OpenDataLoader PDF

PDF Parsing for RAG — Convert to Markdown & JSON, Fast, Local, No GPU

Convert PDFs into LLM-ready Markdown and JSON with accurate reading order, table extraction, and bounding boxes — all running locally on your machine.

Why developers choose OpenDataLoader:

• Deterministic — Same input always produces same output (no LLM hallucinations)
• Fast — Process 100+ pages per second on CPU
• Private — 100% local, zero data transmission
• Accurate — Bounding boxes for every element, correct multi-column reading order

pip install -U opendataloader-pdf

import opendataloader_pdf

# PDF to Markdown for RAG
opendataloader_pdf.convert(
    input_path="document.pdf",
    output_dir="output/",
    format="markdown,json"
)

Why OpenDataLoader?

Building RAG pipelines? You've probably hit these problems:

| Problem | How We Solve It | |---------|-----------------| | Multi-column text reads left-to-right incorrectly | XY-Cut++ algorithm preserves correct reading order | | Tables lose structure | Border + cluster detection keeps rows/columns intact | | Headers/footers pollute context | Auto-filtered before output | | No coordinates for citations | Bounding box for every element | | Cloud APIs = privacy concerns | 100% local, no data leaves your machine | | GPU required | Pure CPU, rule-based — runs anywhere |

Key Features

For RAG & LLM Pipelines

• Structured Output — JSON with semantic types (heading, paragraph, table, list, caption)
• Bounding Boxes — Every element includes [x1, y1, x2, y2] coordinates for citations
• Reading Order — XY-Cut++ algorithm handles multi-column layouts correctly
• Noise Filtering — Headers, footers, hidden text, watermarks auto-removed
• LangChain Integration — Official document loader

Performance & Privacy

• No GPU — Fast, rule-based heuristics
• Local-First — Your documents never leave your machine
• High Throughput — Process thousands of PDFs efficiently
• Multi-Language SDK — Python, Node.js, Java, Docker

Document Understanding

• Tables — Detects borders, handles merged cells
• Lists — Numbered, bulleted, nested
• Headings — Auto-detects hierarchy levels
• Images — Extracts with captions linked
• Tagged PDF Support — Uses native PDF structure when available
• AI Safety — Auto-filters prompt injection content

Output Formats

| Format | Use Case | |--------|----------| | JSON | Structured data with bounding boxes, semantic types | | Markdown | Clean text for LLM context, RAG chunks | | HTML | Web display with styling | | Annotated PDF | Visual debugging — see detected structures (sample) |

JSON Output Example

{
  "type": "heading",
  "id": 42,
  "level": "Title",
  "page number": 1,
  "bounding box": [72.0, 700.0, 540.0, 730.0],
  "heading level": 1,
  "font": "Helvetica-Bold",
  "font size": 24.0,
  "text color": "[0.0]",
  "content": "Introduction"
}

| Field | Description | |-------|-------------| | type | Element type: heading, paragraph, table, list, image, caption | | id | Unique identifier for cross-referencing | | page number | 1-indexed page reference | | bounding box | [left, bottom, right, top] in PDF points | | heading level | Heading depth (1+) | | font, font size | Typography info | | content | Extracted text |

Full JSON Schema →

Quick Start

• Python
• Node.js / TypeScript
• Docker
• Java

Advanced Options

opendataloader_pdf.convert(
    input_path="document.pdf",
    output_dir="output/",
    format="json,markdown,pdf",

    # Image output mode: "off", "embedded" (Base64), or "external" (default)
    image_output="embedded",

    # Image format: "png" or "jpeg"
    image_format="jpeg",

    # Tagged PDF
    use_struct_tree=True,            # Use native PDF structure
)

Full CLI Options Reference →

AI Safety

PDFs can contain hidden prompt injection attacks. OpenDataLoader automatically filters:

• Hidden text (transparent, zero-size)
• Off-page content
• Suspicious invisible layers

This is enabled by default. Learn more →

Tagged PDF Support

Why it matters: The European Accessibility Act (EAA) took effect June 28, 2025, requiring accessible digital documents across the EU. This means more PDFs will be properly tagged with semantic structure.

OpenDataLoader leverages this:

• When a PDF has structure tags, we extract the exact layout the author intended
• Headings, lists, tables, reading order — all preserved from the source
• No guessing, no heuristics needed — pixel-perfect semantic extraction

opendataloader_pdf.convert(
    input_path="accessible_document.pdf",
    use_struct_tree=True  # Use native PDF structure tags
)

Most PDF parsers ignore structure tags entirely. We're one of the few that fully support them.

Learn more about Tagged PDF →

Hybrid Mode

For documents with complex tables or OCR needs, enable hybrid mode to route challenging pages to an AI backend while keeping simple pages fast and local.

Results: Table accuracy jumps from 0.49 → 0.93 (+90%) with acceptable speed trade-off.

pip install -U "opendataloader-pdf[hybrid]"

Terminal 1: Start the backend server

opendataloader-pdf-hybrid --port 5002

Terminal 2: Process PDFs with hybrid mode

opendataloader-pdf --hybrid docling-fast input.pdf

Or use in Python:

opendataloader_pdf.convert(
    input_path="complex_tables.pdf",
    output_dir="output/",
    hybrid="docling-fast"  # Routes complex pages to AI backend
)

• Local-first: Simple pages processed locally, complex pages routed to backend
• Fallback: If backend unavailable, gracefully falls back to local processing
• Privacy: Run the backend locally in Docker for 100% on-premise

Hybrid Mode Guide →

LangChain Integration

OpenDataLoader PDF has an official LangChain integration for seamless RAG pipeline development.

pip install -U langchain-opendataloader-pdf

from langchain_opendataloader_pdf import OpenDataLoaderPDFLoader

loader = OpenDataLoaderPDFLoader(
    file_path=["document.pdf"],
    format="text"
)
documents = loader.load()

# Use with any LangChain pipeline
for doc in documents:
    print(doc.page_content[:100])

• LangChain Documentation
• GitHub Repository
• PyPI Package

Benchmarks

We continuously benchmark against real-world documents.

View full benchmark results →

Quick Comparison

| Engine | Overall | Reading Order | Table | Heading | Speed (s/page) | |-----------------------------|----------|---------------|----------|----------|----------------| | opendataloader | 0.68 | 0.91 | 0.49 | 0.65 | 0.05 | | opendataloader [hybrid] | 0.88 | 0.93 | 0.93 | 0.78 | 0.48 | | docling | 0.86 | 0.90 | 0.89 | 0.80 | 0.73 | | marker | 0.83 | 0.89 | 0.81 | 0.80 | 53.93 | | mineru | 0.82 | 0.86 | 0.87 | 0.74 | 5.96 | | pymupdf4llm | 0.57 | 0.89 | 0.40 | 0.41 | 0.09 | | markitdown | 0.29 | 0.88 | 0.00 | 0.00 | 0.04 |

Scores are normalized to [0, 1]. Higher is better for accuracy metrics; lower is better for speed. Bold indicates best performance.

Visual Comparison

Roadmap

See our upcoming features and priorities →

Documentation

• Quick Start Guide
• JSON Schema Reference
• CLI Options
• Tagged PDF Support
• AI Safety Features

Frequently Asked Questions

What is the best PDF parser for RAG?

For RAG pipelines, you need a parser that preserves document structure, maintains correct reading order, and provides element coordinates for citations. OpenDataLoader is designed specifically for this use case — it outputs structured JSON with bounding boxes, handles multi-column layouts correctly with XY-Cut++, and runs locally without GPU requirements.

How do I extract tables from PDF for LLM?

OpenDataLoader detects tables using both border analysis and text clustering, preserving row/column structure in the output. Tables are exported as structured data in JSON or as formatted Markdown tables, ready for LLM consumption.

Can I use this without sending data to the cloud?

Yes. OpenDataLoader runs 100% locally on your machine. No API calls, no data transmission — your documents never leave your environment. This makes it ideal for sensitive documents in legal, healthcare, and financial industries.

What makes OpenDataLoader unique?

OpenDataLoader takes a different approach from many PDF parsers:

• Rule-based extraction — Deterministic output without GPU requirements
• Bounding boxes for all elements — Essential for citation systems
• XY-Cut++ reading order — Handles multi-column layouts correctly
• Built-in AI safety filters — Protects against prompt injection
• Native Tagged PDF support — Leverages accessibility metadata

This means: consistent output (same input = same output), no GPU required, faster processing, and no model hallucinations.

How do I get better accuracy for complex tables?

Enable hybrid mode with pip install -U "opendataloader-pdf[hybrid]". This routes pages with complex tables to an AI backend (like docling-serve) while keeping simple pages fast and local. Table accuracy improves from 0.49 to 0.93 — matching or exceeding dedicated AI parsers while remaining faster and more cost-effective.

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

License

Mozilla Public License 2.0

Found this useful? Give us a star to help others discover OpenDataLoader.