We don't use to_prompt to generate AI responses anymore, only for events. Grabbing all the comments don't make as much sense, and I'm sure it is quite expensive
So that we can reuse in other scenarios