✨ Highlights
DeepSeek-V4 Series
- DeepSeek-V4-Flash — fast inference model
- DeepSeek-V4-Pro — high-performance model
- vLLM engine now supports DeepSeek-V4 series
Log Center & Monitoring
- New Log Center UI with Elasticsearch integration
- Kibana-style log enhancements
- Monitoring UI: time picker, auto-refresh, dashboard improvements
- Enhanced structured inference logs & monitoring metrics
Advanced Auth System
Experimental- Enhanced authentication & permission control
- Disabled by default — will be fully enabled in Xinference 3.0
New Web UI
Experimental- New UI rolling out gradually
- Collapsible sidebar with state persistence
- Disabled by default — will be fully enabled in Xinference 3.0
🌐 Community Edition
📦 Installation
pip
pip install 'xinference==2.9.0'
Docker
Pull the latest image or pip update inside the container.
🆕 New Models
✨ New Features
- 1 Log Center UI — New Elasticsearch-integrated log management interface
- 2 Monitoring UI — Time range picker and auto-refresh support
- 3 Kibana-style Log Enhancements — Advanced log visualization and filtering
- 4 Advanced Auth System — Authentication & permission control (experimental)
- 5 New Web UI — Modern interface with collapsible sidebar (experimental)
- 6 vLLM Engine — Added DeepSeek-V4 series support
- 7 Collapsible Sidebar — With state persistence
✦ Enhancements
Structured Logs
Added role and address fields
Monitoring Metrics
Enhanced monitoring metrics rolled back
i18n
Added Japanese and Korean language support
Docker Build
Split Docker build process for optimization
Model Configuration
Updated model JSON configurations
⚠ Bug Fixes
DeepSeek OCR Coordinate Parsing
Security fix for coordinate parsing in DeepSeek OCR output
tool_call Message
Fixed tool_call message handling
Docs Workflow Branch Injection
Security fix for branch injection risk in docs workflow
Dockerfile vLLM Update
Updated vLLM to version 0.21.0 in Dockerfile
📚 Documentation
New Release Notes
Added v2.8.0 release notes documentation
🏢 Enterprise Edition
Stability & Performance Updates
- Multiple stability fixes and optimizations
- Enterprise-specific performance improvements