Docling

Free

Great library for ingesting any kind of document for RAG ⭐

FreeFree tier

Outputs: code

Type

Open Source

Visit Website GitHub (40 stars)

About Docling

Docling simplifies document processing by parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the generative AI ecosystem. It supports a wide range of formats such as PDF, DOCX, PPTX, XLSX, HTML, EPUB, WAV, MP3, WebVTT, images, email formats (EML, MSG), LaTeX, plain text, and more. Advanced PDF capabilities include layout analysis, reading order, table structure, code, formulas, and image classification. Docling offers local execution for sensitive data, extensive OCR, support for visual language models (e.g., GraniteDocling), audio automatic speech recognition, and video parsing. It exports to Markdown, HTML, WebVTT, DocLang, DocTags, and JSON, and integrates with LangChain, LlamaIndex, Crew AI, and Haystack. A CLI and API server (docling-serve) are also available.

Key Features

Parses PDF, DOCX, PPTX, XLSX, HTML, EPUB, audio (WAV, MP3), video (MP4, AVI), email (EML, MSG), images, LaTeX, and plain text

Advanced PDF understanding: layout, reading order, table structure, code, formulas, image classification

Unified DoclingDocument representation with export to Markdown, HTML, WebVTT, DocLang, DocTags, and JSON

Local execution for sensitive data and air-gapped environments

Integrations with LangChain, LlamaIndex, Crew AI, and Haystack

Extensive OCR support for scanned PDFs and images

Support for visual language models (e.g., GraniteDocling)

Audio speech recognition (ASR) and video parsing with transcripts and keyframes

Chart understanding (bar, pie, line) converting charts into tables or code

CLI and API server (docling-serve) for running as a service

Pros & Cons

Pros

Open-source and free to use
Supports an extensive variety of document formats (text, audio, video, images)
Advanced PDF layout and structure analysis
Local execution for data privacy and offline scenarios
Seamless integration with popular AI frameworks (LangChain, LlamaIndex, etc.)
Active development with regular updates and new feature releases

Cons

Requires Python 3.10 or higher (dropped Python 3.9 support)
Some features (e.g., metadata extraction, complex chemistry) are still in development

Best For

Ingesting documents for Retrieval-Augmented Generation (RAG) pipelinesExtracting structured data from financial reports (XBRL, XLSX)Processing patent documents (USPTO XML) and academic articles (JATS)Transforming emails and attachments into AI-ready formatsBuilding document processing workflows in air-gapped or sensitive environmentsConverting scanned PDFs and images into machine-readable text via OCR

FAQ

What is Docling?

Docling is an open-source tool that simplifies document processing by parsing diverse formats (PDF, DOCX, images, audio, video, etc.) and providing integrations with the generative AI ecosystem.

How do I install Docling?

Install via pip: 'pip install docling'. Requires Python 3.10 or higher. Works on macOS, Linux, and Windows (x86_64 and arm64).

What document formats does Docling support?

Docling supports PDF, DOCX, PPTX, XLSX, HTML, EPUB, WAV, MP3, WebVTT, Box Notes, email (EML, MSG), images (PNG, TIFF, JPEG), LaTeX, plain text, ODF (ODT, ODS, ODP), XBRL, and video (MP4, AVI, MOV, MKV, WebM), among others.

Can Docling work offline without sending data to external services?

Yes, Docling supports local execution, making it suitable for sensitive data and air-gapped environments.

Does Docling integrate with LangChain or LlamaIndex?

Yes, Docling provides plug-and-play integrations with LangChain, LlamaIndex, Crew AI, and Haystack for agentic AI workflows.