Identifies and labels cognitive distortions in input text (e.g., catastrophizing, black‑and‑white thinking), returning types, text spans, and brief explanations.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.