Routes LLM requests to the best model based on task, cost, and latency needs.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.