Policy-Governed LLM Routing with Intent Matching for Instrument Laboratories
It addresses the tension between assistance and learning in engineering labs by giving instructors policy control over LLM tutoring, with demonstrated improvements in both pedagogical metrics and cost efficiency.
The paper presents a routing and governance system for LLM-based lab assistance that improves challenge-alignment index from 0.90 to 0.98 and reduces token costs by 66% while maintaining perfect canonical hit rate.
AI tutoring systems in engineering labs face a tension between providing sufficient assistance and preserving learning opportunities. Existing systems typically offer instructors limited control over assistance timing, content, or cost. This paper describes a routing and governance system for LLM-based lab assistance comprising two components: Routiium, an OpenAI-compatible gateway that manages multiple LLM backends with configurable prompt modifications and usage logging, and EduRouter, a policy-aware routing service that enforces per-lab budgets, approval workflows, and embedding-based question matching. We evaluated the system using trace-driven simulation calibrated from two engineering labs (LED characterization, RC circuit analysis) and a 100-query replay through live models. In simulations, governed policies (P1/P2) increased challenge-alignment index from 0.90 to 0.98 and overlay-adherence score from 0.69 to 0.87 compared to ungoverned operation (P0). The productive-struggle window metric increased from 1.4 to 3.6 simulated turns before high-scaffold hints appeared. In the 100-query replay, EduRouter routed 75% of queries to a local model, reducing token costs by 66% ($0.087 vs. $0.26 for all-premium routing) while maintaining canonical hit rate of 1.0 for the curated 89-intent question bank. We release Routiium, EduRouter, canonical-task tooling, and simulator configurations to support replication and future classroom studies.