Hand the chat your camera and it sees what you see. Useful for the case where you cannot name the thing — a label, a plate of food, a part on the bench.
From the workspace, add camera as a source.
It is identified from the live feed.
What it finds becomes a subject, with the usual tree.
under the hoodPOST /api/food-detection on frames from the live stream.
Needs a camera and its browser permission; nothing is recorded.