This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle. Small shop owners deal with sales, inventory, daily expenses, and profit calculations every day. Many still manage these tasks manually or use disconnected tools, making it difficult to understand how their business is performing.
DukaanAI is an AI-assisted inventory and business management system built specifically for Pakistani Kiryana stores, general stores, and small retailers.
It brings everyday shop operations into one application:
The system uses Pakistani Rupees, supports locally relevant payment methods such as JazzCash and EasyPaisa, and includes realistic Pakistani retail demo data.
Independent shopkeepers, Kiryana store owners, and small retailers who need an affordable alternative to complicated retail management software.
The business model could be a one-time setup fee or an affordable support and customization package. The goal is to make better business management accessible to shops that cannot justify expensive software subscriptions.
GitHub Repository: https://github.com/riazaslam029/DukaanAi Video Demo: https://drive.google.com/file/d/1a42gF3KGqIwmlvL0sIVoH2X6CD3FU_6S/view?usp=sharing
The project includes sample retail products, sales history, and expenses so viewers can explore the main workflows.
Explore the source code and run DukaanAI locally:
The project uses a React and TypeScript frontend, a Python FastAPI backend, and SQLite for local data storage.
I built DukaanAI using a full-stack architecture designed around practical retail workflows rather than relying on AI for every operation.
Technology stack:
One important design decision was separating deterministic business logic from AI-generated guidance.
Sales totals, stock deductions, expense records, and profit calculations are handled by application code. The AI integration is intended for natural-language assistance and business insights, where a language model can add value without becoming responsible for financial arithmetic.
The application also has an AI service interface that can detect a local Ollama instance and support a Gemma-based integration. The current repository documents the Gemma integration as a Phase 2 capability, so the AI functionality should be understood as a work in progress rather than a fully completed production feature.
For a small shop, recurring AI API costs can make an otherwise affordable tool difficult to sustain. Sending business records to an external service can also raise privacy concerns. Open-weight models such as Gemma offer a different approach.
With local inference through Ollama, DukaanAI can work toward three important goals:
1. Lower operating costs
Once the model is downloaded, local inference avoids per-request charges from a paid AI API. Hardware and electricity still have costs, but there is no requirement to pay for every model call.
2. Greater control over business data
A properly configured local deployment can keep sales, expenses, and inventory data on the shop owner's computer instead of sending that information to a third-party AI service.
3. Freedom to experiment and customize
An open-weight model gives developers more control over deployment and experimentation. It also makes it possible to explore different models and tailor the assistant to the language and workflows of Pakistani retailers.
Open innovation matters here because the people who could benefit from AI the most are not always the people who can afford expensive AI infrastructure.
I want DukaanAI to explore how local AI can make useful business software more accessible, affordable, and privacy-conscious.