Backend API
PDF Price Scraping Pipeline
How Cookest extracts supermarket promotion prices from weekly PDF flyers
PDF Price Scraping Pipeline
Cookest includes an admin-only pipeline that extracts product prices from weekly supermarket promotional flyers and makes them available to Pro users via the shopping list optimizer.
Pipeline overview
Requirements
# macOS
brew install poppler
# Debian/Ubuntu
sudo apt install poppler-utils
# Pull the vision model
ollama pull llavaAdmin endpoints
| Method | Path | Description |
|---|---|---|
POST | /api/admin/stores | Create a new store record |
POST | /api/admin/stores/:id/promotions/upload | Upload weekly promo PDF |
GET | /api/admin/stores/:id/jobs | Check job status |
All admin endpoints require a JWT with is_admin: true verified against the database (not just the token claim).
Extraction prompt
The vision model is sent a structured prompt asking it to extract:
- Product name and brand
- Original price and discounted price
- Unit (per kg, per unit, per litre, etc.)
- Promotion validity dates
The response is parsed as JSON and inserted into store_promotion_candidates for admin review before going live.
Pro user access
Once promotions are live, Pro tier users can:
GET /api/shopping-list/prices— current prices for all items in their shopping listGET /api/shopping-list/optimize— cheapest single-store and cheapest multi-store split
Price data is store-specific and promotion-based — not a real-time price feed. Prices reflect the most recently uploaded weekly flyer for each store.