Context

Implement the model routing engine per spec 9.4. The router determines which LLM model handles each request based on configurable strategies, enabling cost optimization and intelligent model selection.

Acceptance Criteria

• Manual routing strategy: explicit model per agent configuration
• Role-based routing: C-Suite → Opus, Senior → Sonnet, Junior → Haiku
• Cost-optimized routing: selects cheapest capable model for the task
• Smart routing: combines role + task complexity + budget constraints
• Router selects model from config/rules correctly
• Fallback chain works when primary model is unavailable
• Cost-based routing considers remaining budget
• Routing configuration loaded from YAML
• Unit tests for each routing strategy
• Routing decision logging for observability and debugging

Dependencies

• Depends on Implement provider adapter layer (direct API or via LiteLLM) #5 (provider adapter layer implementation)

Design Spec Reference

Section 9.4 — Model Routing Engine