Home

ShelfWise Wiki

Welcome to the ShelfWise wiki — the living documentation for the AI Product Portfolio Builder that turns a spreadsheet of UPCs and SKUs into a complete, market-ready product catalog.

What is ShelfWise?

When someone buys a pallet of inventory, they often receive hundreds of products with nothing but UPC barcodes or internal SKUs. At around 8 minutes per item, a pallet of 400 products can require more than 50 hours of manual research before anything is listed for sale.

ShelfWise automates that work. Upload a CSV of UPCs or SKUs, or paste them directly into the app. The system searches across the web — retailer listings, manufacturer pages, marketplaces, specialty stores, distributor catalogs, and public product databases — and uses a multi-step reasoning agent to build cited product records with verified photos, titles, and descriptions.

Primary use cases: liquidation pallets, wholesale lots, estate sales, POS-to-marketplace listing generation for Uber Eats, DoorDash, Grubhub, Shopify, Amazon, eBay, Etsy, and more.

Key Capabilities

Capability	Description
UPC-to-Listing Automation	Paste UPCs or upload a CSV to get complete product records.
Multi-Source Scraping	10 core sources + up to 271 registry sources queried concurrently (default 10 registry sources). Per-source health tracking.
Reasoning Agent	Deduplicates names, weighted-vote brand/category resolution, normalized attributes, and computes confidence scores.
Verified Imagery	Downloads candidate images and scores them for background, resolution, focus, sharpness, and duplication. Marketplace-aware sizing. Falls back to web image search for local PLUs. Supports manual upload/review.
Optional LLM Enrichment	Azure OpenAI / GitHub Models / Ollama integration with deterministic local fallback.
Live Streaming UI	Real-time SSE progress bars and job status updates.
Multi-Format Export	11 formats with filtering and preview: CSV, JSON, Shopify, Amazon, WooCommerce, eBay, Etsy, BigCommerce, DoorDash, Uber Eats, Grubhub.
Accessibility First	WCAG 2.1 AA compliant, keyboard navigation, screen-reader support, reduced motion & high contrast.

Quick Links

• 🚀 Getting Started
• 🎯 Use Cases
• 🏗️ Architecture
• 📡 API Reference
• ⚙️ Configuration
• 🗺️ Roadmap
• 🧪 Testing
• 🚢 Deployment

Project Status

Current phase: Hackathon-ready MVP with local reasoning, optional Azure OpenAI enrichment, Docker support, and CI/CD via GitHub Actions.

ShelfWise works without any API keys using a deterministic local reasoning engine. Optional cloud AI services can be enabled for higher-confidence enrichment.

Repository

• Source code: agents-league-shelfwise/
• Main docs: README.md, INNOVATION_STUDIO.md, TESTING_PLAN.md
• License: MIT