I noticed that in CALM it is directly the LLM that defines which slot to set based on the information received. It also seems that any extracted Entities are not provided in the prompt as contextual information. I would like to ask if it is possible to define (perhaps via action) a mapping policy for the slot that uses LLM when available otherwise directly an entity?
Also, I would like to ask if it is possible to programmatically set the allowed_values of a slot in such a way as to restrict the choice whether using CALM or NLU-based.