Auxilus08/github-docs-mcp-server

GitHub Docs MCP Server

A Model Context Protocol (MCP) server that provides AI-powered Q&A capabilities for GitHub repositories using OpenAI GPT-4 or GPT-3.5-turbo.

🤖 OpenAI Integration

This server integrates with OpenAI's GPT models to provide intelligent, context-aware answers to questions about GitHub repositories. The AI analyzes repository documentation, code, and issues to generate comprehensive responses.

Features

• AI-Powered Responses: Uses GPT-4 or GPT-3.5-turbo to generate intelligent answers
• Context-Aware: Analyzes repository docs, code snippets, and issues
• Markdown Formatting: Returns well-formatted answers with proper structure
• Source Links: Includes links back to original GitHub files and issues
• Fallback Support: Gracefully falls back to template responses if OpenAI API fails
• Configurable: Support for different OpenAI models and parameters

Setup

1. Get an OpenAI API Key

   • Visit OpenAI API
   • Create a new API key
   • Ensure you have credits available

2. Configure Environment Variables

   # Copy the example environment file
   cp .env.example .env
   
   # Edit .env and add your keys
   OPENAI_API_KEY=your_openai_api_key_here
   GITHUB_TOKEN=your_github_token_here
   
   # Optional: Configure model preferences
   OPENAI_MODEL=gpt-4  # or gpt-3.5-turbo
   OPENAI_MAX_TOKENS=1500
   OPENAI_TEMPERATURE=0.3
   

3. Install Dependencies

   pip install -r requirements.txt
   

4. Validate Installation

   python validate_openai.py
   

🚀 Usage

Start the Server

python -m src.github_docs_mcp.main

The server will start on http://localhost:8000

Check Service Status

curl http://localhost:8000/status

This endpoint shows the status of all services including OpenAI integration:

{
  "server": {
    "name": "github-docs-qa",
    "version": "1.0.0",
    "status": "running"
  },
  "services": {
    "openai_service": {
      "status": "healthy",
      "model": "gpt-4",
      "features": {
        "ai_answers": true,
        "fallback_answers": true
      }
    }
  }
}

Ask Questions

Send POST requests to /ask with repository questions:

curl -X POST http://localhost:8000/ask \
  -H "Content-Type: application/json" \
  -d '{
    "repository": "microsoft/vscode",
    "question": "How do I create a VS Code extension?",
    "include_code": true,
    "include_issues": false,
    "max_results": 5
  }'

Response Format

The AI generates responses in Markdown format with:

• Structured answers with headers and sections
• Code examples with proper syntax highlighting
• Source links to original GitHub files
• Contextual information based on repository content

Example response:

{
  "question": "How do I create a VS Code extension?",
  "answer": "# Creating a VS Code Extension\n\nTo create a VS Code extension, you'll need to...\n\n## Getting Started\n\n1. Install the required tools\n2. Generate the extension scaffold\n3. Configure your extension\n\n## 📚 Sources\n\n1. **Documentation:** [extension-authoring.md](https://github.com/microsoft/vscode/blob/main/docs/extension-authoring.md)\n2. **Code:** [example-extension.js](https://github.com/microsoft/vscode/blob/main/examples/example-extension.js)",
  "repository": "microsoft/vscode",
  "sources": [...],
  "confidence": 0.85,
  "processing_time_ms": 2500
}

🔄 Fallback Behavior

When OpenAI API is not available or fails, the server automatically falls back to template-based responses:

• No API Key: Uses structured templates with source content
• API Errors: Gracefully handles rate limits and other errors
• Network Issues: Continues serving requests with fallback responses

🧪 Testing

Test the OpenAI integration:

# Run validation checks
python validate_openai.py

# Test integration with live server
python test_openai_integration.py

# Run full test suite
python test_runner.py

📊 Monitoring

Monitor OpenAI usage through:

• Service Status: /status endpoint shows AI service health
• Server Logs: Detailed logging of AI interactions
• Response Analysis: Check for AI vs fallback responses

⚙️ Configuration

OpenAI Models

Supported models:

• gpt-4 (recommended, higher quality)
• gpt-3.5-turbo (faster, lower cost)

Parameters

• Max Tokens: Control response length (default: 1500)
• Temperature: Control creativity (default: 0.3 for factual responses)
• Timeout: API call timeout (configured automatically)

Rate Limiting

The server respects OpenAI rate limits and handles:

• Rate limit errors: Automatic fallback to templates
• Token limits: Smart content truncation
• Cost management: Configurable token limits

🔧 Troubleshooting

Common Issues

1. "OpenAI service unavailable"

   • Check API key in .env file
   • Verify OpenAI account has credits
   • Test API key with OpenAI directly

2. Slow responses

   • Try gpt-3.5-turbo for faster responses
   • Reduce max_results in requests
   • Check network connectivity

3. Rate limit errors

   • Upgrade OpenAI plan for higher limits
   • Implement request queuing if needed
   • Monitor usage in OpenAI dashboard

Debug Mode

Enable detailed logging:

DEBUG=true LOG_LEVEL=debug python -m src.github_docs_mcp.main

🛡️ Security

• API Keys: Never commit API keys to version control
• Environment Variables: Store sensitive data in .env files
• Request Validation: All inputs are validated and sanitized
• Error Handling: Sensitive information is not exposed in error messages

📈 Performance

Typical response times:

• GPT-4: 2-5 seconds
• GPT-3.5-turbo: 1-3 seconds
• Fallback: < 1 second

Optimization tips:

• Use gpt-3.5-turbo for speed-critical applications
• Limit max_results to reduce context size
• Cache frequently asked questions
• Monitor token usage to optimize costs

🤝 Contributing

When contributing to OpenAI features:

1. Test with both GPT-4 and GPT-3.5-turbo
2. Ensure fallback behavior works correctly
3. Add appropriate error handling
4. Update documentation and examples
5. Test with various repository types and question formats

📄 License

MIT License - see for details.