TagPipeIn

TagPipeIn is the entry point to this pack's own little data type, the TAGSET. If you've used rgthree or similar packs' "pipe" nodes, you already know the idea: bundle a bunch of related values into one wire so the graph doesn't turn into spaghetti. This pack's version is about as minimal as it gets - a TAGSET is just a key/value dict of strings, and TagPipeIn lets you build one from up to six keyN / valueN pairs.

You type a key and its value, and the node packs them into a single tagset output: a dictionary like {"positive": "masterpiece, 1girl", "negative": "blurry, lowres"}. Every pair is optional, so you can build a two-entry set or a six-entry set; untyped keys just don't appear. There's no validation, no type checking, no magic - it's a dict literal with a custom type label so ComfyUI treats it as one wire instead of six.

Why bother

The payoff is downstream. Instead of six dangling strings fanning out across the graph, one TAGSET wire carries them all. TagPipeOut (six keys), TagPipeOutOne (one key), and TagPipeUpdate (set one key, get a new set) read and modify it, so you can thread a whole "prompt bundle" through a workflow - into a subgraph, out the other side - without reconnecting half a dozen ports every time you change something. It's the same ergonomic argument as any pipe system, scaled down to strings.

The catch

The TAGSET is only understood by this pack's own nodes. TagPipeIn's output won't plug into some other pack's pipe port, and there's no visual "what's inside" in the node UI - you have to remember what keys you put in. The keys are also case-sensitive and matched literally, so Positive and positive are different entries. Keep your key names consistent and lowercase and you'll save yourself a debug session.

And honestly, for a pack this small, the pipe nodes are the part you might skip. If you're already using a general-purpose pipe solution (rgthree, or Impact Pack), this family is redundant with it. TagPipeIn earns its keep in a self-contained workflow that stays inside this pack's tag-tooling world.

Install

Same as the rest of the pack: 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. The whole pipe family is a few dozen lines of Python - a dict in, a dict out, a .get() to read - and it shows. Simple, predictable, and occasionally exactly the tidying you needed.