2 leches

ShelfSight is a cashierless smart store for small Mexican convenience stores (“abarrotes”), built with one phone camera instead of expensive sensors. The problem: corner stores sell a large share of everyday goods in Mexico, but they run on manual cash registers and paper inventory. Stores like Amazon Go solve checkout with weight sensors, proximity sensors and custom-trained vision models, which no small shop can afford. How it works: the customer enters by showing a QR code from our app, which links the visit to their account. An iPhone pointed at the shelf detects when a hand takes or returns a product. It sends before/after photos to Meta’s Muse Spark, which identifies the product and quantity from the store’s catalog. The item goes to the customer’s live cart. When they leave, they tap “Exit and pay”: the sale, card charge and inventory update happen automatically. When the model is unsure, the owner confirms with one tap in a review queue. At the counter, Muse Spark also replaces the barcode scanner: one photo identifies the product. The owner gets a full web dashboard with sales, profit, low-stock alerts, inventory history (kardex) and tickets. Stack: React PWA, Bun + Elysia, PostgreSQL, deployed on Railway. Quité los saltos de línea y espacios sobrantes, y agregué el punto final que faltaba.