route-switch investigates the cost-quality frontier in LLM inference. When should you route a query to a small, fast model versus a large, expensive one? This project implements MIPROv2-based automatic prompt tuning and model selection to answer that question empirically.