I want to customize the responses for each of the low confidence intent that might appear in the chat. For previous versions, I used an customized input map with the function provided in this tutorial:
But keeping this implementation, the ‘Yes’ button is populated in its payload with the ‘nlu_fallback’ intent. Do I have to still declare the FallbackClassifier as a component in the pipeline if I want to keep the two stage fallback from the blog example? Does the rule need a different step for this?