●Change8

v1.95.0-rc.3

📦 litellmView on GitHub →
✨ 2 features🐛 26 fixes🔧 11 symbols

Summary

This release introduces Docker image signature verification using cosign for enhanced security. It also includes numerous bug fixes across various components like Azure AI, guardrails, UI, and proxy, along with CI/CD improvements and dependency updates.

Migration Steps

  1. Verify Docker image signatures using cosign. Recommended method is to use the pinned commit hash for cryptographic immutability.
  2. To verify using the pinned commit hash, use: `cosign verify --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub ghcr.io/berriai/litellm:v1.95.0-rc.3`
  3. To verify using the release tag (easier but relies on tag protection), use: `cosign verify --key https://raw.githubusercontent.com/BerriAI/litellm/v1.95.0-rc.3/cosign.pub ghcr.io/berriai/litellm:v1.95.0-rc.3`

✨ New Features

  • Send RFC 8707 resource indicators on upstream OAuth legs.
  • Mark Cost Optimization as beta in the left navigation.

🐛 Bug Fixes

  • Advertise 1M context window for Claude Opus 4.6+ on Foundry.
  • Resolve judge_model credentials via lazy Router lookup in llm_as_a_judge.
  • Match MCP Servers tabs to the dashboard's line tab pattern.
  • Center vertical toolbar dividers.
  • Restore the Add MCP Server dialog size and header spacing.
  • Truncate long team names in the models table team dropdown.
  • Resolve dashboard base URL from env instead of hardcoding localhost.
  • Preserve cache_control breakpoints in compresr write-back.
  • Pass an explicit Python version request to uv tool install.
  • Match exact class in callback dedup so a custom subclass does not block a built-in logger.
  • Populate cache write token metrics for OpenAI-style usage.
  • Validate default team values in Default User Settings.
  • Sanitize per-key callback config out of logged metadata.
  • Roll up tool spend daily instead of scanning SpendLogs.
  • Keep one chat completion id per stream and always stream completed responses.
  • Compress content-parts messages in headroom guardrail (Anthropic traffic).
  • Release the pre-routing strategy slot when a deployment is replaced or deleted.
  • Classify all 4xx HTTPException guardrail blocks as intervened.
  • Unstake the reasoning-effort grid count and the responses bridge test.
  • Realign Admin UI specs with the MCP dialog and keyless landing.
  • Unblock the ui suite, fix the mcp registration race, park two known product bugs.
  • Make MCP and prometheus e2e tests robust to data-plane sync lag.
  • Fix cache leakage card layout to keep date picker on right.
  • Route /a2a through the gateway component.
  • Poll for both spend rows before asserting the cache-hit contract.
  • Avoid DB outage during planned RDS IAM rotation.

Affected Symbols

azure_aimcpguardrailsllm_as_a_judgeuie2eprometheusrouterproxyresponses_bridgegateway