Ollama-OCR

Why we included this project

Extracting text from images and PDFs usually means either standing up a dedicated OCR engine or writing a lot of glue code. This package skips that by sending each file to a vision language model running locally through Ollama, so you can switch between LLaVA, Llama 3.2 Vision, Moondream, and others without changing your calling code. It returns more than plain text: markdown, JSON, key-value pairs, and table data make it practical for digitizing invoices and forms. Batch processing with progress tracking and custom prompts cover the common automation cases, and the included Streamlit interface gives non-technical users a drag-and-drop way to run the same features. Accuracy varies by model, so it is worth sanity-checking the smaller ones before trusting them with important documents.