Automated VS Code Coding Setup with AI Agents

This project provides a fully automated setup for coding assistance in Visual Studio Code using local AI models, integrated with GitHub for version control and collaboration.

Features

• Local AI Integration: Uses Ollama with models like CodeLlama for code generation
• VS Code Extensions: Continue.dev and Tabnine for AI-powered coding assistance
• Aider Integration: Configured to use Ollama models for automated code editing
• MCP Servers: Model Context Protocol servers for AI agent communication
• GitHub Automation: Automated commit and push workflows
• Agent Communication: AI agents can communicate via shared message files

Installation

1. Prerequisites:

   • Python 3.x with transformers, torch, accelerate
   • Node.js
   • Ollama installed
   • GitHub CLI (gh) authenticated

2. Clone or Download:

   git clone https://github.com/Obayne/CodeSet.git
   cd CodeSet

3. Install Dependencies:

   # For MCP servers
   cd mcp-servers/backend-developer
   npm install
   npm run build

4. Configure VS Code:

   • Install extensions: Continue, Tabnine
   • Copy .continue/config.json to your VS Code Continue config

5. Set up MCP in Cline:

   • The MCP server is configured in your Cline settings

Usage

Automated Workflow

Run the automated coding workflow:

python automate_workflow.py "Add a new API endpoint for user authentication"

This will:

• Use Aider to generate/modify code based on the prompt
• Commit changes
• Push to GitHub

AI Agent Communication

Use the MCP tools in Cline:

• generate_backend_code: Generate code using local AI model
• send_message: Send messages between AI agents

Messages are stored in messages.json for inter-agent communication.

Project Structure

CodeSet/
├── .aider.conf.yml          # Aider configuration
├── .continue/config.json    # Continue extension config
├── .gitignore
├── README.md
├── automate_workflow.py     # Automated coding script
├── messages.json            # Agent communication log
├── mcp-servers/             # MCP servers for agents
│   └── backend-developer/
│       ├── inference.py     # Python script for AI inference
│       ├── package.json
│       ├── src/
│       └── build/
└── AI Local/                # Local AI models (external)

Models

The setup uses CodeLlama models from HuggingFace, loaded locally for privacy and performance.

Supported models:

• CodeLlama-7b-Instruct
• CodeLlama-13b
• And others in D:\DEV\AI Local

GitHub Integration

• Repository: https://github.com/Obayne/CodeSet
• Automated commits and releases
• Use GitHub CLI for additional automation

Extending

To add more AI agents:

1. Create new folders in mcp-servers/
2. Implement inference.py for the specific model
3. Build and add to MCP settings
4. Add communication tools as needed

License

This project is open source. Check individual model licenses for usage restrictions.