Maybe it should be configurable from a dialog dropdown but I don't know how you expect it to work with an LLM that doesn't involve putting the setting in its context, which is what I think you are referring to with "process tokens constantly".