# Model Configuration for SPARKNET
# Maps task types to appropriate Ollama models

models:
  # Large models for complex reasoning
  reasoning:
    - name: "qwen2.5:14b"
      size: "9.0 GB"
      use_cases: ["complex_planning", "advanced_reasoning", "multi_step_tasks"]
      temperature: 0.7

  # Mid-size models for general tasks
  general:
    - name: "llama3.1:8b"
      size: "4.9 GB"
      use_cases: ["general_tasks", "code_generation", "analysis"]
      temperature: 0.7

    - name: "mistral:latest"
      size: "4.4 GB"
      use_cases: ["general_tasks", "creative_writing", "summarization"]
      temperature: 0.7

  # Lightweight models for simple tasks
  lightweight:
    - name: "llama3.2:latest"
      size: "2.0 GB"
      use_cases: ["classification", "routing", "simple_qa"]
      temperature: 0.5

    - name: "phi3:latest"
      size: "2.2 GB"
      use_cases: ["quick_reasoning", "structured_output"]
      temperature: 0.5

  # Embedding models
  embeddings:
    - name: "nomic-embed-text:latest"
      size: "274 MB"
      use_cases: ["text_embeddings", "semantic_search"]

    - name: "mxbai-embed-large:latest"
      size: "669 MB"
      use_cases: ["high_quality_embeddings", "rag"]

# Model routing rules
routing:
  # Map task complexity to model tier
  task_complexity:
    simple: "lightweight"
    moderate: "general"
    complex: "reasoning"

  # Fallback chain if primary model unavailable
  fallback_chain:
    - "llama3.2:latest"
    - "mistral:latest"
    - "llama3.1:8b"