LLM Routing System for Engineering Lab Assistance

A novel governance and routing framework for lab assistance utilizing LLMs has been introduced, featuring two key elements: Routiium, an OpenAI-compatible interface that oversees various LLM backends with customizable prompt adjustments and usage tracking, and EduRouter, a policy-driven routing mechanism that implements lab-specific budgets, approval processes, and question matching based on embeddings. The framework underwent evaluation through a simulation informed by data from two engineering laboratories (LED characterization and RC circuit analysis) and involved a 100-query replay with live models. In these simulations, the regulated policies (P1/P2) enhanced the challenge-alignment index from 0.90 to 0.98 and improved the overlay-adherence score from 0.69 to 0.87, compared to unregulated benchmarks. This system aims to balance adequate AI support with maintaining learning opportunities, granting educators control over the timing, content, and expenses of assistance.