Input String

There's no cleverness here, and that's the point. get_string is a multiline text box that outputs whatever you typed into it, as a STRING. Every graph that needs a hardcoded chunk of text - a fixed instruction, a few-shot example, a block of reference material you're not pulling from a file - needs something exactly this boring, and this pack ships it as its own node rather than making you improvise with something heavier.

How it works

You type text into the node. It comes out the other side unchanged. That's the entire mechanism - no formatting, no template substitution, no parsing.

The inputs and outputs that matter

• input_string - the only input, a multiline text field. Whatever you type here is exactly what comes out.

Output: string (STRING) - wire it anywhere in the graph that expects text: an LLM node's prompt, a json2text-adjacent converter, a system-prompt slot, whatever needs static text rather than something generated at run time.

How to install it

Via ComfyUI Manager: search comfyui_LLM_party, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/heshengtao/comfyui_LLM_party.git

Then pip install -r requirements.txt from inside the pack folder using ComfyUI's own Python, and restart. This node needs nothing beyond the base install - no API keys, no models - it's pure UI plumbing, so if this is the only reason you installed the pack, you're carrying a lot of dependency weight (local-LLM loaders, GGUF support, TTS/ASR wrappers) for one text box. If you don't need any of that, set fast_installed = True in config.ini per the README to skip the heavier GGUF-related install.

Common issues & troubleshooting

Newlines or special formatting look wrong downstream. Multiline fields pass through exactly what's typed, including line breaks - if a downstream LLM call is choking on formatting, check what the raw string actually contains rather than assuming the node is doing something to it.

You keep re-typing the same boilerplate. If you're pasting the same instructions into get_string across multiple workflows, that's usually a sign you want a persona node instead (this pack ships several, like flux_persona and ic_lora_persona) - those bake a template in once rather than making you retype it per graph.

It's not doing anything "smart" and you expected it to. That's by design - if you need variable substitution, JSON parsing, or file loading instead of a literal string, this isn't the node; look at the pack's converter or iterator nodes for that.

Why this exists as its own node at all. ComfyUI already has generic text-input widgets in other packs, but a dedicated one inside comfyui_LLM_party means it shows up in the pack's own node menu next to everything else you're wiring an LLM graph out of, instead of making you hunt through an unrelated pack for a plain text box. Small, but it's the kind of friction that adds up when you're assembling a graph from a dozen different node categories at once - worth using this one over an equivalent from elsewhere purely to keep your node search results predictable.