Mpi Prompt Display

This pack's prompt system is built around MpiPromptList - a node that assembles prompt options with probability, blocking rules, prefixes and suffixes into an opaque PROMPT_LST structure. And that's the problem: an opaque structure is a black box. When your prompt comes out wrong, you can't just read the wire. MpiDisplayPrompt is the debugger for exactly that situation - it takes a PROMPT_LST and exposes its guts as three plain strings you can read, display, or log.

How it works

One required input, prompt_list (type PROMPT_LST), which you wire straight from any prompt-list-producing node. It walks the list and produces three STRING outputs:

• titles - the option titles that were not processed (no positive prompt attached). This is the list of raw, unselected candidates.
• blocks - the blacklist: every blocked option, joined together. The blocking rules are what keep "anime girl" from ever appearing next to "photorealistic," so seeing this tells you whether your blocklist is doing what you think.
• processed_prompts - the titles that do have a positive prompt attached - i.e. the entries that will actually generate.

Those three outputs are ordinary strings, which means they can feed a Show Text node, a logger, a text saver, or anything else that accepts text. It's a transparent window into a structure that otherwise has none.

Where you'd actually use it

Whenever a prompt list behaves wrong and you're guessing. The classic failure: you expected four options to be live, but the generated prompt only ever uses two. processed_prompts tells you immediately which titles made it through the processing and which got dropped or blocked - usually revealing a blocking rule you forgot you wrote, or a title typo that means the processor never matched it.

It's also a cheap way to verify a prompt-list workflow before you spend generations on it: wire a display node in, read the strings, and confirm the composition is what you intended. The output strings are comma-joined, so they read like a prompt preview rather than a JSON dump - useful, not noise.

Where people get burned

MpiDisplayPrompt is a debugging node - it doesn't process or modify the list, and it has no pass-through output. If you splice it into a chain expecting the prompt list to continue on to the processor, it won't; it's a side branch, not an inline node. Also, remember the distinction: titles with a positive prompt appear under processed_prompts, titles without one under titles. If something you expected in the live set shows up in titles instead, the list entry is missing its positive side - that's your bug.

Install

From the MadPonyInteractive/ComfyUi-MpiNodes pack. ComfyUI Manager → search ComfyUi-MpiNodes (publisher mad-pony-interactive), install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/MadPonyInteractive/ComfyUi-MpiNodes

then restart. No dependencies, no model files. The pack is AGPL-3.0 (≤ 1.2.6 MIT) and is the node library behind the author's Cubric Vision app.

The verdict

If you only use this pack's prompt-list nodes occasionally, you'll install it and forget it - until the day a prompt comes out with a word you never wrote, and then it's the fastest way to find out why. It's the difference between debugging prompt lists by staring at them and debugging them by reading them.