Recall
A visual-memory assistant. Photograph your drawers and shelves, a vision model inventories what is inside, and later you ask where something is and get the exact spot with the photo.
The same project, from different angles
One document per perspective, because the roles I played on this are not the same role.
An experiment in whether a vision model is good enough to replace the mental index of where household things live — and in how little of that has to leave the device for it to work.
Capture, downscale on-device, send only the single frame being analysed to whichever vision provider is configured, then review and correct the returned inventory before it is saved. The photos and the index stay in the browser's local database.
The whole thing — a weekend-scale experiment, not a product.
Runs locally, including on a phone over the local network. Deliberately unreleased; it is here as an experiment rather than as something finished.