▶ Live demo: aidvalor.github.io/fieldframe (no install, opens in the browser)
Problem: Field crews take hundreds of jobsite photos, and they end up as an unlabeled camera roll. Finding "the before shot of the north wall" later is slow, and reports get built by hand.
Who it's for: Contractors and field techs documenting work on a phone, often one-handed, with spotty signal.
What it does: A three-screen clickable prototype: capture → AI processing → organized output. AI runs on shutter press (not continuous video) and tags, groups and suggests captions. When confidence is low, it falls back to asking the user instead of guessing. Side annotations explain each design call.
Personal project; not affiliated with or endorsed by any company. Released under the MIT License.