Output Node

The first time you run the main LLM_Node in this pack, you get a string out and then have to hunt for it. The Output Node is the answer to that: it's the pack's own display widget, a big read-only text box that shows whatever text you feed it. It's not doing any magic, but it's genuinely the right way to look at model output instead of wiring it into a random text widget and squinting.

What it actually is

It's an output node - OUTPUT_NODE = True - with exactly one input, text, and it accepts any type. That's deliberate: it doesn't care whether you feed it the STRING from LLM_Node, a number, or a list, it just displays it. It returns nothing downstream; it's the end of the line, which is what makes it show up properly in ComfyUI's UI.

The bundled JavaScript swaps the default node for a read-only multiline text widget and stretches it to three times the normal node height, so you get a real text panel instead of a tiny one-liner. That's the whole thing. Under the hood the Python just returns {"ui": {"text": ...}}, and the frontend fills the widget on execution.

Why you'd reach for it

Two reasons. First, it's the natural sink for LLM_Node's string output - you generate your text, and here's where you read it and copy it. Second, because it takes any type, it doubles as a cheap debug viewer: stuck mid-workflow and not sure what a node actually produced? Run it through the Output Node and see. It's the pack's own equivalent of the various text-display utilities in other packs, and it's already installed with the pack, so you might as well use it.

Installing it

It ships in the LLM Node for ComfyUI pack (Big-Idea-Technology/ComfyUI_LLM_Node). The easy path is ComfyUI Manager - search for "LLM Node" and install the pack, which pulls in its dependencies (transformers, torch, accelerate, llama-cpp-python). Or clone it directly:

cd ComfyUI/custom_nodes
git clone https://github.com/Big-Idea-Technology/ComfyUI_LLM_Node

then restart ComfyUI. Note this node itself needs no model - but whatever LLM_Node is feeding it does, and that's where people get stuck (models go in ComfyUI/models/LLM_checkpoints/<model_folder>).

Gotchas

If you see a plain generic node instead of the big text panel, the pack's web asset didn't load - restart ComfyUI and hard-refresh the browser (Ctrl+Shift+R). And don't try to type in it; the widget is deliberately read-only, which throws a lot of people for a second. Finally, it displays whatever was passed at execution time, so if it's blank, the problem is upstream - LLM_Node prints its errors to the console, so that's the first place to look.