Human memory is excellent at feeling that we know someone and notoriously poor at retrieving why. Where did we meet? What were they working on? Did they mention a sick parent, a new job, a trip to Lisbon? Recall-AI is an ambient perception system that closes that gap. A camera and microphone feed a pipeline that:
Recognizes the face in front of you in near real time.
Listens to the conversation and works out who said what.
Distills each conversation into a living Persona: names, relationship, key facts, and a rolling summary.
Surfaces that persona the next time the same face appears, before you have to ask.
"You're meeting them for the first time." is the empty state.
"Alice. Works at a robotics startup. Training for a marathon. Asked you to send the paper." is the goal.
- SETUP.md — install and run on macOS / Windows
- Browser client: start the backend and open
http://localhost:8000. Select AWS Rekognition (default), local InsightFace, or Mock withRECALL_FACE_BACKEND. - DESIGN.md — architecture · TASKS.md — build plan · RESEARCH.md — prior art