intelligent-router

To route sub-agent tasks to the right LLM model, classify each request using weighted multi-dimension scoring to identify its complexity tier, then map that tier to a configured primary model. This ensures cost-effective model selection for delegated agent workflows.

Task classification for cost-aware model routing scores requests across multiple dimensions to categorize workloads as simple, medium, complex, reasoning, or critical. This classification enables mapping each tier to optimal models to reduce wasted spend on simple work.

Handle LLM model failures during agentic task execution by configuring automatic retry chains and fallback models. The router maps each task tier to a primary model and a fallback chain so sub-agents can recover automatically when a preferred model fails.

Yes, you can estimate token and dollar costs before routing an LLM task. The router validates health and estimates execution costs beforehand, making it easier to compare routing options and avoid overusing premium models for simple checks or code fixes.

Yes, intelligent routing works for security audits and formal reasoning workflows. It applies to sub-agent delegation for critical review workflows, feature builds, code fixes, and monitoring checks, protecting quality for high-stakes requests through tiered model mappings and confidence thresholds.

Before using a multi-tier model routing system, you need to add your model configuration including primary models and fallback chains. Once configured, you can ask the router to classify a task and spawn the right sub-agent model with confidence using weighted scoring.