Skip to content

LLM Integration

gpio includes a skill file that teaches Large Language Models (ChatGPT, Claude, Gemini, etc.) how to work with spatial data using gpio.

Quick Start

# List available skills
gpio skills

# Print skill content to stdout
gpio skills --show

# Copy skill to current directory
gpio skills --copy .

Using with LLMs

Option 1: Paste skill content

Copy the skill content to your clipboard and paste it at the start of a conversation:

gpio skills --show | pbcopy

gpio skills --show | xclip -selection clipboard

gpio skills --show | Set-Clipboard

Then paste into your LLM conversation to give it context about gpio.

Option 2: Reference the file

Tell your LLM to read the skill file:

Read geoparquet_io/skills/geoparquet.md and help me convert my shapefile to GeoParquet.

Option 3: Claude Code

For Claude Code users, the skill is automatically available:

  • Invoke via /geoparquet command
  • Or ask Claude to help with GeoParquet conversions

What the Skill Teaches

The skill teaches LLMs how to:

  1. Ingest spatial data from URLs and local files
  2. Explore data structure, CRS, and schema
  3. Convert to optimized GeoParquet format
  4. Validate against best practices
  5. Recommend partitioning strategies based on size
  6. Publish to cloud storage

Customizing Skills

Copy the skill to customize it for your workflow:

gpio skills --copy .
# Edit geoparquet.md with your preferences

Custom skills can include: - Your preferred compression settings - Default cloud storage paths - Organization-specific workflows - Additional validation requirements

Skill Content

The skill includes:

  • Command reference table (auto-generated from CLI)
  • Compression options and defaults
  • Step-by-step workflow guides
  • Remote file access patterns
  • Example sessions

All generated sections stay in sync with the actual CLI via menard documentation tracking.