Show Text

If you've built any kind of LLM chain in ComfyUI, you already know the problem: the graph runs, the queue finishes, and you have absolutely no idea what the model actually said. STRING outputs don't render anything on their own - they just sit there as a wire with data flowing through it. Show Text is the node that fixes that. Drop it at the end of an LLM call (or anywhere you want to eyeball a string mid-workflow) and it prints the text right on the node itself.

It's about as simple as nodes get, which is exactly the point. Complex LLM agent graphs in this pack - multi-tool calls, RAG lookups, looping debates between two models - produce a lot of intermediate text you'll want to inspect while you're debugging, and you don't want to reach for a heavier logging setup just to read a string. The pack's own author has said as much directly: it's the node they recommend using as the display output for any LLM node in the party ecosystem, precisely because everything else is just wiring.

What it does

The schema is deliberately minimal. One required input, text (a STRING), and one output that mirrors it back out as a list. That output matters more than it looks - because it's a list output, you can chain Show Text mid-graph without dead-ending your pipeline; the text gets displayed and passed along, so you're not forced to choose between "see it" and "use it downstream." It's also flagged as an output node, which is what tells ComfyUI to actually render the text widget on the node face when the queue completes, rather than treating it as a silent pass-through.

There's nothing to configure. You wire a STRING into it - the reply from an LLM API node, the result of a tool call, a RAG retrieval, whatever - and after execution the node shows you the text. That's the entire interface.

Where it fits in an LLM Party workflow

This pack builds full LLM agent graphs: API loaders, local model loaders, tool nodes (web search, GitHub, Wikipedia, workflow-calling), RAG knowledge bases, even looping multi-agent setups where two models argue with each other. All of that produces text at various stages, and almost none of ComfyUI's core nodes are built to display arbitrary strings. Show Text is the connective tissue that makes those graphs debuggable - stick one after your final LLM response, and another after a tool call if you want to confirm the tool actually returned something useful before it gets fed back into the model.

Installing it

Show Text ships as part of the whole comfyui_LLM_party pack - there's no way to install it standalone, and you wouldn't want to; it's meant to sit alongside the loaders and tool nodes it's built to display. Two ways in:

• ComfyUI Manager (recommended): search "comfyui_LLM_party" in the Manager's node browser, install, restart ComfyUI.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then run pip install -r requirements.txt inside your ComfyUI Python environment (on a portable/embedded build, that's python_embeded\python.exe -m pip install -r requirements.txt), then restart.

The pack pulls in a fair amount - transformers for local models, llama-cpp-python for GGUF support, various API and search libraries - so the requirements install can take a minute even though this specific node needs none of it.

Troubleshooting

Show Text itself is close to unbreakable - if it's not displaying anything, the problem is almost always upstream, not in this node. A couple of things worth checking first:

• Nothing shows at all: confirm the upstream node actually ran and produced output - an API call that errored or a tool that returned empty will leave Show Text with nothing to print, and it'll just look blank rather than throwing an obvious error.
• The node fails to appear in the menu, or fails to import: that's a pack-level install problem, not a Show Text problem - people do report the whole comfyui_LLM_party import failing after a Manager install with no obvious cause, usually tied to a CUDA/PyTorch mismatch with the pack's heavier dependencies. Reinstalling requirements.txt by hand inside the ComfyUI environment (not your system Python) is the first thing to try.