Installation

To use Docling, simply install docling from your Python package manager, e.g. pip:

pip install docling

If you are using uv:

uv add docling

Works on macOS, Linux, and Windows, with support for both x86_64 and arm64 architectures.

Alternative PyTorch distributions

The Docling models depend on the PyTorch library. Depending on your architecture, you might want to use a different distribution of torch. For example, you might want support for different accelerator or for a cpu-only version. All the different ways for installing torch are listed on their website https://pytorch.org/.

One common situation is the installation on Linux systems with cpu-only support. In this case, we suggest the installation of Docling with the following options:

# Example for installing on the Linux cpu-only version
pip install docling --extra-index-url https://download.pytorch.org/whl/cpu

For uv users, add the PyTorch CPU index to your pyproject.toml:

[[tool.uv.index]]
name = "pytorch-cpu"
url = "https://download.pytorch.org/whl/cpu"
explicit = true

Then pin torch to that index:

[tool.uv.sources]
torch = [{ index = "pytorch-cpu" }]

Then run:

uv add docling

Installation on macOS Intel (x86_64)

When installing Docling on macOS with Intel processors, you might encounter errors with PyTorch compatibility. This happens because newer PyTorch versions (2.6.0+) no longer provide wheels for Intel-based Macs.

If you're using an Intel Mac, install Docling with compatible PyTorch Note: PyTorch 2.2.2 requires Python 3.12 or lower. Make sure you're not using Python 3.13+.

# For uv users
uv add torch==2.2.2 torchvision==0.17.2 docling

# For pip users
pip install "docling[mac_intel]"

# For Poetry users
poetry add docling

Available extras

The docling package is designed to offer a working solution for the Docling default options. Some Docling functionalities require additional third-party packages and are therefore installed only if selected as extras (or installed independently).

The following table summarizes the extras available in the docling package. They can be activated with: pip install "docling[NAME1,NAME2]"

Extra	Description
`asr`	Installs dependencies for running the ASR pipeline.
`vlm`	Installs dependencies for running the VLM pipeline.
`easyocr`	Installs the EasyOCR OCR engine.
`feat-ocr-nemotron`	Installs NVIDIA Nemotron OCR. Supported only on Linux x86_64 with Python 3.12 and CUDA 13.x.
`tesserocr`	Installs the Tesseract binding for using it as OCR engine.
`ocrmac`	Installs the OcrMac OCR engine.
`htmlrender`	Installs dependencies for HTML page rendering in the HTML backend.
`rapidocr`	Installs the RapidOCR OCR engine with onnxruntime backend.

OCR engines

Docling supports multiple OCR engines for processing scanned documents. The current version provides the following engines.

Engine	Installation	Usage
EasyOCR	`easyocr` extra or via `pip install easyocr`.	`EasyOcrOptions`
Nemotron OCR	`feat-ocr-nemotron` extra. Supported only on Linux x86_64 with Python 3.12 and CUDA 13.x. See installation note below.	`NemotronOcrOptions`
Tesseract	System dependency. See description for Tesseract and Tesserocr below.	`TesseractOcrOptions`
Tesseract CLI	System dependency. See description below.	`TesseractCliOcrOptions`
OcrMac	System dependency. See description below.	`OcrMacOptions`
RapidOCR	`rapidocr` extra can or via `pip install rapidocr onnxruntime`	`RapidOcrOptions`
OnnxTR	Can be installed via the plugin system `pip install "docling-ocr-onnxtr[cpu]"`. Please take a look at docling-OCR-OnnxTR.	`OnnxtrOcrOptions`

The Docling DocumentConverter allows you to choose the OCR engine with the ocr_options settings. For example

from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import (
    TesseractOcrOptions,
    PdfPipelineOptions,
)
from docling.document_converter import DocumentConverter, PdfFormatOption

pipeline_options = PdfPipelineOptions()
pipeline_options.do_ocr = True
pipeline_options.ocr_options = TesseractOcrOptions()  # Use Tesseract

doc_converter = DocumentConverter(
    format_options={InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)}
)

Tesseract installation

Tesseract is a popular OCR engine which is available on most operating systems. For using this engine with Docling, Tesseract must be installed on your system, using the packaging tool of your choice. Below we provide example commands. After installing Tesseract you are expected to provide the path to its language files using the TESSDATA_PREFIX environment variable (note that it must terminate with a slash /).

macOS (via Homebrew)Debian-basedRHEL

brew install tesseract leptonica pkg-config
TESSDATA_PREFIX=/opt/homebrew/share/tessdata/
echo "Set TESSDATA_PREFIX=${TESSDATA_PREFIX}"

apt-get install tesseract-ocr tesseract-ocr-eng libtesseract-dev libleptonica-dev pkg-config
TESSDATA_PREFIX=$(dpkg -L tesseract-ocr-eng | grep tessdata$)
echo "Set TESSDATA_PREFIX=${TESSDATA_PREFIX}"

dnf install tesseract tesseract-devel tesseract-langpack-eng tesseract-osd leptonica-devel
TESSDATA_PREFIX=/usr/share/tesseract/tessdata/
echo "Set TESSDATA_PREFIX=${TESSDATA_PREFIX}"

Linking to Tesseract

The most efficient usage of the Tesseract library is via linking. Docling is using the Tesserocr package for this.

If you get into installation issues of Tesserocr, we suggest using the following installation options:

pip uninstall tesserocr
pip install --no-binary :all: tesserocr

Nemotron OCR installation

Nemotron OCR requires the CUDA 13 PyTorch wheels. Install it with the feat-ocr-nemotron extra, the CUDA 13 PyTorch index, and the unsafe-best-match index strategy so pip resolves the CUDA-enabled torch packages correctly.

pip install "docling[feat-ocr-nemotron]" \
  --extra-index-url https://download.pytorch.org/whl/cu130 \
  --index-strategy unsafe-best-match

Nemotron OCR is currently supported only on Linux x86_64 with Python 3.12 and CUDA 13.x.

Development setup

To develop Docling features, bugfixes etc., install as follows from your local clone's root dir:

uv sync --all-extras --no-extra feat-ocr-nemotron

The feat-ocr-nemotron extra is intentionally excluded from the default development setup because it is only usable on Linux x86_64 with Python 3.12 and CUDA 13.x.