Worn Wear Monitor Bot

01 / Problem

Patagonia's Worn Wear secondhand site lists vintage grails — Synchilla fleeces, Snap-Ts, Retro-X jackets — that sell out within minutes of going live. Manually refreshing is hopeless. The broader problem is structural: the just-added collection has 15,000+ items, and the site's pagination bails after a few clicks with no new results, so naive scraping only sees a few hundred items and misses anything buried deeper.

02 / Approach

Rather than scraping the full catalog, the bot targets specific search URLs (e.g. vintage+synchilla, vintage+snap-t) that return 100–500 results instead of 15,000 — so pagination completes fully and nothing gets missed. Playwright drives a headless Chromium instance. Items are matched against configurable keywords and style-number patterns (STYLE_NUMBERS=*_vintage catches vintage variants by URL). A seen_items.json file persists across restarts so the bot only alerts on genuinely new listings. When a match is found, ntfy.sh fires a push notification to your phone instantly. AUTO_ADD_CART is available but off by default. The whole thing runs as a systemd service on a $4/mo DigitalOcean droplet.

03 / Result

Catches new vintage listings the moment they appear, rather than whenever you happen to check. The priming workflow — running one or two poll cycles with AUTO_ADD_CART off to populate seen_items.json — means enabling auto-cart later only bags items added after you set it up, not the hundreds of existing ones already in the catalog.