Plans context window allocation and compression strategies when content exceeds model limits.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.