Method optimizeByCost

internal/router/rule_router.go:109–150

(tenant config.TenantConfig, req types.AIRequest, base types.RouteDecision, state types.RuntimeState)

}

func (r *RuleRouter) optimizeByCost(tenant config.TenantConfig, req types.AIRequest, base types.RouteDecision, state types.RuntimeState) types.RouteDecision {
	if r.costEngine == nil {
		return base
	}
	if state.OverloadDegrade {
		return base
	}

	best := base
	bestCost := base.EstimatedCost

	for _, backend := range r.cfg.Backends {
		if !backendSelectable(state, backend.Name) {
			continue
		}
		if !supportsTask(backend, req.TaskType) {
			continue
		}

		estimate := r.costEngine.Estimate(req, types.BackendMetadata{
			Name:           backend.Name,
			Type:           backend.Type,
			Capabilities:   backend.Capabilities,
			CostUnit:       backend.Cost.Unit,
			MaxConcurrency: backend.MaxConcurrency,
		})
		if tenant.BudgetPerRequest > 0 && estimate.EstimatedTotal > tenant.BudgetPerRequest {
			continue
		}

		if estimate.EstimatedTotal < bestCost {
			best = types.RouteDecision{
				BackendName:   backend.Name,
				Reason:        "cost-optimized-from-" + base.Reason,
				EstimatedCost: estimate.EstimatedTotal,
			}
			bestCost = estimate.EstimatedTotal
		}
	}

	return best
}

func supportsTask(backend config.BackendConfig, taskType string) bool {
	if taskType == "" {