This seems like an obvious question but I cannot find answers anywhere.
I am using RASA with LLM integration. The idea is that RASA will handle the logical flow, ensuring the conversation stays on track, but hand off to the LLM at certain points for deeper ‘chit chat’.
I have a basic story:
stories:
- story: greet mood
steps:
- intent: greet
- action: utter_greet_morning
- action: utter_how_are_you
- intent: mood_poor # this can be anything!
- action: action_llm_general_feeling
But I don’t want to classify an intent at the part of the story intent: mood_poor. Ideally, I’d like this to be something like intent: * and it just runs the custom action regardless of what the user says.
This is because my LLM will handle the NLU part and provide a response.
How can I achieve this without having to define all possible intents? Surely this is easy to do.