●Change8

litellm_stable_build-test-v0.1

📦 litellmView on GitHub →
✨ 17 features🐛 24 fixes🔧 11 symbols

Summary

This release focuses heavily on stability, bug fixes across various providers (including Bedrock, Cohere, Vertex AI), and significant improvements to the LiteLLM UI and OAuth2 support. New features include Websearch interception and support for Sarvam AI and Perplexity Research API.

Migration Steps

  1. If using semantic filters, be aware of a new warning being issued.
  2. If encountering JSON serialization errors with PostHog, ensure objects being logged are serializable.
  3. If using Vertex AI with Anthropic beta features, ensure extra_headers are propagated correctly.
  4. If using Prometheus, ensure label values are sanitized to prevent scrape failures.
  5. If using shared aiohttp sessions, ensure ssl_verify is respected.
  6. If using callbacks, MAX_CALLBACKS can now be overridden via an environment variable.

✨ New Features

  • Added support for anthropic_messages call type in prompt caching.
  • Added Websearch support using LiteLLM /search via a web search interception hook for chat completion.
  • Added support for MCP M2M Oauth2.
  • Added Sarvam AI STT and TTS support.
  • Added Kimi model pricing by region.
  • Added moonshotai.kimi-k2.5 model support.
  • Added support for managed error file.
  • Added Perplexity research API support.
  • Handle image in assistant message for Gemini.
  • Added output_config as a supported parameter.
  • Added new Vercel AI Anthropic models.
  • Added new Claude 4-6 feature for v1/messages.
  • Added team policy mapping for zguard.
  • Added SDK proxy authentication (OAuth2/JWT auto-refresh) documentation.
  • UI Feature: Logs now show Predefined Error Codes in Filter with User Definable Fallback.
  • UI Feature: Invite User now includes an Email Integration Alert.
  • UI Feature: M2M OAuth2 Flow implemented.

🐛 Bug Fixes

  • Fixed PostHog JSON serialization error for non-serializable objects.
  • Fixed UI issue where Input and Output Copying was broken in Logs.
  • Fixed UI issue regarding Available Teams.
  • Fixed enterprise license check being triggered when guardrails/policies arrays are empty.
  • Fixed critical CVE vulnerabilities.
  • Fixed hook to handle database cases.
  • Fixed Bedrock Claude Opus 4.6 model ID by removing the :0 suffix.
  • Fixed Cohere responseFormat/Pydantic issues for OCI.
  • Fixed showing 'Data Not Available' instead of error details for failed requests.
  • Fixed null guard for models in the API keys table in the UI.
  • Fixed Spend logs pickle error when using Pydantic models and redaction.
  • Fixed Vertex AI propagating extra_headers anthropic-beta to the request body.
  • Fixed preserving tool call argument deltas when streaming ID is omitted.
  • Fixed preserving interleaved thinking/redacted_thinking blocks during streaming.
  • Fixed video list pagination cursors not being encoded with provider metadata.
  • Fixed base_model name for body and deployment name in URL.
  • Fixed Prometheus label values sanitization to prevent metric scrape failures.
  • Fixed OAuth2 root endpoints returning "MCP server not found".
  • Fixed preserving nullable object fields by carrying schema properties.
  • Fixed issue where shared backend model key was being polluted by per-deployment custom pricing.
  • Fixed OpenTelemetry (otel) auto-inferring otlp_http exporter when endpoint is configured.
  • Fixed respecting ssl_verify with shared aiohttp sessions.
  • Fixed PyPI CI Step in CI/CD.
  • Fixed Azure AI/kimi-k2.5 model being added to Azure model DB.

Affected Symbols

PostHogVertex AIPrometheusaiohttpVerificationTokenAnthropicBedrockCohereGeminiOpenTelemetry (otel)langchain_aws