Custom response selector generated by NLG models


I am wondering if there is a way to customize rasa where I can use a language model to handle my utter actions.

For example: if I have action: utter_cheer_up for intent mood_unhappy. Instead of defining a hardcode response, I can give cheer_up as a prompt to language model to generate a response and handle the utter_action.

I would really like to implement this functionality by myself in Rasa. If you can give me guidance on what would be the best way to do this. That would be really helpful.

You could do this via a custom action that calls your generation model or via the Rasa NLG interface.