[V2] re-add %readback or an equivalent to see what the ai detects an image as

[artifishvr]

potential other ideas could be appending what the vision model saw to the end of the prompt, instead of a /readback