A real AI agent watches your hand (eyes) and decides an action (brain) — both in your browser. Static + motion gestures. Toggle Afferens: on, the agent acts on a stable named intent; off, it’s stuck with raw landmarks.
Structured gesture / intent
🤖 AI agent
A real LLM agent that decides an action from your gesture. Loads once, then runs offline — no server, no cost.
What you’re seeing: a real agent — eyes (MediaPipe) + brain (Llama-3.2, in your browser). Afferens OFF = raw landmark coordinates, the agent can’t act. ON = a stable named gesture, it acts decisively. That’s the Afferens layer. Recognition adapted from Saad Kamal’s open-source GestureSense (MIT).