v4.2.4
📦 localaiView on GitHub →
✨ 2 features🐛 4 fixes
Summary
This release focuses on bug fixes across distributed systems, HTTP handling, and agent pool persistence, alongside adding VRAM parsing from vulkaninfo and new Liquid Audio support.
✨ New Features
- Parse VRAM budget/usage from vulkaninfo.
- Add Liquid Audio s2s model and assistant mode on talk page for realtime features.
🐛 Bug Fixes
- Cascade-clean stale node_models rows and filter routing by healthy status in distributed systems.
- Honor X-Forwarded-Prefix header when proxy strips the prefix in HTTP requests.
- Fix truncate-then-read race condition in agent_jobs.json persistence for agent pools.
- Parse OpenAI-spec tool_choice correctly in /v1/chat/completions endpoint.