yoga, guided by voice and vision
Yogi asks for your microphone and camera once, the first time you meet — your browser will show a single permission prompt. Camera video is analyzed on this device and never uploaded; your voice is understood on this device too, by an open Whisper model — nothing extra to allow, ever. Yogi's conversational mind prefers a local open model (Ollama on this machine — see the 🧠 Brain pill), falls back to Anthropic's API with a key, and to built-in rules otherwise.