●Change8

v1.81.10-nightly

📦 litellmView on GitHub →
✨ 12 features🐛 19 fixes🔧 9 symbols

Summary

This release focuses heavily on stability, security fixes (including critical CVEs), and extensive UI improvements and refactoring. New features include support for Websearch interception hooks and expanded model/API support for Anthropic, Sarvam AI, and Perplexity.

Migration Steps

  1. If using semantic filtering, be aware of the new warning related to semantic filter hooks.
  2. If encountering issues with Spend logs serialization involving Pydantic models and redaction, this release contains a fix.
  3. If using Vertex AI with Anthropic beta headers, ensure extra_headers are propagated correctly.
  4. If using Prometheus metrics, label values are now sanitized to prevent scrape failures.

✨ New Features

  • Added support for anthropic_messages call type in prompt caching.
  • Added Websearch support using LiteLLM /search (using web search interception hook) for chat completion.
  • Added support for MCP M2M Oauth2 support.
  • Added Sarvam AI STT and TTS support.
  • Added Kimi model pricing by region.
  • Added moonshotai.kimi-k2.5 model.
  • Added support for managed error file.
  • Added Perplexity research api support.
  • Handle image in assistant message for gemini.
  • Add output_config as supported param.
  • Add new claude 4-6 feat for v1/messages.
  • Add support for langchain_aws via litellm passthrough.

🐛 Bug Fixes

  • Fixed PostHog JSON serialization error for non-serializable objects.
  • Fixed UI - Logs: Input and Output Copying issues.
  • Fixed UI - Teams: Available Teams display issue.
  • Fixed enterprise license check being triggered when guardrails/policies arrays are empty.
  • Fixed critical CVE vulnerabilities.
  • Fixed Bedrock Claude Opus 4.6 model ID (removed :0 suffix).
  • Fixed Cohere responseFormat/Pydantic issues in OCI.
  • Fixed showing error details instead of 'Data Not Available' for failed requests.
  • Fixed null guard for models in API keys table in UI.
  • Fixed Spend logs pickle error with Pydantic models and redaction.
  • Fixed vertex_ai propagating extra_headers anthropic-beta to 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 encoded with provider metadata.
  • Fixed base_model name for body and deployment name in URL.
  • Fixed shared backend model key being polluted by per-deployment custom pricing.
  • Resolved OAuth2 root endpoints returning "MCP server not found".
  • Preserved nullable object fields by carrying schema properties.
  • Fixed get_supported_anthropic_messages_params.

Affected Symbols

PostHogVertex AIBedrock Claude OpusCoherePrometheusVerificationTokenlangchain_awsanthropic_messagesget_supported_anthropic_messages_params