🦙 🤖 Llamautoma

A powerful, TypeScript-based AI agent framework built on top of LangChain for automated code generation, editing, and management. Designed to work seamlessly with modern language models and development workflows.

✨ Features

• 🚀 High-performance TypeScript execution engine
• 🔄 Real-time code generation and editing via chat
• 🧠 Advanced AI agent system with ReAct framework
• 🔒 Built-in safety controls and validation
• 📝 Streaming responses for real-time feedback
• 🛠️ Extensible tool system

🚀 Quick Start

Prerequisites

• Bun runtime installed
• A compatible language model (e.g., Ollama with qwen2.5-coder)

Installation

# Clone the repository
git clone https://github.com/yourusername/llamautoma.git
cd llamautoma

# Install dependencies
bun install

# Start the server
bun run start

🔌 API Reference

The server exposes HTTP endpoints for chat and workspace synchronization. All endpoints accept POST requests and expect JSON payloads.

Base URL

http://localhost:3000/v1

Common Request Properties

All requests support these common properties:

• threadId (optional): Unique identifier for the conversation thread
• safetyConfig (optional): Configuration for safety controls
  • maxInputLength: Maximum allowed input length (default: 8192)
  • requireToolConfirmation: Whether to require confirmation for tool usage
  • requireToolFeedback: Whether to require feedback for tool usage
  • dangerousToolPatterns: Array of patterns to flag as potentially dangerous

Common Response Format

All endpoints return Server-Sent Events (SSE) streams with the following event types:

• start: Initial response start with metadata
• content: Main response content (may be sent multiple times)
• end: Response completion with final results

The specific content of each event varies by endpoint but follows this general structure:

{
  "event": "start" | "content" | "end",
  "threadId": string,
  "data": {
    // Endpoint-specific data structure
  }
}

1. Chat Endpoint

POST /v1/chat

Interactive chat endpoint with streaming responses. Handles all code operations through natural language.

{
  "messages": [
    {
      "role": "user",
      "content": "Create a React component for a user profile"
    }
  ],
  "threadId": "optional-thread-id",
  "safetyConfig": {
    "maxInputLength": 8192
  }
}

Response: Server-Sent Events (SSE) stream with the following events:

• start: Initial response start
• content: Main response content
• end: Response completion

2. Sync Endpoint

POST /v1/sync

Synchronize and analyze codebase structure with progress updates.

{
  "root": "/path/to/project",
  "excludePatterns": ["node_modules", "dist", ".git"]
}

Response: Server-Sent Events (SSE) stream with file content and status updates.

🔒 Safety Features

Llamautoma includes several safety features to prevent potentially harmful operations:

• Input length validation
• Dangerous pattern detection
• Tool execution confirmation
• Execution feedback collection
• Rate limiting

🛠️ Configuration

The server can be configured through environment variables or the config file:

{
  "modelName": "qwen2.5-coder:7b",
  "host": "http://localhost:11434",
  "maxIterations": 10,
  "userInputTimeout": 30000,
  "safetyConfig": {
    "requireToolConfirmation": true,
    "requireToolFeedback": true,
    "maxInputLength": 8192,
    "dangerousToolPatterns": [
      "rm -rf /",
      "DROP TABLE",
      "sudo rm",
      // ... more patterns
    ]
  }
}

📝 License

MIT License - see the LICENSE file for details.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.