FollowupAction and utter_template not working due to KerasPolicy prediction

Hi there,

I’m having trouble with some custom actions in which I want to utter some actions if several conditions are met. I’m trying to do it both through dispatcher.utter_template('utter_name', tracker) and return [events.FollowupAction('utter_name')].

In both cases, the log shows Predicted next action 'utter_name' with confidence 1.00. The problem is that, after that, logs show that KerasPolicy predicts a different next action, so utter_name is not uttered at all and the conversational flow is broken. The same happens if I use EmbeddingPolicy instead of KerasPolicy.

Does anyone know how can I force it so that it utters that action when the conditions are met no matter what KerasPolicy or EmbeddingPolicy predicts?

Thank you in advance.

I am working on a similar issue on another thread, can you share with me your config details and some examples of the training set, I want to confirm if maybe there is a issue with the KerasPolicy and FollowupActions