MiniMax H3 Rule Library (10)

This is the pack's smallest node and also the most underrated. It doesn't touch a model, doesn't load anything, and does no inference at all - it's ten editable text slots with a single active_index switch, and its whole job is to output whichever rule you've selected. That sounds like nothing until you've used the Two-Stage node and found yourself editing a giant system-prompt text box every time you want a different writing style.

If you've written prompt instructions for LLMs, you know the pattern: "you are a professional action-scene prompt writer," "you are a gentle cinematic storyteller," "describe in the style of a 1990s anime." The Rule Library is a place to store those as named, one-click options instead of retyping them. It pairs with the pack's MiniMax H3 Prompt (Qwen) Two-Stage node - its selected_rule output plugs straight into Two-Stage's rule_text input, and that's the intended workflow.

How it works

Nothing clever, which is the point. The node holds rule_1 through rule_10 as multiline strings, and active_index (1–10) picks which one comes out of the single output. Internally it clamps the index to the valid range and returns the matching string. No model loading, no API, no VRAM - it's pure workflow plumbing, and it's instant to run.

The inputs

Just two things to know:

• active_index - which rule (1 to 10) is live. This is the input you'll actually interact with, and you can drive it with a seed node or leave it manual.
• rule_1 … rule_10 - the rules themselves. Each is a multiline text box, so they can be long-form system-prompt-style instructions, not just one-liners.

One output: selected_rule, a plain string carrying the active rule.

Installing it

Shared with the rest of the pack - if you have any of the other four nodes, this one's already there:

cd ComfyUI/custom_nodes
git clone https://github.com/wangminxing2019/ComfyUI-MiniMaxH3-QwenPrompt.git
git clone https://github.com/JamePeng/ComfyUI-llama-cpp_vlm.git

Restart ComfyUI and it's under "MiniMax H3" in the node browser. It needs no model files of its own - nothing in models/LLM required for this node to work, though the Two-Stage node it feeds certainly does.

Gotchas worth knowing

• The slots start empty - there's no default text baked in, so plugging the library into Two-Stage with blank rules falls back to Two-Stage's own built-in default rule. Fill at least one slot before you rely on it.
• The index is 1-based - active_index runs 1–10, matching the rule_1…rule_10 naming, which is just unusual enough to trip people up after years of 0-based indexing everywhere else in ComfyUI.
• Rules are per-workflow state - they live in the node, so save the workflow if you want them to survive. There's no import/export, so a big rule library is worth pasting into a text file as a backup.