I am trying to create custom actions as part of the a chatbot. I have defined multiple custom actions, and in one of the custom actions I am validating a particular text that I am taking as an input. The return I am testing if it is a valid description I am returning it as
When I use rasa interactive, it is working till here and after that, it is triggering rasa fallback. I am kind of confused as to why this is happening.
Just on a side note: When I copied the whole code into a new folder and did a rasa init. Then when I have triggered the same thing, it is working absolutely fine.
I am new to RASA, I am not sure what I am missing in this whole mix. Any help is greatly appreciated.
Rasa Version : 3.6.13
Minimum Compatible Version: 3.5.0
Rasa SDK Version : 3.6.2
Python Version : 3.8.18
Operating System : Linux-5.15.0-88-generic-x86_64-with-glibc2.17
Hi @DarkSURrealYA ,
The “rasa fallback” is generally triggered when you utter something with insufficient confidence.
Could you try executing “rasa interactive” to observe the confidence levels and the slots that have been assigned during a conversation. Try seeing if something looks out of the ordinary to you.
Hi @Harsh2 ,
I am not trying to get any intent from that, so it is the last step and whatever comes there should be thrown to an API.
Anyways, this is what I see as part of the interactive terminal
I know that the default NLU_Fallback is at 0.3, so I am confused.
@stephens could you help us with the issue. Please and thank you
From the screenshot, it looks like there’s an issue with the dialogue prediction. Can you provide an example of how the conversation should go and the rules/stories/forms you’ve written to support it?
Also, the screenshot shows a problem with the slot mapping for
valid_description. You should not be filling it with the user message.
Hi @stephens ,
thanks for the response, yesterday I have re-visited the masterclass on rasa on youtube, and then watched the live coding for rasa 3.0. From that I have redesigned the flow of things and the issue is resolved.
I changed the way I am taking the first intent, I have set the slot, and picked up the intent from there and controled the flow of things. This helped me overcoming the issue of it getting confused with flow of things.
however, when doing so, I just have to make sure that the similar setence doesn’t come in any of the conversation flow so that it doesn’t re-trigger the flow from the start.