This is my submission for the Hacktoberfest Open-Source AI Challenge: Week 2 — Side Hustle.
What I Built
Meet SoloBooks AI — a local-first AI-powered expense and receipt organizer built for freelancers, independent creators, and side hustlers.
When you're running a small side business, keeping track of receipts and expenses can quickly become a tedious task. Receipts get scattered across folders, expense records become disorganized, and preparing information for bookkeeping takes time away from actual work.
I built SoloBooks AI to make this process easier.
What can it do?
- 🧾 Receipt processing: Upload receipts and extract useful expense information using OCR and AI-assisted processing.
- 📂 Expense organization: Keep expense records organized in one place.
- 📊 Expense dashboard: Review recorded expenses and visualize spending.
- 📄 CSV export: Export expense data for further analysis and bookkeeping workflows.
- 🤖 Open-weight AI: Use Gemma 3 4B through an Ollama-compatible inference setup.
The goal is simple: help people spend less time organizing financial paperwork and more time building their businesses.
SoloBooks AI is a prototype, not a replacement for professional accounting or tax advice.
Demo
🎥 Watch the demo:
The video demonstrates the project and its workflow.
Code
💻 GitHub repository:
https://github.com/rudrakshk25060-csds/solobooks-ai
The repository contains the application code, setup instructions, and tests.
Tech stack
- Backend: Python and FastAPI
- Database: SQLite
- AI: Gemma 3 4B through Ollama-compatible inference
- Receipt processing: Tesseract OCR and PDF processing
- Frontend: HTML, CSS, and JavaScript
- Charts: Chart.js
I wanted to build something practical that combines AI with a real administrative problem faced by freelancers and small businesses.
How I Built It
I built SoloBooks AI using an AI-assisted development workflow and open-source technologies.
The application uses FastAPI for its backend, SQLite for storing expense records, and a lightweight web interface built with HTML, CSS, and JavaScript.
For receipt processing, the application combines OCR and AI-assisted extraction. Gemma 3 4B is used through an Ollama-compatible inference setup to help interpret receipt information.
The application then brings the expense information into an organized workflow where users can review their records, explore spending, and export their data.
I also added automated tests to check important application behavior.
One important design decision was to keep financial calculations in application logic rather than relying on an LLM to perform arithmetic. AI can assist with interpreting unstructured information, while deterministic code handles calculations.
Why Does Open Innovation Matter?
For a tool that handles financial documents, control over data and infrastructure matters.
Using an open-weight model such as Gemma gives developers the option to run inference locally instead of depending entirely on a proprietary, hosted AI API.
For SoloBooks AI, this creates several potential advantages:
- More control over data: In a local setup, users can keep their documents and application data on their own machine, subject to the actual configuration.
- Greater flexibility: Developers can inspect the surrounding application code and adapt the workflow to their needs.
- More control over costs: Local inference can reduce dependence on per-request API pricing, although hardware and electricity still have costs.
- Less vendor lock-in: An open model-based architecture gives developers more flexibility to experiment with alternative models and inference setups.
Open innovation also makes it easier for other developers to learn from the implementation, contribute improvements, and adapt the project for different workflows.
Local inference has trade-offs, including hardware requirements and model performance. I see that as part of the engineering challenge rather than something to hide.
My goal is to make useful AI-powered bookkeeping tools more accessible to people who cannot justify expensive software for a small side hustle.
Prize Categories
Best Use of Gemma
Gemma 3 4B is a core part of the AI-assisted receipt-processing workflow, making this project relevant to the Gemma category.
Top comments (0)