LiteLLM is moving to Rust Read the latest updates.
LiteLLM v1.108.11 (LTS) fixes OOM on 429 retries and SSO behind TLS
tip
v1.108.11 is a patch release on the 1.108 LTS line. The lts image tag always points to the latest LTS release, so pull docker.litellm.ai/berriai/litellm:lts to stay on the newest LTS patch. See the LTS pilot.
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:1.108.11
pip install litellm==1.108.11
This release is published as ghcr.io/berriai/litellm:v1.108.11. See the GitHub release and the full releases page.
v1.108.11 is a patch release on top of v1.108.10. There are no new database migrations. The v1.108.11 tag points at c50d78d.
Breaking Changes
What's Changed​
- Azure: Fixed 400 when sending
reasoning_effortto DeepSeek V4 Flash and Pro onazure_ai- PR #52070 - AWS Bedrock: Fixed
guarded_textparts in system and assistant messages being dropped instead of checked by the guardrail - PR #51776 - AWS Bedrock: Fixed aborted Invoke streams on
/v1/messagesthat kept generating billed output tokens upstream - PR #52210 - Vertex AI and Gemini: Fixed 500 when the chat history has an earlier tool call with null or truncated arguments - PR #50020
- Vertex AI and Gemini: Fixed 400 when sending
reasoning_effort: "xhigh"togemini-4-hydrogen- PR #52060 - Other providers: Fixed Cohere embeddings ignoring
COHERE_API_BASEand always calling api.cohere.ai - PR #52194 - Other providers: Fixed OpenRouter streams sending an empty choice after
finish_reason: stopthat broke strict clients - PR #52200 - Chat completions: Fixed LiteLLM's own MCP and
fake_streamparams reaching provider request bodies and causing 400s - PR #52225 - Responses API: Fixed 500 on
/v1/responseswhenallowed_openai_paramsis a string or tuple - PR #52203 - Messages API: Fixed streamed thinking blocks ending without
signature_delta, so clients dropped them on the next tool step - PR #52117 - Embeddings, images and audio: Fixed custom image handlers not receiving
n,size,qualityand the other OpenAI image params - PR #52051 - Keys, teams and auth: Changed team model writes so only proxy admins can set
os.environ/values, team admins now get 403 - PR #51874 - Admin UI: Fixed SSO login detouring through
/sso/key/generate/and redirecting tohttp://behind TLS termination, broken since v1.108.0 - PR #52213 - Deployment: Changed the UI image to slim nginx, which drops
curlbut keepswgetandnc- PR #52298 - SDK: Fixed valid Messages and OCR requests with dict-subclass inputs failing when
LITELLM_RUST=true- PR #51812 - Performance: Fixed proxy pods running out of memory when clients retry 429s with large request bodies - PR #51428
- Other: Added native
generateContentsupport for custom providers registered incustom_provider_map- PR #51083
The changes are backported in PR #51982, PR #52074, PR #52155, PR #52201, PR #52233, PR #52248, PR #52306 and PR #52312.
Full Changelog​
https://github.com/BerriAI/litellm/compare/v1.108.10...v1.108.11