An autonomous AI agent that can improve your codebase. Gets work to do via GitHub issues. This project provides a flexible foundation for creating AI agents that can understand issues, generate solutions, and manage pull requests.

• 🤖 Autonomous Operation: Monitors GitHub issues and automatically handles tasks based on labels
• 🧠 Local LLM Support: Primary support for Ollama models, with optional OpenAI integration
• 📚 RAG Pipeline: Efficient context retrieval using ChromaDB for better task understanding
• 🔄 Feedback Loop: Built-in feedback collection and learning mechanisms
• 🛠️ Modular Architecture: Easy to extend and customize for specific use cases
• 🔒 Type Safety: Full TypeScript support with Zod schema validation

• Bun (v1.0.0 or higher)
• Ollama for local LLM support
• Redis (optional, for persistence)

1. Clone the repository:

   git clone https://github.com/yourusername/ai-agent-scaffold.git
   cd ai-agent-scaffold

2. Install dependencies:

   bun install

3. Copy the environment variables template:

   cp .env.example .env

4. Configure your environment variables in .env

The agent can be configured through environment variables:

• GitHub Configuration

  • GITHUB_TOKEN: Personal Access Token with repo scope
  • GITHUB_OWNER: Repository owner (username or organization)
  • GITHUB_REPO: Repository name

• LLM Configuration

  • LLM_PROVIDER: LLM provider ('ollama' or 'openai')
  • LLM_MODEL: Model name (e.g., 'mistral' for Ollama)
  • OLLAMA_BASE_URL: Ollama API URL
  • OPENAI_API_KEY: OpenAI API Key (optional)

• Storage Configuration

  • REDIS_URL: Redis connection URL (optional)
  • CHROMA_DB_PATH: ChromaDB storage path (optional)

• Agent Configuration

  • MAX_ITERATIONS: Maximum iterations for the planning loop
  • POLL_INTERVAL: GitHub polling interval in milliseconds
  • APPROVED_LABELS: Labels that trigger agent action
  • PORT: HTTP server port
  • AUTO_START: Whether to start automatically

1. Start the agent:

   bun run dev

2. The agent will start monitoring GitHub issues with the configured labels.

3. Access the API endpoints:

   • Health check: GET /health
   • Agent control:
     • Start: POST /agent/start
     • Stop: POST /agent/stop
     • Status: GET /agent/status
   • Task management:
     • Get task: GET /tasks?id=<task_id>
     • Add feedback: POST /tasks/:id/feedback

The project follows a modular architecture with the following components:

• Agent Core: Central orchestrator that manages the autonomous workflow
• GitHub Watcher: Monitors repository issues and creates tasks
• Task Manager: Handles task lifecycle and persistence
• LLM Client: Manages interactions with language models
• RAG Pipeline: Provides relevant context for task understanding
• Feedback System: Collects and processes feedback for improvement
• Learning Module: Adapts agent behavior based on feedback

1. Run in development mode:

   bun run dev

2. Run tests:

   bun test

3. Build for production:

   bun run build

1. Fork the repository
2. Create your feature branch: git checkout -b feature/amazing-feature
3. Commit your changes: git commit -m 'Add amazing feature'
4. Push to the branch: git push origin feature/amazing-feature
5. Open a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

• Built with Bun
• Uses Hono for the HTTP server
• Powered by Ollama for local LLM support