code_analyzer

AI-Powered Analysis

Refactoroscope includes AI-powered code quality suggestions that provide intelligent insights on code readability, performance, potential bugs, and security issues. Note: AI functionality is now integrated directly into the main analyze command rather than being a separate command.

How It Works

The AI analysis feature uses multiple AI providers to analyze your code and provide suggestions for improvement. The analysis is performed on a file-by-file basis, with each file being analyzed separately by the AI model. AI-generated suggestions are integrated into the existing code smell detection system.

Supported AI Providers

Refactoroscope supports multiple AI providers:

1. OpenAI: Supports GPT models (GPT-3.5, GPT-4, etc.)
2. Anthropic: Supports Claude models
3. Google: Supports Gemini models
4. Ollama: Supports locally-run models (no API key required)
5. Qwen: Supports Qwen models

Usage

To analyze your project with AI:

# Enable AI suggestions during regular analysis
uv run refactoroscope analyze /path/to/your/project --ai

# Enable AI suggestions during watching
uv run refactoroscope watch /path/to/your/project --ai

AI-generated suggestions will appear in the “Code Smells Detected” section of the analysis output.

Configuration

AI functionality is configured in the .refactoroscope.yml configuration file. See the Configuration Guide for details.

To analyze with a specific AI provider:

uv run refactoroscope ai /path/to/your/project --provider openai

To enable AI suggestions during regular analysis:

uv run refactoroscope analyze . --ai

To enable AI suggestions during watching:

uv run refactoroscope watch . --ai

Configuration

AI features are configured through the .refactoroscope.yml configuration file:

# AI configuration
ai:
  # Enable AI-powered code suggestions
  enable_ai_suggestions: true
  
  # Maximum file size to analyze with AI (in bytes)
  max_file_size: 50000
  
  # Whether to cache AI analysis results
  cache_results: true
  
  # Cache time-to-live in seconds
  cache_ttl: 3600
  
  # Preference order for AI providers
  provider_preferences:
    - "openai"
    - "anthropic"
    - "google"
    - "ollama"
    - "qwen"
  
  # Provider configurations
  providers:
    openai:
      # API key (can also be set via OPENAI_API_KEY environment variable)
      # api_key: "your-openai-api-key"
      
      # Model to use
      model: "gpt-3.5-turbo"
      
      # Whether this provider is enabled
      enabled: true
    
    anthropic:
      # API key (can also be set via ANTHROPIC_API_KEY environment variable)
      # api_key: "your-anthropic-api-key"
      
      # Model to use
      model: "claude-3-haiku-20240307"
      
      # Whether this provider is enabled
      enabled: true
    
    google:
      # API key (can also be set via GOOGLE_API_KEY environment variable)
      # api_key: "your-google-api-key"
      
      # Model to use
      model: "gemini-pro"
      
      # Whether this provider is enabled
      enabled: true
    
    ollama:
      # Ollama doesn't require API keys
      
      # Model to use
      model: "llama2"
      
      # Base URL for Ollama (default is localhost)
      base_url: "http://localhost:11434"
      
      # Whether this provider is enabled
      enabled: true
    
    qwen:
      # Qwen doesn't require API keys when using local Ollama
      
      # Model to use
      model: "qwen2"
      
      # Base URL for Qwen (default is localhost)
      base_url: "http://localhost:11434"
      
      # Whether this provider is enabled
      enabled: true

Environment Variables

For cloud-based providers, you can set API keys via environment variables:

• OPENAI_API_KEY for OpenAI
• ANTHROPIC_API_KEY for Anthropic
• GOOGLE_API_KEY for Google

Output Format

AI analysis results are integrated into the regular analysis output and include:

• Code quality suggestions
• Performance recommendations
• Potential bug detection
• Security issue identification
• Readability improvements

Each AI suggestion includes a confidence level and a detailed explanation.

Limitations

1. API Costs: Cloud-based AI providers may incur costs based on usage
2. Privacy: Code is sent to external AI providers (except for local providers like Ollama)
3. Response Time: AI analysis may take longer than static analysis
4. Accuracy: AI suggestions may not always be applicable to your specific context

Best Practices

1. Start with Local Models: Use Ollama or similar local providers for initial testing
2. Set File Size Limits: Configure appropriate file size limits to control costs
3. Use Caching: Enable caching to avoid re-analyzing unchanged files
4. Review Suggestions: Always review AI suggestions before implementing them
5. Configure Preferences: Set up provider preferences based on your needs and budget

This site is open source. Improve this page.