avinfohub.vercel.app ↗ LIVE

How I built it

Problem: local info is scattered across WhatsApp groups. Solution: scrape, structure with AI, and publish as a clean feed.

Sources
  ├─ WhatsApp groups (Events, Market, Rentals, Carpool)
  └─ Direct submissions from app (form → HTML)
            │
            ▼
     Supabase Storage — ingestion bucket
     (raw HTML/PDF kept, images as base64)
            │
            ▼
     AI Extraction (Gemma / Gemini via AICredits)
     reads each post → structured row:
       events: title · date · time · location · details · image
       market: item · cost · details · images
       rentals: type · cost · details
       carpool: route · seats · details
            │
            ▼
     Supabase Postgres
     events / buy_sell / houserentals / carpooling
     + all images bucket (public URLs)
            │
            ▼
     Auroville Community Hub (Vercel)
     reads tables → searchable feeds, filters, images

Runs daily on a 24-hour window. Duplicates cleaned by date+contact, old events auto-removed, images deleted only when no row references them.

What visitors get

• One feed instead of 5 WhatsApp groups to check
• Every listing has a real date — Cinema week split into per-day rows, weekly events resolved to next date
• Marketplace with up to 3 photos per item, carpool split into request vs offer
• App submissions appear the same way as WhatsApp posts