shuyu-labs/office-to-md

Office Document to Markdown Converter

Convert various Office document formats to structured Markdown with text, table, and image extraction.

File Description

• enhanced_parser.py - Core document parser
• doc_converter.py - DOC to DOCX converter (requires LibreOffice)
• requirements.txt - Python dependencies

Install Dependencies

pip install -r requirements.txt

Additional Dependencies for DOC Format

.doc format requires LibreOffice:

# Windows: Install LibreOffice from official website
# https://www.libreoffice.org/download/

# Linux
sudo apt install libreoffice

# Mac
brew install --cask libreoffice

Quick Start

Python Code

from enhanced_parser import EnhancedDocumentParser

# Initialize parser
parser = EnhancedDocumentParser(
    image_base_url="http://localhost:5000",
    image_save_dir="./static/images",
    filter_headers_footers=True  # Filter headers and footers
)

# Parse document
result = parser.parse_document("document.docx")

if result["success"]:
    print(result["markdown"])
    print(f"Extracted {result['images_count']} images")

Start API Service

# Start service using app.py from project root
python app.py

# Visit http://localhost:5000/analyzer to upload files

Supported Formats

| Format | Extensions | Notes | |--------|-----------|-------| | Word | .docx, .doc | .doc requires LibreOffice | | Excel | .xlsx, .xls | Supports multiple worksheets and date formats | | PowerPoint | .pptx | Extracts slide text and images | | PDF | .pdf | Auto-detects tables and images |

Features

Word Documents

• Automatic heading level detection
• Convert tables to Markdown tables
• Extract inline images
• Filter headers and footers
• Preserve list formatting

Excel Workbooks

• Support for multiple worksheets
• Automatic date format detection (prevents display as numbers)
• Convert to Markdown tables
• Extract embedded images

PowerPoint Presentations

• Extract content by slide
• Extract images and text boxes
• Preserve slide order

PDF Documents

• Auto-detect tables (line detection + text position detection)
• Extract page images
• Intelligently identify headings and lists
• Output content in original order

Advanced Options

DOC Conversion

# Test LibreOffice configuration
python doc_converter.py

PDF Table Strategy

parser = EnhancedDocumentParser(
    pdf_table_strategy="lines_strict"  # Default: strict line detection, fastest
    # "lines": Normal line detection
    # "text": Based on text position, more accurate but slower
)

Image Processing

parser = EnhancedDocumentParser(
    image_base_url="https://your-domain.com",  # Image access URL
    image_save_dir="./static/images"           # Image save directory
)

Return Format

{
  "success": true,
  "markdown": "# Document Title\n\nContent...",
  "images_count": 2,
  "images": [
    {
      "filename": "uuid.png",
      "url": "http://localhost:5000/static/images/uuid.png",
      "size": 12345
    }
  ],
  "file_type": "docx",
  "file_info": {
    "name": "document.docx",
    "size": 45678,
    "paragraphs": 50,
    "tables": 3
  }
}

Common Issues

DOC Conversion Failed

• Ensure LibreOffice is installed
• Run python doc_converter.py to test configuration

Dates Display as Numbers

• Excel parsing automatically handles date formats
• Ensure you're using the latest version of enhanced_parser.py

PDF Table Recognition Inaccurate

• Try different pdf_table_strategy parameters
• Use "lines_strict" for standard tables
• Use "text" for complex tables

File Limitations

• Maximum file size: 160MB
• Supported extensions: docx, doc, pdf, xlsx, xls, pptx
• Automatic cleanup of temporary files