Hey Sebastian S., it sounds like you want to evaluate an LLM call using structured outputs or tools. That makes sense. We are thinking about how to update our evals APIs, including llm_classify. Adding the ability to pass in tool and structured output definitions makes sense.