Hello everyone,
I am trying to create RASA CALM flows. I am trying to understand how can I use more complex conditions. For example, if i dont want to check a number or something standard how can I check if the user answered something like an intent and based on that to ask the next question. For example:
I want to ask “Do you have any siblings?”
The user may say “yes” or " I have one sister" or “two brothers”. So until now with my old rasa chatbot I would use an intent with all these answers and continue the conversation accordingly. Now with the new installation I have done with CALM there are no entities training so I am not sure how do it.
Can anyone explain or point me to a tutorial or something? Since I haven’t found something similar.
Thank you very much
You can still use entities in CALM, you can add all the nlu related components like tokinizer,…, intent classifier and entity extractor. then chose from entity as a mapping for your slot
If you want to rely on the llm to do the mapping then you can use llm as mapping which is the default one and a good tip is to add descriptions to the slot like here
collect: amount
description: the number of US dollars to send
Then there is the type of the slot, if it is categorical then it is straight forward since you know ahead the possible answers but if you don’t know the answers ahead then you can use text.
it feels like your example it might be text that you need. since you want to keep it flexible then you can rely on the llm to fill in the slot(slot description here is important to give an idea of what is expected to capture as information and store it in the slot)
thank you for your answer! So another question if I use both NLU and Flows how can I create the conversation order? Through flows? I mean like we used to do with stories, in flows I will say which flow calls the next one etc?