Compares LLM responses on relevance, length, clarity, and confidence to rank the winner.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.