●Change8

v1.82.dev1

📦 litellmView on GitHub →
✨ 11 features🐛 41 fixes⚡ 1 deprecations🔧 51 symbols

Summary

This release focuses heavily on bug fixes across various providers (Gemini, Anthropic, OpenAI, Azure, OpenRouter) and internal systems like MCP and caching. It also introduces new model support, performance optimizations for spendlogs and streaming latency, and new UI features for project management.

Migration Steps

  1. If you rely on the structure of ModelResponse, note that StreamingChoices has been removed and you should use ModelResponseStream instead.

✨ New Features

  • Added gemini-3.1-flash-image-preview to the model database for Vertex AI.
  • Added v1 for anthropic responses transformation.
  • Enabled local file support for OCR.
  • Added Access group CRUD: Bidirectional team/key sync functionality.
  • Added gpt-audio-1.5 to the model cost map.
  • Added gpt-realtime-1.5 to the model cost map.
  • Added OpenRouter native models to the model cost map.
  • Included timestamps in /project/list response.
  • Added Projects page in the UI with list and create flows.
  • Added in_flight_requests metric to /health/backlog endpoint and Prometheus.
  • Added Agent RBAC Permission Fix: Ensure Internal Users cannot create agents.

🐛 Bug Fixes

  • Preserved $ref in JSON Schema for Gemini 2.0+.
  • Moved duration to _hidden_params to match OpenAI transcription response spec.
  • Mapped reasoning_effort to output_config for Anthropic Claude 4.6 models.
  • Fixed converse handling for parallel_tool_calls.
  • Preserved forwarding server side called tools.
  • Fixed free models working from UI.
  • Added ChatCompletionImageObject in OpenAIChatCompletionAssistantMessage.
  • Fixed poetry lock issues.
  • Stripped stale mcp-session-id to prevent 400 errors across proxy workers.
  • Forwarded extra_headers on OpenAI code path in image_generation().
  • Passed model_info/metadata in image_edit for custom pricing.
  • Fixed tool_calls streaming indexes for ChatGPT.
  • Corrected supported_openai_params for GPT-5 model family.
  • Corrected supported params for gpt-5-search models.
  • Resolved api_base from env var in Azure AI Document Intelligence OCR.
  • Enabled function calling for PublicAI Apertus models.
  • Passed through native Gemini imageConfig params for image generation in Vertex AI.
  • Used provider-reported usage in OpenRouter streaming without stream_options.
  • Preserved image_url blocks in multimodal messages for Moonshot.
  • Removed StreamingChoices from ModelResponse, using ModelResponseStream instead.
  • Used output_index for parallel tool call streaming indices in responses.
  • Included system and tools in token counting API requests.
  • Added Gemini and Vertex AI support to HeliconeLogger.
  • Handled openrouter models without '/' in name during model registration.
  • Fixed double-stripping of model names with provider-matching prefixes in adapter.
  • Fixed test mocks for renamed filter_server_ids_by_ip_with_info.
  • Added PROXY_ADMIN role to system user for key rotation.
  • Populated user_id and user_info for admin users in /user/info.
  • Stored task references in LLMClientCache._remove_key.
  • Propagated extra_headers to Upstream in image_generation.
  • Passed MCP auth headers from request into tool fetch for /v1/responses and chat completions.
  • Shortened guardrail benchmark result filenames for Windows long path support.
  • Set default available_on_public_internet to true for MCP.
  • Filtered internal json_tool_call when mixed with real tools in Bedrock.
  • Fixed OIDC discovery URLs, roles array handling, and dot-notation error hints in JWT.
  • Fixed Python 3.10 compatibility for OpenAIContainerConfig.
  • Fixed parallel_tool_calls mapping in map_openai_params for Bedrock.
  • Fixed Claude code plugin schema.
  • Added missing migration for LiteLLM_ClaudeCodePluginTable.
  • Updated Azure pass-through test after Responses API routing change.
  • Updated realtime guardrail test assertions for voice violation behavior.

Affected Symbols

geminitranscriptionanthropicvertexspendlogsconverseparallel_tool_callsOpenAIContainerConfigOpenAIChatCompletionAssistantMessagemcp-session-idimage_generation()gpt-audio-1.5gpt-realtime-1.5image_edittool_callssupported_openai_paramsgpt-5gpt-5-searchazure_aiDocument Intelligence OCRPublicAI Apertus modelsGemini imageConfig paramsopenroutermoonshotModelResponseModelResponseStreamoutput_indexcount_tokensHeliconeLoggerregister_modeladapter/project/listClaude code plugin schemaAssemblyAI docsUniversal-3 ProSpeech UnderstandingLLM Gateway/health/backlogPrometheusrealtime guardrail testAzure pass-through testLiteLLM_ClaudeCodePluginTablebedrockfilter_server_ids_by_ip_with_info/user/infoLLMClientCache/v1/responseschat completionsguardrail benchmark result filenamesjwtOIDC discovery URLs

⚡ Deprecations

  • Added deprecation dates for xAI grok-2-vision-1212 and grok-3-mini models.
litellm v1.82.dev1 - Change8