It seems to me that this component is primarily intended to be used for an AI as an interactive person. It would be useful to be able to incorporate other ways that LLMs are used (explaining a concept, providing figures, etc.) in behavioral research that require more than plain text input/output.