AI Engines

Hatcher integrates with multiple AI coding assistants to provide flexibility and leverage the unique strengths of different models.

Supported Engines

Claude Code

Anthropic's Claude is our primary AI engine, optimized for:

• Code Understanding: Excellent comprehension of existing codebases
• Pattern Recognition: Maintains consistency with project conventions
• Safe Refactoring: Conservative approach to code modifications
• Context Awareness: Handles large codebases effectively

Configuration:

{
  "engine": "claude",
  "model": "claude-3-sonnet",
  "maxTokens": 4096,
  "temperature": 0.1
}

Gemini CLI

Google's Gemini provides complementary capabilities:

• Creative Solutions: Alternative approaches to complex problems
• Performance Optimization: Focus on efficient code generation
• Multi-language Support: Strong support for diverse tech stacks
• Rapid Iteration: Fast response times for quick edits

Configuration:

{
  "engine": "gemini",
  "model": "gemini-pro",
  "maxTokens": 2048,
  "temperature": 0.2
}

Engine Selection Strategy

Hatcher automatically selects the best engine based on the task:

Code Generation Tasks

• New Components: Claude for structure, Gemini for creativity
• Bug Fixes: Claude for careful analysis
• Performance: Gemini for optimization suggestions

Context Considerations

• Large Codebases: Claude's superior context handling
• Rapid Prototyping: Gemini's speed advantage
• Complex Refactoring: Claude's conservative approach

Configuration

Global Settings

Set your preferred default engine:

{
  "defaultEngine": "claude",
  "fallbackEngine": "gemini",
  "autoSwitch": true
}

Project-Specific Overrides

Override engines for specific projects:

{
  "engines": {
    "vue": "claude",
    "optimization": "gemini",
    "testing": "claude"
  }
}

API Keys Setup

Claude API Key

1. Visit Anthropic Console
2. Generate a new API key
3. Add to Hatcher settings or environment variable:

   export ANTHROPIC_API_KEY="your-key-here"

Gemini API Key

1. Visit Google AI Studio
2. Create a new API key
3. Configure in Hatcher:

   export GOOGLE_AI_API_KEY="your-key-here"

Engine Comparison

Feature	Claude	Gemini
Code Quality	⭐⭐⭐⭐⭐	⭐⭐⭐⭐
Speed	⭐⭐⭐	⭐⭐⭐⭐⭐
Context Size	⭐⭐⭐⭐⭐	⭐⭐⭐
Creativity	⭐⭐⭐	⭐⭐⭐⭐⭐
Safety	⭐⭐⭐⭐⭐	⭐⭐⭐⭐

Future Engines

We're actively working on support for:

• GitHub Copilot: Direct integration with VS Code workflows
• OpenAI GPT-4: General-purpose coding assistance
• Local Models: Self-hosted options for privacy-conscious teams
• Custom Engines: Plugin system for proprietary AI tools

Best Practices

Prompt Engineering

Hatcher automatically optimizes prompts for each engine:

• Claude: Detailed context and explicit instructions
• Gemini: Concise prompts with clear objectives

Error Handling

Robust fallback systems ensure continuity:

• Primary Engine Down: Automatic switch to fallback
• Rate Limiting: Intelligent request queuing
• Network Issues: Local caching of common patterns

Cost Optimization

• Token Management: Efficient context compression
• Request Batching: Combine multiple small changes
• Caching: Reuse similar code generation results