Reloved

Mobile-first web app for buying and selling preloved women's clothing in Pakistan. Built by Amna and Bismah for the Google Cloud AI Builder Cup. Prototype due 18 October 2026.

Read CLAUDE.md first. It holds the project rules, what is built so far, and the decisions that are easy to undo by accident. The product spec is docs/PRD.md; the seller flow's technical plan is docs/seller-flow-research.md.

Getting set up

You need Node and the gcloud command line tool.

npm install

Gemini runs on Vertex AI and there are no API keys. You sign in instead, once:

gcloud auth login
gcloud auth application-default login
gcloud config set project YOUR_PROJECT_ID
gcloud services enable aiplatform.googleapis.com

Then copy .env.example to .env.local and fill in GOOGLE_CLOUD_PROJECT. .env.local is gitignored and must never be committed.

Running it

npm run dev

• App: http://localhost:3000
• Is Gemini working? http://localhost:3000/api/health/gemini — expect {"ok":true,...,"reply":"ok"}
• Try the seller AI by hand: http://localhost:3000/dev/prefill

If the health check fails, your sign-in has usually expired. Run gcloud auth application-default login again.

Other commands

npm run build                    Production build
npx tsc --noEmit                 Type check
npm run try-prefill -- ./photos  Run a folder of photos through the AI, print a table

What works today

The two seller AI calls, end to end: photos in, draft listing fields and photo quality flags out. Photos are converted from HEIC and resized in the browser first.

There are no screens yet, no Firebase, and no styling. /dev/prefill is a bare test page, not a design — it returns 404 in a production build on purpose.

See "What is actually built" in CLAUDE.md for the current line between plan and code.

Careful with

• Never commit .env.local or any key.
• The AI routes have no sign-in check and no spending limit yet. Laptop only. Do not deploy them publicly.
• Every Gemini call is server-side only, never from the browser.