opendataloader-pdf

OpenDataLoader PDF is an open‑source parser that turns PDF files into formats that large language models can work with, such as Markdown and JSON. The tool runs locally, so no data leaves your machine, and it is released under the Apache‑2.0 license.

The parser focuses on the needs of retrieval‑augmented generation (RAG) pipelines. It extracts reading order, tables, formulas, pictures, charts and bounding‑box coordinates for every element, then outputs a structured JSON schema that includes type, page number, font details and exact coordinates. You can also request HTML, plain PDF or Markdown alongside the JSON. The library is available for Python, Node.js and Docker, and a LangChain integration is provided for building RAG applications. Typical usage involves calling `opendataloader_pdf.convert` with an input path array and an output directory, then selecting one or more output formats such as `json,html,pdf,markdown`.

If you need to improve accessibility, OpenDataLoader PDF can auto‑tag untagged PDFs, producing a screen‑reader‑ready Tagged PDF that complies with PDF/UA specifications and has been validated with veraPDF. The auto‑tagging pipeline works entirely offline and adds the necessary accessibility tags according to the PDF Association standards. For OCR, the tool supports hybrid OCR + AI processing in over 80 languages, and it can enhance complex table extraction with an optional LLM step that claims 93 % table accuracy in the benchmark results shown on the site.

The benchmark tables on the website list OpenDataLoader PDF as the top‑scoring parser across several metrics, including overall score (0.907), reading‑order accuracy (0.902), table extraction (0.928) and heading detection (0.739). Speed figures are also provided, with a full pipeline time of about 0.015 seconds per page. The site includes a live demo, sample extractions and a JSON schema example that shows how each element is represented, for instance a heading object with its bounding box and typography information.

The tool is suited for developers building RAG pipelines, data engineers who need reliable layout extraction, and accessibility teams that must batch‑process PDFs for compliance. It may be less appropriate for casual users who only need simple text extraction without bounding‑box data, or for organizations that require a hosted SaaS solution, since the product is purely local and command‑line driven.

Key features

• Converts PDFs to LLM‑ready Markdown, JSON, HTML and PDF outputs