LiteLLM v1.106.5, v1.105.6, v1.104.8 and v1.103.10 trim failure logs for rejected requests
Deploy this version
- Docker
- Pip
docker run \
-e LITELLM_MASTER_KEY=sk-<paste-a-long-random-key> \
-e DATABASE_URL=postgresql://<user>:<password>@<host>:5432/<dbname> \
-e STORE_MODEL_IN_DB=True \
-p 4000:4000 \
docker.litellm.ai/berriai/litellm:1.105.6
pip install litellm==1.105.6
This release is published as ghcr.io/berriai/litellm:v1.105.6. See the GitHub release and the full releases page
v1.105.6 has a known performance regression compared to v1.104.x when clients send Accept-Encoding: gzip, which most HTTP clients and the OpenAI SDKs do by default. The proxy gzips non-streaming responses over 500 bytes on the event loop, which slows down those requests and every other request on the same worker, including streaming and MCP requests.
Upgrade to v1.105.7 instead, which ships the fix. The fix is backported in PR #48679.
v1.105.6 is a patch release on top of v1.105.5. It changes what failure logs for rejected requests carry and how the GCS logger picks its bucket on stable/1.105.x. There are no new database migrations. The v1.105.6 tag points at 3a05120
What's Changed
- Failure logs for rejected requests keep only core request fields,
tagsand guardrail details - PR #48599 - The GCS logger uses the bucket and credentials configured on the proxy - PR #48599
The change and the version bump are backported in PR #48604
Full Changelog
https://github.com/BerriAI/litellm/compare/v1.105.5...v1.105.6
