Model usage

Configure and optimize LLM usage for your environment.

Supported Models

Claude Models

• Claude Opus 4

• Claude Sonnet 4

• Claude Sonnet 3.7

• Claude Haiku 3.5

Gemini Models

• Gemini Pro 2.5

• Gemini Flash 2.5

Model Selection Strategy

Task Complexity     Recommended Model     Rationale
───────────────     ─────────────────     ─────────
Simple refactor     Haiku/Flash          Cost effective
Complex analysis    Opus/Pro             Higher accuracy
Creative solutions  Opus/Pro             Better reasoning
Bulk operations     Haiku/Flash          Token efficiency

API Configuration

models:
  primary:
    provider: "anthropic"
    model: "claude-4-opus"
    api_key: "${ANTHROPIC_API_KEY}"
    max_tokens: 100000
    temperature: 0.2

Token Optimization

1. Context Window Management

   • Prioritize relevant code

   • Compress historical context

   • Use incremental updates

2. Prompt Engineering

   • Concise, clear instructions

   • Structured output formats

   • Few-shot examples

3. Caching Strategy

   • Cache analysis results

   • Reuse common patterns

   • Share context across steps