There are some situations that I need the bot classify the sentence according to all intents except some, just compare the sentence to these intents and not those?
So that I avoid if there is a probability that 2 intents may affect confidence of each other.
Can I do such a thing like that?
You can do indirectly.
In each node you can specify the matching criteria, i.e., predicates that need to be true. The expression language for intents allows to access their properties. So you could check which of the two intents in question has the higher confidence.
Depending on the exact situation (you did not provide any details) you might need to set alternate_intents
to true to have more detected intents returned.