DukaanAI: Turn Product Photos into a Shop Inventory with Open-Weight AI Tags: #hf26challenge #devchallenge #hacktoberfest #ai

Small retailers often need to maintain product catalogues containing product names, categories, descriptions, prices, and stock quantities. Entering this information manually can be repetitive, especially for shops with many products.

I wanted to explore a simple question: what if a shopkeeper could start creating an inventory entry just by uploading a product photo?

That question led me to build DukaanAI, an AI-assisted inventory tool designed for small retailers and local shopkeepers.

DukaanAI converts a product image into a suggested inventory entry.

AI-generated product names, categories, and descriptions

The goal is not to replace a shopkeeper's judgment. It is to reduce repetitive data entry while keeping the final decision in the shopkeeper's hands.

The frontend is built with React and Vite. A Node.js and Express backend receives uploaded images and communicates with a locally running Gemma 3 model through Ollama.

The model analyzes the image and returns structured JSON containing suggested product details. The shopkeeper reviews these details, enters the actual price and stock quantity, and saves the record.

The initial version stores inventory in browser localStorage, making it a local-first prototype rather than a cloud-based multi-user system.

I chose an open-weight model because the cost and control of AI inference matter for a product aimed at small businesses.

Using Gemma through Ollama allows the model to run locally on compatible hardware. Product images can be analyzed without sending each image to a third-party hosted AI API.

This approach also lets developers experiment with prompts, evaluate different models, and build a workflow around their specific use case.

Local inference has trade-offs: the shopkeeper needs suitable hardware, model inference can be slower than hosted services, and model-generated information can be incorrect. The application therefore asks users to review AI suggestions before saving them.

Building DukaanAI is teaching me how to connect a user interface, backend API, image-processing workflow, and open-weight AI model into one practical application.

It has also highlighted an important product lesson: generating text is only one part of the solution. The application must fit the shopkeeper's existing workflow and produce results that are accurate enough to be useful.

My initial target users are independent grocery shops and small retailers that need a simple way to create product catalogues.

I plan to validate whether the tool saves meaningful time by testing it with local shopkeepers. I will compare manual product entry with AI-assisted entry, collect feedback about accuracy, and ask whether users would pay for a more complete version.

A possible future business model is a monthly subscription, but willingness to pay must be validated before committing to a price.

Future improvements could include cloud synchronization, shop accounts, barcode support, low-stock alerts, and integration with existing billing systems.

GitHub repository: [INSERT YOUR REPOSITORY LINK]

Working demo or video: [INSERT YOUR DEMO LINK]

Screenshots: [INSERT SCREENSHOTS OR LINKS]

I want to test DukaanAI with real shopkeepers and discover which inventory tasks create the most friction in their daily work.

If you run a small retail business, I would appreciate your feedback: would generating a product catalogue from photos save you time, and what would you need before using a tool like this in your shop?

Built for the Hacktoberfest Open-Source AI Challenge: Week 2 — Side Hustle.