v2.13.0 v0.2.11
medium risk
cheaper-inference/OmniRouteAug 3, 2026 → Aug 3, 20260 days between releases
What's New
Adds Chat Completions backend and text-only response capabilities.
Breaking Changes
- Removed deprecated 'melo' option from tts arguments
New Features
- Chat Completions LLM backend (OpenAI /v1/chat/completions)
- Text-only response mode
- Realtime web demo application
- Default Qwen3 TTS to GGML format
Improvements
- Improved realtime web demo app
- Updated README documentation
- Defer client conversation items during active response
- Bumped dependency groups
Bug Fixes
- Fix Windows installation error for faster-qwen3-tts
- Fix mid-generation conversation races (tool call & image)
- Fix Docker image build
Upgrade Notes
The deprecated 'melo' option has been removed from tts arguments. Update configuration to remove references to this option. Ensure Docker image builds correctly after update.
The Two Releases
v2.13.0
v2.13.0
Aug 3, 2026· 3 months ago
A stability-focused release that adds configurable legends to diagrams and improves the robustness of output and release generation.
Full release notesv0.2.11
v0.2.11
Aug 3, 2026· 3 months ago
A release focusing on stability and new backend support, introducing the Chat Completions backend and text-only response capabilities.
Full release notes