Fallback policy calls "utter_default" when more entities are detected

I made a chatbot that answers differently based on detected entity. As showed in the stories.md bellow: ## greet * greet - utter_greet - utter_default

## goodbye
* goodbye
  - utter_goodbye
  - utter_default

## rain
* weather{"rain":"rain"}
  - utter_rain
  - utter_default
  
## sunny
* weather{"sunny":"be sunny"}
  - utter_sunny
  - utter_default

But when weather intents detects a second entity Rasa Core answer with the fallback policy, even if the NLU understanding was correct:

I want it to generalize and answer according with the “sun” or “rain” entities, shouldn’t it work since they are the ones in the stories? How can I fix this behavior?

The full example project can be seen at: https://github.com/devdeca/rasa_forum.git

1 Like

It looks like you have no slots defined in the domain. NLU can parse the entities, but they must be stored in slots to affect action prediction.

See Domain