* feat(proxy): add NO_OPENAPI env var to disable /openapi.json endpoint (#25696)
* feat(proxy): add NO_OPENAPI env var to disable /openapi.json endpoint - Fixes#25538
* test(proxy): add tests for _get_openapi_url
---------
Co-authored-by: Progressive-engg <lov.kumari55@gmail.com>
* feat(prometheus): add api_provider label to spend metric (#25693)
* feat(prometheus): add api_provider label to spend metric
Add `api_provider` to `litellm_spend_metric` labels so users can
build Grafana dashboards that break down spend by cloud provider
(e.g. bedrock, anthropic, openai, azure, vertex_ai).
The `api_provider` label already exists in UserAPIKeyLabelValues and
is populated from `standard_logging_payload["custom_llm_provider"]`,
but was not included in the spend metric's label list.
* add api_provider to requests metric + add test
Address review feedback:
- Add api_provider to litellm_requests_metric too (same call-site as
spend metric, keeps label sets in sync)
- Add test_api_provider_in_spend_and_requests_metrics following the
existing pattern in test_prometheus_labels.py
* fix: ensure `litellm_metadata` is attached to `pre_call` guardrail to align with `post_call` guardrail (#25641)
* fix: ensure `litellm_metadata` is attached to pre_call to align with post_call
* refactor: remove unused BaseTranslation._ensure_litellm_metadata
* refactor: module level imports for ensure_litellm_metadata and CodeQL
* fix: update based off of Codex comment
* revert: undo usage of `_guardrail_litellm_metadata`
* feat: add pricing entry for openrouter/google/gemini-3.1-flash-lite-preview (#25610)
* fix(bedrock): skip synthetic tool injection for json_object with no schema (#25740)
When response_format={"type": "json_object"} is sent without a JSON
schema, _create_json_tool_call_for_response_format builds a tool with an
empty schema (properties: {}). The model follows the empty schema and
returns {} instead of the actual JSON the caller asked for.
This patch:
- Skips synthetic json_tool_call injection when no schema is provided.
The model already returns JSON when the prompt asks for it.
- Fixes finish_reason: after _filter_json_mode_tools strips all
synthetic tool calls, finish_reason stays "tool_calls" instead of
"stop". Callers (like the OpenAI SDK) misinterpret this as a pending
tool invocation.
json_schema requests with an explicit schema are unchanged.
Co-authored-by: Claude <noreply@anthropic.com>
* fix(utils): allowed_openai_params must not forward unset params as None
`_apply_openai_param_overrides` iterated `allowed_openai_params` and
unconditionally wrote `optional_params[param] = non_default_params.pop(param, None)`
for each entry. If the caller listed a param name but did not actually
send that param in the request, the pop returned `None` and `None` was
still written to `optional_params`. The openai SDK then rejected it as
a top-level kwarg:
AsyncCompletions.create() got an unexpected keyword argument 'enable_thinking'
Reproducer (from #25697):
allowed_openai_params = ["chat_template_kwargs", "enable_thinking"]
body = {"chat_template_kwargs": {"enable_thinking": False}}
Here `enable_thinking` is only present nested inside
`chat_template_kwargs`, so the helper should forward
`chat_template_kwargs` and leave `enable_thinking` alone. Instead it
wrote `optional_params["enable_thinking"] = None`.
Fix: only forward a param if it was actually present in
`non_default_params`. Behavior is unchanged for the happy path (param
sent → still forwarded), and the explicit `None` leakage is gone.
Adds a regression test exercising the helper in isolation so the test
does not depend on any provider-specific `map_openai_params` plumbing.
Fixes#25697
---------
Co-authored-by: lovek629 <59618812+lovek629@users.noreply.github.com>
Co-authored-by: Progressive-engg <lov.kumari55@gmail.com>
Co-authored-by: Ori Kotek <ori.k@codium.ai>
Co-authored-by: Alexander Grattan <51346343+agrattan0820@users.noreply.github.com>
Co-authored-by: Mohana Siddhartha Chivukula <103447836+iamsiddhu3007@users.noreply.github.com>
Co-authored-by: Amiram Mizne <amiramm@users.noreply.github.com>
Co-authored-by: Claude <noreply@anthropic.com>
When modify_params is true, Bedrock Converse setup no longer prepends or
appends the default user message if the boundary assistant turn has
prefix: true, so OpenAI-style assistant prefill reaches the API unchanged.
Made-with: Cursor
* fix(proxy): enforce key-level model allowlist for custom auth
custom_auth_run_common_checks only runs common_checks (team/user/project model checks).
Custom auth now also enforces key-level model restrictions via can_key_call_model.
Move the custom-auth key-access regression tests to test_user_api_key_auth.py and keep test_custom_auth_end_user_budget.py focused on end-user budget behavior.
Made-with: Cursor
* fix(proxy): gate custom-auth key model checks behind opt-in
Keep key-level model allowlist enforcement in custom auth behind `custom_auth_run_common_checks` to preserve backwards compatibility, and update tests to verify default non-enforcement and opt-in enforcement behavior.
Made-with: Cursor
* test(proxy): isolate custom auth default check from shared settings state
Patch `proxy_server.general_settings` to an empty dict in the default custom-auth key-access test so it remains deterministic under shared module state.
Made-with: Cursor
* test(proxy): strengthen custom auth post-check assertions
Tighten custom auth regression tests by asserting exact can_key_call_model args and remove an unused common_checks mock from the default behavior path.
Made-with: Cursor
* fix(agentcore): parse A2A JSON-RPC responses in AgentCore provider
* fix(prompt-templates): ensure_alternating_roles handles tool-call chains
* feat(auth): add JWT claim routing overrides for OAuth2 validation
Made-with: Cursor
* docs(auth): document JWT-to-OAuth2 routing overrides
Add generic docs for running JWT and OAuth2 together, including routing_overrides YAML examples and list-based selector behavior for iss/client_id/aud.
Made-with: Cursor
---------
Co-authored-by: Milan <milan@berri.ai>
Co-authored-by: michelligabriele <gabriele.michelli@icloud.com>
* update bedrock models in tests
* updated more tests and model_prices_and_context_window
* fix model id and pricing
* replace more sonnet models
* update tests
* git push
* update pricing
* flaky total cost
* monkey patch
* relax the cost change
* fix and revert some changes
* revert the pricing
* chore: move cost/pricing changes to bedrock-cost-fixes branch
* chore: split Bedrock file-api beta stripping to separate branch
Removes strip_unsupported_file_api_betas_for_bedrock_invoke from this branch;
see litellm_bedrock_invoke_strip_file_api_betas for that fix.
Made-with: Cursor
Addresses Greptile review feedback: replace direct litellm.model_cost
lookup with the standard _supports_factory infrastructure used by
supports_reasoning, supports_native_streaming, etc.
- Add supports_native_structured_output() utility in litellm/utils.py
- Add supports_native_structured_output field to ModelInfoBase type
- Wire field into _get_model_info_helper return dict
- Delegate from Bedrock _supports_native_structured_outputs to utility
- Add field to JSON schema validator in test_utils.py
Wrap cost-map-dependent tests in try/finally to restore
os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] and litellm.model_cost,
preventing test-ordering sensitivity.
Integration testing confirmed gemma-3 (4b/12b/27b) ignores the JSON
schema and returns free text, and nemotron-nano (9b/12b) errors with
"Tool calling is not supported in streaming mode" even on sync calls.
Remove the flag so these models fall back to the tool-call approach.
Also fix test assertions to match (nemotron-nano-3-30b is supported,
gemma-3 and nemotron-nano-12b are not).
Move the source of truth for which Bedrock models support native structured
outputs (outputConfig.textFormat) from a hardcoded substring set
(BEDROCK_NATIVE_STRUCTURED_OUTPUT_MODELS) to the cost JSON via a new
"supports_native_structured_output" flag. This makes it possible to add
support for new models (including Claude Sonnet 4.6, which was missing)
by updating the JSON alone, with no code changes needed.
Add support for {"location": "tool_config"} in cache_control_injection_points,
which appends a cachePoint block to the Bedrock Converse toolConfig.tools array.
This enables prompt caching of tool definitions on Bedrock Claude models.
Also update the cache control hook to pass through non-message injection points
to provider-specific handling instead of silently dropping them.
Fixes#21969
Keep both sets of tests: upstream's OAuth2 token injection test and
our case-insensitive tool matching tests. Use upstream's version of
the bedrock output_config test (more comprehensive).
- MCP tests: set mock_mcp_server.oauth2_flow = None to prevent MagicMock
leaking into Pydantic Literal validation for MCPServer
- AgentCore tests: pass api_key="test-jwt-token" to bypass SigV4 credential
lookup that fails in CI without AWS credentials
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(agentcore): handle JSON responses from agents using sync return
BedrockAgentCoreApp agents that use synchronous `return` (instead of
async `yield`) respond with Content-Type: application/json instead of
text/event-stream. The streaming parser only handles SSE format, silently
discarding the JSON body and returning empty content to the client.
This adds Content-Type detection in both sync and async streaming
wrappers — when application/json is received, the response is parsed
and converted to a single-chunk stream. Also extends _parse_json_response
with a fallback chain supporting multiple agent response schemas (standard
AgentCore, Strands framework, plain string, raw JSON fallback).
* fix(agentcore): add dict-type guard to _parse_json_response
Prevent AttributeError when json.loads() returns a non-dict
(e.g. JSON array or primitive) by adding an isinstance check
at the top of _parse_json_response. Non-dict values fall back
to raw JSON string content.
* fix(agentcore): handle malformed JSON and split streaming chunks
- Wrap json.loads() in try/except in both sync and async streaming
wrappers so malformed JSON bodies raise a structured BedrockError
instead of a raw JSONDecodeError
- Split the JSON-fallback streaming path into two chunks (content
chunk with finish_reason=None, then stop sentinel with empty delta)
to match the SSE path convention
* fix(agentcore): catch IO errors in streaming JSON path + async error test
- Broaden except clause to catch both json.JSONDecodeError and IO-level
exceptions (httpx.ReadError, etc.) from response.read()/aread(), so
all failures surface as structured BedrockError
- Add async malformed-JSON test to mirror the sync test coverage
* fix(bedrock): strip output_config from Bedrock Invoke requests
Bedrock Invoke API does not support the output_config parameter
(added to Anthropic Messages API). Requests with output_config cause
400 errors: 'extraneous key [output_config] is not permitted'.
Strip output_config in both Bedrock Invoke transformation layers
(messages and chat), consistent with how output_format is already
handled and how VertexAI strips both parameters.
Fixes: https://github.com/BerriAI/litellm/issues/22797
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* test(bedrock): add output_config test for chat/invoke path
Addresses review feedback — the chat/invoke_transformations path now has
symmetric test coverage matching the messages/invoke_transformations path.
Fixes: https://github.com/BerriAI/litellm/issues/22797
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: giulio-leone <6887247+giulio-leone@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Fixes#18381: When using both tools and response_format with Bedrock
Converse API, LiteLLM internally adds json_tool_call to handle structured
output. Bedrock may return both this internal tool AND real user-defined
tools, breaking consumers like OpenAI Agents SDK.
Changes:
- Non-streaming: Added _filter_json_mode_tools() to handle 3 scenarios:
only json_tool_call (convert to content), mixed (filter it out), or
no json_tool_call (pass through)
- Streaming: Added json_mode tracking to AWSEventStreamDecoder to suppress
json_tool_call chunks and convert to text content
- Fixed optional_params.pop() mutation issue
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
AgentCore MCP server endpoints require the Accept header to contain
both application/json and text/event-stream per the MCP specification
(Streamable HTTP transport). Without this header, requests are rejected
with a 406 Not Acceptable error (JSON-RPC code -32011).
Sets the Accept header at the top of sign_request() so both JWT/Bearer
and SigV4 authentication paths include it.
- Rename _is_nova_lite_2_model → _is_nova_2_model to match all nova-2-* variants
- Add bedrock/converse/ routing prefix stripping in model detection
- Fix pre-existing test_get_supported_openai_params_bedrock_converse failure
- Remove thinking_blocks tests from Nova 2 test file (not Nova 2 behavior)
- Add end-to-end request, response, multi-turn, and model detection tests
- Parametrize key tests across both nova-2-lite and nova-2-pro model IDs
* fix: SSO PKCE support fails in multi-pod Kubernetes deployments
* fix: virutal key grace period from env/UI
* fix: refactor, race condition handle, fstring sql injection
* fix: add async call to avoid server pauses
* Update tests/test_litellm/proxy/management_endpoints/test_ui_sso.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix: add await in tests
* add modify test to perform async run
* Update tests/test_litellm/proxy/management_endpoints/test_ui_sso.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* Update tests/test_litellm/proxy/management_endpoints/test_ui_sso.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix grace period with better error handling on frontend and as per best practices
* Update tests/test_litellm/proxy/management_endpoints/test_ui_sso.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix: as per request changes
* Update litellm/proxy/utils.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* Fix errors when callbacks are invoked for file delete operations:
* Fix errors when callbacks are invoked for file operations
* Fix: pass deployment credentials to afile_retrieve in managed_files post-call hook
* Fix: bypass managed files access check in batch polling by calling afile_content directly
* Update tests/test_litellm/proxy/management_endpoints/test_ui_sso.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix: afile_retrieve returns unified ID for batch output files
* fix: batch retrieve returns unified input_file_id
* fix(chatgpt): drop unsupported responses params for Codex
Co-authored-by: Cursor <cursoragent@cursor.com>
* test(chatgpt): ensure Codex request filters unsupported params
Co-authored-by: Cursor <cursoragent@cursor.com>
* Fix deleted managed files returning 403 instead of 404
* Add comments
* Update litellm/proxy/utils.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix: thread deployment model_info through batch cost calculation
batch_cost_calculator only checked the global cost map, ignoring
deployment-level custom pricing (input_cost_per_token_batches etc.).
Add optional model_info param through the batch cost chain and pass
it from CheckBatchCost.
* fix(deps): add pytest-postgresql for db schema migration tests
The test_db_schema_migration.py test requires pytest-postgresql but it was
missing from dependencies, causing import errors:
ModuleNotFoundError: No module named 'pytest_postgresql'
Added pytest-postgresql ^6.0.0 to dev dependencies to fix test collection
errors in proxy_unit_tests.
This is a pre-existing issue, not related to PR #21277.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* fix(test): replace caplog with custom handler for parallel execution
The cost calculation log level tests were failing when run with pytest-xdist
parallel execution because caplog doesn't work reliably across worker processes.
This causes "ValueError: I/O operation on closed file" errors.
Solution: Replace caplog fixture with a custom LogRecordHandler that directly
attaches to the logger. This approach works correctly in parallel execution
because each worker process has its own handler instance.
Fixes test failures in PR #21277 when running with --dist=loadscope.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* fix(test): correct async mock for video generation logging test
The test was failing with AuthenticationError because the mock wasn't
intercepting the actual HTTP handler calls. This caused real API calls
with no API key, resulting in 401 errors.
Root cause: The test was patching the wrong target using string path
'litellm.videos.main.base_llm_http_handler' instead of using patch.object
on the actual handler instance. Additionally, it was mocking the sync
method instead of async_video_generation_handler.
Solution: Use patch.object with side_effect pattern on the correct
async handler method, following the same pattern used in
test_video_generation_async().
Fixes test failure in PR #21277 when running with --dist=loadscope.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* fix(test): add cleanup fixture and no_parallel mark for MCP tests
Two MCP server tests were failing when run with pytest-xdist parallel
execution (--dist=loadscope):
- test_mcp_routing_with_conflicting_alias_and_group_name
- test_oauth2_headers_passed_to_mcp_client
Both tests showed assertion failures where mocks weren't being called
(0 times instead of expected 1 time).
Root cause: These tests rely on global_mcp_server_manager singleton
state and complex async mocking that doesn't work reliably with
parallel execution. Each worker process can have different state
and patches may not apply correctly.
Solution:
1. Added autouse fixture to clean up global_mcp_server_manager registry
before and after each test for better isolation
2. Added @pytest.mark.no_parallel to these specific tests to ensure
they run sequentially, avoiding parallel execution issues
This approach maintains test reliability while allowing other tests
in the file to still benefit from parallelization.
Fixes test failures exposed by PR #21277.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* Regenerate poetry.lock with Poetry 2.3.2
Updated lock file to use Poetry 2.3.2 (matching main branch standard).
This addresses Greptile feedback about Poetry version mismatch.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* Remove unused pytest import and add trailing newline
- Removed unused pytest import (caplog fixture was removed)
- Added missing trailing newline at end of file
Addresses Greptile feedback (minor style issues).
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* Remove redundant import inside test method
The module litellm.videos.main is already imported at the top of
the file (line 21), so the import inside the test method is redundant.
Addresses Greptile feedback (minor style issue).
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
* Fix converse anthropic usage object according to v1/messages specs
* Add routing based on if reasoning is supported or not
* add fireworks_ai/accounts/fireworks/models/kimi-k2p5 in model map
* Removed stray .md file
* fix(bedrock): clamp thinking.budget_tokens to minimum 1024
Bedrock rejects thinking.budget_tokens values below 1024 with a 400
error. This adds automatic clamping in the LiteLLM transformation
layer so callers (e.g. router with reasoning_effort="low") don't
need to know about the provider-specific minimum.
Fixes#21297
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: improve Langfuse test isolation to prevent flaky failures (#21093)
The test was creating fresh mocks but not fully isolating from setUp state,
causing intermittent CI failures with 'Expected generation to be called once.
Called 0 times.'
Instead of creating fresh mocks, properly reset the existing setUp mocks to
ensure clean state while maintaining proper mock chain configuration.
* feat(s3): add support for virtual-hosted-style URLs (#21094)
Add s3_use_virtual_hosted_style parameter to support AWS S3 virtual-hosted-style URL format (bucket.endpoint/key) alongside the existing path-style format (endpoint/bucket/key).
This enables compatibility with S3-compatible services like MinIO and aligns with AWS S3 official terminology.
* Addressed greptile comments to extract common helpers and return 404
* Allow effort="max" for Claude Opus 4.6 (#21112)
* fix(aiohttp): prevent closing shared ClientSession in AiohttpTransport (#21117)
When a shared ClientSession is passed to LiteLLMAiohttpTransport,
calling aclose() on the transport would close the shared session,
breaking other clients still using it.
Add owns_session parameter (default True for backwards compatibility)
to AiohttpTransport and LiteLLMAiohttpTransport. When a shared session
is provided in http_handler.py, owns_session=False is set to prevent
the transport from closing a session it does not own.
This aligns AiohttpTransport with the ownership pattern already used
in AiohttpHandler (aiohttp_handler.py).
* perf(spend): avoid duplicate daily agent transaction computation (#21187)
* fix: proxy/batches_endpoints/endpoints.py:309:11: PLR0915 Too many statements (54 > 50)
* fix mypy
* Add doc for OpenAI Agents SDK with LiteLLM
* Add doc for OpenAI Agents SDK with LiteLLM
* Update docs/my-website/sidebars.js
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix mypy
* Update tests/test_litellm/proxy/_experimental/mcp_server/test_mcp_server.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* Add blog fffor Managing Anthropic Beta Headers
* Add blog fffor Managing Anthropic Beta Headers
* correct the time
* Fix: Exclude tool params for models without function calling support (#21125) (#21244)
* Fix tool params reported as supported for models without function calling (#21125)
JSON-configured providers (e.g. PublicAI) inherited all OpenAI params
including tools, tool_choice, function_call, and functions — even for
models that don't support function calling. This caused an inconsistency
where get_supported_openai_params included "tools" but
supports_function_calling returned False.
The fix checks supports_function_calling in the dynamic config's
get_supported_openai_params and removes tool-related params when the
model doesn't support it. Follows the same pattern used by OVHCloud
and Fireworks AI providers.
* Style: move verbose_logger to module-level import, remove redundant try/except
Address review feedback from Greptile bot:
- Move verbose_logger import to top-level (matches project convention)
- Remove redundant try/except around supports_function_calling() since it
already handles exceptions internally via _supports_factory()
* fix(index.md): cleanup str
* fix(proxy): handle missing DATABASE_URL in append_query_params (#21239)
* fix: handle missing database url in append_query_params
* Update litellm/proxy/proxy_cli.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
---------
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix(mcp): revert StreamableHTTPSessionManager to stateless mode (#21323)
PR #19809 changed stateless=True to stateless=False to enable progress
notifications for MCP tool calls. This caused the mcp library to enforce
mcp-session-id headers on all non-initialize requests, breaking MCP
Inspector, curl, and any client without automatic session management.
Revert to stateless=True to restore compatibility with all MCP clients.
The progress notification code already handles missing sessions gracefully
(defensive checks + try/except), so no other changes are needed.
Fixes#20242
* UI - Content Filters, help edit/view categories and 1-click add categories + go to next page (#21223)
* feat(ui/): allow viewing content filter categories on guardrail info
* fix(add_guardrail_form.tsx): add validation check to prevent adding empty content filter guardrails
* feat(ui/): improve ux around adding new content filter categories
easy to skip adding a category, so make it a 1-click thing
* Fix OCI Grok output pricing (#21329)
* fix(proxy): fix master key rotation Prisma validation errors
_rotate_master_key() used jsonify_object() which converts Python dicts
to JSON strings. Prisma's Python client rejects strings for Json-typed
fields — it requires prisma.Json() wrappers or native dicts.
This affected three code paths:
- Model table (create_many): litellm_params and model_info converted to
strings, plus created_at/updated_at were None (non-nullable DateTime)
- Config table (update): param_value converted to string
- Credentials table (update): credential_values/credential_info
converted to strings
Fix: replace jsonify_object() with model_dump(exclude_none=True) +
prisma.Json() wrappers for all Json fields. Wrap model delete+insert
in a Prisma transaction for atomicity. Add try/except around MCP
server rotation to prevent non-critical failures from blocking the
entire rotation.
---------
Co-authored-by: Harshit Jain <harshitjain0562@gmail.com>
Co-authored-by: Harshit Jain <48647625+Harshit28j@users.noreply.github.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
Co-authored-by: Ephrim Stanley <ephrim.stanley@point72.com>
Co-authored-by: Jay Prajapati <79649559+jayy-77@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Julio Quinteros Pro <jquinter@gmail.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Co-authored-by: Sameer Kankute <sameer@berri.ai>
Co-authored-by: mjkam <mjkam@naver.com>
Co-authored-by: Fly <48186978+tuzkiyoung@users.noreply.github.com>
Co-authored-by: Kristoffer Arlind <13228507+KristofferArlind@users.noreply.github.com>
Co-authored-by: Constantine <Runixer@gmail.com>
Co-authored-by: Emerson Gomes <emerson.gomes@thalesgroup.com>
Co-authored-by: Atharva Jaiswal <92455570+AtharvaJaiswal005@users.noreply.github.com>
Co-authored-by: Krrish Dholakia <krrishdholakia@gmail.com>
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
Co-authored-by: Ishaan Jaff <ishaanjaffer0324@gmail.com>
Resolve conflict in test_converse_transformation.py by keeping both
the structured outputs tests (from this branch) and the
TestBedrockMinThinkingBudgetTokens tests (from main).
Bedrock rejects thinking.budget_tokens values below 1024 with a 400
error. This adds automatic clamping in the LiteLLM transformation
layer so callers (e.g. router with reasoning_effort="low") don't
need to know about the provider-specific minimum.
Fixes#21297
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix several tests that fail in CI due to parallel test execution and
module reloading in conftest.py.
1. test_empty_assistant_message_handling:
- Use patch.object on factory_module.litellm instead of direct assignment
- Ensures the correct litellm reference is modified after conftest reloads
2. test_embedding_header_forwarding_with_model_group:
- Use patch.object on pre_call_utils_module.litellm instead of direct assignment
- Same fix for module reloading issue
3. test_embedding_input_array_of_tokens:
- Move mock inside test function (after fixture initializes router)
- Add skip condition if llm_router is None
- Fixes "AttributeError: None does not have 'aembedding'" in parallel execution
Root cause: conftest.py reloads litellm at module scope, which can cause:
- Different litellm references between test code and library code
- Global state (like llm_router) being None at decorator execution time
- isinstance checks failing due to class identity mismatches
- Add `definitions` handling alongside `$defs` in schema normalization
(older JSON Schema drafts use `definitions` instead of `$defs`)
- Fall back to tool-call approach when `response_format: {type: json_object}`
has no explicit schema, since the native API requires one
- Add tests for both cases
- Add whitelist-based filtering for anthropic_beta headers
- Only allow Bedrock-supported beta flags (computer-use, tool-search, etc.)
- Filter out unsupported flags like mcp-servers, structured-outputs
- Remove output_format parameter from Bedrock Invoke requests
- Force tool-based structured outputs when response_format is used
Fixes#16726
* feat(bedrock): add OpenAI-compatible service_tier parameter translation
Translates OpenAI's service_tier parameter (string) to Bedrock's
serviceTier format (object with type field).
* docs(bedrock): add OpenAI-compatible service_tier parameter documentation
Document the automatic translation from OpenAI-style service_tier
parameter to Bedrock's native serviceTier format.
* feat(bedrock): add service_tier to response when present
According to OpenAI's API documentation, when service_tier is sent in the
request, it should be returned in the response. This commit implements
this behavior for Bedrock Converse API to maintain compatibility with
OpenAI's API.
Changes:
- Added serviceTier field to ConverseResponseBlock type definition
- Moved ServiceTierBlock definition before ConverseResponseBlock to fix
type reference order
- Added response transformation to map Bedrock serviceTier (object) to
OpenAI service_tier (string format)
- Added 4 new tests for response transformation with service_tier
The service_tier is only added to the response when present in Bedrock's
response, maintaining backward compatibility.
Add support for the Bedrock Converse API serviceTier parameter to allow
specifying processing tier (priority, default, or flex).
Changes:
- Add ServiceTierBlock type in litellm/types/llms/bedrock.py
- Add serviceTier to CommonRequestObject
- Add serviceTier to get_config_blocks() in AmazonConverseConfig
- Add comprehensive tests for serviceTier functionality
- Add documentation for serviceTier usage
This allows users to configure service tier via:
- litellm_params in proxy config
- optional_params in SDK calls
