Unit Tests: Proxy DB Operations / proxy-db (auth-checks, tests/proxy_unit_tests/test_auth_checks.py tests/proxy_unit_tests/test_user_api_key_auth.py, 20, 8) (push) Has been cancelled
Unit Tests: Proxy DB Operations / proxy-db (remaining, tests/proxy_unit_tests --ignore=tests/proxy_unit_tests/test_key_generate_prisma.py --ignore=tests/proxy_unit_tests/test_auth_checks.py --ignore=tests/proxy_unit_tests/test_user_api_key_auth.py, 20, 8) (push) Has been cancelled
Strip CR/LF from JSON payload strings before verbose_proxy_logger.debug
in _debug_log_request_payload. Add _sanitize_for_log helper and unit tests.
Made-with: Cursor
Extract effort from reasoning payloads without a summary key so chat
completion providers receive a string. Preserve the full dict when
summary is present for the Responses API bridge in main.py.
Made-with: Cursor
- Pass model_id, api_base, and process_response_headers output into streaming
iterators so streamGenerateContent gets the same x-litellm-* headers as
non-streaming paths.
- Drop request_data deployment mutation from build_litellm_proxy_success_headers_from_llm_response.
- Avoid logging raw request key names in oversized debug payload (code scanning).
- Extend tests for streaming iterator shape, metadata fallback, and helper.
Made-with: Cursor
Wire build_litellm_proxy_success_headers_from_llm_response for :generateContent
and :streamGenerateContent so x-litellm-*, rate limit, and provider headers
match the OpenAI-style proxy path. Add unit test.
Annotate httpx.HTTPStatusError branch so pyright accepts .response after optional
exception transform. Remove unused variable in streaming tracer test (Ruff F841).
Made-with: Cursor
- Add pop_vertex_request_labels / vertex_request_labels_from_litellm_params in common_utils
- Vertex embeddings: pass litellm_params, set predict body labels; Gemini uses shared helper
- Imagen: top-level labels from metadata; rerank: userLabels for Discovery Engine Rank API
- Thread litellm_params through rerank handler and all BaseRerankConfig implementations
Made-with: Cursor
* 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
cryptography 46.0.5 excludes Python 3.9.0 and 3.9.1, which conflicts
with pyproject.toml's python = ">=3.9,<4.0" range. Docker still uses
46.0.5 via requirements.txt.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
aioboto3 was listed as a dependency for async sagemaker calls but is not
imported anywhere in the codebase — async calls use httpx + botocore SigV4
instead. Removing it eliminates the unresolvable botocore version conflict
between boto3 and aiobotocore, along with all grep -v / --no-deps workarounds
across Dockerfiles and CI.
Also addresses Greptile review feedback: collapse redundant grpcio
python-version markers, bump pyproject.toml cryptography to 46.0.5 to
match Docker (GHSA-r6ph-v2qm-q3c2), and fix misleading .npmrc comment.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Both are transitive deps of aiobotocore, added to requirements.txt in
the previous commit. aioitertools is MIT, wrapt is BSD.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
boto3==1.42.80 and aioboto3==15.5.0 have incompatible botocore version
ranges. No aioboto3 release supports botocore 1.42.x yet. Both uv and
pip 26.0.1 reject the resolution.
Fix: filter aioboto3 out of requirements.txt at install time, then
install aioboto3+aiobotocore with --no-deps to bypass resolution.
Added wrapt and aioitertools to requirements.txt as pinned transitive
deps of aiobotocore (skipped by --no-deps). Fixed pip stdin handling
(/dev/stdin). Applied to all 5 Dockerfiles and all CircleCI install
paths.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
boto3==1.42.80 and aioboto3==15.5.0 have incompatible botocore ranges.
Both uv and pip 26.0.1 reject the resolution. Fix: filter aioboto3 out
of requirements.txt at install time, then install aioboto3+aiobotocore
separately with --no-deps to bypass resolution. Removes uv-overrides.txt
which only partially addressed the conflict.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
aiobotocore[boto3] pins both boto3<1.40.62 and botocore<1.40.62.
The previous commit only overrode botocore. Added boto3 override too.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
boto3==1.42.80 requires botocore>=1.42.80 but aioboto3==15.5.0 (via
aiobotocore==2.25.1) requires botocore<1.40.62. No aioboto3 release
supports botocore 1.42.x yet. pip's lenient resolver handles this for
Docker builds, but uv's strict resolver rejects it in CI. Added
uv-overrides.txt to force botocore to match boto3 during uv installs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
CircleCI had stale version pins (e.g. boto3==1.36.0, aioboto3==13.4.0) that
conflict with requirements.txt (boto3==1.42.80, aioboto3==15.5.0), causing
uv resolution failures. Updated all mismatched pins across config.yml and
.circleci/requirements.txt to match requirements.txt as the source of truth.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
After merging main, pyproject.toml had updated dependency versions but
requirements.txt still had the old pins. The Dockerfile builds litellm
from source (using pyproject.toml) then installs deps from
requirements.txt, so version mismatches cause pip resolution failures.
Updated 21 packages to match: openai, fastuuid, tiktoken,
importlib-metadata, tokenizers, click, jsonschema, fastapi, pyyaml,
uvicorn, boto3, mcp, orjson, polars, apscheduler, fastapi-sso, pyjwt,
python-multipart, azure-identity, rich, aiohttp.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Resolve conflicts between pinned versions and main's caret ranges,
keeping exact pins. Add pytest-cov==5.0.0 from main. Regenerate
poetry.lock.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
hf-xet is Apache 2.0 licensed but PyPI metadata doesn't expose the
license string, so the automated checker can't determine it.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
pytest-asyncio 1.x no longer provides an implicit event loop in sync
fixtures/tests. Make async-dependent fixtures and tests async, and
replace deprecated asyncio.get_event_loop() in tests. Switch
Dockerfile.build_from_pip from Alpine to Debian slim since
pyroscope-io 0.8.x has no musl wheels.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
asyncio.create_task in CSW.__anext__ scheduled the deferred logging
callback as an independent task that raced with unified_guardrail's
end-of-stream block. For short-stream providers (Vertex AI, Azure,
Anthropic), the logging fired before guardrail_information was written,
causing post_call guardrail entries to be missing from
StandardLoggingPayload.
Move the deferred callback trigger from CSW.__anext__ to
ProxyLogging.async_post_call_streaming_iterator_hook (after the full
streaming pipeline completes). CSW now stores the assembled response
args; the outer consumer fires the callback after all guardrail
end-of-stream blocks finish. Also skip apply_guardrail guardrails in
_run_deferred_stream_guardrails to eliminate duplicate API calls.
- Add route_all_chat_openai_to_responses global flag docs under
'Getting Reasoning Content in /chat/completions' in openai.md
with SDK and proxy examples using gpt-5.4
- Remove the section from responses_api.md (wrong location)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Use .lower() == "true" for LITELLM_ROUTE_ALL_CHAT_OPENAI_TO_RESPONSES to avoid bool("False") == True bug
- Add clarifying comment on early return in responses_api_bridge_check
- Document route_all_chat_openai_to_responses flag in openai/responses_api.md
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds `litellm.route_all_chat_openai_to_responses` (env: `LITELLM_ROUTE_ALL_CHAT_OPENAI_TO_RESPONSES`) to route all OpenAI /chat/completions requests through the Responses API bridge. Also fixes reasoning param dict passthrough in completion transformation.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
