Book Pdf Extract

Example:

# Using environment variable
export LLAMA_CLOUD_API_KEY=your_api_key_here
python pdf_extract.py --pdf input.pdf --output output.md --method llamaparse

# Or using command-line argument
python pdf_extract.py --pdf input.pdf --output output.md --method llamaparse --llamaparse-key your_api_key_here

When to Use

• User has uploaded a PDF textbook that needs to be converted to markdown
• Starting a new textbook project from a PDF source
• User needs to extract content from scanned or digital PDF textbooks

When NOT to Use

• Source content is already in markdown format (use book-scrape instead)
• PDF contains images that need OCR (use book-ocr-extract instead)

Guardrails

⚠️ Must NOT:

• Skip validation of output markdown
• Overwrite existing markdown files without confirmation
• Process corrupted or malformed PDFs without user warning
• Use LlamaParse without API key if using cloud features

Extraction Methods

Method 1: LlamaParse (Best for Textbooks)

• Uses LlamaIndex's LlamaParse model
• Excellent at detecting document structure
• Handles math formulas well
• Preserves table formatting
• Setup: Set LLAMA_CLOUD_API_KEY environment variable or use --llamaparse-key argument
• Install: pip install llama-parse

Method 2: Docling (Primary - Recommended)

• Focuses on layout-aware extraction
• Excellent for textbooks with mixed text/images
• Automatically extracts <!-- image --> placeholders
• Generates proper markdown with ## headers
• Install: pip install docling
• CLI: docling <file.pdf> --output <file.md>

Method 3: Pdfplumber (Fallback)

• Works without API keys
• Simple text extraction with page-by-page processing
• Install: pip install pdfplumber Pillow

Quick Start

1. Verify PDF file exists and is valid
2. Choose extraction method based on availability
3. Run extraction with --verbose for progress
4. Check extracted markdown quality
5. Proceed to next pipeline stage

Workflow

Phase 1: Validate Input

• INPUT: PDF file path
• ACTION: Check file exists, verify it's a PDF, check size (<1GB for local methods)
• OUTPUT: Confirmed PDF file info (pages, size)

Phase 2: Extract Content ( Method selection)

• LlamaParse (preferred): Uses LlamaParse for structure-aware extraction
• Docling (alternative): Layout-aware extraction from docling
• Pdfplumber (fallback): Page-by-page text extraction

Phase 3: Format Output

• INPUT: Extracted text
• ACTION:
  • Convert headings to markdown headers
  • Preserve basic formatting
  • Handle page breaks
  • Extract tables as markdown tables (if available)
• OUTPUT: full.md in extracted directory

Phase 4: Validation

• INPUT: Extracted markdown
• ACTION:
  • Check file size is reasonable (>1KB)
  • Verify headings structure
  • Check for excessive page numbers/artifacts
• OUTPUT: Validation report + extracted file location

Scripts

Error Handling

Common Mistakes