I’m writing a robot for my company’s reimbersement process, and it’s designed to be connected with an OCR tool for reciepts recognition. To write this robot, I need to make judgements about “whether this input is a picture or not” and “is the slot filled” but I don’t know how. Is there any references that I can look up for or is it not possible for rasa to do this? It’s like, connect rasa with an external tool, and make input/output with it.
The slot filled part is handled natively by Rasa since all slot values are available on the conversation tracker.
Re. whether input is a picture or not: This should be handled by your channel connector; the judgment that it is not text should be happening in the frontend of the channel itself.