# OpenDataLoader PDF

An open-source PDF parser that extracts Markdown, JSON with bounding boxes, and HTML from PDFs for AI and RAG pipelines, and automates PDF accessibility tagging.

OpenDataLoader PDF is an open-source PDF parser that extracts Markdown, JSON with bounding boxes, and HTML from PDF documents, aimed at AI data extraction and retrieval-augmented generation. It runs in a deterministic local mode or an AI hybrid mode for complex pages, and includes built-in OCR in more than 80 languages for scanned documents.

The hybrid mode also handles complex or borderless tables, LaTeX formulas, and AI-generated descriptions of pictures and charts. A separate accessibility feature automatically tags untagged PDFs into screen-reader-ready Tagged PDFs; converting a Tagged PDF further to PDF/UA-1 or PDF/UA-2 compliance is a paid add-on. The auto-tagging was built with Dual Lab, the developers of veraPDF, following the Well-Tagged PDF specification.

It requires Java 11 or newer and Python 3.10 or newer, installs with pip, and has Python, Node.js and Java SDKs. The layout-analysis and auto-tagging feature is licensed Apache-2.0.

Website: https://opendataloader.org