🖥️ Log to Workspace

"🖥️ Log to Workspace" is the pack's on-canvas debug output. Feed it any string, and it shows that string on the node itself in a read-only textarea, then passes it through. No scrolling back through a terminal to find which prompt made that image - the answer is sitting on the canvas right where you left it. It's the node you drag into a prompt chain when you want to see what the randomizer actually produced before the sampler does anything with it.

How it works

The node is marked as an output node, which is ComfyUI's way of saying "always run this, even if nothing is wired out of it." That's the whole trick: it can't be skipped by the graph's backward-walking cache, so it fires every run. A bundled JavaScript extension creates the textarea widget, makes it read-only, and auto-scrolls it to the newest lines as they arrive. Text in, text out - the output is just a pass-through, so you can drop it inline in a prompt chain without breaking anything.

The one thing to know: the display is read-only and not part of the workflow you save. The extension deliberately keeps it out of the saved JSON, so when you reload a workflow, the box comes back empty until the next run fills it. That's by design, not a bug.

The inputs that matter

• text - the string to log. Wire it from a randomizer, a concatenator, or a CLIP encode.

Output: text (STRING), unchanged. Run it into whatever was downstream.

Installing it

Pack-standard - no requirements.txt, no models, pure Python plus a small JS file. The README is a "TODO." stub, but there's nothing else to configure.

ComfyUI Manager → Custom Nodes Manager → search "TextRandomizer" → Install → Restart

or cd ComfyUI/custom_nodes && git clone https://github.com/fls-eugene/ComfyUI_TextRandomizer, then restart. The JS extension loads automatically from the pack's web/ folder.

When to reach for it

Any time you're debugging a randomized prompt. The classic move: wire a Weighted Multiline Randomizer into this, wire this into the CLIP encoder, run a batch, and read exactly which prompt produced each image without touching the terminal. It also pairs naturally with the Advanced variant if you want the logged text cleaned up on the way through - this one shows you the raw string, warts and all, which is usually what you want first.