Add detailed latency metrics overlay

Add head tracking with smoothed roll control, auto-start detection, and UI cleanup

Add YOLO26 model support with detection + pose estimation

Fix HF Space OAuth scopes to match allowed values

Prepare for HF Space deployment with OAuth auto-connect

Fix recursive stack overflow in central signaling cleanup

Add remote robot discovery via Grand Central signaling server

Add local object detection with bounding box overlay

Eagerly load Whisper model on page init instead of first mic tap

Add robot host switcher UI and fix TTS language detection

Add describe_scene vision tool for camera-based scene description

V2: Replace keyword parser with OpenAI-compatible LLM API

Detect language from transcribed text instead of Whisper API

Fix crash when releasing mic before recording starts

Auto-detect language for STT, TTS, and command responses

V1: Add voice interaction with STT/TTS and keyword commands

V0: Svelte + Vite app with WebRTC robot connection