Can the knowledge agent get images from the PDFs or webs that he is searching? For example, the user asks for the best wine of 2020, the bot answers and provides an image of that wine. The image is in the KB? Can it be done differently?