Fit Receipt

Fit Receipt is a private digital lookbook and virtual fitting room built for the NUDE lingerie collection, where a shopper can browse the collection, open product detail, and try pieces on virtually before deciding. The experience is framed around a single principle stated on the page: start with your life, then choose the bra. Instead of leading with a product grid, the fitting room first learns how you will actually wear the piece, then moves on to styles and the whole-set budget. Any answer you give can be changed later, and the person using it — not the software — keeps the final decision. On Product Hunt it is described simply as a lingerie fitting room that shows its reasoning. Buying lingerie without a fitting room is a guessing game: sizing, shape and comfort are hard to judge from pictures, and recommendations are often opaque. Fit Receipt addresses that by making reasoning visible rather than hidden. Every pick carries an AI judgment receipt, so the shopper can follow the case from a real life brief to a receipt-backed draft. At the same time, the product treats privacy as a design constraint rather than an afterthought: the text you type stays in the browser's memory only and is never sent to a model, photos stay in your browser, and there is no server photo storage. That combination — transparent reasoning plus local-only photos — is what distinguishes the fitting room from a conventional try-on widget. The first step in the room is a needs-first flow. A prompt asks, "What would you like to solve this time?", inviting the shopper to describe the situation in their own words. That text stays in the browser's memory only and is never sent to a model. The options presented underneath are the needs you confirm, so the system works from what the shopper explicitly agrees to rather than from inferences. Two entry points are offered: Confirm my needs, or Fill in a demo case for people who want to see the flow with example inputs first. Because every answer can be changed, the brief remains a living document that the shopper can revise as their thinking develops. The ordering matters — lifestyle and use case come before styles and the whole-set budget — so the product set under consideration is scoped to how the piece will actually be worn. Once the brief is confirmed, the room supports browsing the collection, viewing product detail, and trying on virtually. Each pick that emerges from this process is attached to an AI judgment receipt, a written trace of how the choice was reached. The Product Hunt listing frames the whole experience as a case that runs from a real life brief to a receipt-backed draft, with the person keeping the final say. The receipt is therefore best understood as an accountability layer: it records what was judged and why, so the shopper is reviewing an explained draft rather than accepting an unexplained recommendation. The reasoning behind those receipts comes from JEV (TypeSafe), described as a judgment model that scores trade-offs with calibrated confidence in roughly 300 milliseconds — explicitly not essays. JEV is invoked by the agent, which knows when to call it. If the model's confidence is low, the result stays unresolved instead of being forced into a definitive answer, and the judgments that are produced get recorded. Importantly, the fitting room works without JEV as well, so the judgment layer is an enhancement rather than a hard dependency. This division of labour is stated plainly: the agent handles chat, while code protects facts and pricing. In other words, conversational flexibility never gets to override the deterministic parts of the experience such as product facts and prices. The privacy model is the other half of the architecture. JEV sees typed fields and never the photo, so the judgment step operates on structured inputs rather than imagery. Photos stay in your browser, and there is no server photo storage of any kind. Text entered into the brief likewise stays in the browser's memory only and is never sent to a model. The API is rate-limited, and the build is hosted on Vercel and released as open source. Fit Receipt is described as a reference implementation, which is why these boundaries — what the agent may decide, what the model may see, and what never leaves the device — are made explicit rather than left to be assumed. For the shopper, the outcomes follow from those choices. Recommendations arrive as a draft with reasoning attached, so decisions can be reviewed rather than merely accepted. Uncertainty is surfaced instead of hidden: when confidence is low, the item is left unresolved rather than dressed up as a conclusion. Judgments are recorded, which means the case can be revisited after any answer is changed. And because photos and typed notes remain on the device, trying styles on virtually does not require handing personal images to a server. Concrete use cases follow the flow described on the page. A shopper can open the fitting room with a real-life brief, describe how they intend to wear the piece, and confirm the needs the room should solve for; from there they browse the collection, open product detail, and try items on virtually. Someone who wants to understand the mechanics first can fill in a demo case and watch a case move from brief to receipt-backed draft. A shopper who is unsure can revise any answer and see how the recorded judgments change. Developers and evaluators have a further use case: Fit Receipt is an open-source reference implementation on Vercel that demonstrates how an agent can call a judgment model such as JEV, keep low-confidence results unresolved, and keep facts and pricing under code control. Cross-border shoppers are served too, since the room ships from Taiwan with pricing shown in both currencies. On the audience and commercial side, the room is aimed at lingerie shoppers — and anyone who wants a fitting experience that shows its reasoning while keeping photos local. The technology stack, as described, includes JEV (TypeSafe) as the judgment model, a rate-limited API, Vercel for hosting, and open-source code. Pricing on the page is presented in Taiwanese dollars, with US dollar amounts shown as approximate conversions at NT$31.76 = US$1 (Bank of Taiwan spot rate, Sep 21, 2026), and prices are charged in NT$ at checkout. Orders ship from Taiwan with free shipping over NT$3,000 in Asia and NT$5,000 to Europe and the Americas, arriving in 7–10 business days; overseas orders cannot be returned. Taken together, Fit Receipt pairs a virtual lingerie fitting room with a receipt for every pick, so the shopper always sees the reasoning behind a recommendation instead of a black box. The brief comes first, the person keeps the final say, low confidence stays unresolved, and photos never leave the browser — a fitting room that shows its work.