Explore if we want to also include the results of the generated code that was executed. It can give us more information about users' data and expectations.

UPD: Tabling it for now in favor of just passing the sample document. Might revisit this idea In the future if we don't meet a desirable accuracy.