Building a Local Receipt Extractor with Open-Source AI (Side Hustle)

This submission is for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle. The author has created a Local Receipt & Invoice Extractor, a privacy-focused tool designed for freelancers and small business owners. Users simply upload a receipt image, and the tool extracts key details like the vendor name, date, total amount, and tax amount.

The primary advantage of this application is its emphasis on data privacy. Freelancers often hesitate to upload sensitive client bills to cloud servers belonging to companies like Google or OpenAI. This new tool eliminates that concern.

The UI of the application is built using Streamlit (Python). The Optical Character Recognition (OCR) process, which converts the image into text, is handled by Tesseract OCR. For AI parsing, the app utilizes Ollama running the Llama 3 model locally.

The author highlights the importance of open innovation. With the use of Ollama to run Llama 3 locally, users can ensure their financial data remains completely offline and private. No API calls are made, and no cloud servers are involved. Furthermore, there are zero API costs, making the application entirely free to run. This model also ensures the app remains free from lock-ins, allowing users to swap models or fine-tune prompts according to their specific needs, such as processing Indian-style GST invoices.

To illustrate the potential user base, the author mentions a friend who runs a small boutique. This friend currently stores all their bills in a shoebox. They expressed interest in paying just ₹200/month for a tool that could securely sort their bills without the need to upload them to the cloud.