Local Receipt Sorter
An open-source receipt organization tool built on the Gemma vision model with local inference via Ollama. It runs entirely on-device to preserve privacy, supports recognition of PDFs, photographs and screenshots, and extracts fields such as vendor, date and amount. Rule-based validation and export functions are built in.
Timeline (last 90 days)
Local Receipt Sorter is a receipt organizing tool that runs entirely on the user's local computer, using the Gemma vision model and the Ollama local inference framework to achieve automatic receipt recognition, field extraction, anomaly detection, and export.
Local Receipt Sorter uses the Gemma 3 vision model (4B parameters) served locally through Ollama, and the model can be switched to the 12B version via an environment variable to improve accuracy.
Local Receipt Sorter's tech stack includes PyMuPDF for parsing PDFs, Streamlit for providing the interactive interface, and pandas for handling exports.
Local Receipt Sorter forces the model to output conforming to a JSON schema, ensuring that the field structure returned is consistent across different input types (PDFs, photos, screenshots).
Rather than having the model report its own confidence, Local Receipt Sorter uses built-in rule checkers to automatically flag missing fields, erroneous or future dates, non-positive totals, and rows where the sum of subtotal and tax does not equal the total.
The creator tested Local Receipt Sorter on 25 documents, including 15 PDFs, 10 photos, and 15 screenshots, and the author acknowledged that this sample size can only demonstrate feasibility, not reliability.
Local Receipt Sorter still worked normally after Wi-Fi was turned off, confirming that the inference process is completed locally.
Local Receipt Sorter is an entry for the second week's "side hustle" theme of the Hacktoberfest challenge, and its code has been open-sourced on GitHub.
Local Receipt Sorter's commercialization ideas include selling a one-time license to freelancers, charging monthly, or offering a batch-processing service on behalf of clients.
Local Receipt Sorter's next-step plans include adding test documents, trying larger models, fine-tuning for niche scenarios, and attracting its first paying user.
All Facts (10)
Local Receipt Sorter is a receipt organizing tool that runs entirely on the user's local computer, using the Gemma vision model and the Ollama local inference framework to achieve automatic receipt recognition, field extraction, anomaly detection, and export.
50%UnverifiedLocal Receipt Sorter uses the Gemma 3 vision model (4B parameters) served locally through Ollama, and the model can be switched to the 12B version via an environment variable to improve accuracy.
50%UnverifiedLocal Receipt Sorter's tech stack includes PyMuPDF for parsing PDFs, Streamlit for providing the interactive interface, and pandas for handling exports.
50%UnverifiedLocal Receipt Sorter forces the model to output conforming to a JSON schema, ensuring that the field structure returned is consistent across different input types (PDFs, photos, screenshots).
50%UnverifiedRather than having the model report its own confidence, Local Receipt Sorter uses built-in rule checkers to automatically flag missing fields, erroneous or future dates, non-positive totals, and rows where the sum of subtotal and tax does not equal the total.
50%UnverifiedLocal Receipt Sorter is an entry for the second week's "side hustle" theme of the Hacktoberfest challenge, and its code has been open-sourced on GitHub.
50%UnverifiedLocal Receipt Sorter's next-step plans include adding test documents, trying larger models, fine-tuning for niche scenarios, and attracting its first paying user.
50%UnverifiedLocal Receipt Sorter still worked normally after Wi-Fi was turned off, confirming that the inference process is completed locally.
50%UnverifiedThe creator tested Local Receipt Sorter on 25 documents, including 15 PDFs, 10 photos, and 15 screenshots, and the author acknowledged that this sample size can only demonstrate feasibility, not reliability.
50%UnverifiedLocal Receipt Sorter's commercialization ideas include selling a one-time license to freelancers, charging monthly, or offering a batch-processing service on behalf of clients.
50%