StockEntry Guard: Local Gemma-Powered Product Onboarding for Small Retailers
Description and use
StockEntry Guard uses a local Gemma 3 model to extract product data from images, checks for duplicates, and lets the retailer manually approve additions or updates to a local SQLite inventory.
Problem
Small retailers struggle with duplicate product entries when catalogs grow, leading to messy inventories.
Outcome
StockEntry Guard reduces repetitive data entry and prevents duplicate products by combining local AI extraction with deterministic duplicate detection, all under human approval.
Workflow
- Import existing inventory CSV
- Upload product photo
- Extract product details locally with Gemma 3 via Ollama
- Review and edit extracted fields
- Compare candidate against existing catalog for duplicates
- Choose to create new product, update existing, or discard
- Save approved product to local SQLite database
- Export updated inventory to CSV
Evidence
StockEntry Guard helps a retailer: Import an existing inventory CSV. Upload a product photo. Extract product details locally with Gemma 3. Review and edit the extracted fields. Compare the product against the existing catalog. See whether there is no match, a possible match, or a likely duplicate. Create a new product or update an existing one only after explicit human approval. Export the updated inventory back to CSV.
The demo shows the complete retailer workflow: Import an existing inventory CSV. Upload a product image. Extract structured product details locally with Gemma. Review and edit the extracted fields. Check the product against the existing inventory for possible duplicates. Approve the product. Save it into the local inventory database. Export the updated inventory back to CSV.
StockEntry Guard runs Gemma 3 locally through Ollama for product-image extraction.
The extracted result is never written directly to inventory. The retailer reviews the fields first, then runs StockEntry Guard's deterministic duplicate check.
- Record date: 2026-10-18T22:41:02.326Z