☁️Mini Long Text Error Corrector

Most of comfyui_LLM_party is about wiring together an agent from several nodes - a loader, a persona, a memory node, a tools node, all feeding one central LLM node. This isn't that. It's part of what the pack calls "mini-party": self-contained, single-node LLM utilities that skip all the assembly and just do one job, with the model connection configured right on the node itself instead of pulled from a separate loader.

That's the actual reason this node is worth knowing about even if you're already comfortable building full agent graphs elsewhere in this pack - sometimes you don't want an agent, you want a one-step text cleanup pass, and wiring up a full LLM+memory+persona chain for that is overkill. This node takes a block of text, sends it through an LLM configured to correct it, and hands back the fixed version plus whatever it flagged as wrong. No system prompt to write, no persona to pick - you feed in text and get out corrected text.

Inputs that matter. input_str is the text to correct - required, no default. model_name defaults to gpt-4o-mini, a genuinely sensible default for this kind of task: it's cheap, fast, and more than capable of catching typos and grammar issues in long text without needing a frontier model. The optional base_url and api_key let you point this node at any OpenAI-compatible endpoint directly on the node itself, rather than pulling from config.ini the way the pack's main LLM node typically does - that's the "mini" part of mini-party, self-contained configuration. is_enable is the standard toggle to disable the node without removing it. Three outputs come back: input_text just echoes what you sent in, handy for side-by-side comparison downstream; output_text is the corrected result; and error reports whatever the correction pass flagged or, if the call itself failed, what went wrong.

Installing it. This ships with the full comfyui_LLM_party pack - no separate install. Get it via ComfyUI Manager (search "comfyui_LLM_party") or cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then pip install -r requirements.txt from inside the project folder using your ComfyUI Python environment, and restart.

Troubleshooting. If error comes back with something about the request itself rather than the text, check base_url first - the README is explicit across this whole pack that any OpenAI-format base_url needs to end in /v1/, and a missing trailing slash is a recurring, easy-to-miss source of failed calls. If you left base_url and api_key blank, this node likely falls back to whatever's set in the project's config.ini - worth checking that file is actually configured if calls are failing silently with nothing obviously wrong on the node itself. And separately from this node: real users have reported the entire comfyui_LLM_party pack failing to import right after a fresh install, which is a dependency conflict in the project's large, heavily bundled requirements.txt rather than anything to do with this specific node - the README's own requirements_fixed.txt is the documented fallback worth trying if that happens.