Repository navigation
First release of Loci: your room as a memory palace.
What it does
- Take or choose a photo of your room. Paste 3 to 12 items, in order.
- An AI finds real objects in the photo. Code picks the stops and joins them into one route.
- Learn: the view walks the route, with one short scene per stop.
- Lights out: the room goes dark. Say or type each item. A right answer turns its stop green and lights that part of the room again.
- Code checks every answer, never an AI. Retry only the misses.
- Palaces stay on your device.
- Three example rooms (AI-generated) and three example lists, so anyone can try it in seconds.
Tests
- 81 unit tests (Vitest)
- 19 browser checks (Playwright, with the AI and the microphone stubbed)
- 7 live checks against the real AI
About the GIF
Recorded from the production build of this tag with npm run demo (scripts/record-demo.mjs, Playwright), then converted with ffmpeg. It uses the prepared example palace, so no AI call is made. The answers are typed by the script, not recalled by a person. One typo ("occulomotor") and one accepted answer ("auditory") show what the checker forgives.
Links
- App: https://loci.edycu.dev (mirror: https://loci.edycu.dev)
- Plan: scope 路 PRD 路 technical spec 路 build checklist
