NO8D-Prompt-view

Here's a scenario you've lived: an LLM node like NO8D-Prompt just spat out an expanded prompt, and you have no idea what it actually said before it got encoded into conditioning. Or you want to hand-edit that generated prompt - trim a clause, add a style - before it reaches the sampler. NO8D-Prompt-view is the inspection window for exactly that. It sits in the text path, shows you what's flowing through, lets you edit it, and forwards the result as a STRING.

The node is designed around the NO8D-Prompt flow, but it accepts any text input, so you can drop it anywhere you want to see and control a prompt string in transit.

What you set

• auto_output - default on. When it's on, incoming text passes through automatically and the node shows you what's flowing. Flip it to edit mode and the node holds output while you work, without clearing the editor content - this is the "temporarily stop automatic output" feature. Very handy when you're mid-iteration and don't want a half-finished graph firing.
• fixed_text - a prefix that gets prepended whenever the node emits text. Put your locked-in style tokens here, the part you never want the LLM to overwrite.
• edited_text - your manual version. When you edit in the node's box, this is what gets sent instead of (or on top of) the incoming text, depending on mode.
• send_seq - a hidden-feeling string field (default "0"). It's the one-click "send now" trigger: set it to a positive number and the node emits immediately. You rarely touch this directly; it's how the frontend's Send button works.

Output is positive - a single STRING you wire into your CLIP encoder. It's an output node, so it always runs and works backward from there in the execution graph.

How the modes actually behave

The logic, from the source: in auto mode the output is fixed_text + incoming text (falling back to your edited text if nothing's coming in). In send mode it's fixed_text + edited text. In edit mode the node emits an empty string and shows you the current composition for review - that's the "stop output without losing my edits" state. The subtle bit is that in auto mode with nothing incoming, your edited_text becomes the output, which is how you use this as a plain prompt editor with a fixed prefix.

Why it's worth it

An LLM prompt node is only useful if you trust it, and trust is hard to build blind. Prompt-view closes the loop: run once, read what the LLM actually produced, fix it in place, and you're not at the mercy of whatever the model decided "detailed" meant. It's a small node with one job, and it does it without any model or API of its own.

Install

Same as every node in this pack:

cd ComfyUI/custom_nodes
git clone https://github.com/no8d/ComfyUI-NO8D-controls.git
python -m pip install -r requirements.txt

ComfyUI Manager → search "NO8D-controls" → install → restart → hard-refresh the browser (the editing UI is frontend JS, and a stale tab is the classic "the node shows but the editor won't open" failure). One honest gotcha: because auto_output gates execution, a workflow that looks "stuck" is often this node parked in edit mode - flip the toggle back on and output resumes.