Skip to main content

Getting Started with PromptPack

PromptPack is an open specification for packaging conversational AI systems. This guide helps you get started based on your goals.

What You Can Do Today​

1. Learn the Specification​

Best for: Understanding what PromptPack is and how it works

Time investment: 30-60 minutes

2. Create a PromptPack​

Best for: Packaging your prompts in PromptPack format

Example - Minimal Valid PromptPack:

{
  "$schema": "https://promptpack.org/schema/latest/promptpack.schema.json",
  "id": "my-assistant",
  "name": "My Assistant",
  "version": "1.0.0",
  "template_engine": {
    "version": "v1",
    "syntax": "{{variable}}"
  },
  "prompts": {
    "default": {
      "id": "default",
      "name": "Default Assistant",
      "version": "1.0.0",
      "system_template": "You are a helpful assistant."
    }
  }
}

Validate against the schema using any JSON Schema validator.

Time investment: 1-2 hours for first pack

3. Test with PromptArena​

Best for: Multi-provider testing and evaluation

PromptArena is a CLI tool for testing conversational AI systems:

# Install (see PromptKit repository for latest instructions)
# Create test scenarios
# Run tests across multiple providers
promptarena run --provider openai,anthropic --format html

PromptArena Guide →

Time investment: 2-4 hours including setup

4. Contribute to the Spec​

Best for: Shaping the future of PromptPack

Time investment: Ongoing participation

What's Coming​

PromptKit Runtime: Reference implementation for loading and executing PromptPacks. When available, will enable:

  • CLI validation and linting
  • SDK libraries for Python and JavaScript
  • Direct execution of PromptPacks

Track PromptKit Development →

Need Help?​

Next Steps​

Choose your path: