PDF To Data

PDF to Data Extractor

Extract tables, text, and form fields from PDFs into structured formats.

Usage

/pdf-to-data invoice.pdf
/pdf-to-data report.pdf --format xlsx
/pdf-to-data forms.pdf --format csv --output data.csv

Arguments

• $0 - Path to PDF file (required)
• --format - Output format: json (default), csv, xlsx
• --output - Output file path (optional, defaults to stdout for json/csv)
• --tables-only - Extract only tables, skip general text
• --page - Extract specific page number (1-indexed)

How It Works

1. Text Extraction: Extracts all text content with layout preservation
2. Table Detection: Uses PyMuPDF's table detection to find and extract tabular data
3. Form Fields: Extracts form field names and values from fillable PDFs
4. Metadata: Includes document metadata (title, author, creation date)

Requirements

Ensure PyMuPDF is installed:

pip install pymupdf openpyxl

Running the Extraction

python ~/.claude/skills/pdf-to-data/scripts/extract.py "$0" $ARGUMENTS

Or if running from the project directory:

python ./scripts/extract.py "$0" $ARGUMENTS

Output Formats

JSON (default)

{
  "metadata": { "title": "...", "pages": 5 },
  "pages": [
    {
      "number": 1,
      "text": "...",
      "tables": [[["Header1", "Header2"], ["row1col1", "row1col2"]]],
      "form_fields": [{"name": "field1", "value": "..."}]
    }
  ]
}

CSV

Exports all tables as CSV. Multiple tables are separated by blank lines.

XLSX

Creates an Excel workbook with:

• Sheet "Text" - Full text content per page
• Sheet "Table_1", "Table_2", etc. - Each detected table
• Sheet "Form_Fields" - All form field data

Examples

Basic extraction:

/pdf-to-data quarterly-report.pdf

Export tables to Excel:

/pdf-to-data financial-data.pdf --format xlsx --output financials.xlsx

Extract specific page as CSV:

/pdf-to-data large-doc.pdf --page 3 --tables-only --format csv

Error Handling

• If PDF is encrypted, the script will report the error
• If no tables found, returns empty tables array
• Invalid page numbers are handled gracefully