mistocr 0.0.4

Simple batch OCR for PDFs using Mistral's state-of-the-art vision model

mistocr

Why mistocr?

Performance: Mistral’s OCR delivers state-of-the-art accuracy on complex documents including tables, charts, and multi-column layouts.

Scale: Process entire folders of PDFs in a single batch job. Upload once, process asynchronously, and retrieve results when ready - perfect for large document sets.

Cost savings: Batch OCR mode reduces costs from $1/1000 pages to $0.50/1000 pages - a 50% reduction compared to synchronous processing.

Simplicity: A single ocr() function handles everything - uploading, batch submission, polling for completion, and saving results as markdown with extracted images. Process one PDF or an entire folder with the same simple interface.

Organized output: Each PDF is automatically saved to its own folder with pages as separate markdown files and images in an img subfolder, making results easy to navigate and process further.

Installation

Install latest from the GitHub repository:

$ pip install git+https://github.com/franckalbinet/mistocr.git

or from pypi

$ pip install mistocr

How to use

Basic usage

Process a single PDF:

from mistocr.core import ocr

fname = 'files/test/attention-is-all-you-need.pdf'
result = ocr(fname)

Or process an entire folder:

results = ocr('files/test')

Output structure

Each PDF is saved to its own folder with pages as separate markdown files and images in an img subfolder:

files/test/md/
├── attention-is-all-you-need/
│   ├── img/
│   │   ├── img-0.jpeg
│   │   ├── img-1.jpeg
│   │   └── ...
│   ├── page_1.md
│   ├── page_2.md
│   └── ...
└── resnet/
    ├── img/
    └── ...

Reading results

Read all pages from a processed PDF:

from mistocr.core import read_pgs

text = read_pgs('files/test/md/attention-is-all-you-need')

Or read a specific page:

text = read_pgs('files/test/md/attention-is-all-you-need', 10)

Customization

Customize output directory, image inclusion, and polling interval:

results = ocr('files/test', out_dir='output', inc_img=False, poll_interval=5)

Parameters:

• path: A single PDF file or folder containing multiple PDFs
• out_dir: Directory name for saving markdown output (default: 'md')
• inc_img: Include extracted images in the output (default: True)
• key: Your Mistral API key (uses MISTRAL_API_KEY environment variable if not provided)
• poll_interval: Seconds between batch job status checks (default: 2)

Returns: List of paths to the generated markdown files

Developer Guide

If you are new to using nbdev here are some useful pointers to get you started.

Install mistocr in Development mode

# make sure mistocr package is installed in development mode
$ pip install -e .

# make changes under nbs/ directory
# ...

# compile to have changes apply to mistocr
$ nbdev_prepare

Documentation

Documentation can be found hosted on this GitHub repository’s pages. Additionally you can find package manager specific guidelines on conda and pypi respectively.