litellm_presidio-dev-v1.81.16
📦 litellmView on GitHub →
✨ 4 features🐛 8 fixes🔧 4 symbols
Summary
This release focuses on stability and UI improvements, introducing pagination for key aliases, fixing OOM issues, and enhancing real-time WebSocket error handling. A previously added /public/supported_endpoints feature was reverted.
Migration Steps
- The feature to add /public/supported_endpoints endpoint was reverted; this functionality is not available in this release.
✨ New Features
- Added control for setting upperbound on chunk processing time.
- Implemented paginated Key Alias Select UI.
- Added /public/endpoints endpoint for provider endpoint support.
- Added tags in project.
🐛 Bug Fixes
- Added pagination and search to /key/aliases endpoint to prevent OOMs.
- Suppressed attr-defined mypy errors on realtime websocket calls.
- Fixed virtual Keys pagination displaying stale totals when filtering.
- Cleaned up LiteLLM prometheus multiproc.
- Fixed guardrails not firing for Gemini/Vertex AI and provider_config realtime WebSocket sessions.
- Fixed Prometheus Metrics Team +Inf Budgets.
- Removed cache eviction close that kills in-use httpx clients.
- Enhanced MidStreamFallbackError to preserve original status code and attributes.