BerriAI/litellm v1.74.9.rc-draft

on GitHub

What's Changed

• Fix SSO Logout | Create Unified Login Page with SSO and Username/Password Options by @NANDINI-star in #12703
• docs(moonshot): correct base url and document CN-specific endpoint by @magicalne in #12804
• Adding HolmesGPT to projects using LiteLLM by @pavangudiwada in #12798
• feat: Add Lambda AI provider support by @colesmcintosh in #12817
• [UI Bug Fix] Show correct guardrails when editing a team by @ishaan-jaff in #12823
• [Docs] Show correct list of vertex ai mistral models by @ishaan-jaff in #12828
• [LLM Translation - GH Copilot] added dynamic endpoint support by @jugaldb in #12827
• feat: add Morph provider support by @colesmcintosh in #12821
• fix: remove deprecated groq/qwen-qwq-32b and add qwen/qwen3-32b by @aholmberg in #12831
• Add Google Cloud Model Armor guardrail documentation by @colesmcintosh in #12814
• [Feat] Add Recraft Image Generation API Support - New LLM Provider by @ishaan-jaff in #12832
• [Bug Fix] - gemini leaking FD for sync calls with litellm.completion by @ishaan-jaff in #12824
• [Feat] Add fireworks - fireworks/models/kimi-k2-instruct by @ishaan-jaff in #12837
• [LLM Translation] add qwen-vl-plus by @jugaldb in #12829
• [Azure OpenAI Feature] - Support DefaultAzureCredential without hard-coded environment variables by @ishaan-jaff in #12841
• Docs - litellm benchmarks by @ishaan-jaff in #12842
• fix(watsonx): IBM Watsonx - use correct parameter name for tool choice by @Tomas2D in #9980
• feat: add Hyperbolic provider support by @colesmcintosh in #12826
• Litellm batch cost tracking debug by @krrishdholakia in #12782
• Passthrough Auth - make Auth checks OSS + Anthropic - only show 'reasoning_effort' for supported models by @krrishdholakia in #12847
• build(deps): bump form-data from 4.0.0 to 4.0.4 in /ui/litellm-dashboard by @dependabot[bot] in #12851
• Fix team_member_budget update logic by @krrishdholakia in #12843
• Fix async retryer on .acompletion() + forward clientside headers - filter out content-type from clientside request (causes llm api call to hang) by @krrishdholakia in #12848
• Openrouter - filter out cache_control flag for non-anthropic models (allows usage with claude code) by @krrishdholakia in #12850
• Improvements on the Regenerate Key Flow by @NANDINI-star in #12788
• Replace non-root Dockerfile base with Alpine multi-stage build; by @mdiloreto in #12707
• build(deps): bump form-data from 4.0.3 to 4.0.4 in /docs/my-website by @dependabot[bot] in #12867
• Bug fix - Azure KeyVault not in image, add azure-keyvault==4.2.0 to Docker img by @ishaan-jaff in #12873
• [Separate Health App] Pass through cmd args via supervisord by @jugaldb in #12871
• [Feat] Add Recraft API - Image Edits Support by @ishaan-jaff in #12874
• [LLM Translation] Litellm gemini 2.0 live support by @jugaldb in #12839
• [Feat] Add cost tracking for new vertex_ai/llama-3 API models by @ishaan-jaff in #12878
• [Feat] - Track cost + add tags for health checks done by LiteLLM Proxy by @ishaan-jaff in #12880
• Fix async retryer on .acompletion() + forward clientside headers - filter out content-type from clientside request (causes llm api call to hang) by @krrishdholakia in #12886
• Request Headers - support x-litellm-num-retries + Usage - support usage by model group by @krrishdholakia in #12890
• [Docs ]Litellm mcp access group doc by @jugaldb in #12883
• [Add health check] add architecture diagram by @jugaldb in #12879
• rm retired anthropic models from model_prices_and_context_window.json by @staeiou in #12864
• feat: extended /v1/models endpoint, now it returns with fallbacks on demand by @murad-khafizov in #12811
• feat: Add Pillar Security guardrail integration by @eagle-p in #12791
• Swagger - bump DOMPurify version + add back Support local swagger bundle (good for airgapped environments) + support local swagger on custom root paths by @krrishdholakia in #12911
• [Feat] LiteLLM x Cloudzero integration - Allow exporting spend to cloudzero by @ishaan-jaff in #12908
• [Feat] LiteLLM CloudZero Integration updates - using LiteLLM_SpendLogs Table by @ishaan-jaff in #12922
• Add GA version of gemini 2.5 flash lite for both vertex and gemini by @apoorv-sharma in #12920
• [UI] Allow setting up CloudZero Usage through LiteLLM UI by @ishaan-jaff in #12923
• Update control_plane_and_data_plane.md by @superpoussin22 in #12939
• Prometheus - tags, fix '[tag]="false"' when tag is set by @krrishdholakia in #12916
• Proxy - specify key_type - allows specifying if key can call LLM API routes vs. Management routes only by @krrishdholakia in #12909
• fix(internal_user_endpoints.py): delete member from team table on /user/delete by @krrishdholakia in #12926
• [Feat] Backend Router - Add Auto-Router powered by semantic-router by @ishaan-jaff in #12955
• [LLM Translation] - Bug fix Anthropic Tool calling by @ishaan-jaff in #12959
• [LLM Translation] added new realtime model for openai by @jugaldb in #12946
• [Feat] UI - Allow Adding LiteLLM Auto Router on UI by @ishaan-jaff in #12960
• [Feat] Edit Auto Router Settings on UI by @ishaan-jaff in #12966
• [Feat] Add inpainting support and corresponding tests for Amazon Nova… by @SantoshDhaladhuli in #12949
• docs: added documentation about metadata exposed over the /v1/models endpoint by @murad-khafizov in #12942
• Fix: Shorten Gemini tool_call_id for Open AI compatibility by @tonga54 in #12941
• feat: add openrouter/qwen/qwen3-coder model configuration by @colesmcintosh in #12910
• fix(auth_utils): make header comparison case-insensitive by @sings-to-bees-on-wednesdays in #12950
• GuardrailsAI: use validatedOutput to allow usage of "fix" guards by @DmitriyAlergant in #12891
• Show global retry policy on UI by @krrishdholakia in #12969
• Guardrails - support model-level guardrails by @krrishdholakia in #12968
• build(github/manual_pypi_publish.yml): manual workflow to publish pip package - used for pushing dev releases by @krrishdholakia in #12985
• [LLM Translation] Add bytedance/ui-tars-1.5-7b on openrouter by @jugaldb in #12882
• clean and verify key before inserting by @jugaldb in #12840
• [LLM Translation] fix query params for realtime api intent by @jugaldb in #12838
• [MCP Gateway] Move cost tracking and permission management to OSS by @jugaldb in #12988
• build: update pip package by @krrishdholakia in #12998
• Fix issue writing db by @emerzon in #13001
• [BUG Fix] Cannot pickle coroutine object by @jugaldb in #13005
• [MCP Gateway] add Litellm mcp alias for prefixing by @jugaldb in #12994
• Fixup ollama model listing (again) by @matthid in #13008
• [Vector Store] make vector store permission management OSS by @jugaldb in #12990
• [FEAT] Model-Guardrails: Add on UI by @krrishdholakia in #13006
• UI SSO - fix reset env var when ui_access_mode is updated by @krrishdholakia in #13011
• [MCP Gateway] Litellm mcp multi header propagation by @jugaldb in #13003

New Contributors

• @magicalne made their first contribution in #12804
• @pavangudiwada made their first contribution in #12798
• @mdiloreto made their first contribution in #12707
• @murad-khafizov made their first contribution in #12811
• @eagle-p made their first contribution in #12791
• @apoorv-sharma made their first contribution in #12920
• @SantoshDhaladhuli made their first contribution in #12949
• @tonga54 made their first contribution in #12941
• @sings-to-bees-on-wednesdays made their first contribution in #12950

Full Changelog: v1.74.7-stable...v1.74.9.rc-draft