Ollama Starter Document OCR

Simple CLI for extracting text from images and PDFs using Ollama.

Prerequisites

1. Install and run Ollama

Usage

# Process explicit files
npx ollama-starter-document-ocr ./receipts/scan1.png ./receipts/scan2.jpg

# Process all images in a folder (non-recursive)
npx ollama-starter-document-ocr ./receipts

# Process a PDF (one .txt per page)
npx ollama-starter-document-ocr ./receipts/statement.pdf

# Override output directory
npx ollama-starter-document-ocr ./receipts --out-dir ./output

# Use a different model 
npx ollama-starter-document-ocr ./receipts --model deepseek-ocr

Output

• Each image has a corresponding .txt file with the extracted text.
• For PDFs: each page is rendered to an image and then processed
• A JSON file is written to the output directory with the full results of every image/page
• Some models will detect text bounding boxes and annotate the images with them

Environment

Use OLLAMA_HOST if your Ollama server is not on the default http://localhost:11434.

OLLAMA_HOST=http://localhost:11444 npx ollama-starter-document-ocr ./receipts