🎯 Consumer Retail Optimization
Canterbury consumers face fragmented grocery prices between Pak’nSave, New World, and Countdown/Woolworths. Existing apps require expensive Google Maps API keys or manual price entry.
🛒 Core Intelligent Capabilities
- Zero-Cost Store Discovery: Queries OpenStreetMap Overpass API with multi-mirror failover (
overpass-api.de,maps.mail.ru,kumi.systems), achieving full local supermarket discovery without recurring map billing. - TSP Multi-Store Deal Optimizer: Solves a Traveling Salesperson Problem route (
Home ➔ Store A ➔ Store B ➔ Home), recommending multi-store trip splits whenever grocery savings exceed the threshold. - Triple-Mode Receipt Vision: Uses
gemini-3.5-flash-litewith GPU-accelerated image downsampling (createImageBitmap) to parse printed paper receipts directly into digital line items.
🧪 Rigorous QA & Container Deployment
- 85 passing unit tests verifying price parsers, spatial deduplication, and currency normalization.
- Packaged with Docker Compose, running Flask + Gunicorn and PostgreSQL 15.
💻 Tech Stack & Repositories
- Backend: Python 3.11, Flask, Gunicorn, PostgreSQL 15, Docker Compose.
- Frontend: PWA, Vanilla JS, IndexedDB, HTML5 Camera API.
- AI OCR: Gemini 3.5 Flash-lite Vision API.
- GitHub Repository: smartshop.git