docling-converter（文档转换器）

📄 Docling Document Converter

Version: 1.0.1 Status: ✨ Production Ready | 🌍 Universal

Convert documents (PDF, DOCX, PPTX, Images, HTML) into structured Markdown and JSON using Docling's intelligent parsing engine.

📋 Overview

Docling Converter is a powerful document processing skill that transforms unstructured files into clean, structured Markdown and JSON. Unlike basic text extractors, it preserves document layout, tables, and hierarchy, making it ideal for RAG (Retrieval-Augmented Generation) pipelines and knowledge base ingestion.

✨ Key Features

• 📄 Multi-Format Support: PDF, DOCX, PPTX, XLSX, HTML, Images, AsciiDoc.
• 🧠 Intelligent Parsing: Preserves tables, headers, and document structure.
• 👁️ OCR Integration: Handles scanned PDFs and images (requires docling[ocr]).
• 📝 Clean Markdown: Generates LLM-ready Markdown output.
• ⚡ Batch Processing: Handles single files or entire directories.

🚀 Quick Start

Invoke the Skill

Use any of these trigger phrases:

copilot> convert this pdf to markdown: report.pdf
copilot> extract tables from: data.xlsx
copilot> docling convert: presentation.pptx
copilot> process document: scanned-contract.pdf --ocr
copilot> convert this pptx to markdown: deck.pptx
copilot> pptx to markdown: strategy-2026.pptx
claude> convert this presentation to markdown: roadmap.pptx
claude> pptx to markdown: proposal.pptx

🛠️ Workflow

Step 0: Discovery & Setup

Objective: Verify Docling installation and dependencies.

Actions:

# Check if docling is installed
if python3 -c "import docling" 2>/dev/null; then
    echo "✅ Docling detected"
else
    echo "⚠️  Docling not found"
    echo "🔧 Installing docling..."
    pip install docling --break-system-packages
fi

# Check for OCR support if requested
if [[ "$OCR_REQUESTED" == "true" ]]; then
    if python3 -c "import easyocr" 2>/dev/null; then
        echo "✅ OCR dependencies detected"
    else
        echo "⚠️  OCR dependencies missing"
        echo "🔧 Installing docling[ocr]..."
        pip install "docling[ocr]" --break-system-packages
    fi
fi

Step 1: Create Conversion Script

Objective: Generate a robust Python script to handle the conversion.

Actions:

Create a temporary script .gemini/tmp/docling_convert.py:

import sys
import json
import os
from pathlib import Path
from docling.document_converter import DocumentConverter, PdfFormatOption, WordFormatOption
from docling.datamodel.pipeline_options import PdfPipelineOptions, TableFormerMode

def convert_document(input_path, output_dir, use_ocr=False):
    input_path = Path(input_path)
    output_dir = Path(output_dir)
    output_dir.mkdir(parents=True, exist_ok=True)

    # Configure Pipeline
    pipeline_options = PdfPipelineOptions()
    pipeline_options.do_ocr = use_ocr
    pipeline_options.do_table_structure = True
    pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE

    converter = DocumentConverter(
        format_options={
            "pdf": PdfFormatOption(pipeline_options=pipeline_options),
            "docx": WordFormatOption(),
            "pptx": WordFormatOption()
        }
    )

    print(f"🔄 Converting: {input_path.name}...")

    try:
        result = converter.convert(input_path)

        # Export Markdown
        md_output = result.document.export_to_markdown()
        md_path = output_dir / f"{input_path.stem}.md"
        with open(md_path, "w", encoding="utf-8") as f:
            f.write(md_output)

        # Export JSON (structure)
        json_output = result.document.export_to_dict()
        json_path = output_dir / f"{input_path.stem}.json"
        with open(json_path, "w", encoding="utf-8") as f:
            json.dump(json_output, f, ensure_ascii=False, indent=2)

        print(f"✅ Success! Saved to: {md_path}")
        return True

    except Exception as e:
        print(f"❌ Error converting {input_path.name}: {str(e)}")
        return False

if __name__ == "__main__":
    if len(sys.argv) < 3:
        print("Usage: python docling_convert.py <input_file> <output_dir> [ocr]")
        sys.exit(1)

    input_file = sys.argv[1]
    output_directory = sys.argv[2]
    ocr_enabled = len(sys.argv) > 3 and sys.argv[3] == "--ocr"

    success = convert_document(input_file, output_directory, ocr_enabled)
    sys.exit(0 if success else 1)

Batch Mode — Parallel Document Processing

When the user provides a directory or multiple files, launch one DoclingConverter agent per file simultaneously in a single block.

Each DoclingConverter agent prompt begins with:

# DoclingConverter — Document Processing Agent
Role: Convert a single document to Markdown and JSON using Docling. Run the conversion script, validate that output files were created, return status and output paths.
Input: File path: {PATH} | Output format: {FORMAT}

Wait for all DoclingConverter agents to complete. Merge all Markdown outputs. Report summary: files processed, conversion time, any failures.

Step 2: Execute Conversion

Objective: Run the conversion script on the user's file.

Actions:

# Define paths
INPUT_FILE="$USER_INPUT_FILE"
OUTPUT_DIR="./converted_docs"

# Execute script
python3 .gemini/tmp/docling_convert.py "$INPUT_FILE" "$OUTPUT_DIR" $OCR_FLAG

Step 3: Result & Validation

Objective: meaningful output to the user.

Actions:

if [ $? -eq 0 ]; then
    echo ""
    echo "🎉 Conversion Complete!"
    echo "📂 Output Directory: $OUTPUT_DIR"
    ls -lh "$OUTPUT_DIR"
else
    echo "❌ Conversion Failed. Please check the error logs above."
fi

Error Handling

📄 Version

v1.0.1 | Agentic Workflow | Auto-Install