Change8

v1.79.3.dev5

📦 litellmView on GitHub →
10 features🐛 10 fixes🔧 11 symbols

Summary

This release adds several UI enhancements, new provider support, and numerous bug fixes, including a Zscaler AI Guard hook and RunwayML video generation provider.

✨ New Features

  • Add Zscaler AI Guard hook.
  • UI: Add Tags to Edit Key flow.
  • Add RunwayML provider for video generation.
  • UI: Improve usage indicator.
  • UI: Add LiteLLM params to Edit Model screen.
  • Add atexit handlers to flush callbacks for async completions.
  • Update model logging format for custom LLM provider.
  • Vertex AI Rerank: safe load Vertex AI credentials.
  • Add support for filtering Bedrock Knowledge Base queries.
  • Bedrock Embeddings: ensure correct aws_region is used when provided dynamically.

🐛 Bug Fixes

  • Removed enterprise restriction from guardrails list endpoint.
  • Litellm tags usage now includes request_id.
  • Use user budget instead of key budget when creating new team.
  • Sanitize null token usage in OpenAI-compatible responses.
  • Added new models, deleted duplicate models, and updated pricing.
  • Fix: Convert SSE stream iterator to async for proper streaming support.
  • Fix: Use vllm passthrough config for hosted vllm provider instead of raising error.
  • Fix: app_roles missing from JWT payload.
  • Fix: Bedrock Knowledge Bases filtering support.
  • Fix: Bedrock Embeddings correct aws_region handling.

🔧 Affected Symbols

litellm.guardrails.list_endpointlitellm.tags.add_request_idlitellm.team.create_new_teamlitellm.callbacks.async_flushlitellm.logging.model_formatlitellm.providers.vertex_ai.reranklitellm.agentcore.sse_stream_iteratorlitellm.providers.vllm.passthrough_configlitellm.auth.jwt_payloadlitellm.bedrock.knowledge_baselitellm.bedrock.embeddings