name: create-table description: > Create PDF tables using ReportLab Table class for proper structure detection. Creates tables that Marker and other extractors can parse. Use when you need detectable tables in PDFs. allowed-tools: Bash, Read, Write triggers:

• create table
• create pdf table
• generate table
• make table
• reportlab table
• detectable table
• table for extraction metadata: short-description: "Create detectable PDF tables (ReportLab)"

create-table

Create PDF tables using ReportLab's Table class that extractors (Marker, Camelot) can detect.

Why This Exists

PyMuPDF raw drawing commands (draw_line, insert_text) create visual tables but not proper PDF table structures. Marker and other extractors cannot detect these as tables.

ReportLab's Table class creates proper PDF table structures with cell boundaries that extraction tools recognize.

Quick Start

cd .pi/skills/create-table

# Generate from preset (simple, medium, complex)
uv run generate.py preset simple
uv run generate.py preset complex --output my_table.pdf

# Generate with custom data
uv run generate.py generate --output custom.pdf \
  --columns "ID,Name,Status" \
  --rows '[["1","Alice","Active"],["2","Bob","Pending"]]'

# Borderless table
uv run generate.py generate --no-border --output borderless.pdf

# Thick borders (3pt)
uv run generate.py generate --line-width 3 --output thick.pdf

# Render as image (for scanned document testing)
uv run generate.py preset medium --as-image --dpi 200

# Generate all examples
uv run generate.py example

Commands

preset - Generate from preset complexity

uv run generate.py preset <name> [options]

generate - Create custom table

uv run generate.py generate [options]

Options:

example - Generate all examples

uv run generate.py example

Creates examples/ directory with all presets, styles, and border variations.

Multi-Index Tables

For hierarchical headers that span columns:

uv run generate.py generate \
  --columns "ID,Q1,Q2,Q3,Q4,Total" \
  --rows '[["A","10","20","30","40","100"]]' \
  --multi-index '[[{"text":"","span":1},{"text":"2024 Quarters","span":4},{"text":"","span":1}]]'

Scanned Document Simulation

Generate table as a rasterized PNG image (for testing OCR/VLM extraction):

# 150 DPI (default)
uv run generate.py preset medium --as-image

# Higher quality (300 DPI)
uv run generate.py preset complex --as-image --dpi 300 --output scanned_table.png

Verification

Test that Camelot detects the table:

uv run generate.py preset simple --output test.pdf

python -c "
import camelot
tables = camelot.read_pdf('test.pdf', pages='1', flavor='lattice')
print(f'Tables found: {len(tables)}')
if tables:
    print(tables[0].df)
"

Dependencies

Self-contained via uv run - no pre-installation needed.

[project]
dependencies = [
    "reportlab>=4.0.0",
    "typer>=0.9.0",
    "pymupdf>=1.23.0",
]