⭐Mind Map

Go_to_image (⭐Mind Map) is the pack's oddest node, and the first thing you should know is what the README says about it in two Chinese characters: 不可推理 - "cannot infer." It is not a functional node. The author uses it as a visual placeholder in sample workflows, a big labeled marker (its display name is "思维导图," mind map) so a workflow document reads like a diagram. If you wire it into a real graph expecting it to do something, you will be disappointed; if you try to execute it, you may well get an error, because the shipped source code doesn't even match its own declared inputs.

Let's be concrete about the mismatch. The schema advertises three inputs - image (a file picker from your input folder), pos_text (a multiline string), and images (an IMAGE tensor) - and three outputs: IMAGE, MASK, CONDITIONING. The actual load_image function in the source takes only one argument, does the standard Load Image routine (EXIF fix, RGB, mask from alpha), and returns two values. So the declared CONDITIONING output has no code behind it, and the function can't receive the inputs the schema promises. The clean way to read this: the author stopped halfway through building a real node and kept it as decoration.

What is it for, then? Look at the sample workflows in the pack - the mind-map JSONs - and you'll see it used as an annotation: "here's where the image comes in," "this is the positive prompt," and so on. It makes a saved workflow legible as a picture, which is a real thing in a pack whose whole pitch is "hand this to someone who doesn't understand nodes." Just don't connect it to anything you plan to run.

Install: it ships with ComfyUI_to_webui - ComfyUI Manager → "ComfyUI_to_webui", or

cd ComfyUI/custom_nodes
git clone https://github.com/kungful/ComfyUI_to_webui.git

then restart. No dependencies beyond the pack's own.

The takeaway, in one sentence: if you're reading this because a workflow you downloaded contains a ⭐Mind Map node and ComfyUI won't run it - that's expected, delete or ignore it, and the rest of the graph is probably fine. It's documentation wearing a node costume. If you want the actual behavior it looks like it should have (image-in, mask and conditioning out), use the pack's GradioInputImage plus a normal conditioning encoder instead. This one is a label, not a loader.