Product Decoder

AI/ML / Full-Stack / Monorepo

Completed / In Development

Oct 2024 - Nov 2025

Description

A visual product recognition and metadata extraction experience: upload a product photo and receive nutrition facts, price history, store availability, and other structured information.

Project Overview

A high-performance full-stack monorepo application for visual product recognition. The architecture features a Next.js App Router frontend with Clerk authentication and shadcn/ui components, communicating with a Bun-powered Elysia.js API backend. The backend integrates with Google Vision (for reverse image search), GNews, Google CSE, and YouTube for comprehensive metadata extraction. Data persistence is managed via Prisma with a MongoDB backend. The project heavily utilizes the Effect-TS ecosystem for declarative data fetching and ArkType for runtime validation, ensuring a robust, type-safe, and observable data flow across the entire stack.

Technologies Used

Next.js 15

Bun

Elysia.js

TypeScript

Tailwind CSS

Prisma

MongoDB

Clerk Auth

Google Vision API

Effect-TS

ArkType

TurboRepo

Cypress

Playwright

OpenTelemetry

Project Gallery

Hover over any image to view it in full screen. Drag horizontally to browse through the gallery.

Project image

Back to All Projects