This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What I Built
OrderMint — Local-First AI Order Management for Small Sellers
Small businesses often receive orders through WhatsApp messages, Instagram DMs, and other informal conversations. Turning these messages into structured orders, checking quantities, calculating totals, and preparing bills manually takes time and can lead to mistakes.
I built OrderMint to simplify that workflow.
A seller can turn an informal customer message into a structured order draft, review the extracted items, correct details, and approve the order before generating a bill.
Key features include:
- AI-assisted extraction of products, quantities, and order notes.
- Product catalog matching and deterministic validation.
- Human review before order approval.
- SQLite-backed order persistence.
- Printable bills and daily order summaries.
- Local AI inference with a fallback when the model is unavailable.
Who would pay for it? Home bakers, tiffin services, cloud kitchens, and Instagram sellers who want a simpler way to manage incoming orders without adopting an expensive, complicated system.
Demo
Hosted application: https://ordermint-nw7h.onrender.com
The hosted frontend is currently experiencing a deployment issue, so the live interface may not load correctly. I’m working on resolving it rather than presenting a broken demo as functional.
For now, you can inspect the source code below. If a working local demo recording is available, I’ll add it here.
Code
GitHub repository: https://github.com/ohm-7643/OrderMint
The project is built with React, Vite, Node.js, Express, and SQLite. The repository contains the application code and setup instructions.
How I Built It
I built OrderMint using Qwen3-8B, an open-weight model running locally through Ollama.
The architecture consists of:
- Frontend: React + Vite
- Backend: Node.js + Express
- Database: SQLite with better-sqlite3
- AI inference: Qwen3-8B through Ollama
- Validation: Deterministic checks and fallback logic
The model assists with interpreting customer messages, while application logic validates the output and checks it against the product catalog. The seller remains responsible for reviewing and approving the final order.
I also added automated tests and synthetic order-evaluation cases to check application behavior.
One important design decision was to avoid making the AI model the sole authority for business-critical data. The model helps interpret the message, but validation and human approval remain part of the workflow.
Why Does Open Innovation Matter?
Open innovation made it possible for me to experiment with a capable language model locally without building the entire product around a paid, proprietary inference API.
Using an open-weight model through Ollama gave me more control over the inference environment and allowed me to explore a local-first approach to customer-order processing.
It also made the project easier to experiment with as a student developer working with limited resources. I could focus on the actual product workflow — extraction, validation, persistence, and review — instead of depending entirely on a closed API.
Open-source tools across the stack made it possible to build and test the application using technologies that other developers can inspect, learn from, and extend.
Prize Categories
I’m submitting OrderMint for the Side Hustle challenge.
The project explores a practical business use case for open-source AI: helping small sellers spend less time managing orders and more time serving customers.
What’s Next
My next priorities are to resolve the hosted frontend issue, test the workflow with actual small-business sellers, improve extraction accuracy using real feedback, and evaluate whether the tool can save enough time to justify a paid product.
I’d especially appreciate feedback from people who manage orders through chat messages and want a simpler workflow.
Top comments (0)