A local web app that gathers restaurant search results, stores them in CSV, and lets you filter restaurants and check reservation availability.
I started with a practical spec: a laptop-friendly restaurant research app with a UI, manual refresh, CSV storage, and filters for cuisine and rating. The AI scaffolded the FastAPI app, scraper, templates, styling, CSV merge logic, and reservation-checking flow. I made product decisions along the way, like starting with Google only, keeping storage as CSV, deduplicating records, and assuming party size of two. We iterated on real runtime errors and messy scraped output until the app behaved more like a usable local tool.
The first scraping pass produced duplicate restaurant rows because Google’s result markup exposed both outer cards and inner detail blocks. Reservation availability also hit the limits of scraping: OpenTable and Resy rely on dynamic pages and browser behavior, so Playwright became necessary and still required graceful failure handling.
I started with a rough product idea, then narrowed scope through specific constraints: local app, manual refresh, CSV instead of a database, and Dallas,TX as the first test city. After the first version worked, I prompted one feature or bug at a time, using pasted output and stack traces to guide each fix.