Right. I’ve had a lot of success using structured output to force LLMs to make Boolean choices, like can they reply or not.