Fit Receipt keeps the person in control: code protects product facts and pricing, while JEV weighs bounded trade-offs and shows uncertainty when the evidence is weak.
Where should an AI fitting agent explain its reasoning, and where should it leave room for personal preference?
Try on lingerie virtually; every pick has an AI judgment receipt.
Fit Receipt is a reference implementation. The agent handles chat; code protects facts and pricing. JEV (TypeSafe) is a judgment model that scores trade-offs with calibrated confidence in ~300ms—not essays. Low-confidence results stay unresolved; judgments are recorded; works without JEV.
Photos stay in your browser; no server photo storage. JEV sees typed fields, never the photo. Rate-limited API. Vercel + open source.
