Skip to main content
Docling simplifies document processing, parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the gen AI ecosystem.

​
What is Docling?

Docling is a powerful SDK and CLI tool that converts documents from various formats (PDF, DOCX, PPTX, HTML, images, audio, and more) into a unified, structured representation perfect for downstream AI workflows. With state-of-the-art PDF understanding capabilities, OCR support, and native integrations with popular AI frameworks, Docling accelerates your document AI applications.

Installation

Get started with pip, uv, or your preferred package manager

Quick start

Convert your first document in minutes with simple Python code

Core concepts

Learn about Docling’s architecture and key components

Usage guides

Explore detailed guides for common workflows

​
Key features

​
Multi-format support

Docling handles multiple document formats out of the box:
  • PDF with advanced layout understanding, reading order, table structure, formulas, and more
  • Office documents: DOCX, PPTX, XLSX
  • Web formats: HTML
  • Images: PNG, TIFF, JPEG with OCR support
  • Audio: WAV, MP3 with Automatic Speech Recognition (ASR)
  • Specialized formats: LaTeX, WebVTT, XBRL financial reports, JATS articles, USPTO patents

​
Advanced PDF understanding

Docling excels at PDF processing with:
  • Page layout analysis - Detect headers, paragraphs, lists, tables, figures, and formulas
  • Reading order detection - Preserve logical document flow
  • Table structure recognition - Extract tables with accurate cell relationships
  • Code block detection - Identify and preserve code snippets
  • Formula understanding - Recognize mathematical equations
  • Image classification - Categorize figures and diagrams

​
Unified document representation

All documents are converted to the expressive DoclingDocument format, providing:
  • Structured content with hierarchical organization
  • Metadata preservation including fonts, styles, and layout information
  • Relationship tracking between document elements
  • Export flexibility to Markdown, HTML, JSON, DocTags, and more

​
AI framework integrations

Plug-and-play integrations with popular frameworks:

LangChain

Build RAG applications with LangChain document loaders

LlamaIndex

Index and query documents with LlamaIndex readers

Haystack

Create search pipelines with Haystack converters

Crew AI

Power AI agents with structured document data

​
Privacy and security

Docling supports local execution for sensitive data and air-gapped environments - all processing happens on your infrastructure with no external API calls required.

​
What’s new

Docling is actively developed with frequent improvements and new features.
  • Structured information extraction - Extract specific data using schemas [beta]
  • Heron layout model - Faster PDF parsing by default
  • MCP server - Connect to agentic applications via Model Context Protocol
  • XBRL parsing - Process financial reports in eXtensible Business Reporting Language
  • WebVTT support - Parse and export Web Video Text Tracks
  • LaTeX parsing - Convert LaTeX documents to structured format

​
Use cases

Docling powers a wide range of document AI applications:
  • RAG (Retrieval-Augmented Generation) - Convert documents for vector databases and semantic search
  • Document Q&A - Extract structured content for question answering systems
  • Data extraction - Pull tables, figures, and structured data from reports
  • Archive digitization - OCR and structure legacy documents
  • Compliance & analysis - Process financial reports, patents, and regulatory documents
  • Content migration - Convert documents between formats while preserving structure

​
Technical foundation

Docling leverages cutting-edge AI models including:
  • Layout models (Heron) for page segmentation
  • Table structure models for accurate table extraction
  • Visual Language Models (GraniteDocling) for enhanced understanding
  • OCR engines (EasyOCR, Tesseract, RapidOCR, macOS Vision)
  • ASR models (Whisper with MLX acceleration on Apple Silicon)
For technical details, see the Docling Technical Report (arXiv:2408.09869).

​
Get started

1

Install Docling

Install via pip or your preferred package manager
pip install docling
2

Convert a document

Run your first conversion with 3 lines of code
from docling.document_converter import DocumentConverter

converter = DocumentConverter()
result = converter.convert("document.pdf")
print(result.document.export_to_markdown())
3

Explore capabilities

Check out usage guides and examples to learn more

​
Next steps

Installation guide

Detailed installation instructions for all platforms

Quick start tutorial

Complete walkthrough with working examples

Join the community

Get help and share your projects on Discord

GitHub repository

Star the project and contribute on GitHub

Build docs developers (and LLMs) love

Get started for freeTalk to us