docling-project

docling: Parse Documents for Generative AI

Docling is a Python document-processing library that turns PDFs and many other formats into structured content for generative AI workflows. It supports local processing on macOS, Linux, and Windows, with integrations for popular AI frameworks.

ActivePythonMITAILibraryPDF
Stars
69.8k
Forks
5.2k
Stars, last 30 days
+1,644
Last commit
today
Contributors
100+

Our take

Activeregular commits and releases

Health score

79/100

How it is scored
Activity
90
Community
92
Issues
73
Pull requests
59

+1,644 stars in the last 30 days

  • Releases more often than 93% of the projects we track
  • Merges more pull requests than 84% of the projects we track
  • More contributors than 78% of the projects we track

Docling is an actively maintained, broad document-processing library with a practical fit for teams that need structured inputs for AI applications.

Good fit if

  • You need to parse PDFs and a range of other formats into structured, exportable content.
  • Local processing, OCR, or integrations with the named AI frameworks matter to your workflow.
  • Your environment can run Python 3.10 or later and you want an MIT-licensed core library.

Look elsewhere if

  • You need metadata extraction such as document authors, references, or language, which the README lists as coming soon.
  • Your workflow depends on complex chemistry understanding, also listed as coming soon.
  • You need a single fixed model or service: Docling supports multiple models and pipelines rather than framing one as its sole engine.
All health signals
Last commit2026-10-28 (today)
Commits, last 90 days100+
Releases, last 12 months100+ (latest v2.144.0, 2026-10-28)
Contributors100+ (top contributor: 18% of commits)
Issues closed, last 90 days24+ (typically closed in 2 days)
Pull requests merged, last 90 days76 (typically merged in 1 day)
Project age2 years

Checked on 2026-10-29 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

70,673

Forks

5,364

Change since Nov 16, 2026: +46 stars·+3 forks

Overview

Docling is a Python library for preparing documents for generative AI: it parses varied file formats and provides advanced understanding of PDFs, including layout, reading order, and tables. Developed by IBM Research Zurich's AI for knowledge team, it uses a unified document representation to export structured content for downstream applications.

It runs on macOS, Linux, and Windows, on both x86_64 and arm64. Its local execution option can suit sensitive-data workflows, while integrations with frameworks such as LangChain, LlamaIndex, Crew AI, and Haystack connect document processing to AI applications.

Key Features

  • Parse PDFs, office files, HTML, EPUB, images, audio, video, email, and other formats.
  • Analyze PDF layout, reading order, tables, code, formulas, and images.
  • Export to Markdown, HTML, WebVTT, DocLang, DocTags, and lossless JSON.
  • Apply OCR to scanned PDFs and images.
  • Use visual language models, including IBM's GraniteDocling, and audio transcription models.
  • Connect workflows through integrations, an MCP server, an API server, or the CLI.
  • Run locally for sensitive data or air-gapped environments.

Use Cases

  • Build retrieval or generative AI pipelines from PDFs and office documents, with structured text and tables instead of raw files.
  • Process scanned documents when OCR and page-layout understanding are needed.
  • Connect document ingestion to an existing LangChain, LlamaIndex, Crew AI, or Haystack application.
  • Convert mixed-format collections locally when documents should not be sent to an external service.
  • Extract transcripts and representative frames from video, or transcribe supported audio files for downstream workflows.

What you need

Detected in the repository

  • Python >=3.10,<4.0 (from pyproject.toml)
  • A Dockerfile, so it can run in a container
  • A test suite and automated checks on GitHub Actions

License in plain words

MITpermissive

  • Commercial use: yes
  • Modify and redistribute: yes
  • You must keep: the copyright and license notice
  • Share your changes: no

A summary, not legal advice: the LICENSE file is what applies.

Getting Started

Install the Python package:

pip install docling

See the README and documentation for setup and usage details.

Alternatives

  • xberg: Xberg uses a Rust core and offers bindings for 15 languages plus CLI, REST API, and MCP interfaces.
  • marker: Marker focuses on converting documents to Markdown, JSON, HTML, or chunks while preserving tables, equations, and images.
  • unstructured: Unstructured parses documents into structured elements for downstream workflows, with format-specific dependencies for some inputs.
  • pdf-inspector: PDF Inspector focuses on classifying PDFs and extracting positioned text, routing scanned pages to OCR.
ProjectLanguageLicenseStarsStatus
doclingPythonMIT69.8kActive
xbergRustMIT9.4kActive
markerPythonApache-2.040.2kActive
unstructuredHTMLApache-2.015.5kActive
pdf-inspectorRustMIT20kActive

Considerations

The project is active, with frequent commits and releases. Its contributor base is broad rather than concentrated around one person, and issue and pull-request handling is recent. The package requires Python 3.10 or later; model usage may carry separate license terms from the MIT-licensed codebase. The README also identifies metadata extraction and complex chemistry understanding as features that are not yet available.

Found this useful?

Share it with someone who would like docling.

Comparisons