MarkItDown on Google Colab

This repository demonstrates how to use Microsoft's MarkItDown library to convert PDF files into structured Markdown format, directly in Google Colab. This conversion facilitates document analysis and integration with natural language models.

Features

• PDF to Markdown conversion
• Plugin support, including OCR for scanned PDFs
• File upload and download in the Google Colab environment
• Display of converted content directly in the notebook

How to Use

1. Install the MarkItDown library with all plugins:

   !pip install 'markitdown[all]' --quiet

2. Import the necessary libraries:

   from markitdown import MarkItDown
   from google.colab import files
   import os

3. Upload the PDF file you want to convert:

   print("Please select the PDF file to convert:")
   uploaded = files.upload()

4. Initialize the converter with plugins enabled:

   md = MarkItDown(enable_plugins=True)

5. Convert the PDF file to Markdown:

   for filename in uploaded.keys():
       print(f"\nConverting file: {filename}")
       result = md.convert(filename)
   
       # Display the converted content
       print("\nConverted Markdown content:")
       print(result.text_content)
   
       # Save the content to a .md file
       output_filename = os.path.splitext(filename)[0] + ".md"
       with open(output_filename, "w", encoding="utf-8") as f:
           f.write(result.text_content)
   
       # Make the file available for download
       print(f"\nDownloading the converted file: {output_filename}")
       files.download(output_filename)

Notes

• Integrated OCR: If the PDF contains images with text (e.g., scanned documents), MarkItDown can extract this text automatically, as long as plugins are enabled.

• Support for Multiple Formats: In addition to PDFs, MarkItDown supports Word, PowerPoint, Excel, images, audio, and more.

• Stream Conversion: If working with in-memory files (e.g., using io.BytesIO), MarkItDown also supports stream-based conversion.

Example Notebook

You can access an example notebook in Google Colab with all the steps above:

Example Notebook on Google Colab

License

This project is licensed under the MIT License.