Show Last Text

Show Last Text is the other half of the pack's pitch. The author's own repo description is literally "view last seed or prompt" - Show Last Seed covers the seed, and this node covers the prompt. Feed it any text, run, and that text gets drawn on the node body so you can see it without opening anything. It also hands the same text back unchanged as a string output, which makes it easy to drop into the middle of a workflow without changing your data.

What it's actually for

The natural wiring is: take the positive prompt out of your CLIP Text Encode node and feed it into Show Last Text. After each run, the exact prompt that generated the image is sitting there on the canvas. That's surprisingly useful when you're juggling a queue of variations and you want to remember which wording actually produced the result you're looking at - the workflow metadata has it, but this puts it where your eyes already are.

It's not limited to prompts. Negative prompts, a generated filename, a LoRA stack passed around as text - anything that flows through the graph as a string can be displayed. And because the output is a pass-through (it returns exactly what it received), you can splice it into a chain as an output node without worrying about it altering your data. Think of it as a debug readout that happens to be a node.

How it works

Same machinery as its sibling. The backend sends the text to the browser over ComfyUI's WebSocket and the frontend extension stores it as next. When the run finishes - detected via the progress event hitting 100% - the value gets committed to what's drawn on the node. So like everything in this pack, it displays the value from the last completed run, not the one currently in progress.

The drawing itself word-wraps the text to fit the node's width and paints it in a small serif font. Long strings wrap onto multiple lines automatically; numbers and short strings just sit there.

Inputs and outputs

• text (STRING) - the only input you set. Any string: a prompt, a filename, a label.
• Output STRING - the same text, passed through unchanged.

Behind the scenes the node also reports its own internal id so the extension knows which canvas node to paint - none of that is visible to you, and it's the same design across all three nodes in the pack. Show Last Text is an output node, so it sits at the end of a branch and forces that branch to actually run.

Installing it

Identical to the rest of the pack. ComfyUI Manager → search "Remembering utils" (registry-published v1.0.2), or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/bombless/comfyUI-RememberingUtils

Restart ComfyUI. No Python dependencies, no requirements.txt, no model files - this pack is a few short Python files plus one JavaScript file, which also makes it one of the easier custom nodes to audit before you install, since custom nodes run arbitrary code with your user's permissions.

Where people get burned

• Long text gets clipped. The word-wrap stops drawing once it passes the node's height, and the node won't auto-resize. A big prompt can trail off mid-sentence - drag the node bigger to see more of it.
• Line breaks in your source text flatten out. The wrapping splits on spaces, so a multi-line prompt renders as continuous wrapped text; you lose your paragraph structure on screen (though the node's output still contains it).
• It's blank until a run completes. The first value only commits when a run finishes, so a fresh node shows nothing, and a cancelled queue leaves it showing the previous text.
• Nobody's searching for it yet - and that's fine. It's a niche readout, not a headline feature. But if you've ever re-run a prompt and lost track of what you actually sent, it earns its place.

One honest caveat before you rely on it: the pack ships with zero documentation (the GitHub README is empty, and there's an open issue asking the author for even a sentence of explanation). The good news is that reading the source yourself takes about two minutes, and at this size there's nothing hidden in it. It's a display node with a pass-through output. That's the whole story.