Skip to content
PromptKit Logo PromptKit

PackC

Compiler for the PromptPack open standard

What is PackC?

Section titled “What is PackC?”

PackC compiles prompt source files into PromptPack-compliant packages—a vendor-neutral, framework-agnostic format that works with any AI runtime or provider.

Why PromptPack?

Section titled “Why PromptPack?”

Today’s AI prompt development is fragmented. Each framework has its own format for prompts, tools, conversations, and test scenarios. When teams switch providers or frameworks, they rebuild their entire prompt infrastructure from scratch.

PromptPack solves this with an open specification built on three principles:

  • Vendor Neutrality: A framework-agnostic JSON format that works across any runtime
  • Completeness: Prompts, tools, guardrails, and resources in a single file
  • Discipline: Treating prompts as version-controlled, testable engineering artifacts

PackC is the reference compiler for this standard.

What PackC Does

Section titled “What PackC Does”
  • Compiles YAML/JSON sources into .pack.json files conforming to PromptPack spec
  • Validates structure against the official schema
  • Optimizes for production with minification and preprocessing
  • Versions packages for distribution and deployment

Quick Start

Section titled “Quick Start”
Terminal window
# Install PromptKit (includes PackC)
brew install promptkit


# Or with Go
go install github.com/AltairaLabs/PromptKit/tools/packc@latest


# Create a prompt source file
cat > greeting.yaml <<EOF
apiVersion: promptkit.altairalabs.ai/v1alpha1
kind: PromptConfig
spec:
  task_type: greeting
  name: Greeting Assistant
  description: A friendly assistant that greets users
  system_prompt: |
    You are a friendly assistant. Greet the user warmly.
  parameters:
    temperature: 0.7
    max_tokens: 150
EOF


# Compile to PromptPack format
packc compile-prompt greeting.yaml --output greeting.pack.json


# Validate against the spec
packc validate greeting.pack.json

The resulting .pack.json can be used with any PromptPack-compatible runtime—not just PromptKit.

Next: Your First Pack Tutorial

Framework Agnostic

Section titled “Framework Agnostic”

A key benefit of PromptPack is portability. Packs compiled with PackC work with:

  • PromptKit SDK (Go)
  • Any PromptPack-compatible runtime in other languages
  • Custom integrations that read the standard JSON format

Build once, deploy everywhere. No vendor lock-in.

Diagram

Documentation by Type

Section titled “Documentation by Type”

📚 Tutorials (Learn by Doing)

Section titled “📚 Tutorials (Learn by Doing)”

Step-by-step guides for learning PackC:

  1. First Pack - Create your first PromptPack
  2. Multi-Prompt Packs - Bundle multiple prompts
  3. Validation Workflow - Ensure pack quality
  4. Pack Management - Organize and version packs
  5. CI/CD Pipeline - Automate pack builds

🔧 How-To Guides (Accomplish Specific Tasks)

Section titled “🔧 How-To Guides (Accomplish Specific Tasks)”

Focused guides for specific tasks:

💡 Explanation (Understand the Concepts)

Section titled “💡 Explanation (Understand the Concepts)”

Deep dives into PackC and PromptPack:

📖 Reference (Look Up Details)

Section titled “📖 Reference (Look Up Details)”

Complete command and format specifications:

Key Features

Section titled “Key Features”

Compilation to Open Standard

Section titled “Compilation to Open Standard”

Transform YAML/JSON prompts into PromptPack-compliant packages:

Terminal window
packc compile prompts/my-app.yaml \
  --output dist/my-app.pack.json \
  --optimize \
  --version 1.0.0

Schema Validation

Section titled “Schema Validation”

Ensure packs conform to the PromptPack specification:

Terminal window
packc validate my-app.pack.json --strict

Checks:

  • ✅ Schema compliance with PromptPack spec
  • ✅ Required fields present
  • ✅ Template syntax valid
  • ✅ Tool definitions complete

Production Optimization

Section titled “Production Optimization”

Production-ready output:

Terminal window
packc compile prompts/*.yaml --optimize
  • Minify JSON output
  • Remove comments and whitespace
  • Validate templates
  • Check for common errors

PromptPack Format

Section titled “PromptPack Format”

A compiled pack follows the PromptPack specification:

{
  "apiVersion": "promptkit.altairalabs.ai/v1alpha1",
  "kind": "PromptPack",
  "metadata": {
    "name": "my-app",
    "version": "1.0.0"
  },
  "prompts": [
    {
      "id": "greeting",
      "system": "You are helpful.",
      "template": "Greet the user."
    }
  ]
}

This format is:

  • Self-contained: Everything needed to run the prompt
  • Portable: Works with any compatible runtime
  • Versionable: Track changes with semantic versioning
  • Testable: Include test metadata for validation

Learn more at promptpack.org.

CI/CD Integration

Section titled “CI/CD Integration”

Automate pack builds in your pipeline:

GitHub Actions

Section titled “GitHub Actions”
- name: Compile PromptPacks
  run: |
    packc compile prompts/*.yaml --output-dir dist/
    packc validate dist/*.pack.json --strict

Makefile

Section titled “Makefile”
.PHONY: build-packs
build-packs:
    packc compile prompts/*.yaml --output-dir dist/packs/


.PHONY: validate-packs
validate-packs:
    packc validate dist/packs/*.pack.json --strict

Best Practices

Section titled “Best Practices”

Use the Open Standard

Section titled “Use the Open Standard”
  • Follow the PromptPack specification for maximum portability
  • Don’t add custom fields outside the spec
  • Test packs with multiple runtimes if possible

Version Management

Section titled “Version Management”
  • Use semantic versioning (MAJOR.MINOR.PATCH)
  • Update version on breaking changes
  • Keep changelog of prompt changes

Quality Assurance

Section titled “Quality Assurance”
  • Always validate after compilation
  • Use --strict mode in CI/CD
  • Test packs with Arena before distribution

Resources

Section titled “Resources”
Section titled “Related Tools”