What's Changed
- Add new Azure AI models with pricing details by @emerzon in #15387
- [Feat] Add EnkryptAI Guardrails on LiteLLM by @ishaan-jaff in #15390
- [Fix] Fix dynamic Rate limiter v3 - inserting litellm_model_saturation by @ishaan-jaff in #15394
- [Fix] - Sessions not being shared by @AlexsanderHamir in #15388
- fix: Prevents DB from accidentally overriding config file values if they are empty in DB by @ARajan1084 in #15340
- Temporarily relax ResponsesAPIResponse parsing to support custom backends (e.g., vLLM) by @ashengstd in #15362
- Fix - OpenRouter cache_control to only apply to last content block by @ishaan-jaff in #15395
- [Fix]: remove panic from hot path by @AlexsanderHamir in #15396
- [Feat] Support for Vertex AI Gemma Models on Custom Endpoints by @ishaan-jaff in #15397
- [Perf]: optimize SSL/TLS handshake performance with prioritized cipher by @AlexsanderHamir in #15398
- fix: Ensure MCP client stays open during tool call by @uc4w6c in #15391
- Teams page: new column "Your Role" on the teams table by @ARajan1084 in #15384
- Add new together models by @zainhas in #15383
- feat: posthog per request api key by @carlos-marchal-ph in #15379
- Implement Shared Health Check State Across Pods by @Sameerlite in #15380
- Remove hardcoded "public" schema in 20250918083359_drop_spec_version_column_from_mcp_table/migration.sql by @vkolehmainen in #15363
- Minimal fix: gpt5 models should not go on cooldown when called with temperature!=1 by @jlan-nl in #15330
- Add W&B Inference documentation by @xprilion in #15278
- Add OCI Cohere support with tool calling and streaming capabilities by @Sameerlite in #15365
- Updates guardrail provider logos by @ARajan1084 in #15421
- LiteLLM Dashboard Teams UI refactor by @ARajan1084 in #15418
- Enforces removal of unused imports from UI by @ARajan1084 in #15416
- fix: usage page >> Model Activity >> spend per day graph: y-axis clipping on large spend values by @ARajan1084 in #15389
- [Fix] VertexAI - gemma model family support (custom endpoints) by @ishaan-jaff in #15419
- fix lint errors by @Sameerlite in #15406
- feat: add Bedrock AU Cross-Region Inference for Claude Sonnet 4.5 by @BCook98 in #15402
- [Feat] VertexAI Gemma model family streaming support + Added MedGemma by @ishaan-jaff in #15427
- Models & Endpoints Initial Refactor by @ARajan1084 in #15435
- [Feature]: Include server_name in /v1/mcp/server/health endpoint response by @Copilot in #15431
- [Fix] - SensitiveDataMasker converts lists to string by @AlexsanderHamir in #15420
- Deletion of docker-compose buggy comment that cause
config.yamlbased startup fail by @PabloGmz96 in #15425 - Litellm UI API Reference page updates by @ARajan1084 in #15438
- [Feat] Tag Management - Add support for setting tag based budgets by @ishaan-jaff in #15433
- [Fix] - shared session parsing and usage issue by @AlexsanderHamir in #15440
- [Fix]: handle closed aiohttp sessions by @AlexsanderHamir in #15442
- [Fix]: prevent session leaks when recreating aiohttp sessions by @AlexsanderHamir in #15443
- fix mapped tests 1 by @Sameerlite in #15445
New Contributors
- @ashengstd made their first contribution in #15362
- @vkolehmainen made their first contribution in #15363
- @jlan-nl made their first contribution in #15330
- @BCook98 made their first contribution in #15402
- @PabloGmz96 made their first contribution in #15425
Full Changelog: v1.77.7.dev10...v1.78.0-nightly