For handling wrong intent classification

Can you suggest about rasa chatbot, how to handle an unknown question which enters by the user which would be not trained and getting random intent with it?

Hey there @MaulishShah, for this we suggest having an out_of_scope intent where you put lots of training data for irrelevant questions and comments. Then when your model predicts the intent as out_of_scope, you can utter something like utter_out_of_scope “Sorry, I can’t help you with that.” or whatever you’d like to do (maybe tell the user what you can help with). You can check out our demobot which has the out_of_scope intent data in the nlu training file and the utterances in the domain file.

I already know about that but want to know if there is another way or not. Also can you suggest how to store whole chat in database or text file when we are create personal UI.?