TagPipeOutOne

TagPipeOutOne is the minimal read node for this pack's TAGSET type: one TAGSET in, one key1 field, one value1 string out. Where TagPipeOut unpacks six values at once, this pulls exactly one. It exists because the overwhelmingly common case is "I bundled this set earlier and now I just need the positive prompt back" - and a one-port read is the smallest node that does that cleanly.

Under the hood it's a single dict .get(): look up key1 in the TAGSET, return the value, and return an empty string if the key isn't there. No state, no side effects, and reading doesn't consume anything - you can hang as many TagPipeOutOne nodes on the same TAGSET wire as you like, each pulling a different key.

When you'd reach for it

The usual shape: TagPipeIn packs {"positive": "...", "negative": "...", "lora_tag": "..."} once at the top of your workflow. Downstream, wherever you need the positive prompt, you drop a TagPipeOutOne, type positive into key1, and wire value1 into your text encoder. If that's the only value you need, why would you use the six-port TagPipeOut? You wouldn't - this is the tidy version.

The one thing that bites: the key must match exactly, case included. positive and Positive are different keys, and a typo yields a silent empty string. If you're seeing blanks, check the key spelling against what TagPipeIn actually stored - there's no visual inspection of a TAGSET's contents in the UI, so this is a "remember your keys" system, not a "discover your keys" one.

Install

The same pack as all of these, obviously: ComfyUI Manager search "comfyui_tag_filter", or

cd ComfyUI/custom_nodes
git clone https://github.com/sugarkwork/comfyui_tag_fillter

Restart. No dependencies, no models. It's a companion utility - it does one thing, does it in a line and a half of Python, and fades into the background of a graph, which is exactly what a read node should do.