellm.REPEATED_STREAMING_CHUNK_LIMIT - 1), + False, + id="below_threshold_no_raise", + ), + pytest.param( + [None] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + False, + id="none_content_no_raise", + ), + pytest.param( + [""] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + False, + id="empty_content_no_raise", + ), + # Short content (len <= 2) should not raise + pytest.param( + ["##"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + False, + id="short_content_2chars_no_raise", + ), + pytest.param( + ["{"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + False, + id="short_content_1char_no_raise", + ), + pytest.param( + ["ab"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + False, + id="short_content_2chars_ab_no_raise", + ), + # All different chunks + pytest.param( + ["diff"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT, + False, + id="all_different_no_raise", + ), + # One chunk different at various positions + pytest.param( + ["different_first"] + ["same"] * (litellm.REPEATED_STREAMING_CHUNK_LIMIT - 1), + False, + id="first_chunk_different_no_raise", + ), + pytest.param( + ["same"] * (litellm.REPEATED_STREAMING_CHUNK_LIMIT - 1) + ["different_last"], + False, + id="last_chunk_different_no_raise", + ), + pytest.param( + ["same"] * (litellm.REPEATED_STREAMING_CHUNK_LIMIT // 2 + 1) + ["different_mid"] + ["same"] * (litellm.REPEATED_STREAMING_CHUNK_LIMIT - litellm.REPEATED_STREAMING_CHUNK_LIMIT // 2 + 1), + False, + id="middle_chunk_different_no_raise", + ), + pytest.param( + ["same"] * (litellm.REPEATED_STREAMING_CHUNK_LIMIT - 2) + ["diff", "diff"], + False, + id="last_two_different_no_raise", + ), + pytest.param( + ["diff"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT + ["same"] * litellm.REPEATED_STREAMING_CHUNK_LIMIT + ["diff"], + True, + id="in_between_same_and_diff_raise", + ), +] + + +@pytest.mark.parametrize("chunks_pattern,should_raise", _REPETITION_TEST_CASES) +def test_raise_on_model_repetition( + initialized_custom_stream_wrapper: CustomStreamWrapper, + chunks_pattern: list, + should_raise: bool, +): + wrapper = initialized_custom_stream_wrapper + chunks = _build_chunks(chunks_pattern, len(chunks_pattern)) + + if should_raise: + with pytest.raises(litellm.InternalServerError) as exc_info: + for chunk in chunks: + wrapper.chunks.append(chunk) + wrapper.raise_on_model_repetition() + assert "repeating the same chunk" in str(exc_info.value) + else: + for chunk in chunks: + wrapper.chunks.append(chunk) + wrapper.raise_on_model_repetition() def test_usage_chunk_after_finish_reason_updates_hidden_params(logging_obj): """ Test that provider-reported usage from a post-finish_reason chunk diff --git a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py index d9f513d8d1d..20427e8cc94 100644 --- a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py +++ b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py @@ -6,6 +6,7 @@ from litellm.types.llms.openai import ( ChatCompletionToolCallChunk, ChatCompletionToolCallFunctionChunk, ) +from litellm.types.responses.main import OutputCodeInterpreterCall def test_redacted_thinking_content_block_delta(): @@ -479,14 +480,22 @@ def test_partial_json_chunk_accumulation(): # First partial chunk should return None (still accumulating) result1 = iterator._parse_sse_data(f"data:{partial_chunk_1}") assert result1 is None, "First partial chunk should return None while accumulating" - assert iterator.chunk_type == "accumulated_json", "Should switch to accumulated_json mode" - assert iterator.accumulated_json == partial_chunk_1, "Should have accumulated first part" + assert ( + iterator.chunk_type == "accumulated_json" + ), "Should switch to accumulated_json mode" + assert ( + iterator.accumulated_json == partial_chunk_1 + ), "Should have accumulated first part" # Second partial chunk should complete the JSON and return a parsed result result2 = iterator._parse_sse_data(f"data:{partial_chunk_2}") assert result2 is not None, "Second chunk should return parsed result" - assert iterator.accumulated_json == "", "Buffer should be cleared after successful parse" - assert result2.choices[0].delta.content == "Hello", f"Expected 'Hello', got '{result2.choices[0].delta.content}'" + assert ( + iterator.accumulated_json == "" + ), "Buffer should be cleared after successful parse" + assert ( + result2.choices[0].delta.content == "Hello" + ), f"Expected 'Hello', got '{result2.choices[0].delta.content}'" def test_complete_json_chunk_no_accumulation(): @@ -503,7 +512,9 @@ def test_complete_json_chunk_no_accumulation(): assert result is not None, "Complete chunk should return parsed result immediately" assert iterator.chunk_type == "valid_json", "Should remain in valid_json mode" assert iterator.accumulated_json == "", "Buffer should remain empty" - assert result.choices[0].delta.content == "Hello", f"Expected 'Hello', got '{result.choices[0].delta.content}'" + assert ( + result.choices[0].delta.content == "Hello" + ), f"Expected 'Hello', got '{result.choices[0].delta.content}'" def test_multiple_partial_chunks_accumulation(): @@ -620,7 +631,9 @@ def test_web_search_tool_result_no_extra_tool_calls(): # Should have exactly 2 tool calls: # 1. From content_block_start (server_tool_use) with id and name # 2. From content_block_delta with the actual query - assert len(tool_calls_emitted) == 2, f"Expected 2 tool calls, got {len(tool_calls_emitted)}" + assert ( + len(tool_calls_emitted) == 2 + ), f"Expected 2 tool calls, got {len(tool_calls_emitted)}" # First tool call should have the id and name assert tool_calls_emitted[0]["id"] == "srvtoolu_01ABC123" @@ -722,7 +735,10 @@ def test_web_search_tool_result_captured_in_provider_specific_fields(): { "type": "content_block_delta", "index": 0, - "delta": {"type": "input_json_delta", "partial_json": '{"query": "otter facts"}'}, + "delta": { + "type": "input_json_delta", + "partial_json": '{"query": "otter facts"}', + }, }, # 4. content_block_stop for server_tool_use {"type": "content_block_stop", "index": 0}, @@ -822,7 +838,10 @@ def test_web_fetch_tool_result_captured_in_provider_specific_fields(): { "type": "content_block_delta", "index": 0, - "delta": {"type": "input_json_delta", "partial_json": '{"url": "https://example.com"}'}, + "delta": { + "type": "input_json_delta", + "partial_json": '{"url": "https://example.com"}', + }, }, # 4. content_block_stop for server_tool_use {"type": "content_block_stop", "index": 0}, @@ -946,7 +965,7 @@ def test_web_fetch_tool_result_no_extra_tool_calls(): def test_container_in_provider_specific_fields_streaming(): """ Test that container is captured in provider_specific_fields for streaming responses. - + When container with skills is used, the container field should be present in the provider_specific_fields of the message_delta chunk. """ @@ -1025,7 +1044,9 @@ def test_container_in_provider_specific_fields_streaming(): ] # Verify container was captured - assert container_field is not None, "container should be captured in provider_specific_fields" + assert ( + container_field is not None + ), "container should be captured in provider_specific_fields" assert ( container_field["id"] == "container_011CW9hA9zpZ8xD3bjjShy4p" ), "container id should match" @@ -1033,18 +1054,14 @@ def test_container_in_provider_specific_fields_streaming(): container_field["expires_at"] == "2025-12-16T04:57:16.913181Z" ), "expires_at should match" assert len(container_field["skills"]) == 1, "Should have 1 skill" - assert ( - container_field["skills"][0]["skill_id"] == "pptx" - ), "skill_id should be pptx" - assert ( - container_field["skills"][0]["version"] == "20251013" - ), "version should match" + assert container_field["skills"][0]["skill_id"] == "pptx", "skill_id should be pptx" + assert container_field["skills"][0]["version"] == "20251013", "version should match" def test_container_in_provider_specific_fields_non_streaming(): """ Test that container is captured in provider_specific_fields for non-streaming responses. - + When container with skills is used in non-streaming, the container field should be present in the provider_specific_fields of the response. """ @@ -1106,7 +1123,7 @@ def test_container_in_provider_specific_fields_non_streaming(): def test_container_absent_when_not_provided(): """ Test that container is not added to provider_specific_fields when not provided. - + This ensures we don't add empty or None container fields. """ iterator = ModelResponseIterator( @@ -1133,3 +1150,434 @@ def test_container_absent_when_not_provided(): assert ( "container" not in model_response.choices[0].delta.provider_specific_fields ), "container should not be present when not provided in delta" + + +def test_streaming_code_execution_produces_code_interpreter_results(): + """ + Test that bash_code_execution_tool_result content blocks in streaming + produce code_interpreter_results in provider_specific_fields, so the + Responses API layer can use them without Anthropic-specific knowledge. + """ + + chunks = [ + { + "type": "message_start", + "message": { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "content": [], + "usage": {"input_tokens": 100, "output_tokens": 1}, + }, + }, + { + "type": "content_block_start", + "index": 0, + "content_block": { + "type": "text", + "text": "", + }, + }, + { + "type": "content_block_delta", + "index": 0, + "delta": {"type": "text_delta", "text": "Running code..."}, + }, + {"type": "content_block_stop", "index": 0}, + { + "type": "content_block_start", + "index": 1, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01ABC", + "name": "bash_code_execution", + "input": {"command": "echo hello"}, + }, + }, + {"type": "content_block_stop", "index": 1}, + { + "type": "content_block_start", + "index": 2, + "content_block": { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01ABC", + "content": { + "type": "bash_code_execution_result", + "stdout": "hello\n", + "stderr": "", + "return_code": 0, + }, + }, + }, + {"type": "content_block_stop", "index": 2}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn"}, + "usage": {"output_tokens": 50}, + }, + ] + + iterator = ModelResponseIterator(None, sync_stream=True) + + found_code_interpreter_results = False + for chunk in chunks: + parsed = iterator.chunk_parser(chunk) + psf = None + if parsed.choices and parsed.choices[0].delta: + psf = getattr(parsed.choices[0].delta, "provider_specific_fields", None) + if psf and "code_interpreter_results" in psf: + found_code_interpreter_results = True + results = psf["code_interpreter_results"] + assert len(results) == 1 + assert isinstance(results[0], OutputCodeInterpreterCall) + assert results[0].type == "code_interpreter_call" + assert results[0].id == "srvtoolu_01ABC" + assert results[0].code == "echo hello" + assert results[0].outputs is not None + assert len(results[0].outputs) == 1 + assert results[0].outputs[0].logs == "hello\n" + + assert found_code_interpreter_results, ( + "code_interpreter_results should appear in provider_specific_fields " + "when bash_code_execution_tool_result is streamed" + ) + + +def test_streaming_multiple_code_executions_no_duplicates(): + """ + Test that multiple code executions in a single streaming response emit + cumulative code_interpreter_results on each chunk (matching stream_chunk_builder's + "last value wins" contract). The final emission must contain ALL results. + """ + chunks = [ + { + "type": "message_start", + "message": { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "content": [], + "usage": {"input_tokens": 100, "output_tokens": 1}, + }, + }, + # First code execution + { + "type": "content_block_start", + "index": 0, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01AAA", + "name": "bash_code_execution", + "input": {"command": "echo first"}, + }, + }, + {"type": "content_block_stop", "index": 0}, + { + "type": "content_block_start", + "index": 1, + "content_block": { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01AAA", + "content": { + "type": "bash_code_execution_result", + "stdout": "first\n", + "stderr": "", + "return_code": 0, + }, + }, + }, + {"type": "content_block_stop", "index": 1}, + # Second code execution + { + "type": "content_block_start", + "index": 2, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01BBB", + "name": "bash_code_execution", + "input": {"command": "echo second"}, + }, + }, + {"type": "content_block_stop", "index": 2}, + { + "type": "content_block_start", + "index": 3, + "content_block": { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01BBB", + "content": { + "type": "bash_code_execution_result", + "stdout": "second\n", + "stderr": "", + "return_code": 0, + }, + }, + }, + {"type": "content_block_stop", "index": 3}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn"}, + "usage": {"output_tokens": 50}, + }, + ] + + iterator = ModelResponseIterator(None, sync_stream=True) + + # Collect each emission of code_interpreter_results + emissions = [] + for chunk in chunks: + parsed = iterator.chunk_parser(chunk) + psf = None + if parsed.choices and parsed.choices[0].delta: + psf = getattr(parsed.choices[0].delta, "provider_specific_fields", None) + if psf and "code_interpreter_results" in psf: + emissions.append(psf["code_interpreter_results"]) + + # Should have 2 emissions (one per tool_result block) + assert len(emissions) == 2, f"Expected 2 emissions, got {len(emissions)}" + + # First emission: cumulative list with 1 result + assert len(emissions[0]) == 1 + assert emissions[0][0].id == "srvtoolu_01AAA" + assert emissions[0][0].code == "echo first" + assert emissions[0][0].outputs[0].logs == "first\n" + + # Second (final) emission: cumulative list with BOTH results + # This is what stream_chunk_builder will pick as "last value wins" + assert len(emissions[1]) == 2, ( + f"Expected final emission to have 2 results, got {len(emissions[1])}. " + f"IDs: {[r.id for r in emissions[1]]}" + ) + assert emissions[1][0].id == "srvtoolu_01AAA" + assert emissions[1][0].code == "echo first" + assert emissions[1][0].outputs[0].logs == "first\n" + assert emissions[1][1].id == "srvtoolu_01BBB" + assert emissions[1][1].code == "echo second" + assert emissions[1][1].outputs[0].logs == "second\n" + + +def test_streaming_code_execution_input_assembled_from_deltas(): + """ + In real Anthropic streaming, content_block_start for server_tool_use has + input: {}. The actual input arrives via input_json_delta deltas and must + be assembled at content_block_stop so the code field is populated. + + This test uses realistic chunk shapes (empty input in start, partial JSON + in deltas) to exercise the input assembly path. + """ + chunks = [ + { + "type": "message_start", + "message": { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "content": [], + "usage": {"input_tokens": 100, "output_tokens": 1}, + }, + }, + # server_tool_use with empty input (real streaming behaviour) + { + "type": "content_block_start", + "index": 0, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01AAA", + "name": "code_execution", + "input": {}, + }, + }, + # Input arrives via deltas, split across two chunks + { + "type": "content_block_delta", + "index": 0, + "delta": { + "type": "input_json_delta", + "partial_json": '{"comma', + }, + }, + { + "type": "content_block_delta", + "index": 0, + "delta": { + "type": "input_json_delta", + "partial_json": 'nd": "echo hello"}', + }, + }, + {"type": "content_block_stop", "index": 0}, + # Tool result + { + "type": "content_block_start", + "index": 1, + "content_block": { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01AAA", + "content": { + "type": "bash_code_execution_result", + "stdout": "hello\n", + "stderr": "", + "return_code": 0, + }, + }, + }, + {"type": "content_block_stop", "index": 1}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn"}, + "usage": {"output_tokens": 50}, + }, + ] + + iterator = ModelResponseIterator(None, sync_stream=True) + + code_results = None + for chunk in chunks: + parsed = iterator.chunk_parser(chunk) + psf = None + if parsed.choices and parsed.choices[0].delta: + psf = getattr(parsed.choices[0].delta, "provider_specific_fields", None) + if psf and "code_interpreter_results" in psf: + code_results = psf["code_interpreter_results"] + + # The code field must contain the assembled input, not be empty + assert code_results is not None, "No code_interpreter_results emitted" + assert len(code_results) == 1 + assert code_results[0].id == "srvtoolu_01AAA" + assert code_results[0].code == "echo hello" + assert code_results[0].outputs[0].logs == "hello\n" + + +def test_empty_output_produces_null_outputs(): + """ + When both stdout and stderr are empty, outputs should be None + (matching OpenAI's native behavior) rather than [{logs: ""}]. + """ + chunks = [ + { + "type": "message_start", + "message": { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "content": [], + "usage": {"input_tokens": 100, "output_tokens": 1}, + }, + }, + { + "type": "content_block_start", + "index": 0, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01AAA", + "name": "bash_code_execution", + "input": {"command": "true"}, + }, + }, + {"type": "content_block_stop", "index": 0}, + { + "type": "content_block_start", + "index": 1, + "content_block": { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01AAA", + "content": { + "type": "bash_code_execution_result", + "stdout": "", + "stderr": "", + "return_code": 0, + }, + }, + }, + {"type": "content_block_stop", "index": 1}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn"}, + "usage": {"output_tokens": 50}, + }, + ] + + iterator = ModelResponseIterator(None, sync_stream=True) + + code_results = None + for chunk in chunks: + parsed = iterator.chunk_parser(chunk) + psf = None + if parsed.choices and parsed.choices[0].delta: + psf = getattr(parsed.choices[0].delta, "provider_specific_fields", None) + if psf and "code_interpreter_results" in psf: + code_results = psf["code_interpreter_results"] + + assert code_results is not None, "No code_interpreter_results emitted" + assert len(code_results) == 1 + assert code_results[0].id == "srvtoolu_01AAA" + assert ( + code_results[0].outputs is None + ), f"Expected outputs=None for empty execution, got {code_results[0].outputs}" + + +def test_non_bash_tool_result_skipped(): + """ + Tool result types other than bash_code_execution_tool_result (e.g. + text_editor_code_execution_tool_result) should be skipped and NOT + produce code_interpreter_call items. + """ + chunks = [ + { + "type": "message_start", + "message": { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "content": [], + "usage": {"input_tokens": 100, "output_tokens": 1}, + }, + }, + { + "type": "content_block_start", + "index": 0, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01AAA", + "name": "text_editor", + "input": {"command": "view", "path": "/tmp/test.py"}, + }, + }, + {"type": "content_block_stop", "index": 0}, + # text_editor result — should NOT become a code_interpreter_call + { + "type": "content_block_start", + "index": 1, + "content_block": { + "type": "text_editor_code_execution_tool_result", + "tool_use_id": "srvtoolu_01AAA", + "content": [ + {"type": "text", "text": "file contents here"}, + ], + }, + }, + {"type": "content_block_stop", "index": 1}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn"}, + "usage": {"output_tokens": 50}, + }, + ] + + iterator = ModelResponseIterator(None, sync_stream=True) + + code_results = None + for chunk in chunks: + parsed = iterator.chunk_parser(chunk) + psf = None + if parsed.choices and parsed.choices[0].delta: + psf = getattr(parsed.choices[0].delta, "provider_specific_fields", None) + if psf and "code_interpreter_results" in psf: + code_results = psf["code_interpreter_results"] + + # code_interpreter_results should be emitted but empty (no bash results) + assert ( + code_results is not None + ), "Expected code_interpreter_results key to be emitted" + assert ( + len(code_results) == 0 + ), f"Expected 0 code_interpreter_results for text_editor result, got {len(code_results)}" 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 a95b9413b9d..10a3c107367 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 @@ -183,7 +183,9 @@ def test_extract_response_content_with_citations(): }, } - _, citations, _, _, _, _, _, _ = config.extract_response_content(completion_response) + _, citations, _, _, _, _, _, _ = config.extract_response_content( + completion_response + ) assert citations == [ [ { @@ -305,7 +307,7 @@ def test_web_search_tool_result_extraction(): "type": "server_tool_use", "id": "srvtoolu_01ABC123", "name": "web_search", - "input": {"query": "average weight african elephant kg"} + "input": {"query": "average weight african elephant kg"}, }, { "type": "web_search_tool_result", @@ -317,32 +319,39 @@ def test_web_search_tool_result_extraction(): "title": "African Elephant Facts", "encrypted_content": "encrypted_data_here", "page_age": "2024-01-15", - "snippet": "Adult African elephants weigh between 4,000-6,000 kg..." + "snippet": "Adult African elephants weigh between 4,000-6,000 kg...", } - ] + ], }, { "type": "text", - "text": "Based on my search, African elephants weigh around 5,000 kg." + "text": "Based on my search, African elephants weigh around 5,000 kg.", }, { "type": "tool_use", "id": "toolu_01XYZ789", "name": "add_numbers", - "input": {"a": 5000, "b": 100} - } + "input": {"a": 5000, "b": 100}, + }, ], "stop_reason": "tool_use", "usage": { "input_tokens": 100, "output_tokens": 50, - "server_tool_use": {"web_search_requests": 1} - } + "server_tool_use": {"web_search_requests": 1}, + }, } - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) # Verify text extraction assert "Based on my search" in text @@ -388,7 +397,7 @@ def test_web_search_tool_result_in_provider_specific_fields(): "type": "server_tool_use", "id": "srvtoolu_provider_test", "name": "web_search", - "input": {"query": "test query"} + "input": {"query": "test query"}, }, { "type": "web_search_tool_result", @@ -398,21 +407,18 @@ def test_web_search_tool_result_in_provider_specific_fields(): "type": "web_search_result", "url": "https://example.com/test", "title": "Test Result", - "snippet": "Test snippet content" + "snippet": "Test snippet content", } - ] + ], }, - { - "type": "text", - "text": "Here is the result." - } + {"type": "text", "text": "Here is the result."}, ], "stop_reason": "end_turn", "usage": { "input_tokens": 50, "output_tokens": 25, - "server_tool_use": {"web_search_requests": 1} - } + "server_tool_use": {"web_search_requests": 1}, + }, } raw_response = httpx.Response(status_code=200, headers={}) @@ -432,7 +438,10 @@ def test_web_search_tool_result_in_provider_specific_fields(): assert "web_search_results" in provider_fields assert len(provider_fields["web_search_results"]) == 1 assert provider_fields["web_search_results"][0]["type"] == "web_search_tool_result" - assert provider_fields["web_search_results"][0]["tool_use_id"] == "srvtoolu_provider_test" + assert ( + provider_fields["web_search_results"][0]["tool_use_id"] + == "srvtoolu_provider_test" + ) def test_multiple_web_search_tool_results(): @@ -447,34 +456,52 @@ def test_multiple_web_search_tool_results(): "type": "server_tool_use", "id": "srvtoolu_search1", "name": "web_search", - "input": {"query": "african elephant weight"} + "input": {"query": "african elephant weight"}, }, { "type": "web_search_tool_result", "tool_use_id": "srvtoolu_search1", - "content": [{"type": "web_search_result", "url": "https://example1.com", "title": "Result 1", "snippet": "First result"}] + "content": [ + { + "type": "web_search_result", + "url": "https://example1.com", + "title": "Result 1", + "snippet": "First result", + } + ], }, { "type": "server_tool_use", "id": "srvtoolu_search2", "name": "web_search", - "input": {"query": "asian elephant weight"} + "input": {"query": "asian elephant weight"}, }, { "type": "web_search_tool_result", "tool_use_id": "srvtoolu_search2", - "content": [{"type": "web_search_result", "url": "https://example2.com", "title": "Result 2", "snippet": "Second result"}] + "content": [ + { + "type": "web_search_result", + "url": "https://example2.com", + "title": "Result 2", + "snippet": "Second result", + } + ], }, - { - "type": "text", - "text": "Found information about both elephants." - } + {"type": "text", "text": "Found information about both elephants."}, ] } - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) # Verify both web_search_tool_results are extracted assert web_search_results is not None @@ -751,7 +778,7 @@ def test_anthropic_beta_header_merging_with_output_format(): optional_params = { "output_format": { "type": "json_schema", - "schema": {"type": "object", "properties": {}} + "schema": {"type": "object", "properties": {}}, } } @@ -761,10 +788,12 @@ def test_anthropic_beta_header_merging_with_output_format(): # Both beta headers should be present beta_value = result_headers["anthropic-beta"] - assert "context-1m-2025-08-07" in beta_value, \ - f"User's context-1m beta header missing from: {beta_value}" - assert "structured-outputs-2025-11-13" in beta_value, \ - f"Structured output beta header missing from: {beta_value}" + assert ( + "context-1m-2025-08-07" in beta_value + ), f"User's context-1m beta header missing from: {beta_value}" + assert ( + "structured-outputs-2025-11-13" in beta_value + ), f"Structured output beta header missing from: {beta_value}" def test_anthropic_beta_header_merging_with_multiple_features(): @@ -780,10 +809,10 @@ def test_anthropic_beta_header_merging_with_multiple_features(): optional_params = { "output_format": { "type": "json_schema", - "schema": {"type": "object", "properties": {}} + "schema": {"type": "object", "properties": {}}, }, "context_management": _sample_context_management_payload(), - "tools": [{"type": "web_fetch_20250910", "name": "web_fetch"}] + "tools": [{"type": "web_fetch_20250910", "name": "web_fetch"}], } result_headers = config.update_headers_with_optional_anthropic_beta( @@ -950,20 +979,12 @@ def test_tool_search_regex_detection(): # Test with tool search regex tool tools = [ - { - "type": "tool_search_tool_regex_20251119", - "name": "tool_search_tool_regex" - } + {"type": "tool_search_tool_regex_20251119", "name": "tool_search_tool_regex"} ] assert config.is_tool_search_used(tools) is True # Test without tool search - tools = [ - { - "type": "function", - "function": {"name": "get_weather"} - } - ] + tools = [{"type": "function", "function": {"name": "get_weather"}}] assert config.is_tool_search_used(tools) is False @@ -975,10 +996,7 @@ def test_tool_search_bm25_detection(): # Test with tool search BM25 tool tools = [ - { - "type": "tool_search_tool_bm25_20251119", - "name": "tool_search_tool_bm25" - } + {"type": "tool_search_tool_bm25_20251119", "name": "tool_search_tool_bm25"} ] assert config.is_tool_search_used(tools) is True @@ -1002,10 +1020,7 @@ def test_tool_search_regex_mapping(): """Test that tool search regex tools are properly mapped""" config = AnthropicConfig() - tool = { - "type": "tool_search_tool_regex_20251119", - "name": "tool_search_tool_regex" - } + tool = {"type": "tool_search_tool_regex_20251119", "name": "tool_search_tool_regex"} mapped_tool, mcp_server = config._map_tool_helper(tool) @@ -1019,10 +1034,7 @@ def test_tool_search_bm25_mapping(): """Test that tool search BM25 tools are properly mapped""" config = AnthropicConfig() - tool = { - "type": "tool_search_tool_bm25_20251119", - "name": "tool_search_tool_bm25" - } + tool = {"type": "tool_search_tool_bm25_20251119", "name": "tool_search_tool_bm25"} mapped_tool, mcp_server = config._map_tool_helper(tool) @@ -1037,20 +1049,17 @@ def test_deferred_tools_separation(): config = AnthropicConfig() tools = [ - { - "type": "tool_search_tool_regex_20251119", - "name": "tool_search_tool_regex" - }, + {"type": "tool_search_tool_regex_20251119", "name": "tool_search_tool_regex"}, { "type": "function", "function": {"name": "get_weather"}, - "defer_loading": True + "defer_loading": True, }, { "type": "function", "function": {"name": "search_files"}, - "defer_loading": False - } + "defer_loading": False, + }, ] non_deferred, deferred = config._separate_deferred_tools(tools) @@ -1069,14 +1078,21 @@ def test_server_tool_use_in_response(): "type": "server_tool_use", "id": "srvtoolu_01ABC123", "name": "tool_search_tool_regex", - "input": {"query": "weather"} + "input": {"query": "weather"}, } ] } - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) assert len(tool_calls) == 1 assert tool_calls[0]["id"] == "srvtoolu_01ABC123" @@ -1091,9 +1107,7 @@ def test_tool_search_usage_tracking(): usage_object = { "input_tokens": 100, "output_tokens": 50, - "server_tool_use": { - "tool_search_requests": 2 - } + "server_tool_use": {"tool_search_requests": 2}, } usage = config.calculate_usage(usage_object=usage_object, reasoning_content=None) @@ -1109,16 +1123,13 @@ def test_tool_reference_expansion(): deferred_tools = [ { "type": "function", - "function": { - "name": "get_weather", - "description": "Get weather" - } + "function": {"name": "get_weather", "description": "Get weather"}, } ] content = [ {"type": "text", "text": "I'll search for tools"}, - {"type": "tool_reference", "tool_name": "get_weather"} + {"type": "tool_reference", "tool_name": "get_weather"}, ] expanded = config._expand_tool_references(content, deferred_tools) @@ -1140,13 +1151,11 @@ def test_defer_loading_preserved_in_transformation(): "description": "Get weather information", "parameters": { "type": "object", - "properties": { - "location": {"type": "string"} - }, - "required": ["location"] - } + "properties": {"location": {"type": "string"}}, + "required": ["location"], + }, }, - "defer_loading": True + "defer_loading": True, } mapped_tool, mcp_server = config._map_tool_helper(tool) @@ -1166,45 +1175,51 @@ def test_tool_search_complete_response_parsing(): "content": [ { "type": "text", - "text": "I'll search for weather-related tools that can help you." + "text": "I'll search for weather-related tools that can help you.", }, { "type": "server_tool_use", "id": "srvtoolu_015i6aVA2niwzv4RG4DtnxDJ", "name": "tool_search_tool_regex", "input": {"pattern": "weather", "limit": 5}, - "caller": {"type": "direct"} + "caller": {"type": "direct"}, }, { "type": "tool_search_tool_result", "tool_use_id": "srvtoolu_015i6aVA2niwzv4RG4DtnxDJ", "content": { "type": "tool_search_tool_search_result", - "tool_references": [{"type": "tool_reference", "tool_name": "get_weather"}] - } - }, - { - "type": "text", - "text": "Great! I found a weather tool." + "tool_references": [ + {"type": "tool_reference", "tool_name": "get_weather"} + ], + }, }, + {"type": "text", "text": "Great! I found a weather tool."}, { "type": "tool_use", "id": "toolu_01CrCNx4ntSaeeV9iArT4JfQ", "name": "get_weather", - "input": {"location": "San Francisco"} - } + "input": {"location": "San Francisco"}, + }, ], "usage": { "input_tokens": 1639, "output_tokens": 170, - "server_tool_use": {"web_search_requests": 0} - } + "server_tool_use": {"web_search_requests": 0}, + }, } # Extract content - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) # Verify text extraction (should concatenate both text blocks) assert "I'll search for weather-related tools" in text @@ -1222,12 +1237,14 @@ def test_tool_search_complete_response_parsing(): usage = config.calculate_usage( usage_object=completion_response["usage"], reasoning_content=None, - completion_response=completion_response + completion_response=completion_response, ) assert usage.server_tool_use is not None assert usage.server_tool_use.web_search_requests == 0 - assert usage.server_tool_use.tool_search_requests == 1 # Counted from server_tool_use blocks + assert ( + usage.server_tool_use.tool_search_requests == 1 + ) # Counted from server_tool_use blocks def test_allowed_callers_field_preservation(): @@ -1242,13 +1259,11 @@ def test_allowed_callers_field_preservation(): "description": "Execute a SQL query", "parameters": { "type": "object", - "properties": { - "sql": {"type": "string"} - }, - "required": ["sql"] - } + "properties": {"sql": {"type": "string"}}, + "required": ["sql"], + }, }, - "allowed_callers": ["code_execution_20250825"] + "allowed_callers": ["code_execution_20250825"], } transformed_tool, _ = config._map_tool_helper(tool_with_allowed_callers) @@ -1265,19 +1280,16 @@ def test_programmatic_tool_calling_beta_header(): # Test detection with allowed_callers tools = [ - { - "type": "code_execution_20250825", - "name": "code_execution" - }, + {"type": "code_execution_20250825", "name": "code_execution"}, { "type": "function", "function": { "name": "query_database", "description": "Execute a SQL query", - "parameters": {"type": "object", "properties": {}} + "parameters": {"type": "object", "properties": {}}, }, - "allowed_callers": ["code_execution_20250825"] - } + "allowed_callers": ["code_execution_20250825"], + }, ] is_programmatic = model_info.is_programmatic_tool_calling_used(tools) @@ -1285,8 +1297,7 @@ def test_programmatic_tool_calling_beta_header(): # Test header generation headers = model_info.get_anthropic_headers( - api_key="test-key", - programmatic_tool_calling_used=True + api_key="test-key", programmatic_tool_calling_used=True ) assert "anthropic-beta" in headers @@ -1303,10 +1314,7 @@ def test_caller_field_in_response(): "type": "message", "role": "assistant", "content": [ - { - "type": "text", - "text": "I'll query the database." - }, + {"type": "text", "text": "I'll query the database."}, { "type": "tool_use", "id": "toolu_123", @@ -1314,15 +1322,24 @@ def test_caller_field_in_response(): "input": {"sql": "SELECT * FROM users"}, "caller": { "type": "code_execution_20250825", - "tool_id": "srvtoolu_abc" - } - } + "tool_id": "srvtoolu_abc", + }, + }, ], "stop_reason": "tool_use", - "usage": {"input_tokens": 100, "output_tokens": 50} + "usage": {"input_tokens": 100, "output_tokens": 50}, } - text, citations, thinking, reasoning, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content(completion_response) + ( + text, + citations, + thinking, + reasoning, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) assert len(tool_calls) == 1 assert tool_calls[0]["id"] == "toolu_123" @@ -1337,10 +1354,7 @@ def test_code_execution_20250825_tool_type(): """Test that code_execution_20250825 tool type is handled correctly.""" config = AnthropicConfig() - tool = { - "type": "code_execution_20250825", - "name": "code_execution" - } + tool = {"type": "code_execution_20250825", "name": "code_execution"} transformed_tool, _ = config._map_tool_helper(tool) assert transformed_tool is not None @@ -1360,13 +1374,11 @@ def test_allowed_callers_in_function_field(): "description": "Execute a SQL query", "parameters": { "type": "object", - "properties": { - "sql": {"type": "string"} - }, - "required": ["sql"] + "properties": {"sql": {"type": "string"}}, + "required": ["sql"], }, - "allowed_callers": ["code_execution_20250825"] - } + "allowed_callers": ["code_execution_20250825"], + }, } transformed_tool, _ = config._map_tool_helper(tool) @@ -1389,15 +1401,15 @@ def test_input_examples_field_preservation(): "type": "object", "properties": { "location": {"type": "string"}, - "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]} + "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}, }, - "required": ["location"] - } + "required": ["location"], + }, }, "input_examples": [ {"location": "San Francisco, CA", "unit": "fahrenheit"}, - {"location": "Tokyo, Japan", "unit": "celsius"} - ] + {"location": "Tokyo, Japan", "unit": "celsius"}, + ], } transformed_tool, _ = config._map_tool_helper(tool_with_examples) @@ -1420,11 +1432,9 @@ def test_input_examples_beta_header(): "function": { "name": "get_weather", "description": "Get weather information", - "parameters": {"type": "object", "properties": {}} + "parameters": {"type": "object", "properties": {}}, }, - "input_examples": [ - {"location": "San Francisco, CA"} - ] + "input_examples": [{"location": "San Francisco, CA"}], } ] @@ -1433,8 +1443,7 @@ def test_input_examples_beta_header(): # Test header generation headers = model_info.get_anthropic_headers( - api_key="test-key", - input_examples_used=True + api_key="test-key", input_examples_used=True ) assert "anthropic-beta" in headers @@ -1453,16 +1462,14 @@ def test_input_examples_in_function_field(): "description": "Get weather information", "parameters": { "type": "object", - "properties": { - "location": {"type": "string"} - }, - "required": ["location"] + "properties": {"location": {"type": "string"}}, + "required": ["location"], }, "input_examples": [ {"location": "Paris, France"}, - {"location": "London, UK"} - ] - } + {"location": "London, UK"}, + ], + }, } transformed_tool, _ = config._map_tool_helper(tool) @@ -1483,17 +1490,13 @@ def test_input_examples_with_other_features(): "description": "Execute a SQL query", "parameters": { "type": "object", - "properties": { - "sql": {"type": "string"} - }, - "required": ["sql"] - } + "properties": {"sql": {"type": "string"}}, + "required": ["sql"], + }, }, - "input_examples": [ - {"sql": "SELECT * FROM users WHERE id = 1"} - ], + "input_examples": [{"sql": "SELECT * FROM users WHERE id = 1"}], "defer_loading": True, - "allowed_callers": ["code_execution_20250825"] + "allowed_callers": ["code_execution_20250825"], } transformed_tool, _ = config._map_tool_helper(tool) @@ -1517,19 +1520,20 @@ def test_input_examples_empty_list_not_added(): "description": "Get weather information", "parameters": { "type": "object", - "properties": { - "location": {"type": "string"} - }, - "required": ["location"] - } + "properties": {"location": {"type": "string"}}, + "required": ["location"], + }, }, - "input_examples": [] + "input_examples": [], } transformed_tool, _ = config._map_tool_helper(tool) assert transformed_tool is not None # Empty list should not be added - assert "input_examples" not in transformed_tool or len(transformed_tool.get("input_examples", [])) == 0 + assert ( + "input_examples" not in transformed_tool + or len(transformed_tool.get("input_examples", [])) == 0 + ) # ============ Effort Parameter Tests ============ @@ -1540,18 +1544,14 @@ def test_effort_output_config_preservation(): config = AnthropicConfig() messages = [{"role": "user", "content": "Analyze this code"}] - optional_params = { - "output_config": { - "effort": "medium" - } - } + optional_params = {"output_config": {"effort": "medium"}} result = config.transform_request( model="claude-opus-4-5-20251101", messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) assert "output_config" in result @@ -1565,18 +1565,13 @@ def test_effort_beta_header_injection(): model_info = AnthropicModelInfo() # Test with effort parameter - optional_params = { - "output_config": { - "effort": "low" - } - } + optional_params = {"output_config": {"effort": "low"}} effort_used = model_info.is_effort_used(optional_params=optional_params) assert effort_used is True headers = model_info.get_anthropic_headers( - api_key="test-key", - effort_used=effort_used + api_key="test-key", effort_used=effort_used ) assert "anthropic-beta" in headers @@ -1597,7 +1592,7 @@ def test_effort_validation(): messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) assert result["output_config"]["effort"] == effort @@ -1609,7 +1604,7 @@ def test_effort_validation(): messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) @@ -1618,18 +1613,14 @@ def test_effort_with_claude_opus_45(): config = AnthropicConfig() messages = [{"role": "user", "content": "Complex analysis task"}] - optional_params = { - "output_config": { - "effort": "high" - } - } + optional_params = {"output_config": {"effort": "high"}} result = config.transform_request( model="claude-opus-4-5-20251101", messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) assert "output_config" in result @@ -1650,7 +1641,7 @@ def test_effort_validation_with_opus_46(): messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) assert result["output_config"]["effort"] == effort @@ -1661,14 +1652,16 @@ def test_max_effort_rejected_for_opus_45(): messages = [{"role": "user", "content": "Test"}] - with pytest.raises(ValueError, match="effort='max' is only supported by Claude Opus 4.6"): + with pytest.raises( + ValueError, match="effort='max' is only supported by Claude Opus 4.6" + ): optional_params = {"output_config": {"effort": "max"}} config.transform_request( model="claude-opus-4-5-20251101", messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) @@ -1685,23 +1678,16 @@ def test_effort_with_other_features(): "description": "Get data", "parameters": { "type": "object", - "properties": { - "query": {"type": "string"} - }, - "required": ["query"] - } - } + "properties": {"query": {"type": "string"}}, + "required": ["query"], + }, + }, } ] optional_params = { - "output_config": { - "effort": "low" - }, + "output_config": {"effort": "low"}, "tools": tools, - "thinking": { - "type": "enabled", - "budget_tokens": 1000 - } + "thinking": {"type": "enabled", "budget_tokens": 1000}, } result = config.transform_request( @@ -1709,7 +1695,7 @@ def test_effort_with_other_features(): messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) # Verify all features are present @@ -1752,11 +1738,14 @@ def test_translate_system_message_skips_empty_list_content(): # Test list content with empty text block messages = [ - {"role": "system", "content": [ - {"type": "text", "text": ""}, - {"type": "text", "text": "Valid content"}, - {"type": "text", "text": ""}, - ]}, + { + "role": "system", + "content": [ + {"type": "text", "text": ""}, + {"type": "text", "text": "Valid content"}, + {"type": "text", "text": ""}, + ], + }, {"role": "user", "content": "Hello"}, ] @@ -1794,9 +1783,16 @@ def test_translate_system_message_preserves_cache_control(): # Test list content with cache_control messages = [ - {"role": "system", "content": [ - {"type": "text", "text": "Cached content", "cache_control": {"type": "ephemeral"}}, - ]}, + { + "role": "system", + "content": [ + { + "type": "text", + "text": "Cached content", + "cache_control": {"type": "ephemeral"}, + }, + ], + }, {"role": "user", "content": "Hello"}, ] @@ -1938,7 +1934,7 @@ def test_transform_request_uses_dynamic_max_tokens(): messages=messages, optional_params={}, # No max_tokens provided litellm_params={}, - headers={} + headers={}, ) assert result["max_tokens"] == 64000 @@ -1959,7 +1955,7 @@ def test_transform_request_respects_user_max_tokens(): messages=messages, optional_params={"max_tokens": 1000}, litellm_params={}, - headers={} + headers={}, ) assert result["max_tokens"] == 1000 @@ -2006,11 +2002,12 @@ def test_calculate_usage_completion_tokens_details_with_reasoning(): "output_tokens": 500, } # Simulating reasoning content that would count as ~50 tokens - reasoning_content = "Let me think about this step by step. " * 10 # Roughly 50 tokens + reasoning_content = ( + "Let me think about this step by step. " * 10 + ) # Roughly 50 tokens usage = config.calculate_usage( - usage_object=usage_object, - reasoning_content=reasoning_content + usage_object=usage_object, reasoning_content=reasoning_content ) # completion_tokens_details should be populated with both reasoning and text tokens @@ -2051,7 +2048,7 @@ def test_reasoning_effort_maps_to_adaptive_thinking_for_claude_4_6_models(): non_default_params=non_default_params, optional_params=optional_params, model=model, - drop_params=False + drop_params=False, ) # Should map to adaptive thinking type @@ -2062,7 +2059,9 @@ def test_reasoning_effort_maps_to_adaptive_thinking_for_claude_4_6_models(): # reasoning_effort should not be in the result (it's transformed to thinking) assert "reasoning_effort" not in result # Should set output_config with the mapped effort value - assert "output_config" in result, f"output_config missing for {model} with effort={effort}" + assert ( + "output_config" in result + ), f"output_config missing for {model} with effort={effort}" assert result["output_config"]["effort"] == effort_map[effort] @@ -2123,10 +2122,10 @@ def test_reasoning_effort_maps_to_budget_thinking_for_non_opus_4_6(): # Test with Claude Sonnet 4.5 (non-Opus 4.6 model) test_cases = [ - ("low", 1024), # DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET - ("medium", 2048), # DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET - ("high", 4096), # DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET - ("minimal", 128), # DEFAULT_REASONING_EFFORT_MINIMAL_THINKING_BUDGET + ("low", 1024), # DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET + ("medium", 2048), # DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET + ("high", 4096), # DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET + ("minimal", 128), # DEFAULT_REASONING_EFFORT_MINIMAL_THINKING_BUDGET ] for effort, expected_budget in test_cases: @@ -2137,7 +2136,7 @@ def test_reasoning_effort_maps_to_budget_thinking_for_non_opus_4_6(): non_default_params=non_default_params, optional_params=optional_params, model="claude-sonnet-4-5-20250929", - drop_params=False + drop_params=False, ) # Should map to enabled thinking type with budget_tokens @@ -2166,9 +2165,9 @@ def test_reasoning_effort_sets_output_config_for_46_models(): drop_params=False, ) - assert "output_config" in result, ( - f"output_config missing for {model} with effort={effort}" - ) + assert ( + "output_config" in result + ), f"output_config missing for {model} with effort={effort}" assert result["output_config"]["effort"] == effort @@ -2207,9 +2206,9 @@ def test_reasoning_effort_does_not_set_output_config_for_older_models(): drop_params=False, ) - assert "output_config" not in result, ( - f"output_config should not be set for {model}" - ) + assert ( + "output_config" not in result + ), f"output_config should not be set for {model}" def test_max_effort_rejected_for_sonnet_46(): @@ -2217,7 +2216,9 @@ def test_max_effort_rejected_for_sonnet_46(): config = AnthropicConfig() messages = [{"role": "user", "content": "Test"}] - with pytest.raises(ValueError, match="effort='max' is only supported by Claude Opus 4.6"): + with pytest.raises( + ValueError, match="effort='max' is only supported by Claude Opus 4.6" + ): config.transform_request( model="claude-sonnet-4-6-20260219", messages=messages, @@ -2260,9 +2261,7 @@ def test_effort_beta_header_not_injected_for_46_models(): optional_params={"output_config": {"effort": "high"}}, model=model, ) - assert result is False, ( - f"is_effort_used should return False for {model}" - ) + assert result is False, f"is_effort_used should return False for {model}" def test_effort_beta_header_still_injected_for_older_models(): @@ -2302,17 +2301,12 @@ def test_code_execution_tool_results_extraction(): "role": "assistant", "model": "claude-sonnet-4-5-20250929", "content": [ - { - "type": "text", - "text": "I'll calculate that for you." - }, + {"type": "text", "text": "I'll calculate that for you."}, { "type": "server_tool_use", "id": "srvtoolu_01ABC", "name": "bash_code_execution", - "input": { - "command": "python3 << 'EOF'\nprint(2 + 2)\nEOF\n" - } + "input": {"command": "python3 << 'EOF'\nprint(2 + 2)\nEOF\n"}, }, { "type": "bash_code_execution_tool_result", @@ -2321,8 +2315,8 @@ def test_code_execution_tool_results_extraction(): "type": "bash_code_execution_result", "stdout": "4\n", "stderr": "", - "return_code": 0 - } + "return_code": 0, + }, }, { "type": "server_tool_use", @@ -2331,28 +2325,22 @@ def test_code_execution_tool_results_extraction(): "input": { "command": "create", "path": "test.txt", - "file_text": "Hello" - } + "file_text": "Hello", + }, }, { "type": "text_editor_code_execution_tool_result", "tool_use_id": "srvtoolu_01DEF", "content": { "type": "text_editor_code_execution_result", - "is_file_update": False - } + "is_file_update": False, + }, }, - { - "type": "text", - "text": "Done!" - } + {"type": "text", "text": "Done!"}, ], "stop_reason": "stop", "stop_sequence": None, - "usage": { - "input_tokens": 100, - "output_tokens": 50 - } + "usage": {"input_tokens": 100, "output_tokens": 50}, } # Create mock HTTP response @@ -2377,11 +2365,17 @@ def test_code_execution_tool_results_extraction(): # Verify first tool call assert transformed_response.choices[0].message.tool_calls[0].id == "srvtoolu_01ABC" - assert transformed_response.choices[0].message.tool_calls[0].function.name == "bash_code_execution" + assert ( + transformed_response.choices[0].message.tool_calls[0].function.name + == "bash_code_execution" + ) # Verify second tool call assert transformed_response.choices[0].message.tool_calls[1].id == "srvtoolu_01DEF" - assert transformed_response.choices[0].message.tool_calls[1].function.name == "text_editor_code_execution" + assert ( + transformed_response.choices[0].message.tool_calls[1].function.name + == "text_editor_code_execution" + ) # Verify tool results are in provider_specific_fields provider_fields = transformed_response.choices[0].message.provider_specific_fields @@ -2404,10 +2398,83 @@ def test_code_execution_tool_results_extraction(): assert editor_result["content"]["is_file_update"] is False # Verify text content is properly concatenated - assert "I'll calculate that for you." in transformed_response.choices[0].message.content + assert ( + "I'll calculate that for you." + in transformed_response.choices[0].message.content + ) assert "Done!" in transformed_response.choices[0].message.content +def test_code_execution_tool_results_in_hidden_params(): + """ + Test that tool_results reaches _hidden_params so the Responses API adapter + can surface them via provider_specific_fields. + + The Responses API adapter reads _hidden_params.get("provider_specific_fields") + to set provider_specific_fields on the response. Without this, server-side + code execution results (stdout/stderr) are lost when using responses.create(). + """ + import httpx + + from litellm.types.utils import ModelResponse + + config = AnthropicConfig() + + mock_anthropic_response = { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "model": "claude-sonnet-4-5-20250929", + "content": [ + {"type": "text", "text": "Here's the result."}, + { + "type": "server_tool_use", + "id": "srvtoolu_01ABC", + "name": "bash_code_execution", + "input": {"command": "echo hello"}, + }, + { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01ABC", + "content": { + "type": "bash_code_execution_result", + "stdout": "hello\n", + "stderr": "", + "return_code": 0, + }, + }, + ], + "stop_reason": "stop", + "stop_sequence": None, + "usage": {"input_tokens": 100, "output_tokens": 50}, + } + + mock_raw_response = MagicMock(spec=httpx.Response) + mock_raw_response.json.return_value = mock_anthropic_response + mock_raw_response.status_code = 200 + mock_raw_response.headers = {} + + model_response = ModelResponse() + + transformed_response = config.transform_parsed_response( + completion_response=mock_anthropic_response, + raw_response=mock_raw_response, + model_response=model_response, + json_mode=False, + prefix_prompt=None, + ) + + # Verify tool_results is in _hidden_params for the Responses API adapter + hidden = transformed_response._hidden_params + assert "provider_specific_fields" in hidden + assert "tool_results" in hidden["provider_specific_fields"] + assert len(hidden["provider_specific_fields"]["tool_results"]) == 1 + assert ( + hidden["provider_specific_fields"]["tool_results"][0]["content"]["stdout"] + == "hello\n" + ) + + def test_tool_search_tool_result_not_in_tool_results(): """ Test that tool_search_tool_result is NOT included in tool_results @@ -2425,21 +2492,12 @@ def test_tool_search_tool_result_not_in_tool_results(): "role": "assistant", "model": "claude-sonnet-4-5-20250929", "content": [ - { - "type": "text", - "text": "Found tools." - }, - { - "type": "tool_search_tool_result", - "tool_references": ["tool1", "tool2"] - } + {"type": "text", "text": "Found tools."}, + {"type": "tool_search_tool_result", "tool_references": ["tool1", "tool2"]}, ], "stop_reason": "stop", "stop_sequence": None, - "usage": { - "input_tokens": 100, - "output_tokens": 50 - } + "usage": {"input_tokens": 100, "output_tokens": 50}, } mock_raw_response = MagicMock(spec=httpx.Response) @@ -2479,22 +2537,16 @@ def test_web_search_tool_result_backwards_compatibility(): "role": "assistant", "model": "claude-sonnet-4-5-20250929", "content": [ - { - "type": "text", - "text": "Here are the results." - }, + {"type": "text", "text": "Here are the results."}, { "type": "web_search_tool_result", "search_query": "test query", - "results": [{"title": "Result 1", "url": "https://example.com"}] - } + "results": [{"title": "Result 1", "url": "https://example.com"}], + }, ], "stop_reason": "stop", "stop_sequence": None, - "usage": { - "input_tokens": 100, - "output_tokens": 50 - } + "usage": {"input_tokens": 100, "output_tokens": 50}, } mock_raw_response = MagicMock(spec=httpx.Response) @@ -2540,24 +2592,28 @@ def test_compaction_block_extraction(): "content": [ { "type": "compaction", - "content": "Summary of the conversation: The user requested help building a web scraper..." + "content": "Summary of the conversation: The user requested help building a web scraper...", }, { "type": "text", - "text": "I don't have access to real-time data, so I can't provide the current weather in San Francisco." - } + "text": "I don't have access to real-time data, so I can't provide the current weather in San Francisco.", + }, ], "stop_reason": "max_tokens", "stop_sequence": None, - "usage": { - "input_tokens": 86, - "output_tokens": 100 - } + "usage": {"input_tokens": 86, "output_tokens": 100}, } - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) # Verify compaction blocks are extracted assert compaction_blocks is not None @@ -2587,18 +2643,12 @@ def test_compaction_block_in_provider_specific_fields(): "content": [ { "type": "compaction", - "content": "Summary of the conversation: The user requested help building a web scraper..." + "content": "Summary of the conversation: The user requested help building a web scraper...", }, - { - "type": "text", - "text": "Here is the response." - } + {"type": "text", "text": "Here is the response."}, ], "stop_reason": "end_turn", - "usage": { - "input_tokens": 50, - "output_tokens": 25 - } + "usage": {"input_tokens": 50, "output_tokens": 25}, } raw_response = httpx.Response(status_code=200, headers={}) @@ -2618,7 +2668,10 @@ def test_compaction_block_in_provider_specific_fields(): assert "compaction_blocks" in provider_fields assert len(provider_fields["compaction_blocks"]) == 1 assert provider_fields["compaction_blocks"][0]["type"] == "compaction" - assert "Summary of the conversation" in provider_fields["compaction_blocks"][0]["content"] + assert ( + "Summary of the conversation" + in provider_fields["compaction_blocks"][0]["content"] + ) def test_multiple_compaction_blocks(): @@ -2629,24 +2682,22 @@ def test_multiple_compaction_blocks(): completion_response = { "content": [ - { - "type": "compaction", - "content": "First summary..." - }, - { - "type": "text", - "text": "Some text." - }, - { - "type": "compaction", - "content": "Second summary..." - } + {"type": "compaction", "content": "First summary..."}, + {"type": "text", "text": "Some text."}, + {"type": "compaction", "content": "Second summary..."}, ] } - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) # Verify both compaction blocks are extracted assert compaction_blocks is not None @@ -2665,37 +2716,26 @@ def test_compaction_block_request_transformation(): ) messages = [ - { - "role": "user", - "content": "What is the weather in San Francisco?" - }, + {"role": "user", "content": "What is the weather in San Francisco?"}, { "role": "assistant", "content": [ - { - "type": "text", - "text": "I don't have access to real-time data." - } + {"type": "text", "text": "I don't have access to real-time data."} ], "provider_specific_fields": { "compaction_blocks": [ { "type": "compaction", - "content": "Summary of the conversation: The user requested help building a web scraper..." + "content": "Summary of the conversation: The user requested help building a web scraper...", } ] - } + }, }, - { - "role": "user", - "content": "What about New York?" - } + {"role": "user", "content": "What about New York?"}, ] result = anthropic_messages_pt( - messages=messages, - model="claude-opus-4-6", - llm_provider="anthropic" + messages=messages, model="claude-opus-4-6", llm_provider="anthropic" ) # Find the assistant message @@ -2727,14 +2767,8 @@ def test_compaction_with_context_management(): messages = [{"role": "user", "content": "Hello"}] optional_params = { - "context_management": { - "edits": [ - { - "type": "compact_20260112" - } - ] - }, - "max_tokens": 100 + "context_management": {"edits": [{"type": "compact_20260112"}]}, + "max_tokens": 100, } result = config.transform_request( @@ -2742,7 +2776,7 @@ def test_compaction_with_context_management(): messages=messages, optional_params=optional_params, litellm_params={}, - headers={} + headers={}, ) # Verify context_management is included @@ -2758,30 +2792,28 @@ def test_compaction_block_with_other_content_types(): completion_response = { "content": [ - { - "type": "compaction", - "content": "Summary of previous conversation..." - }, - { - "type": "thinking", - "thinking": "Let me think about this..." - }, - { - "type": "text", - "text": "Based on my analysis..." - }, + {"type": "compaction", "content": "Summary of previous conversation..."}, + {"type": "thinking", "thinking": "Let me think about this..."}, + {"type": "text", "text": "Based on my analysis..."}, { "type": "tool_use", "id": "toolu_123", "name": "get_weather", - "input": {"location": "San Francisco"} - } + "input": {"location": "San Francisco"}, + }, ] } - text, citations, thinking_blocks, reasoning_content, tool_calls, web_search_results, tool_results, compaction_blocks = config.extract_response_content( - completion_response - ) + ( + text, + citations, + thinking_blocks, + reasoning_content, + tool_calls, + web_search_results, + tool_results, + compaction_blocks, + ) = config.extract_response_content(completion_response) # Verify all content types are extracted assert compaction_blocks is not None @@ -2798,11 +2830,11 @@ def test_map_openai_context_management_to_anthropic(): Test mapping OpenAI Responses API context_management format to Anthropic format. """ config = AnthropicConfig() - + # Test OpenAI list format with compaction openai_format = [{"type": "compaction", "compact_threshold": 200000}] result = config.map_openai_context_management_to_anthropic(openai_format) - + assert result is not None assert "edits" in result assert len(result["edits"]) == 1 @@ -2811,26 +2843,32 @@ def test_map_openai_context_management_to_anthropic(): assert result["edits"][0]["trigger"]["value"] == 200000 # Test OpenAI format with instructions - openai_format_with_instructions = [{ - "type": "compaction", - "compact_threshold": 150000, - "instructions": "Focus on preserving code snippets" - }] - result = config.map_openai_context_management_to_anthropic(openai_format_with_instructions) - + openai_format_with_instructions = [ + { + "type": "compaction", + "compact_threshold": 150000, + "instructions": "Focus on preserving code snippets", + } + ] + result = config.map_openai_context_management_to_anthropic( + openai_format_with_instructions + ) + assert result is not None assert result["edits"][0]["trigger"]["value"] == 150000 assert result["edits"][0]["instructions"] == "Focus on preserving code snippets" - + # Test Anthropic format (should pass through) anthropic_format = { - "edits": [{ - "type": "compact_20260112", - "trigger": {"type": "input_tokens", "value": 150000} - }] + "edits": [ + { + "type": "compact_20260112", + "trigger": {"type": "input_tokens", "value": 150000}, + } + ] } result = config.map_openai_context_management_to_anthropic(anthropic_format) - + assert result == anthropic_format @@ -2839,46 +2877,51 @@ def test_map_openai_params_with_context_management(): Test that map_openai_params correctly transforms context_management from OpenAI to Anthropic format. """ config = AnthropicConfig() - + # Test with OpenAI list format non_default_params = { "context_management": [{"type": "compaction", "compact_threshold": 200000}] } optional_params = {} - + result = config.map_openai_params( non_default_params=non_default_params, optional_params=optional_params, model="claude-opus-4-6", - drop_params=False + drop_params=False, ) - + assert "context_management" in result assert "edits" in result["context_management"] assert result["context_management"]["edits"][0]["type"] == "compact_20260112" assert result["context_management"]["edits"][0]["trigger"]["value"] == 200000 - + # Test with Anthropic dict format (should pass through) non_default_params_anthropic = { "context_management": { - "edits": [{ - "type": "compact_20260112", - "trigger": {"type": "input_tokens", "value": 150000}, - "instructions": "Focus on preserving code" - }] + "edits": [ + { + "type": "compact_20260112", + "trigger": {"type": "input_tokens", "value": 150000}, + "instructions": "Focus on preserving code", + } + ] } } optional_params = {} - + result = config.map_openai_params( non_default_params=non_default_params_anthropic, optional_params=optional_params, model="claude-opus-4-6", - drop_params=False + drop_params=False, ) - + assert "context_management" in result - assert result["context_management"] == non_default_params_anthropic["context_management"] + assert ( + result["context_management"] + == non_default_params_anthropic["context_management"] + ) def test_cache_control_in_supported_params(): @@ -2897,9 +2940,7 @@ def test_map_openai_params_with_cache_control(): """ config = AnthropicConfig() - non_default_params = { - "cache_control": {"type": "ephemeral"} - } + non_default_params = {"cache_control": {"type": "ephemeral"}} optional_params = {} result = config.map_openai_params( @@ -2919,9 +2960,7 @@ def test_map_openai_params_cache_control_ignored_when_not_dict(): """ config = AnthropicConfig() - non_default_params = { - "cache_control": "ephemeral" - } + non_default_params = {"cache_control": "ephemeral"} optional_params = {} result = config.map_openai_params( @@ -2974,17 +3013,9 @@ def test_compaction_block_empty_list_not_added(): "type": "message", "role": "assistant", "model": "claude-opus-4-6", - "content": [ - { - "type": "text", - "text": "Just a regular response." - } - ], + "content": [{"type": "text", "text": "Just a regular response."}], "stop_reason": "end_turn", - "usage": { - "input_tokens": 10, - "output_tokens": 5 - } + "usage": {"input_tokens": 10, "output_tokens": 5}, } raw_response = httpx.Response(status_code=200, headers={}) @@ -3001,7 +3032,10 @@ def test_compaction_block_empty_list_not_added(): # Verify compaction_blocks is not in provider_specific_fields when there are none provider_fields = result.choices[0].message.provider_specific_fields if provider_fields: - assert "compaction_blocks" not in provider_fields or provider_fields.get("compaction_blocks") is None + assert ( + "compaction_blocks" not in provider_fields + or provider_fields.get("compaction_blocks") is None + ) def test_fast_mode_beta_header(): @@ -3014,8 +3048,7 @@ def test_fast_mode_beta_header(): optional_params = {"speed": "fast"} result_headers = config.update_headers_with_optional_anthropic_beta( - headers=headers, - optional_params=optional_params + headers=headers, optional_params=optional_params ) assert "anthropic-beta" in result_headers @@ -3029,14 +3062,10 @@ def test_fast_mode_with_other_beta_headers(): config = AnthropicConfig() headers = {} - optional_params = { - "speed": "fast", - "output_format": {"type": "json_object"} - } + optional_params = {"speed": "fast", "output_format": {"type": "json_object"}} result_headers = config.update_headers_with_optional_anthropic_beta( - headers=headers, - optional_params=optional_params + headers=headers, optional_params=optional_params ) assert "anthropic-beta" in result_headers @@ -3056,9 +3085,7 @@ def test_fast_mode_usage_calculation(): } usage = config.calculate_usage( - usage_object=usage_object, - reasoning_content=None, - speed="fast" + usage_object=usage_object, reasoning_content=None, speed="fast" ) assert usage.prompt_tokens == 1000 @@ -3171,7 +3198,7 @@ def test_fast_mode_parameter_mapping(): non_default_params=non_default_params, optional_params=optional_params, model="claude-opus-4-6", - drop_params=False + drop_params=False, ) assert "speed" in result @@ -3236,9 +3263,9 @@ def test_map_tool_helper_enforces_object_type_when_missing(): assert "properties" in result["input_schema"] assert "query" in result["input_schema"]["properties"] # Original parameters dict must not be modified in place - assert tool["function"]["parameters"] == original_params, ( - "parameters dict was mutated; _map_tool_helper should not modify caller data" - ) + assert ( + tool["function"]["parameters"] == original_params + ), "parameters dict was mutated; _map_tool_helper should not modify caller data" def test_map_tool_helper_enforces_object_type_when_wrong_type(): @@ -3264,13 +3291,13 @@ def test_map_tool_helper_enforces_object_type_when_wrong_type(): result, _ = config._map_tool_helper(tool) assert result is not None assert result["input_schema"]["type"] == "object" - assert result["input_schema"].get("properties") == {}, ( - "properties should be injected as {} when schema has non-object type and no properties key" - ) + assert ( + result["input_schema"].get("properties") == {} + ), "properties should be injected as {} when schema has non-object type and no properties key" # Original parameters dict must not be modified in place - assert tool["function"]["parameters"] == original_params, ( - "parameters dict was mutated; _map_tool_helper should not modify caller data" - ) + assert ( + tool["function"]["parameters"] == original_params + ), "parameters dict was mutated; _map_tool_helper should not modify caller data" def test_map_tool_helper_preserves_valid_object_schema(): @@ -3321,3 +3348,54 @@ def test_map_tool_helper_empty_parameters_get_default(): assert result is not None assert result["input_schema"]["type"] == "object" assert result["input_schema"].get("properties") == {} + + +def test_extract_response_content_thinking_block_null_thinking(): + """ + Test that thinking blocks are not dropped when the 'thinking' field is null + or missing. Regression test for https://github.com/BerriAI/litellm/issues/24026 + """ + config = AnthropicConfig() + + # Case 1: thinking key is explicitly null + completion_response_null = { + "content": [ + {"type": "thinking", "thinking": None, "signature": "sig123"}, + {"type": "text", "text": "Hello"}, + ] + } + text, _, thinking_blocks, _, _, _, _, _ = config.extract_response_content( + completion_response_null + ) + assert thinking_blocks is not None, "thinking blocks should not be None when thinking=null" + assert len(thinking_blocks) == 1 + assert "Hello" in text + + # Case 2: thinking key is absent entirely + completion_response_missing = { + "content": [ + {"type": "thinking", "signature": "sig456"}, + {"type": "text", "text": "World"}, + ] + } + text, _, thinking_blocks, _, _, _, _, _ = config.extract_response_content( + completion_response_missing + ) + assert thinking_blocks is not None, "thinking blocks should not be None when thinking key is absent" + assert len(thinking_blocks) == 1 + assert "World" in text + + # Case 3: thinking key has actual content (should still work) + completion_response_text = { + "content": [ + {"type": "thinking", "thinking": "Let me think...", "signature": "sig789"}, + {"type": "text", "text": "Done"}, + ] + } + text, _, thinking_blocks, _, _, _, _, _ = config.extract_response_content( + completion_response_text + ) + assert thinking_blocks is not None + assert len(thinking_blocks) == 1 + assert thinking_blocks[0]["thinking"] == "Let me think..." + assert "Done" in text diff --git a/tests/test_litellm/llms/anthropic/chat/test_code_interpreter_results_extraction.py b/tests/test_litellm/llms/anthropic/chat/test_code_interpreter_results_extraction.py new file mode 100644 index 00000000000..60e45c9b8ce --- /dev/null +++ b/tests/test_litellm/llms/anthropic/chat/test_code_interpreter_results_extraction.py @@ -0,0 +1,268 @@ +""" +Tests for the Responses API _extract_tool_result_output_items path, +the non-streaming _hidden_params propagation of code_interpreter_results, +and mock end-to-end streaming integration. +""" + +from unittest.mock import MagicMock + +from litellm.llms.anthropic.chat.handler import ModelResponseIterator +from litellm.main import stream_chunk_builder +from litellm.responses.litellm_completion_transformation.transformation import ( + LiteLLMCompletionResponsesConfig, +) +from litellm.types.responses.main import ( + OutputCodeInterpreterCall, + OutputCodeInterpreterCallLog, +) +from litellm.types.utils import Choices, Message, ModelResponse + + +def _make_model_response(code_interpreter_results=None, provider_specific_fields=None): + """Helper to build a ModelResponse with provider_specific_fields on the message.""" + psf = provider_specific_fields or {} + if code_interpreter_results is not None: + psf["code_interpreter_results"] = code_interpreter_results + msg = Message(content="test", provider_specific_fields=psf if psf else None) + choice = Choices(index=0, message=msg, finish_reason="stop") + resp = ModelResponse() + resp.choices = [choice] + return resp + + +def test_extract_tool_result_output_items_from_pydantic_objects(): + """Non-streaming path: code_interpreter_results are Pydantic OutputCodeInterpreterCall objects.""" + items = [ + OutputCodeInterpreterCall( + type="code_interpreter_call", + id="srvtoolu_01AAA", + code="echo hello", + container_id=None, + status="completed", + outputs=[OutputCodeInterpreterCallLog(type="logs", logs="hello\n")], + ), + OutputCodeInterpreterCall( + type="code_interpreter_call", + id="srvtoolu_01BBB", + code="echo world", + container_id=None, + status="completed", + outputs=[OutputCodeInterpreterCallLog(type="logs", logs="world\n")], + ), + ] + resp = _make_model_response(code_interpreter_results=items) + result = LiteLLMCompletionResponsesConfig._extract_tool_result_output_items(resp) + assert len(result) == 2 + assert result[0].id == "srvtoolu_01AAA" + assert result[1].id == "srvtoolu_01BBB" + + +def test_extract_tool_result_output_items_from_dicts(): + """Streaming path: after model_dump(), code_interpreter_results are plain dicts. + _extract_tool_result_output_items reconstructs them as Pydantic objects.""" + items = [ + { + "type": "code_interpreter_call", + "id": "srvtoolu_01AAA", + "code": "echo hello", + "container_id": None, + "status": "completed", + "outputs": [{"type": "logs", "logs": "hello\n"}], + }, + ] + resp = _make_model_response(code_interpreter_results=items) + result = LiteLLMCompletionResponsesConfig._extract_tool_result_output_items(resp) + assert len(result) == 1 + assert isinstance(result[0], OutputCodeInterpreterCall) + assert result[0].id == "srvtoolu_01AAA" + + +def test_extract_tool_result_output_items_empty(): + """No code_interpreter_results → empty list.""" + resp = _make_model_response() + result = LiteLLMCompletionResponsesConfig._extract_tool_result_output_items(resp) + assert result == [] + + +def test_extract_tool_result_output_items_no_provider_specific_fields(): + """Message with no provider_specific_fields → empty list.""" + msg = Message(content="test") + choice = Choices(index=0, message=msg, finish_reason="stop") + resp = ModelResponse() + resp.choices = [choice] + result = LiteLLMCompletionResponsesConfig._extract_tool_result_output_items(resp) + assert result == [] + + +def test_in_place_substitution_preserves_ordering(): + """ + function_call items matching code_interpreter_results should be replaced + in-place, preserving the original output ordering. + + Simulates: [message, function_call(exec1), function_call(regular), function_call(exec2)] + Expected: [message, code_interpreter_call(exec1), function_call(regular), code_interpreter_call(exec2)] + """ + code_results = [ + OutputCodeInterpreterCall( + type="code_interpreter_call", + id="srvtoolu_01AAA", + code="echo first", + container_id=None, + status="completed", + outputs=[OutputCodeInterpreterCallLog(type="logs", logs="first\n")], + ), + OutputCodeInterpreterCall( + type="code_interpreter_call", + id="srvtoolu_01CCC", + code="echo third", + container_id=None, + status="completed", + outputs=[OutputCodeInterpreterCallLog(type="logs", logs="third\n")], + ), + ] + resp = _make_model_response(code_interpreter_results=code_results) + + # Build a mock responses_output list with interleaved items + class MockItem: + def __init__(self, type, call_id=None): + self.type = type + self.call_id = call_id + + msg_item = MockItem(type="message") + fc_exec1 = MockItem(type="function_call", call_id="srvtoolu_01AAA") + fc_regular = MockItem(type="function_call", call_id="srvtoolu_01BBB") + fc_exec2 = MockItem(type="function_call", call_id="srvtoolu_01CCC") + + responses_output = [msg_item, fc_exec1, fc_regular, fc_exec2] + + # Apply the same logic as _transform_chat_completion_choices_to_responses_output + tool_result_items = ( + LiteLLMCompletionResponsesConfig._extract_tool_result_output_items(resp) + ) + if tool_result_items: + result_by_id = { + (item.get("id") if isinstance(item, dict) else item.id): item + for item in tool_result_items + } + replaced_ids = set(result_by_id.keys()) + responses_output = [ + ( + result_by_id[getattr(item, "call_id", None)] + if ( + getattr(item, "type", None) == "function_call" + and getattr(item, "call_id", None) in replaced_ids + ) + else item + ) + for item in responses_output + ] + + # Verify ordering: message, code_interpreter(AAA), function_call(BBB), code_interpreter(CCC) + assert len(responses_output) == 4 + assert responses_output[0].type == "message" + assert responses_output[1].type == "code_interpreter_call" + assert responses_output[1].id == "srvtoolu_01AAA" + assert responses_output[2].type == "function_call" + assert responses_output[2].call_id == "srvtoolu_01BBB" + assert responses_output[3].type == "code_interpreter_call" + assert responses_output[3].id == "srvtoolu_01CCC" + + +def test_end_to_end_streaming_chunks_to_code_interpreter_output(): + """ + Mock end-to-end test: Anthropic SSE chunks → ModelResponseIterator → + stream_chunk_builder → _extract_tool_result_output_items → final output + with code_interpreter_call items replacing function_call items. + + This exercises the full streaming data flow without a live server. + """ + # Realistic Anthropic streaming chunks for a single code execution + raw_chunks = [ + { + "type": "message_start", + "message": { + "id": "msg_01XYZ", + "type": "message", + "role": "assistant", + "content": [], + "usage": {"input_tokens": 100, "output_tokens": 1}, + }, + }, + { + "type": "content_block_start", + "index": 0, + "content_block": { + "type": "server_tool_use", + "id": "srvtoolu_01AAA", + "name": "bash_code_execution", + "input": {}, + }, + }, + { + "type": "content_block_delta", + "index": 0, + "delta": { + "type": "input_json_delta", + "partial_json": '{"command": "echo e2e_test"}', + }, + }, + {"type": "content_block_stop", "index": 0}, + { + "type": "content_block_start", + "index": 1, + "content_block": { + "type": "bash_code_execution_tool_result", + "tool_use_id": "srvtoolu_01AAA", + "content": { + "type": "bash_code_execution_result", + "stdout": "e2e_test\n", + "stderr": "", + "return_code": 0, + }, + }, + }, + {"type": "content_block_stop", "index": 1}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn"}, + "usage": {"output_tokens": 50}, + }, + ] + + # Step 1: Parse chunks through ModelResponseIterator (Anthropic handler) + iterator = ModelResponseIterator(None, sync_stream=True) + parsed_chunks = [] + for chunk in raw_chunks: + parsed = iterator.chunk_parser(chunk) + d = parsed.model_dump() + # In production, CustomStreamWrapper sets the model on each chunk; + # stream_chunk_builder requires it. + d["model"] = "claude-sonnet-4-20250514" + parsed_chunks.append(d) + + # Step 2: Assemble via stream_chunk_builder (simulates end-of-stream) + assembled = stream_chunk_builder(chunks=parsed_chunks) + assert assembled is not None + + # Verify stream_chunk_builder picked up code_interpreter_results via last-value-wins + psf = assembled.choices[0].message.provider_specific_fields + assert psf is not None + assert "code_interpreter_results" in psf + code_results = psf["code_interpreter_results"] + assert len(code_results) == 1 + # After model_dump + stream_chunk_builder, results are plain dicts + assert code_results[0]["id"] == "srvtoolu_01AAA" + assert code_results[0]["code"] == "echo e2e_test" + + # Step 3: Extract via _extract_tool_result_output_items (Responses API layer) + tool_result_items = ( + LiteLLMCompletionResponsesConfig._extract_tool_result_output_items(assembled) + ) + assert len(tool_result_items) == 1 + item = tool_result_items[0] + # Items are reconstructed as Pydantic OutputCodeInterpreterCall objects + assert isinstance(item, OutputCodeInterpreterCall) + assert item.type == "code_interpreter_call" + assert item.id == "srvtoolu_01AAA" + assert item.code == "echo e2e_test" + assert item.outputs[0].logs == "e2e_test\n" diff --git a/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py b/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py index b4f2629f8f4..22470b93540 100644 --- a/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py +++ b/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py @@ -1,20 +1,27 @@ """ -Tests for Anthropic OAuth token handling in common_utils. +Tests for Anthropic authentication and environment variable handling in common_utils. -Verifies that OAuth tokens (sk-ant-oat*) are sent via Authorization: Bearer -instead of x-api-key, per Anthropic's OAuth specification. +Verifies that: +- OAuth tokens (sk-ant-oat*) produce Authorization: Bearer headers with OAuth beta flags. +- Regular API keys produce x-api-key headers. +- ANTHROPIC_AUTH_TOKEN produces Authorization: Bearer headers, + matching the official Anthropic SDK behavior. +- ANTHROPIC_BASE_URL is used as a fallback for base URL resolution. +- ANTHROPIC_API_KEY / ANTHROPIC_API_BASE take precedence over their aliases. """ import os import sys +from unittest.mock import patch sys.path.insert( 0, os.path.abspath(os.path.join(os.path.dirname(__file__), "../../../../..")) ) -# Fake OAuth token for testing (not a real secret) +# Fake tokens for testing (not real secrets) FAKE_OAUTH_TOKEN = "sk-ant-oat01-fake-token-for-testing-123456789abcdef" FAKE_REGULAR_KEY = "sk-ant-api03-regular-key-for-testing-123456789" +FAKE_AUTH_TOKEN = "sk-ant-aut01-fake-auth-token-for-testing-123456789" class TestOptionallyHandleAnthropicOAuth: @@ -697,3 +704,430 @@ class TestProxyOAuthHeaderForwarding: assert cleaned["authorization"] == oauth_token # Proxy key must be stripped assert "x-litellm-api-key" not in cleaned + + +class TestGetAnthropicHeadersWithAuthToken: + """Tests for get_anthropic_headers with auth_token parameter.""" + + def test_auth_token_uses_bearer_header(self): + """auth_token should produce Authorization: Bearer header.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + headers = config.get_anthropic_headers( + api_key=None, + auth_token=FAKE_AUTH_TOKEN, + computer_tool_used=False, + prompt_caching_set=False, + pdf_used=False, + is_vertex_request=False, + ) + + assert headers["authorization"] == f"Bearer {FAKE_AUTH_TOKEN}" + assert "x-api-key" not in headers + # auth_token should NOT set OAuth-specific flags + assert "anthropic-dangerous-direct-browser-access" not in headers + + def test_auth_token_includes_standard_headers(self): + """auth_token path should include standard Anthropic headers.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + headers = config.get_anthropic_headers( + api_key=None, + auth_token=FAKE_AUTH_TOKEN, + computer_tool_used=False, + prompt_caching_set=False, + pdf_used=False, + is_vertex_request=False, + ) + + assert headers["anthropic-version"] == "2023-06-01" + assert headers["accept"] == "application/json" + assert headers["content-type"] == "application/json" + + def test_api_key_takes_precedence_over_auth_token(self): + """When both api_key and auth_token are provided, api_key wins.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + headers = config.get_anthropic_headers( + api_key=FAKE_REGULAR_KEY, + auth_token=FAKE_AUTH_TOKEN, + computer_tool_used=False, + prompt_caching_set=False, + pdf_used=False, + is_vertex_request=False, + ) + + assert headers["x-api-key"] == FAKE_REGULAR_KEY + assert "authorization" not in headers + + +class TestValidateEnvironmentAuthToken: + """Tests for validate_environment with auth_token resolution.""" + + def test_auth_token_env_var_produces_bearer_header(self): + """validate_environment should use Bearer auth when only ANTHROPIC_AUTH_TOKEN is set.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, + clear=True, + ): + headers = config.validate_environment( + headers={}, + model="claude-sonnet-4-5-20250929", + messages=[{"role": "user", "content": "Hello"}], + optional_params={}, + litellm_params={}, + api_key=None, + api_base=None, + ) + + assert headers["authorization"] == f"Bearer {FAKE_AUTH_TOKEN}" + assert "x-api-key" not in headers + assert "anthropic-dangerous-direct-browser-access" not in headers + + def test_api_key_param_takes_precedence_over_auth_token_env_var(self): + """validate_environment should prefer explicit api_key over ANTHROPIC_AUTH_TOKEN.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, + clear=True, + ): + headers = config.validate_environment( + headers={}, + model="claude-sonnet-4-5-20250929", + messages=[{"role": "user", "content": "Hello"}], + optional_params={}, + litellm_params={}, + api_key=FAKE_REGULAR_KEY, + api_base=None, + ) + + assert headers["x-api-key"] == FAKE_REGULAR_KEY + assert "authorization" not in headers + + def test_raises_when_no_credentials(self): + """validate_environment should raise when neither API key nor auth token is available.""" + from unittest.mock import patch as mock_patch + + import pytest + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + with mock_patch.dict("os.environ", {}, clear=True): + with pytest.raises( + Exception, match="ANTHROPIC_API_KEY.*ANTHROPIC_AUTH_TOKEN" + ): + config.validate_environment( + headers={}, + model="claude-sonnet-4-5-20250929", + messages=[{"role": "user", "content": "Hello"}], + optional_params={}, + litellm_params={}, + api_key=None, + api_base=None, + ) + + def test_resolves_api_key_from_env_when_param_is_none(self): + """validate_environment should resolve ANTHROPIC_API_KEY from env when api_key param is None.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + config = AnthropicModelInfo() + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_API_KEY": FAKE_REGULAR_KEY}, + clear=True, + ): + headers = config.validate_environment( + headers={}, + model="claude-sonnet-4-5-20250929", + messages=[{"role": "user", "content": "Hello"}], + optional_params={}, + litellm_params={}, + api_key=None, + api_base=None, + ) + + assert headers["x-api-key"] == FAKE_REGULAR_KEY + assert "authorization" not in headers + + + + +class TestGetAuthToken: + """Tests for AnthropicModelInfo.get_auth_token() static method.""" + + def test_returns_env_var_value(self): + """get_auth_token returns the ANTHROPIC_AUTH_TOKEN env var value.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", {"ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, clear=True + ): + assert AnthropicModelInfo.get_auth_token() == FAKE_AUTH_TOKEN + + def test_returns_none_when_not_set(self): + """get_auth_token returns None when ANTHROPIC_AUTH_TOKEN is not set.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict("os.environ", {}, clear=True): + assert AnthropicModelInfo.get_auth_token() is None + + def test_explicit_param_takes_precedence(self): + """Explicit auth_token param takes precedence over env var.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + explicit_token = "sk-ant-aut01-explicit-token-override-123456789" + assert AnthropicModelInfo.get_auth_token(explicit_token) == explicit_token + + +class TestGetAuthHeader: + """Tests for AnthropicModelInfo.get_auth_header() centralized helper.""" + + def test_returns_x_api_key_when_api_key_provided(self): + """Explicit api_key param should return x-api-key header.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + result = AnthropicModelInfo.get_auth_header(api_key=FAKE_REGULAR_KEY) + assert result == {"x-api-key": FAKE_REGULAR_KEY} + + def test_returns_x_api_key_from_env(self): + """ANTHROPIC_API_KEY env var should return x-api-key header.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_API_KEY": FAKE_REGULAR_KEY}, + clear=True, + ): + result = AnthropicModelInfo.get_auth_header() + assert result == {"x-api-key": FAKE_REGULAR_KEY} + + def test_returns_bearer_from_auth_token_env(self): + """ANTHROPIC_AUTH_TOKEN env var should return Authorization: Bearer header.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, + clear=True, + ): + result = AnthropicModelInfo.get_auth_header() + assert result == {"authorization": f"Bearer {FAKE_AUTH_TOKEN}"} + + def test_api_key_takes_precedence_over_auth_token(self): + """ANTHROPIC_API_KEY should take precedence over ANTHROPIC_AUTH_TOKEN.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + { + "ANTHROPIC_API_KEY": FAKE_REGULAR_KEY, + "ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN, + }, + clear=True, + ): + result = AnthropicModelInfo.get_auth_header() + assert result == {"x-api-key": FAKE_REGULAR_KEY} + + def test_explicit_api_key_overrides_env_auth_token(self): + """Explicit api_key param should override ANTHROPIC_AUTH_TOKEN env var.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, + clear=True, + ): + result = AnthropicModelInfo.get_auth_header(api_key=FAKE_REGULAR_KEY) + assert result == {"x-api-key": FAKE_REGULAR_KEY} + + def test_returns_none_when_no_credentials(self): + """Should return None when neither api_key nor auth_token is available.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict("os.environ", {}, clear=True): + result = AnthropicModelInfo.get_auth_header() + assert result is None + + def test_oauth_token_uses_bearer_not_x_api_key(self): + """OAuth token (sk-ant-oat*) should return Authorization: Bearer, not x-api-key.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + result = AnthropicModelInfo.get_auth_header(api_key=FAKE_OAUTH_TOKEN) + assert result == {"authorization": f"Bearer {FAKE_OAUTH_TOKEN}"} + + def test_oauth_token_from_env_uses_bearer(self): + """OAuth token in ANTHROPIC_API_KEY env var should return Authorization: Bearer.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_API_KEY": FAKE_OAUTH_TOKEN}, + clear=True, + ): + result = AnthropicModelInfo.get_auth_header() + assert result == {"authorization": f"Bearer {FAKE_OAUTH_TOKEN}"} + + +class TestGetApiBaseFallbackChain: + """Tests for AnthropicModelInfo.get_api_base() fallback to ANTHROPIC_BASE_URL.""" + + def test_explicit_param_takes_precedence(self): + """Explicit api_base param takes precedence over all env vars.""" + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + assert ( + AnthropicModelInfo.get_api_base("https://explicit.example.com") + == "https://explicit.example.com" + ) + + def test_defaults_to_anthropic_api(self): + """get_api_base returns the default Anthropic API base when no env vars are set.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict("os.environ", {}, clear=True): + assert AnthropicModelInfo.get_api_base() == "https://api.anthropic.com" + + def test_api_base_env_preferred_over_base_url_env(self): + """ANTHROPIC_API_BASE takes precedence over ANTHROPIC_BASE_URL.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + { + "ANTHROPIC_API_BASE": "https://api-base.example.com", + "ANTHROPIC_BASE_URL": "https://base-url.example.com", + }, + clear=True, + ): + assert AnthropicModelInfo.get_api_base() == "https://api-base.example.com" + + def test_falls_back_to_base_url_env(self): + """get_api_base falls back to ANTHROPIC_BASE_URL when ANTHROPIC_API_BASE is not set.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.common_utils import AnthropicModelInfo + + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_BASE_URL": "https://base-url.example.com"}, + clear=True, + ): + assert AnthropicModelInfo.get_api_base() == "https://base-url.example.com" + + +class TestPassthroughAuthToken: + """Tests for passthrough messages endpoint with ANTHROPIC_AUTH_TOKEN.""" + + def test_passthrough_auth_token_uses_bearer_header(self): + """Passthrough endpoint should use Bearer auth when only ANTHROPIC_AUTH_TOKEN is set.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.experimental_pass_through.messages.transformation import ( + AnthropicMessagesConfig, + ) + + config = AnthropicMessagesConfig() + with mock_patch.dict( + "os.environ", {"ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, clear=True + ): + updated_headers, _ = config.validate_anthropic_messages_environment( + headers={}, + model="claude-sonnet-4-5-20250929", + messages=[{"role": "user", "content": "Hello"}], + optional_params={}, + litellm_params={}, + api_key=None, + api_base=None, + ) + + assert updated_headers["authorization"] == f"Bearer {FAKE_AUTH_TOKEN}" + assert "x-api-key" not in updated_headers + assert "anthropic-dangerous-direct-browser-access" not in updated_headers + + def test_passthrough_api_key_takes_precedence(self): + """Passthrough endpoint should prefer ANTHROPIC_API_KEY over ANTHROPIC_AUTH_TOKEN.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.experimental_pass_through.messages.transformation import ( + AnthropicMessagesConfig, + ) + + config = AnthropicMessagesConfig() + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_API_KEY": FAKE_REGULAR_KEY, "ANTHROPIC_AUTH_TOKEN": FAKE_AUTH_TOKEN}, + clear=True, + ): + updated_headers, _ = config.validate_anthropic_messages_environment( + headers={}, + model="claude-sonnet-4-5-20250929", + messages=[{"role": "user", "content": "Hello"}], + optional_params={}, + litellm_params={}, + api_key=None, + api_base=None, + ) + + assert updated_headers["x-api-key"] == FAKE_REGULAR_KEY + assert "authorization" not in updated_headers + + def test_passthrough_get_complete_url_honours_base_url_env(self): + """get_complete_url should use ANTHROPIC_BASE_URL when api_base is None.""" + from unittest.mock import patch as mock_patch + + from litellm.llms.anthropic.experimental_pass_through.messages.transformation import ( + AnthropicMessagesConfig, + ) + + config = AnthropicMessagesConfig() + with mock_patch.dict( + "os.environ", + {"ANTHROPIC_BASE_URL": "https://custom.example.com"}, + clear=True, + ): + url = config.get_complete_url( + api_base=None, + api_key=FAKE_REGULAR_KEY, + model="claude-sonnet-4-5-20250929", + optional_params={}, + litellm_params={}, + ) + + assert url == "https://custom.example.com/v1/messages" diff --git a/tests/test_litellm/llms/anthropic/test_anthropic_files_and_batches.py b/tests/test_litellm/llms/anthropic/test_anthropic_files_and_batches.py index 47571220175..0755c189afc 100644 --- a/tests/test_litellm/llms/anthropic/test_anthropic_files_and_batches.py +++ b/tests/test_litellm/llms/anthropic/test_anthropic_files_and_batches.py @@ -336,7 +336,7 @@ class TestAnthropicFilesHandler: "extra_body": None } - with patch.object(handler.anthropic_model_info, "get_api_key", return_value=None): + with patch.object(handler.anthropic_model_info, "get_auth_header", return_value=None): with pytest.raises(ValueError, match="Missing Anthropic API Key"): await handler.afile_content( file_content_request=file_content_request, diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_discoverable_endpoints.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_discoverable_endpoints.py index 700ba86b108..954f2703e3b 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_discoverable_endpoints.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_discoverable_endpoints.py @@ -1666,3 +1666,141 @@ async def test_oauth_authorize_prefers_request_scope_over_server_config(): redirect_url = response.headers["location"] assert "scope=custom_scope1+custom_scope2" in redirect_url or "scope=custom_scope1%20custom_scope2" in redirect_url assert "default_scope" not in redirect_url + + +@pytest.mark.asyncio +async def test_token_endpoint_refresh_token_grant(): + """Test that token endpoint supports refresh_token grant type.""" + try: + from fastapi import Request + + from litellm.proxy._experimental.mcp_server.discoverable_endpoints import ( + token_endpoint, + ) + from litellm.proxy._experimental.mcp_server.mcp_server_manager import ( + global_mcp_server_manager, + ) + from litellm.proxy._types import MCPTransport + from litellm.types.mcp import MCPAuth + from litellm.types.mcp_server.mcp_server_manager import MCPServer + except ImportError: + pytest.skip("MCP discoverable endpoints not available") + + # Clear registry + global_mcp_server_manager.registry.clear() + + # Create mock OAuth2 server + oauth2_server = MCPServer( + server_id="google_mcp", + name="google_mcp", + server_name="google_mcp", + alias="google_mcp", + transport=MCPTransport.http, + auth_type=MCPAuth.oauth2, + client_id="test_client_id", + client_secret="test_secret", + authorization_url="https://accounts.google.com/o/oauth2/v2/auth", + token_url="https://oauth2.googleapis.com/token", + scopes=["openid", "email"], + ) + global_mcp_server_manager.registry[oauth2_server.server_id] = oauth2_server + + mock_request = MagicMock(spec=Request) + mock_request.base_url = "https://proxy.litellm.example/" + mock_request.headers = {} + + # Mock httpx client response with new tokens + mock_response = MagicMock() + mock_response.json.return_value = { + "access_token": "new_access_token", + "token_type": "Bearer", + "expires_in": 3599, + "refresh_token": "new_refresh_token", + } + mock_response.raise_for_status = MagicMock() + + mock_async_client = MagicMock() + mock_async_client.post = AsyncMock(return_value=mock_response) + + with patch( + "litellm.proxy._experimental.mcp_server.discoverable_endpoints.get_async_httpx_client" + ) as mock_get_client: + mock_get_client.return_value = mock_async_client + + response = await token_endpoint( + request=mock_request, + grant_type="refresh_token", + code=None, + redirect_uri=None, + client_id="test_client_id", + mcp_server_name="google_mcp", + client_secret="test_secret", + refresh_token="rt-test", + scope="openid email", + ) + + # Verify the POST was called with refresh_token grant data + mock_async_client.post.assert_called_once() + call_args = mock_async_client.post.call_args + + assert call_args[1]["data"]["grant_type"] == "refresh_token" + assert call_args[1]["data"]["refresh_token"] == "rt-test" + assert call_args[1]["data"]["client_id"] == "test_client_id" + assert call_args[1]["data"]["client_secret"] == "test_secret" + assert call_args[1]["data"]["scope"] == "openid email" + + # Verify response contains the new tokens + import json + + token_data = json.loads(response.body) + assert token_data["access_token"] == "new_access_token" + assert token_data["refresh_token"] == "new_refresh_token" + + +@pytest.mark.asyncio +async def test_token_endpoint_authorization_code_missing_code(): + """Test that authorization_code grant rejects missing code param.""" + try: + from litellm.proxy._experimental.mcp_server.discoverable_endpoints import ( + exchange_token_with_server, + ) + from litellm.proxy._experimental.mcp_server.mcp_server_manager import ( + global_mcp_server_manager, + ) + from litellm.proxy._types import MCPTransport + from litellm.types.mcp import MCPAuth + from litellm.types.mcp_server.mcp_server_manager import MCPServer + except ImportError: + pytest.skip("MCP discoverable endpoints not available") + + global_mcp_server_manager.registry.clear() + + server = MCPServer( + server_id="test_server", + name="test_server", + server_name="test_server", + alias="test_server", + transport=MCPTransport.http, + auth_type=MCPAuth.oauth2, + client_id="cid", + token_url="https://example.com/token", + ) + global_mcp_server_manager.registry[server.server_id] = server + + mock_request = MagicMock() + mock_request.base_url = "https://proxy.example/" + mock_request.headers = {} + + with pytest.raises(HTTPException) as exc_info: + await exchange_token_with_server( + request=mock_request, + mcp_server=server, + grant_type="authorization_code", + code=None, + redirect_uri="https://example.com/cb", + client_id="cid", + client_secret=None, + code_verifier=None, + ) + assert exc_info.value.status_code == 400 + assert "code is required" in str(exc_info.value.detail) diff --git a/tests/test_litellm/proxy/discovery_endpoints/test_ui_discovery_endpoints.py b/tests/test_litellm/proxy/discovery_endpoints/test_ui_discovery_endpoints.py index f15960a6072..54a127f435b 100644 --- a/tests/test_litellm/proxy/discovery_endpoints/test_ui_discovery_endpoints.py +++ b/tests/test_litellm/proxy/discovery_endpoints/test_ui_discovery_endpoints.py @@ -1,6 +1,6 @@ import os import sys -from unittest.mock import patch +from unittest.mock import MagicMock, patch import pytest from fastapi import FastAPI @@ -11,6 +11,7 @@ sys.path.insert( ) from litellm.proxy.discovery_endpoints.ui_discovery_endpoints import router +from litellm.types.proxy.control_plane_endpoints import WorkerRegistryEntry def test_ui_discovery_endpoints_with_defaults(): @@ -245,9 +246,9 @@ def test_ui_discovery_endpoints_with_admin_ui_enabled(): patch("litellm.proxy.utils.get_proxy_base_url", return_value=None), \ patch("litellm.proxy.auth.auth_utils._has_user_setup_sso", return_value=False), \ patch.dict(os.environ, {"DISABLE_ADMIN_UI": "false"}, clear=False): - + response = client.get("/.well-known/litellm-ui-config") - + assert response.status_code == 200 data = response.json() assert data["server_root_path"] == "/" @@ -256,3 +257,53 @@ def test_ui_discovery_endpoints_with_admin_ui_enabled(): assert data["admin_ui_disabled"] is False assert data["sso_configured"] is False + +def test_ui_discovery_endpoints_is_control_plane_true_when_workers_configured(): + app = FastAPI() + app.include_router(router) + client = TestClient(app) + + mock_config = MagicMock() + mock_config.worker_registry = [ + WorkerRegistryEntry( + worker_id="team-a", name="Team A", url="https://worker-1:4001" + ), + ] + + with patch("litellm.proxy.utils.get_server_root_path", return_value="/"), \ + patch("litellm.proxy.utils.get_proxy_base_url", return_value=None), \ + patch("litellm.proxy.auth.auth_utils._has_user_setup_sso", return_value=False), \ + patch("litellm.proxy.proxy_server.proxy_config", mock_config), \ + patch.dict(os.environ, {"DISABLE_ADMIN_UI": "false"}, clear=False): + + response = client.get("/.well-known/litellm-ui-config") + + assert response.status_code == 200 + data = response.json() + assert data["is_control_plane"] is True + assert len(data["workers"]) == 1 + assert data["workers"][0]["worker_id"] == "team-a" + assert data["workers"][0]["name"] == "Team A" + assert data["workers"][0]["url"] == "https://worker-1:4001" + + +def test_ui_discovery_endpoints_is_control_plane_false_when_no_workers(): + app = FastAPI() + app.include_router(router) + client = TestClient(app) + + mock_config = MagicMock() + mock_config.worker_registry = [] + + with patch("litellm.proxy.utils.get_server_root_path", return_value="/"), \ + patch("litellm.proxy.utils.get_proxy_base_url", return_value=None), \ + patch("litellm.proxy.auth.auth_utils._has_user_setup_sso", return_value=False), \ + patch("litellm.proxy.proxy_server.proxy_config", mock_config), \ + patch.dict(os.environ, {"DISABLE_ADMIN_UI": "false"}, clear=False): + + response = client.get("/.well-known/litellm-ui-config") + + assert response.status_code == 200 + data = response.json() + assert data["is_control_plane"] is False + assert data["workers"] == [] diff --git a/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py index a3bca77ae3a..12ec79d3e0b 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py @@ -1458,10 +1458,6 @@ async def test_unblock_key_supports_both_sk_and_hashed_tokens(monkeypatch): return_value=mock_key_record ) - # Mock get_key_object and _cache_key_object functions - mock_key_object = MagicMock() - mock_key_object.blocked = True # Initially blocked - # Mock hash_token function def mock_hash_token(token): if token == "sk-test123456789": @@ -1482,19 +1478,12 @@ async def test_unblock_key_supports_both_sk_and_hashed_tokens(monkeypatch): ) # Disable audit logs for simpler test # Mock get_key_object and _cache_key_object - async def mock_get_key_object(**kwargs): - return mock_key_object - - async def mock_cache_key_object(**kwargs): + async def mock_delete_cache_key_object(**kwargs): pass monkeypatch.setattr( - "litellm.proxy.management_endpoints.key_management_endpoints.get_key_object", - mock_get_key_object, - ) - monkeypatch.setattr( - "litellm.proxy.management_endpoints.key_management_endpoints._cache_key_object", - mock_cache_key_object, + "litellm.proxy.management_endpoints.key_management_endpoints._delete_cache_key_object", + mock_delete_cache_key_object, ) # Create mock request and user auth @@ -1519,11 +1508,9 @@ async def test_unblock_key_supports_both_sk_and_hashed_tokens(monkeypatch): ) assert result == mock_key_record - assert mock_key_object.blocked == False # Should be updated to unblocked # Reset mocks for second test mock_prisma_client.db.litellm_verificationtoken.update.reset_mock() - mock_key_object.blocked = True # Reset to blocked state # Test Case 2: Using already hashed token hashed_token_request = BlockKeyRequest(key=test_hashed_token) @@ -1541,7 +1528,6 @@ async def test_unblock_key_supports_both_sk_and_hashed_tokens(monkeypatch): ) assert result == mock_key_record - assert mock_key_object.blocked == False # Should be updated to unblocked @pytest.mark.asyncio @@ -1579,6 +1565,249 @@ async def test_unblock_key_invalid_key_format(monkeypatch): assert "Invalid key format" in str(exc_info.value.message) +@pytest.mark.asyncio +async def test_block_key_nonexistent_key_returns_404(monkeypatch): + """ + Test that block_key returns 404 (not misleading 401) when the key + doesn't exist in the database, even when the caller is authenticated + as a proxy admin. + + Previously, block_key would call get_key_object() for cache refresh, + which raised a 401 ProxyException with 'Authentication Error' — making + it look like an auth failure when it was really a missing-key error. + """ + from litellm.proxy._types import BlockKeyRequest + from litellm.proxy.management_endpoints.key_management_endpoints import block_key + + mock_prisma_client = AsyncMock() + mock_user_api_key_cache = MagicMock() + mock_proxy_logging_obj = MagicMock() + + # find_unique returns None → key does not exist + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=None + ) + + def mock_hash_token(token): + return "abcd1234" * 8 # 64-char hex + + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + monkeypatch.setattr( + "litellm.proxy.proxy_server.user_api_key_cache", mock_user_api_key_cache + ) + monkeypatch.setattr( + "litellm.proxy.proxy_server.proxy_logging_obj", mock_proxy_logging_obj + ) + monkeypatch.setattr("litellm.proxy.proxy_server.hash_token", mock_hash_token) + monkeypatch.setattr("litellm.store_audit_logs", False) + + mock_request = MagicMock() + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, api_key="sk-admin", user_id="admin_user" + ) + + data = BlockKeyRequest(key="sk-does-not-exist-key") + + with pytest.raises(ProxyException) as exc_info: + await block_key( + data=data, + http_request=mock_request, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + ) + + assert exc_info.value.code == "404" + assert "not found" in str(exc_info.value.message).lower() + # Must NOT contain "Authentication Error" + assert "Authentication Error" not in str(exc_info.value.message) + # update should never be called since the key doesn't exist + mock_prisma_client.db.litellm_verificationtoken.update.assert_not_called() + + +@pytest.mark.asyncio +async def test_unblock_key_nonexistent_key_returns_404(monkeypatch): + """ + Test that unblock_key returns 404 (not misleading 401) when the key + doesn't exist in the database. + """ + from litellm.proxy._types import BlockKeyRequest + from litellm.proxy.management_endpoints.key_management_endpoints import ( + unblock_key, + ) + + mock_prisma_client = AsyncMock() + mock_user_api_key_cache = MagicMock() + mock_proxy_logging_obj = MagicMock() + + # find_unique returns None → key does not exist + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=None + ) + + def mock_hash_token(token): + return "abcd1234" * 8 + + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + monkeypatch.setattr( + "litellm.proxy.proxy_server.user_api_key_cache", mock_user_api_key_cache + ) + monkeypatch.setattr( + "litellm.proxy.proxy_server.proxy_logging_obj", mock_proxy_logging_obj + ) + monkeypatch.setattr("litellm.proxy.proxy_server.hash_token", mock_hash_token) + monkeypatch.setattr("litellm.store_audit_logs", False) + + mock_request = MagicMock() + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, api_key="sk-admin", user_id="admin_user" + ) + + data = BlockKeyRequest(key="sk-does-not-exist-key") + + with pytest.raises(ProxyException) as exc_info: + await unblock_key( + data=data, + http_request=mock_request, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + ) + + assert exc_info.value.code == "404" + assert "not found" in str(exc_info.value.message).lower() + assert "Authentication Error" not in str(exc_info.value.message) + mock_prisma_client.db.litellm_verificationtoken.update.assert_not_called() + + +@pytest.mark.asyncio +async def test_update_key_nonexistent_key_returns_404(monkeypatch): + """ + Test that update_key_fn returns 404 (not misleading 401) when the body + key doesn't exist in the database, even when the caller is authenticated + as a proxy admin via the Authorization header. + """ + from litellm.proxy.management_endpoints.key_management_endpoints import ( + update_key_fn, + ) + + mock_prisma_client = AsyncMock() + mock_user_api_key_cache = MagicMock() + mock_proxy_logging_obj = MagicMock() + + # find_unique returns None → key does not exist + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=None + ) + + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + monkeypatch.setattr( + "litellm.proxy.proxy_server.user_api_key_cache", mock_user_api_key_cache + ) + monkeypatch.setattr( + "litellm.proxy.proxy_server.proxy_logging_obj", mock_proxy_logging_obj + ) + monkeypatch.setattr("litellm.proxy.proxy_server.llm_router", None) + monkeypatch.setattr("litellm.proxy.proxy_server.premium_user", True) + + mock_request = MagicMock() + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, api_key="sk-admin", user_id="admin_user" + ) + + data = UpdateKeyRequest(key="sk-does-not-exist-key") + + with pytest.raises(ProxyException) as exc_info: + await update_key_fn( + request=mock_request, + data=data, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + ) + + assert exc_info.value.code == "404" + assert "not found" in str(exc_info.value.message).lower() + assert "Authentication Error" not in str(exc_info.value.message) + + +@pytest.mark.asyncio +async def test_block_key_existing_key_succeeds(monkeypatch): + """ + Test that block_key successfully blocks an existing key and + invalidates the cache entry. + """ + from litellm.proxy._types import BlockKeyRequest + from litellm.proxy.management_endpoints.key_management_endpoints import block_key + + mock_prisma_client = AsyncMock() + mock_user_api_key_cache = MagicMock() + mock_proxy_logging_obj = MagicMock() + + test_hashed_token = "a1b2c3d4e5f6789012345678901234567890123456789012345678901234abcd" + + mock_key_record = MagicMock() + mock_key_record.token = test_hashed_token + mock_key_record.blocked = False + mock_key_record.model_dump_json.return_value = ( + f'{{"token": "{test_hashed_token}", "blocked": false}}' + ) + + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=mock_key_record + ) + mock_updated_record = MagicMock() + mock_updated_record.token = test_hashed_token + mock_updated_record.blocked = True + mock_prisma_client.db.litellm_verificationtoken.update = AsyncMock( + return_value=mock_updated_record + ) + + def mock_hash_token(token): + if token.startswith("sk-"): + return test_hashed_token + return token + + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + monkeypatch.setattr( + "litellm.proxy.proxy_server.user_api_key_cache", mock_user_api_key_cache + ) + monkeypatch.setattr( + "litellm.proxy.proxy_server.proxy_logging_obj", mock_proxy_logging_obj + ) + monkeypatch.setattr("litellm.proxy.proxy_server.hash_token", mock_hash_token) + monkeypatch.setattr("litellm.store_audit_logs", False) + + # Mock _delete_cache_key_object + async def mock_delete_cache_key_object(**kwargs): + pass + + monkeypatch.setattr( + "litellm.proxy.management_endpoints.key_management_endpoints._delete_cache_key_object", + mock_delete_cache_key_object, + ) + + mock_request = MagicMock() + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, api_key="sk-admin", user_id="admin_user" + ) + + data = BlockKeyRequest(key="sk-test123456789") + + result = await block_key( + data=data, + http_request=mock_request, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + ) + + # Verify the key was found and updated + mock_prisma_client.db.litellm_verificationtoken.find_unique.assert_called_once_with( + where={"token": test_hashed_token} + ) + mock_prisma_client.db.litellm_verificationtoken.update.assert_called_once_with( + where={"token": test_hashed_token}, data={"blocked": True} + ) + assert result == mock_updated_record + + @pytest.mark.asyncio async def test_validate_key_team_change_with_member_permissions(): """ @@ -4871,14 +5100,16 @@ async def test_validate_max_budget(): async def test_get_and_validate_existing_key(): """ Test _get_and_validate_existing_key helper function. - + Tests: 1. Successfully retrieve existing key - 2. Key not found raises HTTPException + 2. Key not found raises ProxyException 3. Database not connected raises HTTPException """ from fastapi import HTTPException + from litellm.proxy._types import ProxyException + # Test Case 1: Successfully retrieve existing key mock_prisma_client = AsyncMock() mock_key = LiteLLM_VerificationToken( @@ -4887,39 +5118,49 @@ async def test_get_and_validate_existing_key(): models=["gpt-4"], team_id=None, ) - mock_prisma_client.get_data = AsyncMock(return_value=mock_key) - - result = await _get_and_validate_existing_key( - token="test-key-123", - prisma_client=mock_prisma_client, + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=mock_key ) - - assert result == mock_key - mock_prisma_client.get_data.assert_called_once_with( - token="test-key-123", - table_name="key", - query_type="find_unique", - ) - - # Test Case 2: Key not found raises HTTPException - mock_prisma_client.get_data = AsyncMock(return_value=None) - - with pytest.raises(HTTPException) as exc_info: - await _get_and_validate_existing_key( - token="non-existent-key", + + with patch( + "litellm.proxy.management_endpoints.key_management_endpoints._hash_token_if_needed", + return_value="hashed-test-key-123", + ): + result = await _get_and_validate_existing_key( + token="test-key-123", prisma_client=mock_prisma_client, ) - - assert exc_info.value.status_code == 404 - assert "Key not found" in str(exc_info.value.detail) - + + assert result == mock_key + mock_prisma_client.db.litellm_verificationtoken.find_unique.assert_called_once_with( + where={"token": "hashed-test-key-123"} + ) + + # Test Case 2: Key not found raises ProxyException + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=None + ) + + with patch( + "litellm.proxy.management_endpoints.key_management_endpoints._hash_token_if_needed", + return_value="hashed-non-existent-key", + ): + with pytest.raises(ProxyException) as exc_info: + await _get_and_validate_existing_key( + token="non-existent-key", + prisma_client=mock_prisma_client, + ) + + assert str(exc_info.value.code) == "404" + assert "Key not found" in exc_info.value.message + # Test Case 3: Database not connected raises HTTPException with pytest.raises(HTTPException) as exc_info: await _get_and_validate_existing_key( token="test-key-123", prisma_client=None, ) - + assert exc_info.value.status_code == 500 assert "Database not connected" in str(exc_info.value.detail) @@ -4960,75 +5201,82 @@ async def test_process_single_key_update(): "tags": ["production"], } - mock_prisma_client.get_data = AsyncMock(return_value=existing_key) + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( + return_value=existing_key + ) mock_updated_key_obj = MagicMock() mock_updated_key_obj.model_dump.return_value = updated_key_data mock_prisma_client.update_data = AsyncMock( return_value={"data": mock_updated_key_obj} ) - + # Mock prepare_key_update_data with patch( "litellm.proxy.management_endpoints.key_management_endpoints.prepare_key_update_data" ) as mock_prepare: mock_prepare.return_value = {"max_budget": 100.0, "tags": ["production"]} - + # Mock TeamMemberPermissionChecks with patch( "litellm.proxy.management_endpoints.key_management_endpoints.TeamMemberPermissionChecks.can_team_member_execute_key_management_endpoint" ) as mock_permission_check: mock_permission_check.return_value = None - + # Mock _delete_cache_key_object with patch( "litellm.proxy.management_endpoints.key_management_endpoints._delete_cache_key_object" ) as mock_delete_cache: mock_delete_cache.return_value = None - + # Mock hash_token (imported from litellm.proxy._types) with patch( "litellm.proxy._types.hash_token" ) as mock_hash: mock_hash.return_value = "hashed-test-key-123" - - # Mock KeyManagementEventHooks + + # Mock _hash_token_if_needed with patch( - "litellm.proxy.management_endpoints.key_management_endpoints.KeyManagementEventHooks.async_key_updated_hook" + "litellm.proxy.management_endpoints.key_management_endpoints._hash_token_if_needed", + return_value="hashed-test-key-123", ): - # Create update request - key_update_item = BulkUpdateKeyRequestItem( - key="test-key-123", - max_budget=100.0, - tags=["production"], - ) - - user_api_key_dict = UserAPIKeyAuth( - user_role=LitellmUserRoles.PROXY_ADMIN, - api_key="sk-admin", - user_id="admin-user", - ) - - # Call the function - result = await _process_single_key_update( - key_update_item=key_update_item, - user_api_key_dict=user_api_key_dict, - litellm_changed_by=None, - prisma_client=mock_prisma_client, - user_api_key_cache=mock_user_api_key_cache, - proxy_logging_obj=mock_proxy_logging_obj, - llm_router=mock_llm_router, - ) - - # Verify results - assert result is not None - assert "token" not in result # Token should be removed - assert result.get("max_budget") == 100.0 - assert result.get("tags") == ["production"] - - # Verify mocks were called - mock_prisma_client.get_data.assert_called_once() - mock_prisma_client.update_data.assert_called_once() - mock_delete_cache.assert_called_once() + # Mock KeyManagementEventHooks + with patch( + "litellm.proxy.management_endpoints.key_management_endpoints.KeyManagementEventHooks.async_key_updated_hook" + ): + # Create update request + key_update_item = BulkUpdateKeyRequestItem( + key="test-key-123", + max_budget=100.0, + tags=["production"], + ) + + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, + api_key="sk-admin", + user_id="admin-user", + ) + + # Call the function + result = await _process_single_key_update( + key_update_item=key_update_item, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + prisma_client=mock_prisma_client, + user_api_key_cache=mock_user_api_key_cache, + proxy_logging_obj=mock_proxy_logging_obj, + llm_router=mock_llm_router, + ) + + # Verify results + assert result is not None + assert "token" not in result # Token should be removed + assert result.get("max_budget") == 100.0 + assert result.get("tags") == ["production"] + + # Verify mocks were called + mock_prisma_client.db.litellm_verificationtoken.find_unique.assert_called_once() + mock_prisma_client.update_data.assert_called_once() + mock_delete_cache.assert_called_once() @pytest.mark.asyncio @@ -5090,7 +5338,7 @@ async def test_bulk_update_keys_success(monkeypatch): "tags": ["staging"], } - mock_prisma_client.get_data = AsyncMock( + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( side_effect=[existing_key_1, existing_key_2] ) mock_updated_key_1_obj = MagicMock() @@ -5103,7 +5351,7 @@ async def test_bulk_update_keys_success(monkeypatch): {"data": mock_updated_key_2_obj}, ] ) - + # Patch dependencies monkeypatch.setattr( "litellm.proxy.proxy_server.prisma_client", mock_prisma_client @@ -5115,7 +5363,7 @@ async def test_bulk_update_keys_success(monkeypatch): "litellm.proxy.proxy_server.proxy_logging_obj", mock_proxy_logging_obj ) monkeypatch.setattr("litellm.proxy.proxy_server.llm_router", mock_llm_router) - + # Mock helper functions with patch( "litellm.proxy.management_endpoints.key_management_endpoints.prepare_key_update_data" @@ -5124,7 +5372,7 @@ async def test_bulk_update_keys_success(monkeypatch): {"max_budget": 100.0, "tags": ["production"]}, {"max_budget": 200.0, "tags": ["staging"]}, ] - + with patch( "litellm.proxy.management_endpoints.key_management_endpoints.TeamMemberPermissionChecks.can_team_member_execute_key_management_endpoint" ): @@ -5135,45 +5383,49 @@ async def test_bulk_update_keys_success(monkeypatch): "litellm.proxy._types.hash_token" ) as mock_hash: mock_hash.side_effect = ["hashed-key-1", "hashed-key-2"] - + with patch( - "litellm.proxy.management_endpoints.key_management_endpoints.KeyManagementEventHooks.async_key_updated_hook" + "litellm.proxy.management_endpoints.key_management_endpoints._hash_token_if_needed", + side_effect=["hashed-key-1", "hashed-key-2"], ): - # Create request - request_data = BulkUpdateKeyRequest( - keys=[ - BulkUpdateKeyRequestItem( - key="test-key-1", - max_budget=100.0, - tags=["production"], - ), - BulkUpdateKeyRequestItem( - key="test-key-2", - max_budget=200.0, - tags=["staging"], - ), - ] - ) - - user_api_key_dict = UserAPIKeyAuth( - user_role=LitellmUserRoles.PROXY_ADMIN, - api_key="sk-admin", - user_id="admin-user", - ) - - # Call endpoint - response = await bulk_update_keys( - data=request_data, - user_api_key_dict=user_api_key_dict, - litellm_changed_by=None, - ) - - # Verify response - assert response.total_requested == 2 - assert len(response.successful_updates) == 2 - assert len(response.failed_updates) == 0 - assert response.successful_updates[0].key == "test-key-1" - assert response.successful_updates[1].key == "test-key-2" + with patch( + "litellm.proxy.management_endpoints.key_management_endpoints.KeyManagementEventHooks.async_key_updated_hook" + ): + # Create request + request_data = BulkUpdateKeyRequest( + keys=[ + BulkUpdateKeyRequestItem( + key="test-key-1", + max_budget=100.0, + tags=["production"], + ), + BulkUpdateKeyRequestItem( + key="test-key-2", + max_budget=200.0, + tags=["staging"], + ), + ] + ) + + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, + api_key="sk-admin", + user_id="admin-user", + ) + + # Call endpoint + response = await bulk_update_keys( + data=request_data, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + ) + + # Verify response + assert response.total_requested == 2 + assert len(response.successful_updates) == 2 + assert len(response.failed_updates) == 0 + assert response.successful_updates[0].key == "test-key-1" + assert response.successful_updates[1].key == "test-key-2" @pytest.mark.asyncio @@ -5218,7 +5470,7 @@ async def test_bulk_update_keys_partial_failures(monkeypatch): } # First key exists, second key doesn't exist - mock_prisma_client.get_data = AsyncMock( + mock_prisma_client.db.litellm_verificationtoken.find_unique = AsyncMock( side_effect=[existing_key_1, None] # Second key not found ) mock_updated_key_1_obj = MagicMock() @@ -5226,7 +5478,9 @@ async def test_bulk_update_keys_partial_failures(monkeypatch): mock_prisma_client.update_data = AsyncMock( return_value={"data": mock_updated_key_1_obj} ) - + # Mock get_data for the error handler path (used to fetch key_info on failure) + mock_prisma_client.get_data = AsyncMock(return_value=None) + # Patch dependencies monkeypatch.setattr( "litellm.proxy.proxy_server.prisma_client", mock_prisma_client @@ -5238,13 +5492,13 @@ async def test_bulk_update_keys_partial_failures(monkeypatch): "litellm.proxy.proxy_server.proxy_logging_obj", mock_proxy_logging_obj ) monkeypatch.setattr("litellm.proxy.proxy_server.llm_router", mock_llm_router) - + # Mock helper functions with patch( "litellm.proxy.management_endpoints.key_management_endpoints.prepare_key_update_data" ) as mock_prepare: mock_prepare.return_value = {"max_budget": 100.0, "tags": ["production"]} - + with patch( "litellm.proxy.management_endpoints.key_management_endpoints.TeamMemberPermissionChecks.can_team_member_execute_key_management_endpoint" ): @@ -5255,46 +5509,50 @@ async def test_bulk_update_keys_partial_failures(monkeypatch): "litellm.proxy._types.hash_token" ) as mock_hash: mock_hash.return_value = "hashed-key-1" - + with patch( - "litellm.proxy.management_endpoints.key_management_endpoints.KeyManagementEventHooks.async_key_updated_hook" + "litellm.proxy.management_endpoints.key_management_endpoints._hash_token_if_needed", + side_effect=["hashed-key-1", "hashed-non-existent-key"], ): - # Create request with one valid and one invalid key - request_data = BulkUpdateKeyRequest( - keys=[ - BulkUpdateKeyRequestItem( - key="test-key-1", - max_budget=100.0, - tags=["production"], - ), - BulkUpdateKeyRequestItem( - key="non-existent-key", - max_budget=200.0, - tags=["staging"], - ), - ] - ) - - user_api_key_dict = UserAPIKeyAuth( - user_role=LitellmUserRoles.PROXY_ADMIN, - api_key="sk-admin", - user_id="admin-user", - ) - - # Call endpoint - response = await bulk_update_keys( - data=request_data, - user_api_key_dict=user_api_key_dict, - litellm_changed_by=None, - ) - - # Verify response - assert response.total_requested == 2 - assert len(response.successful_updates) == 1 - assert len(response.failed_updates) == 1 - assert response.successful_updates[0].key == "test-key-1" - assert response.failed_updates[0].key == "non-existent-key" - assert "Key not found" in response.failed_updates[0].failed_reason + with patch( + "litellm.proxy.management_endpoints.key_management_endpoints.KeyManagementEventHooks.async_key_updated_hook" + ): + # Create request with one valid and one invalid key + request_data = BulkUpdateKeyRequest( + keys=[ + BulkUpdateKeyRequestItem( + key="test-key-1", + max_budget=100.0, + tags=["production"], + ), + BulkUpdateKeyRequestItem( + key="non-existent-key", + max_budget=200.0, + tags=["staging"], + ), + ] + ) + + user_api_key_dict = UserAPIKeyAuth( + user_role=LitellmUserRoles.PROXY_ADMIN, + api_key="sk-admin", + user_id="admin-user", + ) + + # Call endpoint + response = await bulk_update_keys( + data=request_data, + user_api_key_dict=user_api_key_dict, + litellm_changed_by=None, + ) + + # Verify response + assert response.total_requested == 2 + assert len(response.successful_updates) == 1 + assert len(response.failed_updates) == 1 + assert response.successful_updates[0].key == "test-key-1" + assert response.failed_updates[0].key == "non-existent-key" + assert "Key not found" in response.failed_updates[0].failed_reason @pytest.mark.parametrize( @@ -7379,19 +7637,12 @@ def _setup_block_unblock_mocks(monkeypatch, mock_key_team_id=None): monkeypatch.setattr("litellm.proxy.proxy_server.hash_token", mock_hash_token) monkeypatch.setattr("litellm.store_audit_logs", False) - async def mock_get_key_object(**kwargs): - return mock_key_object - - async def mock_cache_key_object(**kwargs): + async def mock_delete_cache_key_object(**kwargs): pass monkeypatch.setattr( - "litellm.proxy.management_endpoints.key_management_endpoints.get_key_object", - mock_get_key_object, - ) - monkeypatch.setattr( - "litellm.proxy.management_endpoints.key_management_endpoints._cache_key_object", - mock_cache_key_object, + "litellm.proxy.management_endpoints.key_management_endpoints._delete_cache_key_object", + mock_delete_cache_key_object, ) return mock_prisma_client, test_hashed_token @@ -7638,16 +7889,9 @@ async def test_update_key_non_budget_fields_allowed_for_internal_user(monkeypatc monkeypatch.setattr("litellm.proxy.proxy_server.hash_token", mock_hash_token) - async def mock_cache_key_object(**kwargs): - pass - async def mock_delete_cache_key_object(**kwargs): pass - monkeypatch.setattr( - "litellm.proxy.management_endpoints.key_management_endpoints._cache_key_object", - mock_cache_key_object, - ) monkeypatch.setattr( "litellm.proxy.management_endpoints.key_management_endpoints._delete_cache_key_object", mock_delete_cache_key_object, diff --git a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py index eeaeb498328..77ac3a040ac 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py @@ -1519,6 +1519,8 @@ class TestTemporaryMCPSessionEndpoints: client_id="client", client_secret="secret", code_verifier="verifier", + refresh_token=None, + scope=None, ) assert result is exchange_response @@ -1532,6 +1534,56 @@ class TestTemporaryMCPSessionEndpoints: client_id="client", client_secret="secret", code_verifier="verifier", + refresh_token=None, + scope=None, + ) + + @pytest.mark.asyncio + async def test_mcp_token_proxies_refresh_token_grant(self): + from litellm.proxy.management_endpoints.mcp_management_endpoints import ( + mcp_token, + ) + + request = MagicMock() + server = generate_mock_mcp_server_config_record(server_id="server-1") + exchange_response = {"access_token": "new-token", "refresh_token": "new-rt"} + + with ( + patch( + "litellm.proxy.management_endpoints.mcp_management_endpoints._get_cached_temporary_mcp_server_or_404", + return_value=server, + ) as get_server, + patch( + "litellm.proxy.management_endpoints.mcp_management_endpoints.exchange_token_with_server", + AsyncMock(return_value=exchange_response), + ) as exchange_mock, + ): + result = await mcp_token( + request=request, + server_id="server-1", + grant_type="refresh_token", + code=None, + redirect_uri=None, + client_id="client", + client_secret="secret", + code_verifier=None, + refresh_token="rt-123", + scope=None, + ) + + assert result is exchange_response + get_server.assert_called_once_with("server-1") + exchange_mock.assert_awaited_once_with( + request=request, + mcp_server=server, + grant_type="refresh_token", + code=None, + redirect_uri=None, + client_id="client", + client_secret="secret", + code_verifier=None, + refresh_token="rt-123", + scope=None, ) @pytest.mark.asyncio diff --git a/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py index c325b0b6fce..366f659bdab 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py @@ -6441,3 +6441,53 @@ async def test_list_team_v1_batches_key_queries(): assert result[0].keys == [key1, key2] assert result[1].team_id == "team-2" assert result[1].keys == [key3] + + +def test_new_team_request_accepts_team_member_budget_duration(): + """Test that NewTeamRequest does not silently drop team_member_budget_duration.""" + from litellm.proxy._types import NewTeamRequest + + request = NewTeamRequest( + team_member_budget=20.0, + team_member_budget_duration="30d", + ) + assert request.team_member_budget == 20.0 + assert request.team_member_budget_duration == "30d" + + +@pytest.mark.asyncio +async def test_create_team_member_budget_table_with_duration(): + """Verify that create_team_member_budget_table passes budget_duration + through to the new_budget call when team_member_budget_duration is provided.""" + from litellm.proxy._types import NewTeamRequest, UserAPIKeyAuth, LitellmUserRoles + from litellm.proxy.management_endpoints.team_endpoints import TeamMemberBudgetHandler + + mock_budget_response = MagicMock(budget_id="budget-abc") + mock_admin = UserAPIKeyAuth( + user_id="admin", user_role=LitellmUserRoles.PROXY_ADMIN + ) + + data = NewTeamRequest( + team_alias="test-team", + team_member_budget=20.0, + team_member_budget_duration="30d", + ) + + with patch( + "litellm.proxy.management_endpoints.budget_management_endpoints.new_budget", + new_callable=AsyncMock, + return_value=mock_budget_response, + ) as mock_new_budget: + result = await TeamMemberBudgetHandler.create_team_member_budget_table( + data=data, + new_team_data_json={"metadata": None}, + user_api_key_dict=mock_admin, + team_member_budget=20.0, + team_member_budget_duration="30d", + ) + + mock_new_budget.assert_awaited_once() + budget_request = mock_new_budget.call_args.kwargs["budget_obj"] + assert budget_request.budget_duration == "30d" + assert budget_request.max_budget == 20.0 + assert result["metadata"]["team_member_budget_id"] == "budget-abc" diff --git a/tests/test_litellm/proxy/management_endpoints/test_ui_sso.py b/tests/test_litellm/proxy/management_endpoints/test_ui_sso.py index d43b2c4ba05..fc9c37b7f84 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_ui_sso.py +++ b/tests/test_litellm/proxy/management_endpoints/test_ui_sso.py @@ -6,7 +6,7 @@ from unittest.mock import AsyncMock, MagicMock, patch import httpx import pytest -from fastapi import Request +from fastapi import HTTPException, Request from litellm._uuid import uuid @@ -5160,3 +5160,99 @@ def test_generic_response_convertor_extra_attributes_missing_field(monkeypatch): assert result.extra_fields["missing_field"] is None assert result.extra_fields["another_missing"] is None + +class TestValidateReturnTo: + """Tests for SSOAuthenticationHandler._validate_return_to""" + + def test_rejects_when_no_control_plane_url_configured(self, monkeypatch): + """return_to should be rejected if control_plane_url is not in general_settings.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", {} + ) + with pytest.raises(HTTPException) as exc_info: + SSOAuthenticationHandler._validate_return_to("https://cp.example.com/ui") + assert exc_info.value.status_code == 400 + assert "not configured" in exc_info.value.detail + + def test_allows_matching_origin(self, monkeypatch): + """return_to matching the configured control_plane_url origin should pass.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + # Should not raise + SSOAuthenticationHandler._validate_return_to("https://cp.example.com/ui?page=models") + + def test_allows_matching_origin_with_trailing_slash(self, monkeypatch): + """Trailing slash on control_plane_url should not affect origin comparison.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com/"}, + ) + SSOAuthenticationHandler._validate_return_to("https://cp.example.com/ui") + + def test_rejects_prefix_attack(self, monkeypatch): + """return_to like cp.example.com.evil.com must be rejected (not just prefix match).""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + with pytest.raises(HTTPException) as exc_info: + SSOAuthenticationHandler._validate_return_to("https://cp.example.com.evil.com/steal") + assert exc_info.value.status_code == 400 + + def test_rejects_different_origin(self, monkeypatch): + """return_to pointing to a completely different domain should be rejected.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + with pytest.raises(HTTPException) as exc_info: + SSOAuthenticationHandler._validate_return_to("https://evil.com/phish") + assert exc_info.value.status_code == 400 + + def test_case_insensitive_hostname(self, monkeypatch): + """Hostname comparison should be case-insensitive per RFC 3986.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://CP.Example.COM"}, + ) + # Should not raise + SSOAuthenticationHandler._validate_return_to("https://cp.example.com/ui") + + def test_rejects_scheme_mismatch(self, monkeypatch): + """http:// must be rejected when control_plane_url uses https://.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + with pytest.raises(HTTPException) as exc_info: + SSOAuthenticationHandler._validate_return_to("http://cp.example.com/ui") + assert exc_info.value.status_code == 400 + + def test_rejects_port_mismatch(self, monkeypatch): + """Non-default port must be rejected.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + with pytest.raises(HTTPException) as exc_info: + SSOAuthenticationHandler._validate_return_to("https://cp.example.com:8443/ui") + assert exc_info.value.status_code == 400 + + def test_allows_explicit_default_port(self, monkeypatch): + """https://host:443 should match https://host (default port normalisation).""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + SSOAuthenticationHandler._validate_return_to("https://cp.example.com:443/ui") + + def test_allows_matching_custom_port(self, monkeypatch): + """Both sides on the same custom port should match.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com:3000"}, + ) + SSOAuthenticationHandler._validate_return_to("https://cp.example.com:3000/ui") + diff --git a/tests/test_litellm/proxy/test_proxy_cli.py b/tests/test_litellm/proxy/test_proxy_cli.py index c5d6c45f9a5..349fe76ed71 100644 --- a/tests/test_litellm/proxy/test_proxy_cli.py +++ b/tests/test_litellm/proxy/test_proxy_cli.py @@ -280,6 +280,47 @@ class TestProxyInitializationHelpers: assert result.exit_code == 0, f"exit_code={result.exit_code}, output={result.output}" mock_uvicorn_run.assert_called_once() + @patch("uvicorn.run") + @patch("atexit.register") + @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database") + @patch("litellm.proxy.db.prisma_client.should_update_prisma_schema", return_value=False) + def test_proxy_default_api_version_uses_azure_default( + self, mock_should_update, mock_setup_db, mock_atexit_register, mock_uvicorn_run + ): + """Proxy default api_version should match litellm.AZURE_DEFAULT_API_VERSION for consistency.""" + from click.testing import CliRunner + + import litellm + from litellm.proxy.proxy_cli import run_server + + runner = CliRunner() + mock_proxy_module = MagicMock( + app=MagicMock(), + ProxyConfig=MagicMock(), + KeyManagementSettings=MagicMock(), + save_worker_config=MagicMock(), + ) + clean_env = {k: v for k, v in os.environ.items() if k not in ("DATABASE_URL", "DIRECT_URL")} + with patch.dict(os.environ, clean_env, clear=True), patch.dict( + "sys.modules", + { + "proxy_server": mock_proxy_module, + "litellm.proxy.proxy_server": mock_proxy_module, + }, + ), patch( + "litellm.proxy.proxy_cli.ProxyInitializationHelpers._get_default_unvicorn_init_args" + ) as mock_get_args: + mock_get_args.return_value = { + "app": "litellm.proxy.proxy_server:app", + "host": "localhost", + "port": 8000, + } + result = runner.invoke(run_server, ["--local", "--skip_server_startup"]) + assert result.exit_code == 0, f"exit_code={result.exit_code}, output={result.output}" + mock_proxy_module.save_worker_config.assert_called_once() + call_kwargs = mock_proxy_module.save_worker_config.call_args[1] + assert call_kwargs["api_version"] == litellm.AZURE_DEFAULT_API_VERSION + @patch("uvicorn.run") @patch("builtins.print") def test_keepalive_timeout_flag(self, mock_print, mock_uvicorn_run): diff --git a/tests/test_litellm/proxy/test_proxy_server.py b/tests/test_litellm/proxy/test_proxy_server.py index 112a06b1731..bd6162f225a 100644 --- a/tests/test_litellm/proxy/test_proxy_server.py +++ b/tests/test_litellm/proxy/test_proxy_server.py @@ -236,6 +236,217 @@ def test_login_v2_returns_json_on_invalid_json_body(monkeypatch): assert isinstance(data["error"], dict) +def test_login_v3_rejected_without_control_plane_url(monkeypatch): + """v3/login returns 404 when control_plane_url is not configured.""" + mock_prisma_client = MagicMock() + monkeypatch.setattr("litellm.proxy.proxy_server.master_key", "test-master-key") + monkeypatch.setattr("litellm.proxy.proxy_server.general_settings", {}) + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + + client = TestClient(app) + response = client.post( + "/v3/login", + json={"username": "alice", "password": "secret"}, + ) + + assert response.status_code == 404 + assert "control_plane_url" in response.json()["error"]["message"] + + +def test_login_v3_returns_code(monkeypatch): + """v3/login returns an opaque code, not the JWT directly.""" + mock_prisma_client = MagicMock() + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.authenticate_user", + AsyncMock(return_value={"user_id": "test-user"}), + ) + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.create_ui_token_object", + MagicMock(return_value={"user_id": "test-user"}), + ) + monkeypatch.setattr("jwt.encode", MagicMock(return_value="signed-token")) + monkeypatch.setattr("litellm.proxy.proxy_server.master_key", "test-master-key") + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + monkeypatch.setattr("litellm.proxy.proxy_server.premium_user", False) + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + mock_config = MagicMock() + mock_config.worker_registry = [] + monkeypatch.setattr("litellm.proxy.proxy_server.proxy_config", mock_config) + monkeypatch.setattr("litellm.proxy.utils.get_server_root_path", lambda: "") + monkeypatch.setattr("litellm.proxy.utils.get_proxy_base_url", lambda: None) + + client = TestClient(app) + response = client.post( + "/v3/login", + json={"username": "alice", "password": "secret"}, + ) + + assert response.status_code == 200 + data = response.json() + assert "code" in data + assert data["expires_in"] == 60 + assert "token" not in data + + +def test_login_v3_exchange_happy_path(monkeypatch): + """Full flow: v3/login returns code, v3/login/exchange redeems it for JWT.""" + mock_prisma_client = MagicMock() + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.authenticate_user", + AsyncMock(return_value={"user_id": "test-user"}), + ) + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.create_ui_token_object", + MagicMock(return_value={"user_id": "test-user"}), + ) + monkeypatch.setattr("jwt.encode", MagicMock(return_value="signed-token")) + monkeypatch.setattr("litellm.proxy.proxy_server.master_key", "test-master-key") + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + monkeypatch.setattr("litellm.proxy.proxy_server.premium_user", False) + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + mock_config = MagicMock() + mock_config.worker_registry = [] + monkeypatch.setattr("litellm.proxy.proxy_server.proxy_config", mock_config) + monkeypatch.setattr("litellm.proxy.utils.get_server_root_path", lambda: "") + monkeypatch.setattr("litellm.proxy.utils.get_proxy_base_url", lambda: None) + + client = TestClient(app) + + # Step 1: login — get code + login_response = client.post( + "/v3/login", + json={"username": "alice", "password": "secret"}, + ) + assert login_response.status_code == 200 + code = login_response.json()["code"] + + # Step 2: exchange — get JWT + exchange_response = client.post( + "/v3/login/exchange", + json={"code": code}, + ) + assert exchange_response.status_code == 200 + exchange_data = exchange_response.json() + assert exchange_data["token"] == "signed-token" + assert "redirect_url" in exchange_data + assert exchange_response.cookies.get("token") == "signed-token" + + +def test_login_v3_exchange_single_use(monkeypatch): + """Code can only be redeemed once.""" + mock_prisma_client = MagicMock() + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.authenticate_user", + AsyncMock(return_value={"user_id": "test-user"}), + ) + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.create_ui_token_object", + MagicMock(return_value={"user_id": "test-user"}), + ) + monkeypatch.setattr("jwt.encode", MagicMock(return_value="signed-token")) + monkeypatch.setattr("litellm.proxy.proxy_server.master_key", "test-master-key") + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + monkeypatch.setattr("litellm.proxy.proxy_server.premium_user", False) + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + mock_config = MagicMock() + mock_config.worker_registry = [] + monkeypatch.setattr("litellm.proxy.proxy_server.proxy_config", mock_config) + monkeypatch.setattr("litellm.proxy.utils.get_server_root_path", lambda: "") + monkeypatch.setattr("litellm.proxy.utils.get_proxy_base_url", lambda: None) + + client = TestClient(app) + + login_response = client.post( + "/v3/login", + json={"username": "alice", "password": "secret"}, + ) + code = login_response.json()["code"] + + # First exchange succeeds + first = client.post("/v3/login/exchange", json={"code": code}) + assert first.status_code == 200 + + # Second exchange fails + second = client.post("/v3/login/exchange", json={"code": code}) + assert second.status_code == 401 + + +def test_login_v3_exchange_invalid_code(monkeypatch): + """Random code returns 401.""" + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + client = TestClient(app) + response = client.post( + "/v3/login/exchange", + json={"code": "nonexistent-code"}, + ) + assert response.status_code == 401 + + +def test_login_v3_exchange_rejected_without_control_plane_url(monkeypatch): + """v3/login/exchange returns 404 when control_plane_url is not configured.""" + monkeypatch.setattr("litellm.proxy.proxy_server.general_settings", {}) + + client = TestClient(app) + response = client.post( + "/v3/login/exchange", + json={"code": "some-code"}, + ) + + assert response.status_code == 404 + assert "control_plane_url" in response.json()["error"]["message"] + + +def test_login_v3_returns_json_on_proxy_exception(monkeypatch): + """Test that /v3/login returns JSON error when ProxyException is raised""" + from litellm.proxy._types import ProxyErrorTypes, ProxyException + + mock_prisma_client = MagicMock() + mock_authenticate_user = AsyncMock( + side_effect=ProxyException( + message="Invalid credentials", + type=ProxyErrorTypes.auth_error, + param="password", + code=401, + ) + ) + + monkeypatch.setattr( + "litellm.proxy.auth.login_utils.authenticate_user", + mock_authenticate_user, + ) + monkeypatch.setattr("litellm.proxy.proxy_server.master_key", "test-master-key") + monkeypatch.setattr( + "litellm.proxy.proxy_server.general_settings", + {"control_plane_url": "https://cp.example.com"}, + ) + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", mock_prisma_client) + + client = TestClient(app) + response = client.post( + "/v3/login", + json={"username": "alice", "password": "wrong"}, + ) + + assert response.status_code == 401 + assert response.headers["content-type"] == "application/json" + data = response.json() + assert "error" in data + assert data["error"]["message"] == "Invalid credentials" + assert data["error"]["type"] == "auth_error" + + def test_fallback_login_has_no_deprecation_banner(client_no_auth): response = client_no_auth.get("/fallback/login") diff --git a/ui/litellm-dashboard/src/app/(dashboard)/api-reference/page.tsx b/ui/litellm-dashboard/src/app/(dashboard)/api-reference/page.tsx index f9f26c0eac5..02bed1adbe5 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/api-reference/page.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/api-reference/page.tsx @@ -1,16 +1,10 @@ "use client"; import APIReferenceView from "@/app/(dashboard)/api-reference/APIReferenceView"; -import { useState } from "react"; - -interface ProxySettings { - PROXY_BASE_URL: string; - PROXY_LOGOUT_URL: string; - LITELLM_UI_API_DOC_BASE_URL?: string | null; -} +import useProxySettings from "@/app/(dashboard)/hooks/proxySettings/useProxySettings"; const APIReferencePage = () => { - const [proxySettings, setProxySettings] = useState({ PROXY_BASE_URL: "", PROXY_LOGOUT_URL: "" }); + const proxySettings = useProxySettings(); return ; }; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/components/Sidebar2.tsx b/ui/litellm-dashboard/src/app/(dashboard)/components/Sidebar2.tsx index 18ab475f228..27a6e6c13be 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/components/Sidebar2.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/components/Sidebar2.tsx @@ -195,7 +195,7 @@ const menuItems: MenuItemCfg[] = [ icon: , roles: all_admin_roles, }, - { key: "14", page: "api_ref", label: "API Reference", icon: }, + { key: "14", page: "api-reference", label: "API Reference", icon: }, { key: "16", page: "model-hub-table", diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/common/queryKeysFactory.test.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/common/queryKeysFactory.test.ts new file mode 100644 index 00000000000..39afd044097 --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/common/queryKeysFactory.test.ts @@ -0,0 +1,34 @@ +import { describe, it, expect } from "vitest"; +import { createQueryKeys } from "./queryKeysFactory"; + +describe("createQueryKeys", () => { + const keys = createQueryKeys("books"); + + it("should return the resource name as the base key", () => { + expect(keys.all).toEqual(["books"]); + }); + + it("should generate a lists key", () => { + expect(keys.lists()).toEqual(["books", "list"]); + }); + + it("should generate a list key with params", () => { + expect(keys.list({ page: 1, limit: 10 })).toEqual([ + "books", + "list", + { params: { page: 1, limit: 10 } }, + ]); + }); + + it("should generate a list key with undefined params when none provided", () => { + expect(keys.list()).toEqual(["books", "list", { params: undefined }]); + }); + + it("should generate a details key", () => { + expect(keys.details()).toEqual(["books", "detail"]); + }); + + it("should generate a detail key for a specific ID", () => { + expect(keys.detail("123")).toEqual(["books", "detail", "123"]); + }); +}); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/login/useLogin.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/login/useLogin.ts index a15b4a06d13..be53b1c80a8 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/hooks/login/useLogin.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/login/useLogin.ts @@ -3,8 +3,8 @@ import { loginCall, LoginRequest } from "@/components/networking"; export const useLogin = () => { return useMutation({ - mutationFn: async ({ username, password }: LoginRequest) => { - const result = await loginCall(username, password); + mutationFn: async ({ username, password, useV3 }: LoginRequest) => { + const result = await loginCall(username, password, useV3); return result; }, }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/proxySettings/useProxySettings.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/proxySettings/useProxySettings.ts new file mode 100644 index 00000000000..d4fb3073856 --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/proxySettings/useProxySettings.ts @@ -0,0 +1,21 @@ +import { useState, useEffect } from "react"; +import { fetchProxySettings } from "@/utils/proxyUtils"; +import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; + +export default function useProxySettings() { + const { accessToken } = useAuthorized(); + const [proxySettings, setProxySettings] = useState({ + PROXY_BASE_URL: "", + PROXY_LOGOUT_URL: "", + LITELLM_UI_API_DOC_BASE_URL: null as string | null, + }); + + useEffect(() => { + if (!accessToken) return; + fetchProxySettings(accessToken).then((settings) => { + if (settings) setProxySettings(settings); + }); + }, [accessToken]); + + return proxySettings; +} diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/uiConfig/useUIConfig.test.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/uiConfig/useUIConfig.test.ts index aba5dddf13d..b05bae1e818 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/hooks/uiConfig/useUIConfig.test.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/uiConfig/useUIConfig.test.ts @@ -28,6 +28,8 @@ const mockUIConfig: LiteLLMWellKnownUiConfig = { proxy_base_url: "https://proxy.example.com", auto_redirect_to_sso: true, admin_ui_disabled: false, + is_control_plane: false, + workers: [], }; describe("useUIConfig", () => { @@ -102,6 +104,8 @@ describe("useUIConfig", () => { auto_redirect_to_sso: false, sso_configured: false, admin_ui_disabled: true, + is_control_plane: false, + workers: [], }; // Mock successful API call with different data diff --git a/ui/litellm-dashboard/src/app/(dashboard)/teams/components/TeamsHeaderTabs.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/teams/components/TeamsHeaderTabs.test.tsx new file mode 100644 index 00000000000..50a7f10f047 --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/teams/components/TeamsHeaderTabs.test.tsx @@ -0,0 +1,54 @@ +import { render, screen } from "@testing-library/react"; +import React from "react"; +import { describe, expect, it, vi } from "vitest"; +import TeamsHeaderTabs from "./TeamsHeaderTabs"; + +vi.mock("@tremor/react", () => ({ + TabGroup: ({ children, ...props }: any) =>
{children}
, + TabList: ({ children, ...props }: any) =>
{children}
, + Tab: ({ children, ...props }: any) => , + TabPanels: ({ children, ...props }: any) =>
{children}
, + Text: ({ children, ...props }: any) => {children}, + Icon: ({ onClick, ...props }: any) => + + ); + } + + return ( + + columns={teamColumns} + dataSource={displayTeams} + rowKey="team_id" + pagination={false} + onChange={handleTableSort} + locale={{ + emptyText: ( +
+ +
+ No teams yet +
+
+ + Create your first team to organize members and manage access to models. + +
+ {canCreateOrManageTeams(userRole, userID, organizations) && ( + + )} +
+ ), + }} + scroll={{ x: 1000 }} + size="middle" + /> + ); + }; + + const tabItems = [ + { + key: "your-teams", + label: "Your Teams", + children: ( + <> + + + + } + suffix={isSearching ? : null} + placeholder="Search teams by name..." + onChange={(e) => handleSearchChange(e.target.value)} + allowClear + style={{ maxWidth: 400 }} + /> + handleFilterChange("organization_id", value || "")} + loading={isLoading} + /> + + { + setCurrentPage(page); + setPageSize(size); + fetchTeamsV2({ page, size }); + }} + size="small" + showTotal={(total) => `${total} teams`} + showSizeChanger + pageSizeOptions={["10", "20", "50"]} + /> + + + {renderTeamsContent()} + + + + + ), + }, + { + key: "available-teams", + label: "Available Teams", + children: , + }, + ...(isProxyAdminRole(userRole || "") + ? [ + { + key: "default-settings", + label: "Default Team Settings", + children: , + }, + ] + : []), + ]; + return ( -
- - - {canCreateOrManageTeams(userRole, userID, organizations) && ( - - )} - {selectedTeamId ? ( - { - setTeams((teams) => { - if (teams == null) { - return teams; - } - const updated = teams.map((team) => { - if (data.team_id === team.team_id) { - return updateExistingKeys(team, data); - } - return team; - }); - // Minimal fix: refresh the full team list after an update - if (accessToken) { - fetchTeams(accessToken, userID, userRole, currentOrg, setTeams); - } - return updated; - }); - }} - onClose={() => { - setSelectedTeamId(null); - setEditTeam(false); - }} - accessToken={accessToken} - is_team_admin={is_team_admin(teams?.find((team) => team.team_id === selectedTeamId))} - is_proxy_admin={userRole == "Admin"} - userModels={userModels} - editTeam={editTeam} - premiumUser={premiumUser} - /> - ) : ( - - -
- Your Teams - Available Teams - {isProxyAdminRole(userRole || "") && Default Team Settings} -
-
- {lastRefreshed && Last Refreshed: {lastRefreshed}} - -
- - - - - Click on “Team ID” to view team details and manage team members. - - - - -
-
- {/* Search and Filter Controls */} -
- {/* Team Alias Search */} - handleFilterChange("team_alias", value)} - icon={Search} - /> + + {selectedTeamId ? ( + { + setTeams((teams) => { + if (teams == null) { + return teams; + } + return teams.map((team) => { + if (data.team_id === team.team_id) { + return updateExistingKeys(team, data); + } + return team; + }); + }); + fetchTeamsV2(); + }} + onClose={() => { + setSelectedTeamId(null); + setEditTeam(false); + }} + accessToken={accessToken} + is_team_admin={is_team_admin(teams?.find((team) => team.team_id === selectedTeamId))} + is_proxy_admin={userRole == "Admin"} + userModels={userModels} + editTeam={editTeam} + premiumUser={premiumUser} + /> + ) : ( + <> + + + + <TeamOutlined style={{ marginRight: 8 }} /> + Teams + + + Manage teams, members, and their access to models and budgets + + + {canCreateOrManageTeams(userRole, userID, organizations) && ( + + )} + - {/* Filter Button */} - setShowFilters(!showFilters)} - active={showFilters} - hasActiveFilters={!!(filters.team_id || filters.team_alias || filters.organization_id)} - /> + + + )} - {/* Reset Filters Button */} - -
- - {/* Additional Filters */} - {showFilters && ( -
- {/* Team ID Search */} - handleFilterChange("team_id", value)} - icon={User} - /> - - {/* Organization Dropdown */} -
- -
-
- )} -
-
- - - - Team Name - Team ID - Created - Spend (USD) - Budget (USD) - Models - Organization - Info - Actions - - - - - {teams && teams.length > 0 ? ( - teams - .filter((team) => { - if (!currentOrg) return true; - return team.organization_id === currentOrg.organization_id; - }) - .sort((a, b) => new Date(b.created_at).getTime() - new Date(a.created_at).getTime()) - .map((team: any) => ( - - - {team["team_alias"]} - - -
- - - -
- - - {team.created_at ? new Date(team.created_at).toLocaleDateString() : "N/A"} - - - {formatNumberWithCommas(team["spend"], 4)} - - - {team["max_budget"] !== null && team["max_budget"] !== undefined - ? team["max_budget"] - : "No limit"} - - 3 ? "px-0" : ""} - > -
- {Array.isArray(team.models) ? ( -
- {team.models.length === 0 ? ( - - All Proxy Models - - ) : ( - <> -
- {team.models.length > 3 && ( -
- { - setExpandedAccordions((prev) => ({ - ...prev, - [team.team_id]: !prev[team.team_id], - })); - }} - /> -
- )} -
- {team.models.slice(0, 3).map((model: string, index: number) => - model === "all-proxy-models" ? ( - - All Proxy Models - - ) : ( - - - {model.length > 30 - ? `${getModelDisplayName(model).slice(0, 30)}...` - : getModelDisplayName(model)} - - - ), - )} - {team.models.length > 3 && !expandedAccordions[team.team_id] && ( - - - +{team.models.length - 3}{" "} - {team.models.length - 3 === 1 ? "more model" : "more models"} - - - )} - {expandedAccordions[team.team_id] && ( -
- {team.models.slice(3).map((model: string, index: number) => - model === "all-proxy-models" ? ( - - All Proxy Models - - ) : ( - - - {model.length > 30 - ? `${getModelDisplayName(model).slice(0, 30)}...` - : getModelDisplayName(model)} - - - ), - )} -
- )} -
-
- - )} -
- ) : null} -
- - - - {getOrganizationAlias(team.organization_id, organizationsData || organizations)} - - - - {perTeamInfo && - team.team_id && - perTeamInfo[team.team_id] && - perTeamInfo[team.team_id].keys && - perTeamInfo[team.team_id].keys.length}{" "} - Keys - - - {perTeamInfo && - team.team_id && - perTeamInfo[team.team_id] && - perTeamInfo[team.team_id].team_info && - perTeamInfo[team.team_id].team_info.members_with_roles && - perTeamInfo[team.team_id].team_info.members_with_roles.length}{" "} - Members - - - - {userRole == "Admin" ? ( - <> - { - setSelectedTeamId(team.team_id); - setEditTeam(true); - }} - dataTestId="edit-team-button" - tooltipText="Edit team" - /> - handleDelete(team)} - dataTestId="delete-team-button" - tooltipText="Delete team" - /> - - ) : null} - - - )) - ) : ( - - -
- No teams found - Adjust your filters or create a new team -
- - - )} - -
- - - - - - - - - {isProxyAdminRole(userRole || "") && ( - - - - )} - - - )} - {canCreateOrManageTeams(userRole, userID, organizations) && ( + {canCreateOrManageTeams(userRole, userID, organizations) && ( = ({ : "" } > - = ({ optionFilterProp="children" > {adminOrgs?.map((org) => ( - + {org.organization_alias}{" "} ({org.organization_id}) - + ))} - + {/* Show message when org admin needs to select organization */} {isOrgAdmin && !isSingleOrg && adminOrgs.length > 1 && (
- + Please select an organization to create a team for. You can only create teams within organizations where you are an admin. @@ -1190,11 +1211,11 @@ const Teams: React.FC = ({ - - daily - weekly - monthly - + @@ -1313,7 +1334,7 @@ const Teams: React.FC = ({ className="mt-8" help="Select existing guardrails or enter new ones" > - = ({ className="mt-8" help="Select existing policies or enter new ones" > - = ({
- + Create custom aliases for models that can be used by team members in API calls. This allows you to create shortcuts for specific models. @@ -1548,14 +1569,12 @@ const Teams: React.FC = ({
- Create Team +
)} - - -
+ ); }; diff --git a/ui/litellm-dashboard/src/components/Projects/ProjectModals/CreateProjectModal.tsx b/ui/litellm-dashboard/src/components/Projects/ProjectModals/CreateProjectModal.tsx index e490f89303f..bbf56e4930a 100644 --- a/ui/litellm-dashboard/src/components/Projects/ProjectModals/CreateProjectModal.tsx +++ b/ui/litellm-dashboard/src/components/Projects/ProjectModals/CreateProjectModal.tsx @@ -1,5 +1,6 @@ -import { Modal, Form, Button, Typography, message } from "antd"; +import { Modal, Form, Button, Typography } from "antd"; import { FolderAddOutlined } from "@ant-design/icons"; +import MessageManager from "@/components/molecules/message_manager"; import { useCreateProject, ProjectCreateParams, @@ -32,12 +33,12 @@ export function CreateProjectModal({ createMutation.mutate(params, { onSuccess: () => { - message.success("Project created successfully"); + MessageManager.success("Project created successfully"); form.resetFields(); onClose(); }, onError: (error) => { - message.error(error.message || "Failed to create project"); + MessageManager.error(error.message || "Failed to create project"); }, }); } catch (error) { diff --git a/ui/litellm-dashboard/src/components/Projects/ProjectModals/EditProjectModal.tsx b/ui/litellm-dashboard/src/components/Projects/ProjectModals/EditProjectModal.tsx index 75f56b1373f..dc3b43ef73c 100644 --- a/ui/litellm-dashboard/src/components/Projects/ProjectModals/EditProjectModal.tsx +++ b/ui/litellm-dashboard/src/components/Projects/ProjectModals/EditProjectModal.tsx @@ -1,6 +1,7 @@ import { useEffect } from "react"; -import { Modal, Form, Button, Typography, message } from "antd"; +import { Modal, Form, Button, Typography } from "antd"; import { SaveOutlined } from "@ant-design/icons"; +import MessageManager from "@/components/molecules/message_manager"; import { ProjectResponse } from "@/app/(dashboard)/hooks/projects/useProjects"; import { useUpdateProject, @@ -80,12 +81,12 @@ export function EditProjectModal({ { projectId: project.project_id, params }, { onSuccess: () => { - message.success("Project updated successfully"); + MessageManager.success("Project updated successfully"); onSuccess?.(); onClose(); }, onError: (error) => { - message.error(error.message || "Failed to update project"); + MessageManager.error(error.message || "Failed to update project"); }, }, ); diff --git a/ui/litellm-dashboard/src/components/SearchTools/SearchToolTester.tsx b/ui/litellm-dashboard/src/components/SearchTools/SearchToolTester.tsx index 28b34b06089..d1cb5077b18 100644 --- a/ui/litellm-dashboard/src/components/SearchTools/SearchToolTester.tsx +++ b/ui/litellm-dashboard/src/components/SearchTools/SearchToolTester.tsx @@ -1,5 +1,6 @@ import React, { useState } from "react"; -import { Button, Input, Typography, Spin, message } from "antd"; +import { Button, Input, Typography, Spin } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { SearchOutlined, LoadingOutlined } from "@ant-design/icons"; import { searchToolQueryCall } from "../networking"; import NotificationsManager from "../molecules/notifications_manager"; @@ -39,7 +40,7 @@ export const SearchToolTester: React.FC = ({ searchToolNa const handleSearch = async () => { if (!query.trim()) { - message.warning("Please enter a search query"); + MessageManager.warning("Please enter a search query"); return; } diff --git a/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/AddFallbacks.tsx b/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/AddFallbacks.tsx index 34f059516a6..e55089d27dc 100644 --- a/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/AddFallbacks.tsx +++ b/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/AddFallbacks.tsx @@ -5,8 +5,9 @@ */ import { Button as TremorButton } from "@tremor/react"; -import { Button, message } from "antd"; +import { Button } from "antd"; import React, { useEffect, useState } from "react"; +import MessageManager from "@/components/molecules/message_manager"; import NotificationManager from "../../../molecules/notifications_manager"; import { fetchAvailableModels, ModelGroup } from "../../../playground/llm_calls/fetch_models"; import { AddFallbacksModal } from "./AddFallbacksModal"; @@ -90,7 +91,7 @@ export default function AddFallbacks({ (g) => !g.primaryModel || g.fallbackModels.length === 0, ); if (invalidGroups.length > 0) { - message.error( + MessageManager.error( `Please complete configuration for all groups. ${invalidGroups.length} group(s) incomplete.`, ); return; diff --git a/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/FallbackSelectionForm.tsx b/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/FallbackSelectionForm.tsx index 08b031c6835..0482161bc6f 100644 --- a/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/FallbackSelectionForm.tsx +++ b/ui/litellm-dashboard/src/components/Settings/RouterSettings/Fallbacks/FallbackSelectionForm.tsx @@ -5,9 +5,10 @@ */ import { Button } from "@tremor/react"; -import { message, Tabs } from "antd"; +import { Tabs } from "antd"; import { Plus } from "lucide-react"; import React, { useEffect, useState } from "react"; +import MessageManager from "@/components/molecules/message_manager"; import { FallbackGroup, FallbackGroupConfig } from "./FallbackGroupConfig"; interface FallbackSelectionFormProps { @@ -60,7 +61,7 @@ export function FallbackSelectionForm({ const handleRemoveGroup = (targetId: string) => { if (groups.length === 1) { - message.warning("At least one group is required"); + MessageManager.warning("At least one group is required"); return; } const newGroups = groups.filter((g) => g.id !== targetId); diff --git a/ui/litellm-dashboard/src/components/TeamSSOSettings.tsx b/ui/litellm-dashboard/src/components/TeamSSOSettings.tsx index a9c07cdbcf8..98745d388a2 100644 --- a/ui/litellm-dashboard/src/components/TeamSSOSettings.tsx +++ b/ui/litellm-dashboard/src/components/TeamSSOSettings.tsx @@ -26,6 +26,10 @@ const PERMISSION_OPTIONS = [ "/key/unblock", "/key/bulk_update", "/key/{key_id}/reset_spend", + "/key/info", + "/key/list", + "/key/aliases", + "/team/daily/activity", ]; interface SettingRowProps { diff --git a/ui/litellm-dashboard/src/components/agents/add_agent_form.tsx b/ui/litellm-dashboard/src/components/agents/add_agent_form.tsx index c5518596b81..b6d96f0445b 100644 --- a/ui/litellm-dashboard/src/components/agents/add_agent_form.tsx +++ b/ui/litellm-dashboard/src/components/agents/add_agent_form.tsx @@ -1,5 +1,6 @@ import React, { useState, useEffect } from "react"; -import { Modal, Form, message, Select, Input, Steps, Radio, Tag, Divider, Switch, InputNumber, Collapse } from "antd"; +import { Modal, Form, Select, Input, Steps, Radio, Tag, Divider, Switch, InputNumber, Collapse } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { Button } from "@tremor/react"; import { CheckCircleFilled, KeyOutlined, RobotOutlined, AppstoreOutlined, InfoCircleOutlined } from "@ant-design/icons"; import CreatedKeyDisplay from "../shared/CreatedKeyDisplay"; @@ -216,7 +217,7 @@ const AddAgentForm: React.FC = ({ const handleCreateAgent = async () => { if (!accessToken) { - message.error("No access token available"); + MessageManager.error("No access token available"); return; } @@ -226,7 +227,7 @@ const AddAgentForm: React.FC = ({ const values = { ...form.getFieldsValue(true) }; const agentData = buildAgentData(values); if (!agentData) { - message.error("Failed to build agent data"); + MessageManager.error("Failed to build agent data"); setIsSubmitting(false); return; } @@ -301,7 +302,7 @@ const AddAgentForm: React.FC = ({ setCreatedKeyValue(keyResponse.key || null); } else if (keyAssignOption === "existing_key") { if (!selectedExistingKey) { - message.error("Please select an existing key to assign"); + MessageManager.error("Please select an existing key to assign"); setIsSubmitting(false); return; } @@ -318,7 +319,7 @@ const AddAgentForm: React.FC = ({ } catch (error) { console.error("Error creating agent:", error); const errorMessage = error instanceof Error ? error.message : String(error); - message.error(errorMessage ? `Failed to create agent: ${errorMessage}` : "Failed to create agent"); + MessageManager.error(errorMessage ? `Failed to create agent: ${errorMessage}` : "Failed to create agent"); } finally { setIsSubmitting(false); } diff --git a/ui/litellm-dashboard/src/components/agents/agent_info.tsx b/ui/litellm-dashboard/src/components/agents/agent_info.tsx index b41e318a766..d543be8356a 100644 --- a/ui/litellm-dashboard/src/components/agents/agent_info.tsx +++ b/ui/litellm-dashboard/src/components/agents/agent_info.tsx @@ -1,6 +1,7 @@ import React, { useState, useEffect } from "react"; import { Card, Title, Text, Button as TremorButton, Tab, TabGroup, TabList, TabPanel, TabPanels} from "@tremor/react"; -import { Form, Input, InputNumber, Button as AntButton, message, Spin, Descriptions, Divider } from "antd"; +import { Form, Input, InputNumber, Button as AntButton, Spin, Descriptions, Divider } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { ArrowLeftIcon } from "@heroicons/react/outline"; import { getAgentInfo, patchAgentCall, getAgentCreateMetadata, AgentCreateInfo } from "../networking"; import { Agent } from "./types"; @@ -72,7 +73,7 @@ const AgentInfoView: React.FC = ({ } } catch (error) { console.error("Error fetching agent info:", error); - message.error("Failed to load agent information"); + MessageManager.error("Failed to load agent information"); } finally { setIsLoading(false); } @@ -111,12 +112,12 @@ const AgentInfoView: React.FC = ({ } await patchAgentCall(accessToken, agentId, updateData); - message.success("Agent updated successfully"); + MessageManager.success("Agent updated successfully"); setIsEditing(false); fetchAgentInfo(); } catch (error) { console.error("Error updating agent:", error); - message.error("Failed to update agent"); + MessageManager.error("Failed to update agent"); } finally { setIsSaving(false); } diff --git a/ui/litellm-dashboard/src/components/chat/ChatPage.tsx b/ui/litellm-dashboard/src/components/chat/ChatPage.tsx index ccf39d2147e..b547f749175 100644 --- a/ui/litellm-dashboard/src/components/chat/ChatPage.tsx +++ b/ui/litellm-dashboard/src/components/chat/ChatPage.tsx @@ -1,7 +1,8 @@ "use client"; import React, { useCallback, useEffect, useRef, useState, useLayoutEffect } from "react"; -import { Tooltip, Skeleton, Popover, message } from "antd"; +import { Tooltip, Skeleton, Popover } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { SettingOutlined, PlusOutlined, @@ -212,7 +213,7 @@ const ChatPage: React.FC = ({ accessToken, userRole, userId, user localStorage.setItem(LOCALSTORAGE_MODEL_KEY, JSON.stringify([names[0]])); } }) - .catch(() => message.error("Could not load models")) + .catch(() => MessageManager.error("Could not load models")) .finally(() => setIsLoadingModels(false)); }, [accessToken]); diff --git a/ui/litellm-dashboard/src/components/chat/MCPAppsPanel.tsx b/ui/litellm-dashboard/src/components/chat/MCPAppsPanel.tsx index b805cab71ef..d25db73ae46 100644 --- a/ui/litellm-dashboard/src/components/chat/MCPAppsPanel.tsx +++ b/ui/litellm-dashboard/src/components/chat/MCPAppsPanel.tsx @@ -5,7 +5,7 @@ import { Spin, Input, Button, Skeleton } from "antd"; import { SearchOutlined, ArrowLeftOutlined, RightOutlined, ToolOutlined, CheckCircleOutlined } from "@ant-design/icons"; import { deleteMCPOAuthUserCredential, fetchMCPServers, getMCPOAuthUserCredentialStatus, listMCPTools } from "../networking"; import { AUTH_TYPE, MCPServer, MCPTool, handleTransport } from "../mcp_tools/types"; -import { message } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { useUserMcpOAuthFlow } from "@/hooks/useUserMcpOAuthFlow"; // ── OAuth2 connect button ───────────────────────────────────────────────────── @@ -198,7 +198,7 @@ const MCPAppsPanel: React.FC = ({ accessToken, selectedServers, onChange const idToFetch = serverId ?? serverName; const result = await listMCPTools(accessToken, idToFetch); if (result?.error) { - message.warning(`Could not load tools for ${serverName}`); + MessageManager.warning(`Could not load tools for ${serverName}`); return; } // Use the ref so we read the most up-to-date list; guard against duplicates @@ -207,7 +207,7 @@ const MCPAppsPanel: React.FC = ({ accessToken, selectedServers, onChange onChange([...selectedServersRef.current, serverName]); } } catch { - message.warning(`Could not load tools for ${serverName}`); + MessageManager.warning(`Could not load tools for ${serverName}`); } finally { setTogglingOn((prev) => { const next = new Set(prev); diff --git a/ui/litellm-dashboard/src/components/chat/MCPConnectPicker.tsx b/ui/litellm-dashboard/src/components/chat/MCPConnectPicker.tsx index a52ce181568..6ef3aecc469 100644 --- a/ui/litellm-dashboard/src/components/chat/MCPConnectPicker.tsx +++ b/ui/litellm-dashboard/src/components/chat/MCPConnectPicker.tsx @@ -1,5 +1,6 @@ import React, { useEffect, useState } from "react"; -import { Switch, Spin, message } from "antd"; +import { Switch, Spin } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { fetchMCPServers, listMCPTools } from "../networking"; import { MCPServer } from "../mcp_tools/types"; @@ -57,7 +58,7 @@ const MCPConnectPicker: React.FC = ({ accessToken, selectedServers, onCha const result = await listMCPTools(accessToken, serverName); // listMCPTools never throws; it returns { tools, error, message } on failure if (result?.error) { - message.warning( + MessageManager.warning( `Could not load tools for ${serverName} — it will be excluded from this message.` ); // Do not add to selectedServers @@ -65,7 +66,7 @@ const MCPConnectPicker: React.FC = ({ accessToken, selectedServers, onCha } onChange([...selectedServers, serverName]); } catch { - message.warning( + MessageManager.warning( `Could not load tools for ${serverName} — it will be excluded from this message.` ); // Do not add to selectedServers diff --git a/ui/litellm-dashboard/src/components/chat/MCPCredentialsTab.tsx b/ui/litellm-dashboard/src/components/chat/MCPCredentialsTab.tsx index 363ec8c0e40..e2d879b22a8 100644 --- a/ui/litellm-dashboard/src/components/chat/MCPCredentialsTab.tsx +++ b/ui/litellm-dashboard/src/components/chat/MCPCredentialsTab.tsx @@ -8,7 +8,8 @@ */ import React, { useCallback, useEffect, useState } from "react"; -import { Spin, message } from "antd"; +import { Spin } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { DeleteOutlined, LinkOutlined } from "@ant-design/icons"; import { Badge, Table, TableBody, TableCell, TableHead, TableHeaderCell, TableRow } from "@tremor/react"; import { @@ -77,7 +78,7 @@ const MCPCredentialsTab: React.FC = ({ accessToken }) => { await deleteMCPOAuthUserCredential(accessToken, serverId); setCredentials((prev) => prev.filter((c) => c.server_id !== serverId)); } catch { - message.error("Failed to revoke connection. Please try again."); + MessageManager.error("Failed to revoke connection. Please try again."); } finally { setRevoking((prev) => { const n = new Set(prev); n.delete(serverId); return n; }); } diff --git a/ui/litellm-dashboard/src/components/claude_code_plugins/add_plugin_form.tsx b/ui/litellm-dashboard/src/components/claude_code_plugins/add_plugin_form.tsx index 217851f128f..d5e417a9a85 100644 --- a/ui/litellm-dashboard/src/components/claude_code_plugins/add_plugin_form.tsx +++ b/ui/litellm-dashboard/src/components/claude_code_plugins/add_plugin_form.tsx @@ -1,5 +1,6 @@ import React, { useState } from "react"; -import { Modal, Form, Input, Select, message } from "antd"; +import { Modal, Form, Input, Select } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { Button } from "@tremor/react"; import { registerClaudeCodePlugin } from "../networking"; import { @@ -43,13 +44,13 @@ const AddPluginForm: React.FC = ({ const handleSubmit = async (values: any) => { if (!accessToken) { - message.error("No access token available"); + MessageManager.error("No access token available"); return; } // Validate plugin name if (!validatePluginName(values.name)) { - message.error( + MessageManager.error( "Plugin name must be kebab-case (lowercase letters, numbers, and hyphens only)" ); return; @@ -57,7 +58,7 @@ const AddPluginForm: React.FC = ({ // Validate semantic version if provided if (values.version && !isValidSemanticVersion(values.version)) { - message.error( + MessageManager.error( "Version must be in semantic versioning format (e.g., 1.0.0)" ); return; @@ -65,13 +66,13 @@ const AddPluginForm: React.FC = ({ // Validate email if provided if (values.authorEmail && !isValidEmail(values.authorEmail)) { - message.error("Invalid email format"); + MessageManager.error("Invalid email format"); return; } // Validate homepage URL if provided if (values.homepage && !isValidUrl(values.homepage)) { - message.error("Invalid homepage URL format"); + MessageManager.error("Invalid homepage URL format"); return; } @@ -119,14 +120,14 @@ const AddPluginForm: React.FC = ({ } await registerClaudeCodePlugin(accessToken, pluginData); - message.success("Plugin registered successfully"); + MessageManager.success("Plugin registered successfully"); form.resetFields(); setSourceType("github"); onSuccess(); onClose(); } catch (error) { console.error("Error registering plugin:", error); - message.error("Failed to register plugin"); + MessageManager.error("Failed to register plugin"); } finally { setIsSubmitting(false); } diff --git a/ui/litellm-dashboard/src/components/common_components/IconActionButton/TableIconActionButtons/TableIconActionButton.tsx b/ui/litellm-dashboard/src/components/common_components/IconActionButton/TableIconActionButtons/TableIconActionButton.tsx index 488913a734a..2f146aab723 100644 --- a/ui/litellm-dashboard/src/components/common_components/IconActionButton/TableIconActionButtons/TableIconActionButton.tsx +++ b/ui/litellm-dashboard/src/components/common_components/IconActionButton/TableIconActionButtons/TableIconActionButton.tsx @@ -6,6 +6,7 @@ import { ChevronUpIcon, ChevronDownIcon, ExternalLinkIcon, + ClipboardCopyIcon, } from "@heroicons/react/outline"; import { Tooltip } from "antd"; import BaseActionButton from "../BaseActionButton"; @@ -32,6 +33,7 @@ export const TableIconActionButtonMap: Record void; disabled?: boolean; loading?: boolean; + style?: React.CSSProperties; } const OrganizationDropdown: React.FC = ({ @@ -16,16 +19,18 @@ const OrganizationDropdown: React.FC = ({ onChange, disabled, loading, + style, }) => { return ( diff --git a/ui/litellm-dashboard/src/components/key_team_helpers/filter_helpers.test.ts b/ui/litellm-dashboard/src/components/key_team_helpers/filter_helpers.test.ts new file mode 100644 index 00000000000..e3f6a5989f4 --- /dev/null +++ b/ui/litellm-dashboard/src/components/key_team_helpers/filter_helpers.test.ts @@ -0,0 +1,90 @@ +import { describe, expect, it, vi } from "vitest"; +import { fetchTeamFilterOptions } from "./filter_helpers"; + +const mockKeyListCall = vi.fn(); + +vi.mock("@/components/networking", () => ({ + keyListCall: (...args: unknown[]) => mockKeyListCall(...args), + teamListCall: vi.fn(), + organizationListCall: vi.fn(), +})); + +describe("fetchTeamFilterOptions", () => { + it("should return empty arrays when accessToken is null", async () => { + const result = await fetchTeamFilterOptions(null, "team-1"); + + expect(result).toEqual({ keyAliases: [], organizationIds: [], userIds: [] }); + expect(mockKeyListCall).not.toHaveBeenCalled(); + }); + + it("should return empty arrays when teamId is empty", async () => { + const result = await fetchTeamFilterOptions("tok-123", ""); + + expect(result).toEqual({ keyAliases: [], organizationIds: [], userIds: [] }); + expect(mockKeyListCall).not.toHaveBeenCalled(); + }); + + it("should return sorted key aliases from fetched keys", async () => { + mockKeyListCall.mockResolvedValue({ + keys: [ + { key_alias: "zeta-key" }, + { key_alias: "alpha-key" }, + { key_alias: "mid-key" }, + ], + total_pages: 1, + }); + + const result = await fetchTeamFilterOptions("tok-123", "team-1"); + + expect(result.keyAliases).toEqual(["alpha-key", "mid-key", "zeta-key"]); + }); + + it("should deduplicate organization IDs across pages", async () => { + mockKeyListCall + .mockResolvedValueOnce({ + keys: [ + { organization_id: "org-b" }, + { organization_id: "org-a" }, + ], + total_pages: 2, + }) + .mockResolvedValueOnce({ + keys: [ + { organization_id: "org-a" }, + { organization_id: "org-c" }, + ], + total_pages: 2, + }); + + const result = await fetchTeamFilterOptions("tok-123", "team-1"); + + expect(result.organizationIds).toEqual(["org-a", "org-b", "org-c"]); + }); + + it("should map user IDs with email addresses", async () => { + mockKeyListCall.mockResolvedValue({ + keys: [ + { user_id: "u1", user: { user_email: "alice@example.com" } }, + { user_id: "u2", user: { user_email: "bob@example.com" } }, + ], + total_pages: 1, + }); + + const result = await fetchTeamFilterOptions("tok-123", "team-1"); + + expect(result.userIds).toEqual( + expect.arrayContaining([ + { id: "u1", email: "alice@example.com" }, + { id: "u2", email: "bob@example.com" }, + ]), + ); + }); + + it("should handle API errors gracefully and return empty arrays", async () => { + mockKeyListCall.mockRejectedValue(new Error("Network error")); + + const result = await fetchTeamFilterOptions("tok-123", "team-1"); + + expect(result).toEqual({ keyAliases: [], organizationIds: [], userIds: [] }); + }); +}); diff --git a/ui/litellm-dashboard/src/components/key_team_helpers/transform_key_info.test.ts b/ui/litellm-dashboard/src/components/key_team_helpers/transform_key_info.test.ts new file mode 100644 index 00000000000..a1139addbf6 --- /dev/null +++ b/ui/litellm-dashboard/src/components/key_team_helpers/transform_key_info.test.ts @@ -0,0 +1,62 @@ +import { describe, it, expect } from "vitest"; +import { transformKeyInfo } from "./transform_key_info"; + +describe("transformKeyInfo", () => { + it("should combine key and info fields into a single object", () => { + const apiResponse = { + key: "sk-abc123", + info: { + token_id: "tok_1", + key_name: "my-key", + spend: 10.5, + }, + }; + const result = transformKeyInfo(apiResponse); + expect(result).toEqual({ + token: "sk-abc123", + token_id: "tok_1", + key_name: "my-key", + spend: 10.5, + }); + }); + + it("should set the token field from the key property", () => { + const apiResponse = { + key: "sk-xyz789", + info: { key_name: "test" }, + }; + const result = transformKeyInfo(apiResponse); + expect(result.token).toBe("sk-xyz789"); + }); + + it("should preserve all info fields in the result", () => { + const apiResponse = { + key: "sk-abc", + info: { + token_id: "tok_2", + key_name: "prod-key", + spend: 42, + models: ["gpt-4"], + team_id: "team-1", + metadata: { env: "production" }, + }, + }; + const result = transformKeyInfo(apiResponse); + expect(result.token_id).toBe("tok_2"); + expect(result.key_name).toBe("prod-key"); + expect(result.spend).toBe(42); + expect(result.models).toEqual(["gpt-4"]); + expect(result.team_id).toBe("team-1"); + expect(result.metadata).toEqual({ env: "production" }); + }); + + it("should handle empty info object", () => { + const apiResponse = { + key: "sk-empty", + info: {}, + }; + const result = transformKeyInfo(apiResponse); + expect(result.token).toBe("sk-empty"); + expect(Object.keys(result)).toContain("token"); + }); +}); diff --git a/ui/litellm-dashboard/src/components/leftnav.tsx b/ui/litellm-dashboard/src/components/leftnav.tsx index d01fc06bc05..09ab3809427 100644 --- a/ui/litellm-dashboard/src/components/leftnav.tsx +++ b/ui/litellm-dashboard/src/components/leftnav.tsx @@ -13,6 +13,7 @@ import { CreditCardOutlined, DatabaseOutlined, ExperimentOutlined, + ExportOutlined, FileTextOutlined, FolderOutlined, KeyOutlined, @@ -231,8 +232,8 @@ const menuGroups: MenuGroup[] = [ groupLabel: "DEVELOPER TOOLS", items: [ { - key: "api_ref", - page: "api_ref", + key: "api-reference", + page: "api-reference", label: "API Reference", icon: , }, @@ -400,7 +401,7 @@ const Sidebar: React.FC = ({ setPage, defaultSelectedKey, collapse onClick={(e) => e.stopPropagation()} style={{ color: "inherit", textDecoration: "none" }} > - {label} + {label} ); } diff --git a/ui/litellm-dashboard/src/components/mcp_tools/ByokCredentialModal.tsx b/ui/litellm-dashboard/src/components/mcp_tools/ByokCredentialModal.tsx index 473918c1267..58c2a965e03 100644 --- a/ui/litellm-dashboard/src/components/mcp_tools/ByokCredentialModal.tsx +++ b/ui/litellm-dashboard/src/components/mcp_tools/ByokCredentialModal.tsx @@ -1,7 +1,8 @@ "use client"; import React, { useState } from "react"; -import { Modal, Input, Switch, message } from "antd"; +import { Modal, Input, Switch } from "antd"; +import MessageManager from "@/components/molecules/message_manager"; import { KeyOutlined, LockOutlined, @@ -46,7 +47,7 @@ export const ByokCredentialModal: React.FC = ({ const handleAuthorize = async () => { if (!apiKey.trim()) { - message.error("Please enter your API key"); + MessageManager.error("Please enter your API key"); return; } setLoading(true); @@ -63,11 +64,11 @@ export const ByokCredentialModal: React.FC = ({ const err = await response.json(); throw new Error(err?.detail?.error || "Failed to save credential"); } - message.success(`Connected to ${serverDisplayName}`); + MessageManager.success(`Connected to ${serverDisplayName}`); onSuccess(server.server_id); handleClose(); } catch (e: any) { - message.error(e.message || "Failed to connect"); + MessageManager.error(e.message || "Failed to connect"); } finally { setLoading(false); } diff --git a/ui/litellm-dashboard/src/components/molecules/message_manager.tsx b/ui/litellm-dashboard/src/components/molecules/message_manager.tsx new file mode 100644 index 00000000000..e4c1552d5ee --- /dev/null +++ b/ui/litellm-dashboard/src/components/molecules/message_manager.tsx @@ -0,0 +1,38 @@ +import { message as staticMessage } from "antd"; +import type { MessageInstance } from "antd/es/message/interface"; + +let messageInstance: MessageInstance | null = null; + +export const setMessageInstance = (instance: MessageInstance) => { + messageInstance = instance; +}; + +const getMessageApi = () => messageInstance || staticMessage; + +const MessageManager = { + success(content: string, duration?: number) { + getMessageApi().success(content, duration); + }, + + error(content: string, duration?: number) { + getMessageApi().error(content, duration); + }, + + warning(content: string, duration?: number) { + getMessageApi().warning(content, duration); + }, + + info(content: string, duration?: number) { + getMessageApi().info(content, duration); + }, + + loading(content: string, duration?: number) { + return getMessageApi().loading(content, duration); + }, + + destroy() { + getMessageApi().destroy(); + }, +}; + +export default MessageManager; diff --git a/ui/litellm-dashboard/src/components/navbar.tsx b/ui/litellm-dashboard/src/components/navbar.tsx index c46a3af5a6e..96d6ce613bf 100644 --- a/ui/litellm-dashboard/src/components/navbar.tsx +++ b/ui/litellm-dashboard/src/components/navbar.tsx @@ -4,6 +4,7 @@ import { getProxyBaseUrl } from "@/components/networking"; import { useUIConfig } from "@/app/(dashboard)/hooks/uiConfig/useUIConfig"; import { useTheme } from "@/contexts/ThemeContext"; import { clearTokenCookies } from "@/utils/cookieUtils"; +import { clearStoredReturnUrl } from "@/utils/returnUrlUtils"; import { fetchProxySettings } from "@/utils/proxyUtils"; import { MenuFoldOutlined, MenuUnfoldOutlined, MessageOutlined, MoonOutlined, SunOutlined } from "@ant-design/icons"; import { Button, Switch, Tag } from "antd"; @@ -12,6 +13,7 @@ import React, { useEffect, useState } from "react"; import { BlogDropdown } from "./Navbar/BlogDropdown/BlogDropdown"; import { CommunityEngagementButtons } from "./Navbar/CommunityEngagementButtons/CommunityEngagementButtons"; import UserDropdown from "./Navbar/UserDropdown/UserDropdown"; +import WorkerDropdown from "./Navbar/WorkerDropdown/WorkerDropdown"; interface NavbarProps { userID: string | null; @@ -77,9 +79,19 @@ const Navbar: React.FC = ({ const handleLogout = () => { clearTokenCookies(); + localStorage.removeItem("litellm_selected_worker_id"); + localStorage.removeItem("litellm_worker_url"); window.location.href = logoutUrl; }; + const handleWorkerSwitch = (workerId: string) => { + clearTokenCookies(); + clearStoredReturnUrl(); + localStorage.removeItem("litellm_selected_worker_id"); + localStorage.removeItem("litellm_worker_url"); + window.location.href = `/ui/login?worker=${encodeURIComponent(workerId)}`; + }; + return (