v1.79.dev.1
📦 litellmView on GitHub →
✨ 9 features🐛 13 fixes🔧 11 symbols
Summary
This release adds several UI enhancements, new provider and routing features, and numerous bug fixes across management endpoints, streaming, and budget handling.
✨ New Features
- UI - Model Info Page Health Check
- Add softgen to projects that are using litellm
- Add Kimi K2 thinking support
- UI - Test Key Page shows models based on selected endpoint
- Add GET list of providers endpoint
- UI - Invite User searchable team select
- Add SDK‑focused examples for custom prompt management
- Router now supports default fallbacks for unknown models
- AI Gateway – End User Budgets can point max_end_user budget to an ID, applying the default ID to all end users
🐛 Bug Fixes
- Fix container API link in release page
- Docs: fix streaming example in README
- Management Endpoints – inconsistent error responses fixed; non‑existent user now returns proper 404 with consistent schema
- Allow internal users to access video generation routes
- LiteLLM Usage now correctly displays key_hash
- Removed strict master_key check in add_deployment
- Proxy: corrected date range filtering in /spend/logs endpoint
- Updated model_cost_map_url to use environment variable
- Langfuse: handle null usage values to prevent validation errors
- Apply provided timeout value to ClientTimeout.total
- Fixed bug where updated spend was not sent to CloudZero
- Fixed inability to delete MCP server from permission settings
- Bedrock Knowledge bases – ensure users can access `search_results` for both streaming and non‑streaming responses to /chat/completions
🔧 Affected Symbols
add_deploymentClientTimeout.totalmodel_cost_map_urlroutermax_end_user_budgetvideo_generation_routesLiteLLM_UsageBedrockKnowledgeBase.search_resultsGET /providers endpointGET /spend/logs endpointcustomer_management_endpoints