●Change8

v1.79.0.rc.1

📦 litellm
✨ 18 features🐛 19 fixes⚡ 1 deprecations🔧 25 symbols

Summary

This release introduces significant feature enhancements, particularly around Search APIs, Vector Stores (Vertex AI), and Guardrail integration across various endpoints. Numerous bug fixes address issues related to error handling, configuration mapping, and security for the Responses API.

✨ New Features

  • Added DataforSEO Search API.
  • Added Google PSE Search Provider.
  • Added cost tracking for Search API requests: Google PSE, Tavily, Parallel AI, Exa AI.
  • Vector Stores: support Vertex AI Search API as vector store through LiteLLM.
  • Backend: Allow storing configured Search APIs in DB.
  • UI: Allow adding search tools on UI + testing search.
  • UI: Add logos for search providers.
  • Proxy: support absolute RPM/TPM in priority_reservation.
  • Implement Bedrock Guardrail apply_guardrail endpoint support.
  • Lasso Security Guardrail: Add v3 API Support.
  • Add mistral medium 3 and Codestral 2 on vertex.
  • New Guardrail: Dynamo AI Guardrail.
  • Organizations: allow org admins to create teams on UI + IBM Guardrails support.
  • Azure AI - Search Vector Stores support.
  • VertexAI Search Vector Store - Passthrough endpoint support + Vector store search Cost tracking support.
  • Guardrails: Responses API, Image Gen, Text completions, Audio transcriptions, Audio Speech, Rerank, Anthropic Messages API support via the unified `apply_guardrails` function.
  • Org level tpm/rpm limits + Team tpm/rpm validation when assigned to org.
  • Bedrock: add titan image generations w/ cost tracking.

🐛 Bug Fixes

  • Fixed ErrorEvent ValidationError when OpenAI Responses API returns nested error structure.
  • Azure AI Speech - Ensure `voice` is mapped from request body -> SSML body, allow sending `role` and `style`.
  • Use proper bedrock model name in health check.
  • Simplify reasoning item handling for gpt-5-codex.
  • Hooks broken on /bedrock passthrough due to missing metadata.
  • Exa Search API - ensure request params are sent to Exa AI.
  • Fixed cost tracking for Vertex AI search spend.
  • Fixed routing for custom server root path in UI.
  • Pass user-defined headers and extra_headers to image-edit calls.
  • Minimize the occurrence of deadlocks.
  • Fix MLFlow tags - split request_tags into (key, val) if request_tag has colon.
  • Fixed lasso import config, redis cluster hash tags for test keys.
  • Update vertex ai gemini costs.
  • Managed Files: don't raise error if managed object is not found.
  • Batches: Fixed “User default_user_id does not have access to the object” when object not in db.
  • Vector Stores: show config.yaml vector stores on UI.
  • Responses API - support tags in metadata.
  • Responses API security: prevent User A from retrieving User B's response, if response.id is leaked.
  • opentelemetry.py: fix issue where headers were not being split correctly.

🔧 Affected Symbols

Azure AI SpeechbedrockDataforSEO Search APIGoogle PSE Search ProviderVertex AI Search APIgpt-5-codexgemini-1.5-pro-preview-0514Exa Search APIcustom server root pathimage-edit callspriority_reservationBedrock Guardrail apply_guardrail endpointLasso Security GuardrailMLFlow tagsmistral medium 3Codestral 2Dynamo AI GuardrailIBM Guardrailsmanaged_files.pyBatchesconfig.yamlResponses APIapply_guardrails functionopentelemetry.pytitan image generations

⚡ Deprecations

  • Replaced deprecated model name `gemini-1.5-pro-preview-0514`.