Last released Sep 2, 2026
Adaptive routing, a quality gate, and a correctness-aware cache on top of Virginia Tech ARC's LLM API