How we source live event data

LeekDuck is the de facto canonical source for PoGo event news — their team posts within minutes of any official Niantic announcement. ScrapedDuck is an open data project that scrapes LeekDuck's events feed into structured JSON and hosts it on GitHub Pages.

Our /api/events route pulls the ScrapedDuck JSON every hour, normalizes it into our internal Event shape, and serves the locale-translated variant to clients based on the ?locale= query.

1. Hourly cron → fetch https://scraped-duck.github.io/events.json
2. Diff against last snapshot → identify new + changed events
3. For each new event × each of 10 locales → call Gemini Flash for translation
4. INSERT into event_translations table ON CONFLICT DO NOTHING
5. Bust the /api/events?locale=* cache

The events page must never fail just because the data pipeline does:

• ▸If ScrapedDuck is down → serve the last cached JSON from Postgres (up to 24h stale).
• ▸If Postgres is down → serve a static fallback shipped with the build.
• ▸If Gemini translation fails for a single locale → serve English in that locale, log the failure, retry on next cron.
• ▸If a single event has malformed dates → drop just that event, serve the rest.

ScrapedDuck has a stable contract (a JSON file shape), is updated by a community team we coordinate with, and is rate-limited friendly. Hitting LeekDuck directly would require parsing HTML that may change without notice and would add load on their server.

If ScrapedDuck were ever to disappear, we have a direct LeekDuck scraper fallback at scripts/scrape-leekduck.ts but have never had to enable it.