Add ability to have "follow-up" conversations with generated responses #12
Description
We want to be able to have a "follow-up" conversation with a LLM given its initial response to some prompt. This would hopefully allow us to try to do stuff like interrogate the model on why it might have chosen certain websites, why did it mention specific competitors / us, etc.
To do this, I think we'll need to save the raw JSON response object we get from the LLM into the DB (and not just its text), and then on follow-up, we would just feed this back into the LLM as the starting messages, and have a table dedicated to a follow-up.