LLM Completion

This is the no-fuss version of LLM Chat. Instead of building a conversation with roles and turns, you just type a prompt and get text back. For the common case - "take this prompt and expand it," "summarize this," "turn these keywords into a description" - that's all you need, and skipping the message-building step keeps the graph small. If you don't need a system prompt, multi-turn context, or images going in, this is the node to grab.

It's the shortest path from "I want an LLM to write something" to a STRING you can wire into your positive prompt.

Where it shines

The classic use is prompt engineering on autopilot. You feed a short idea into prompt, ask the model to flesh it out into a rich description, and pipe the text output straight into your CLIP text encode. Now every run gets a fuller, more varied prompt without you writing it by hand. Same for cleaning up messy input, translating, or generating captions. Anywhere a text step would help but a full chat is overkill, Completion fits.

How it works

Two of the same three legs as LLM Chat, minus the message-builder. You supply the prompt (plain text), an api handle for the connection, and a config for the model and settings; the node runs the call and returns the result. Under the hood it's still a model request - the "completion" framing just means single-shot rather than conversational.

The inputs and outputs

• prompt (multiline STRING) - what you're asking the model to do or continue.
• api (LLM_API) - the provider connection, from a node like Gemini API or Bedrock Claude.
• config (LLM_CONFIG) - model and settings, from LLM API Config.
• seed (INT) - for reproducibility, with the same caveat as everywhere: LLMs aren't fully deterministic, so a fixed seed isn't a guarantee.
• text (output, STRING) - the model's reply.
• (optional) images (output, IMAGE) - present for image-capable models, though for image work the Chat node with a message and images is the more natural fit.

Installing it

ComfyUI Manager → search comfyui-art-venture → install → restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/sipherxyz/comfyui-art-venture

then restart ComfyUI. Pre-installed on comfy.icu.

Common issues

Missing-node red means the pack isn't installed - Manager → "Install Missing Custom Nodes."

Like the Chat node, the failures usually come from upstream: a bad key or wrong model on the api/config nodes, a stale model ID, or Bedrock model access not enabled for your region. This node is just where the error shows up because it's the one making the call. If output gets cut off mid-sentence, that's the max_token budget on your Config, not a fault here - raise it. And remember it's a paid, networked call: dropping an LLM Completion inside a large image batch means one API request per item, which adds up in both time and tokens. Use it where the text step earns its cost.