Scores training datasets for fine-tuning readiness by checking size, completeness, output variance, and class balance.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.