2 ms·
Show HN: DriftGuard – response drift detection for LangGraph agents
Detects when an LLM starts answering outside its intended domain (a legal assistant drifting into cooking advice, a medical chatbot wandering into finance) without ground-truth labels or a separate classifier.