Docling Docker Container

This repository provides a Docker container for converting PDF files to Markdown using Docling. The container supports both single file conversion and recursive directory conversion with an option to remove embedded images from the output Markdown files.

Features

• Single File Conversion: Convert a single PDF file located in the mounted documents folder.
• Recursive Directory Conversion: Set the DIRECTORY environment variable to process PDFs recursively from a given directory, preserving folder structure.
• Image Removal: Optionally remove embedded images (e.g., ![Image](...)) from the generated Markdown files by setting IMAGELESS to True.
• Host-side Script: A docker_run.sh script is provided to simplify passing arguments and environment variables.

Prerequisites

• Docker must be installed.

Building the Docker Image

Run the following command in the project root (where the Dockerfile is located):

bash docker build -t aimilefth/docling:latest --push .

Using the Container

You can run the container in two ways: directly using docker run or by using the provided host-side script docker_run.sh.

Using Docker Run Directly

Single File Conversion

docker run --rm \ -v $(pwd)/documents:/app/documents \ -v $(pwd)/outputs:/app/outputs \ aimilefth/docling myfile.pdf

Single File Conversion with Image Removal

docker run --rm \ -e IMAGELESS=True \ -v $(pwd)/documents:/app/documents \ -v $(pwd)/outputs:/app/outputs \ aimilefth/docling myfile.pdf

Recursive Directory Conversion

docker run --rm \ -e DIRECTORY=True \ -v $(pwd)/documents:/app/documents \ -v $(pwd)/outputs:/app/outputs \ aimilefth/docling my_subdir

Recursive Conversion with Image Removal

docker run --rm \ -e DIRECTORY=True \ -e IMAGELESS=True \ -v $(pwd)/documents:/app/documents \ -v $(pwd)/outputs:/app/outputs \ aimilefth/docling my_subdir

Using the Host Script: docker_run.sh

The docker_run.sh script simplifies usage by accepting flags:

• -p / --path: Relative path (file or directory) inside the documents folder (required).
• -d / --directory: Process recursively if present.
• -i / --imageless: Remove embedded images after conversion if present.
• --docs: Optional path to the documents folder (default: $(pwd)/documents).
• --outputs: Optional path to the outputs folder (default: $(pwd)/outputs).

For example, to convert a single file (with default options): bash docker_run.sh --path myfile.pdf

To process a directory recursively: bash docker_run.sh --path my_subdir --directory

To process a directory recursively with image removal: bash docker_run.sh --path my_subdir --directory --imageless

Or, using custom documents and outputs paths: bash docker_run.sh --path my_subdir --directory --imageless --docs /path/to/my/docs --outputs /path/to/my/outputs

Project Structure

• Dockerfile: Defines the container image.
• entrypoint.sh: The container entrypoint that handles both single file and recursive directory conversion based on environment variables.
• removeimagesmd.sh: Script used to remove embedded images from generated Markdown files.
• docker_run.sh: Host-side script that simplifies running the container with the proper arguments and volume mounts.
• documents/: Directory containing source PDF files.
• outputs/: Directory where converted Markdown files are saved.

Summary

This project enables flexible PDF to Markdown conversion using Docker and Docling. Adjust the environment variables and script arguments as needed for single file or recursive directory processing and optional image removal.