Display UI 988

ComfyUI is bad at showing you what's actually flowing through a wire. You connect a node, the image comes out fine, but the text - the prompt, the ratio, the diagnostic string - just disappears into the graph. Display UI 988 is the fix: a single input of any type, and it renders whatever arrives as a read-only text box right on the node.

It's a shamelessly simple utility, but it earns its place in this pack because several of the pack's own nodes produce text outputs that are otherwise invisible - the matrix grid from the DaC nodes, the info summaries, LM Studio 988's response and troubleshooting strings. This is how you see any of them without adding a full note-node or saving to a file.

How it works

The input is typed * (the wildcard "ANY"), so it accepts strings, numbers, images, latents - anything a wire can carry. The node converts what it gets to text (integers, floats and booleans become their string form; anything else gets str()-ed) and pushes it to the UI as a text box. It's an output node, which means it always executes, which is exactly what you want from a debugging probe: it can't be skipped by caching.

The input that matters

Just one: value. What you wire in decides what you see. The genuinely useful pattern is to treat it as a probe: temporarily wire it onto any output you're curious about, read the result, then delete the wire. Because it's an output node it forces evaluation of everything upstream of it, so it's also a crude "make this branch always run" tool - though that's what it is, not what it's for.

Where people get burned

• It shows [object] or a tensor shape, not what you expected - you fed it an image or latent, and it did its best (str() of a tensor is not useful). Wire it text outputs and numeric outputs, not pixels.
• You can't find the text - the box only appears once the node has run. Run the workflow, then read.
• It produces no output socket, so you can't chain it - it's a terminal node. That's by design.

When to reach for it

Debugging and confirmation. The KB's plumbing essay is explicit that the invisible middle layer is where workflows become illegible - a display node is the cheapest legibility tool there is. Every time you find yourself asking "did the DaC matrix actually come out 3×2?" or "what did the LLM actually return?", this is the node that answers in one run instead of a file dump.

Install

Part of ComfyUI-988: Manager (search "ComfyUI-988") or:

cd ComfyUI/custom_nodes
git clone https://github.com/kajan988/ComfyUI-988
cd ComfyUI-988
pip install -r requirements.txt

Then restart. No dependencies beyond the pack's light set, and the display logic runs via the pack's own frontend extension (in js/steudio/) - if the box never shows after install, a restart or a browser hard-refresh usually fixes the extension not having loaded.