●Change8

1.84.0-dev.2

📦 litellmView on GitHub →
✨ 9 features🐛 20 fixes⚡ 1 deprecations🔧 15 symbols

Summary

This release introduces Docker image signing verification via cosign and adds support for new models like gpt-image-2 and providers like AIHubMix. Numerous security hardening measures and bug fixes were applied across proxy, authentication, and vector store integrations.

Migration Steps

  1. If you rely on the /ui/chat page, note that it has been removed.

✨ New Features

  • Added support for gpt-image-2 model.
  • Added AIHubMix as an OpenAI-compatible provider.
  • Added support for Azure Entra discovery endpoint.
  • Added durable agent workflow run tracking via /v1/workflows/runs in the proxy.
  • Added team-level search provider credentials in the proxy.
  • Added opt-in short-ID tool prefix for MCP to keep tool names under the 60-character limit.
  • Added support for Bedrock retrievalConfiguration passthrough in vector stores.
  • Added LiteLLM headers on Google native generateContent routes.
  • Implemented lazy loaded imports and lazy loaded front page.

🐛 Bug Fixes

  • Fixed scope of ERROR log assertion to LiteLLM logger in test_model_alias_map.
  • Fixed Bedrock/Anthropic to translate OpenAI file content on tool-result path.
  • Fixed proxy self-heal for Prisma read paths and hardened reconnect state machine.
  • Redacted spend logs error message.
  • Fixed proxy reconnect Prisma DB without blocking the event loop.
  • Dropped sensitive locals from re-raised error messages.
  • Fixed Vertex batches by setting is_redirect=False on mocked retrieve response.
  • Redacted credentials in vector store list/info/update responses and gated update by per-store access.
  • Substituted alias for master key on UserAPIKeyAuth.
  • Fixed MCP public-route detection and OAuth2 fallback gating.
  • Fixed team member null budget fallback.
  • Fixed proxy to inherit caller identity in passthrough batch managed-object.
  • Fixed proxy/auth guardrail modification permission check.
  • Corrected min-release-age value in .npmrc files for CI/Tooling.
  • Added 1-hour cache write pricing for Claude 4.5/4.6/4.7 (Global, US) for Bedrock.
  • Fixed proxy/batches to forward model to retrieve_batch for Bedrock.
  • Fixed passthrough to track spend for interrupted Bedrock streams.
  • Validated AWS region name.
  • Dropped dbName and partitionNames from MILVUS_OPTIONAL_PARAMS for Milvus.
  • Replaced subprocess startup-import diff with static source scan.

Affected Symbols

gpt-image-2AIHubMixazure entra discovery endpoint/v1/workflows/runsBedrock retrievalConfigurationGoogle native generateContent routestest_model_alias_mapbedrockanthropicPrismavertex-batchesmilvusUserAPIKeyAuthMCPClaude 4.5/4.6/4.7

⚡ Deprecations

  • The /ui/chat page has been removed.