v1.89.1 - DB Resilience, MCP & Model-Info Backports

Deploy this version

Docker

docker run \
-e STORE_MODEL_IN_DB=True \
-p 4000:4000 \
docker.litellm.ai/berriai/litellm:1.89.1

Pip

pip install litellm==1.89.1

v1.89.1 is a patch release on top of v1.89.0. It brings the 1.84.8 database-resilience set onto the 1.89 line, hardens the MCP OAuth and credential paths, aligns the /v1/model/info and /v2/model/info surfaces, adds the budget-reservation toggle, and refreshes dependencies.

What's Changed

• fix(proxy): recover from cached-plan errors by reconnecting the Prisma client - PR #29983
• feat(proxy): add option to disable server-side prepared statements for DB lookups - PR #29984
• fix(proxy): return 5xx on DB infra errors during auth; reserve 401 for genuine auth failures - PR #29986
• feat(proxy): add disable_budget_reservation general setting - PR #29493
• fix(proxy): align /v1/model/info with router deployments - PR #30025
• fix(proxy): populate access_via_team_ids on /v1/model/info - PR #30274
• feat(proxy): publish /v2/model/info in the Swagger OpenAPI spec - PR #29900
• fix(proxy): return deprecated-key lookup result directly in the get_data combined view - PR #30327
• fix(mcp): drop phantom 401 span on delegated OAuth2 tool calls - PR #30494
• fix(mcp): drop orphaned per-user credential rows when an MCP server is deleted - PR #30141
• fix(mcp): allow team access-group grants in the OAuth authorize/token access check - PR #30041
• fix(ui/mcp): reset OAuth state on create-server modal close so a prior server's token no longer leaks into the next add-server session - PR #30000
• fix(mcp): load MCP tool configuration tools via the OBO/passthrough-aware GET path - PR #29960
• fix(mcp): let non-creator users OAuth into OBO-mode MCP servers from the Tools page - PR #29867
• fix(passthrough): resolve costing model when body model is unknown - PR #30160
• fix(passthrough): skip [DONE] sentinels and non-JSON SSE frames in Anthropic streaming logging - PR #30202
• fix(cost): resolve completion_cost AttributeError on streaming Anthropic web_search responses, including the server_tool_use type-coercion prerequisite - PR #27346
• fix(proxy): authorize batch files using upload target_model_names - PR #30009
• fix(proxy): atomic merge for team model aliases and team.models on BYOK create - PR #29528
• feat(datadog): add team-scoped Datadog callback support - PR #29947
• feat(bedrock_mantle): add SigV4/IAM auth to the Responses API route - PR #29788
• fix(guardrails): read CrowdStrike AIDR identity from both metadata bags - PR #29991
• fix(slack): use the actual Slack return details for the expires_in default - PR #29951
• chore(deps): bump vitest, brace-expansion, pypdf, and tornado - PR #30220

Full Changelog

https://github.com/BerriAI/litellm/compare/v1.89.0...v1.89.1