opendataloader-project/opendataloader-pdf is a Python and Node library for PDF parsing and accessibility tagging. It processes digital, scanned, and tagged PDFs into structured formats to address parsing issues such as broken tables and incorrect reading order.
Project overview
It combines deterministic local parsing with an optional AI hybrid mode for high-accuracy extraction and offers open-source auto-tagging for screen-reader-ready Tagged PDFs.
Project type
RAG
Use cases
Documents & Office · Knowledge Q&A
Deployment
Refer to project documentation
License
Apache-2.0
Best for
Developers and enterprise teams who need to convert PDF files into structured formats while preserving layout coordinates and generating accessible Tagged PDFs.
Workflows requiring local execution of PDF parsing without mandatory GPU hardware.
Key capabilities
Generates screen-reader-ready Tagged PDFs from untagged inputs by applying structure tags for headings, paragraphs, lists, tables, and reading order.
Extracts exact layout using native PDF structure tags when available, preserving headings, lists, tables, and reading order without guessing.
Verifies PDF tag trees. This capability is currently planned and not yet available.
Limitations and risks
Repeated convert() calls are slow because each invocation spawns a new JVM process.
The system optionally uses SmolVLM (256M), a lightweight vision model, for AI picture descriptions. This optional AI mode involves a model dependency.
An end-to-end local-only data boundary is not documented. Optional AI features and cost dependencies exist that may involve external services.
Getting started
Setup is rated as easy. You must ensure Java 11 or later is pre-installed on your system. You can then install the library by running pip install -U opendataloader-pdf.
Alternatives and comparisons
Converts PDF documents and images into structured JSON or Markdown using a vision-language model for RAG and agentic applications.
Parses PDF, image, DOCX, PPTX, and XLSX inputs into Markdown and JSON, handling complex layouts, formulas, and OCR.
A graph-based RAG framework offering incremental updates and multiple query modes without full rebuilds.
README: solves: [PDF to structured data for RAG/LLM pipelines, accelerate PDF accessibility remediation — layout analysis + auto-tagging to Tagged PDF as foundation for PDF/UA (first open…