get config data

This one will confuse you if you drop it into a graph expecting it to feed something downstream - it has zero outputs. Nothing in ComfyUI ever receives what this node produces through a wire. It's marked as an output node, same category as SaveImage: it does its job by pushing data back through the execution API, not by handing anything to the next node in your graph.

What it's actually returning

Two things, and neither is meant for your image pipeline. First, embeddings - a filesystem path (the schema's own example shows /tmp/ComfyUI/models/embeddings) so an external caller can see where your embeddings folder lives and list what's available, the kind of thing a UI needs to build an autocomplete dropdown. Second, an optional controlnet_config - and this is the genuinely useful part, even if you never touch the Photoshop plugin.

controlnet_config's default value is a complete reference table. It's a JSON dictionary covering every preprocessor this pack's ControlNet nodes support - Canny, OpenPose, HED, FakeScribble, Inpaint, LeReS depth, AnimeLineArt, LineArt, Manga2Anime lineart, MediaPipe FaceMesh, MiDaS normal and depth, M-LSD, BAE normal, OneFormer (COCO and ADE20K variants), PiDiNet, Scribble and Scribble XDoG, Shuffle, Tile, UniFormer, SemSeg, and Zoe depth - and for each one, its default parameters and, where relevant, exactly what the generically-named threshold_a/threshold_b fields on ControlnetUnit and ControlNetScript actually mean for that specific preprocessor. A few concrete examples straight from the schema: CannyEdgePreprocessor maps threshold_a to low_threshold (0–255, default 100) and threshold_b to high_threshold (0–255, default 200). MediaPipe-FaceMeshPreprocessor maps them to max_faces (1–50, default 10) and min_confidence (0.01–1, default 0.5). M-LSDPreprocessor maps them to score_threshold and dist_threshold. If you've ever wired up ControlnetUnit, picked a preprocessor, and wondered what those two optional threshold fields were supposed to do for it, this node's own default payload is the answer key.

Why it exists as a ComfyUI node at all

The Photoshop plugin needs to build a settings panel - sliders and toggles matched to whichever preprocessor you pick - without hardcoding controlnet_aux's parameter names and ranges into the Photoshop-side JavaScript. Querying this node's config instead means the plugin's UI stays correct even if the underlying preprocessor list changes. It's infrastructure for a UI you're probably not building, not a node that belongs in a normal generation workflow.

When you'd actually use it

Basically never, if you're just running ComfyUI workflows through the standard UI - there's nothing here to wire anywhere. It's worth knowing about in two situations: you're building your own client against the ComfyUI API and want a ready-made reference for controlnet_aux's per-preprocessor parameters, or you're troubleshooting one of this pack's ControlNet nodes and need to know what threshold_a/threshold_b are actually supposed to mean for the preprocessor you picked.

Installing it

ComfyUI Manager: search "Comfy-Photoshop-SD", install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/AbdullahAlfaraj/Comfy-Photoshop-SD

Where people get burned

Really just one thing: expecting it to do something visible in your graph. It won't, and that's correct - a node with no outputs that does nothing downstream isn't broken, it's working exactly as designed. If you're chasing a bug and this node is anywhere in the suspect list, it almost certainly isn't the culprit; the nodes actually reading its config live on the Photoshop plugin's side, outside ComfyUI entirely.