Monitors AI pipeline health across error, latency, success, cost, and hallucination metrics versus thresholds.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.