backends:
  anthropic:
    api_key_env: ANTHROPIC_API_KEY
    enabled: false
    model: claude-sonnet-4-6
    type: anthropic
    supports_images: true
  claude_code:
    api_key: any
    base_url: http://localhost:3009/v1
    enabled: false
    model: claude-code-terminal
    type: openai_compat
    supports_images: true
  github_copilot:
    api_key: any
    base_url: http://localhost:3010/v1
    enabled: false
    model: gpt-4o
    type: openai_compat
    supports_images: false
  ollama:
    api_key: ollama
    base_url: http://localhost:11434/v1
    enabled: true
    model: llama3.2:3b   # replace with your fine-tuned cover letter model if you have one
    type: openai_compat
    supports_images: false
  ollama_research:
    api_key: ollama
    base_url: http://localhost:11434/v1
    enabled: true
    model: llama3.1:8b
    type: openai_compat
    supports_images: false
  vllm:
    api_key: ''
    base_url: http://localhost:8000/v1
    enabled: true
    model: __auto__
    type: openai_compat
    supports_images: false
  vision_service:
    base_url: http://localhost:8002
    enabled: false
    type: vision_service
    supports_images: true
fallback_order:
- ollama
- claude_code
- vllm
- github_copilot
- anthropic
research_fallback_order:
- claude_code
- vllm
- ollama_research
- github_copilot
- anthropic
vision_fallback_order:
- vision_service
- claude_code
- anthropic
# Note: 'ollama' intentionally excluded from research order — research
# must never use the cover letter model, and this also avoids evicting
# the writer from GPU memory while a cover letter task is in flight.