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