🍓 Berrry Committer

AI-powered GitHub automation tool that processes issues and generates code automatically using Claude.

✨ Features

• Automatic Code Generation: Responds to GitHub issues with complete code solutions
• Smart File Management: Creates, updates, and organizes files based on issue context
• Iterative Development: Uses evaluation scripts to refine solutions across multiple steps
• Safe Local Testing: Interactive mode for local development with manual confirmations
• GitHub Actions Integration: Fully automated workflow for production use

🚀 Quick Start

GitHub Actions Setup

1. Add Repository Secrets:

   • ANTHROPIC_API_KEY or OPENROUTER_API_KEY

2. Create Workflow (.github/workflows/berrry-committer.yml):

name: Berrry Committer
on:
  issues:
    types: [opened, edited]
  issue_comment:
    types: [created]

jobs:
  berrry-committer:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '18'
      - name: Run Berrry Committer
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        run: node src/main.js

3. Create an Issue: Describe your coding request to trigger Berrry Committer

Local Development

# 1. Set up environment
export ANTHROPIC_API_KEY="your-api-key"

# 2. Direct prompting mode (new!)
berrry --prompt "Create a login component with email/password validation"

# 3. GitHub event mode
export GITHUB_EVENT_PATH="/tmp/test_event.json"
cat > /tmp/test_event.json << 'EOF'
{
  "issue": {
    "number": 1,
    "title": "Add user authentication",
    "body": "Please create a login component with email/password validation"
  }
}
EOF
berrry

# 4. Or run in YOLO mode (auto-execute)
berrry --yolo

🛠️ Configuration

Environment Variables

Variable	Required	Default	Description
ANTHROPIC_API_KEY	Yes*	-	Anthropic API key
OPENROUTER_API_KEY	Yes*	-	OpenRouter API key (alternative)
MODEL	No	anthropic/claude-sonnet-4	Main LLM model
COMMIT_MODEL	No	anthropic/claude-3.5-haiku	Model for commit messages
TEST_COMMAND	No	-	Command to run for validation
YOLO	No	false	Skip confirmations (auto true in Actions)

*One of the API keys is required

Advanced Configuration

# Custom model configuration
export MODEL="anthropic/claude-3-opus"
export COMMIT_MODEL="anthropic/claude-3-haiku"

# Test integration
export TEST_COMMAND="npm test && npm run lint"

# API endpoint override (for proxies)
export API_URL="https://your-proxy.com/v1/chat/completions"

📋 How It Works

1. Issue Processing: Parses GitHub issue/comment for coding requests
2. Context Gathering: Automatically includes relevant files and repository structure
3. Code Generation: Uses AI models to generate complete solutions with analysis
4. File Creation: Writes files using the === FILENAME: path === format
5. Evaluation: Runs bash scripts to validate and determine if more work is needed
6. Iteration: Continues refining until solution is complete (max 5 steps)
7. PR Creation: Automatically creates pull request with all changes

Code Generation Format

The AI outputs files in this format:

## CODE
=== FILENAME: src/components/Login.tsx ===
import React from 'react';

export const Login = () => {
  // Component implementation
};
=== END: src/components/Login.tsx ===

## EVAL
```bash
#!/bin/bash
set -euo pipefail
# Validation script
npx tsc --noEmit || exit 1
exit 0

🎯 Usage Examples

Simple Feature Request

Title: Add dark mode toggle
Body: Please add a dark mode toggle to the settings page

Complex Implementation

Title: Implement user authentication system
Body: Need a complete auth system with:
- Login/signup forms
- JWT token handling  
- Protected routes
- User profile management

See existing code in `src/auth/` for context.

Bug Fix

Title: Fix memory leak in data processing
Body: The data processor in `src/utils/processor.js` is causing memory leaks. 
Please refactor to use streaming and proper cleanup.

🔒 Security & Safety

• Local Mode: Manual confirmation for all script execution
• No Automatic Pushes: Creates commits locally, manual push in local mode
• Script Preview: Shows generated evaluation scripts before execution
• Git Safety: Easy to reset changes with standard git commands
• API Key Protection: Uses environment variables, never logged

🛟 Troubleshooting

Common Issues

"GITHUB_EVENT_PATH not found"

# Create mock event for local testing
export GITHUB_EVENT_PATH="/tmp/test_event.json"

"API key required"

export ANTHROPIC_API_KEY="your-key-here"

Permission denied on eval.sh

# Ensure you're in a writable directory
chmod +x eval.sh

Debug Mode

Add verbose logging:

berrry --verbose

📖 Documentation

• CLAUDE.md - Technical architecture guide for Berrry Committer
• dev.md - Detailed local development guide

🤖 AI-Assisted Development

This project uses itself for development! You can request AI-generated code:

Quick Start

1. Create an issue describing your coding request
2. Berrry Committer will analyze, code, test, and create a PR automatically

Examples

Add --version flag support
Please improve error handling in the parser
Add Docker configuration for easy deployment

📖 Full AI Development Guide

🤝 Contributing

1. Fork the repository
2. Create your feature branch (or use Berrry Committer)
3. Test locally with npm test
4. Submit a pull request

📄 License

MIT License - see LICENSE file for details.

---

Made with 🍓 by Strawberry Computer