"NLG service" endpoint response from external model

Rasa Open Source
1

Currently in the docs, it says that if you expose an external server for NLG, the response back needs to be formatted in this way:

{ “text”: “hey there”, “buttons”: , “image”: null, “elements”: , “attachments”: }

Is there a way to send a response back from the NLG model that triggers a custom action or a an action in the domain file… For example, if the intent for a user response is over a certain threshold —> do not produce nlg response → consider intent/entity classification response from “what ever nlu pipeline used”

Thank you for your help, please let me know if I should elaborate … Fyi I am using rasa core 11.2 but am more than willing to hear out answers for newer verisons. Just want a guide before I dive into the source code

2

Currently using the rasa.agent as a workaround/hack… there should be a better way to compare multimodel outputs before responding with an utterance

3

Still looking for a solution… if I connect an NLG model to the bot engine, it automatically query’s it for all responses and feeds it all user uttertances. I would like to have a switch that only considers NLG for specifics intents/actions that I can turn off and on.