💀Prepack GetPipe

💀Prepack GetPipe is the receiving end of the Prepack pipe pattern. Its sibling SetPipe takes a dozen components and folds them into a single pipe object; GetPipe takes that one wire and unfolds it back into the twelve original pieces on the other side of your graph. You use these two as a pair to stop a wall of parallel wires from crossing the canvas - bundle at one end, unbundle at the other, and everything between stays tidy.

It's the same trick the broader ecosystem uses in its efficiency/pipe packs, and it pays off most when you're routing one full "generation configuration" through a section of the graph where you don't care about the individual pieces - say, past a bypass switch or a comparison branch. GetPipe is where that bundled state becomes usable again.

How it works

The mechanism is about as simple as a node gets: it takes the pipe input, reads the twelve keys off the dict that SetPipe produced, and returns them in a fixed order as twelve typed outputs:

• model (MODEL), clip (CLIP), vae (VAE)
• lora_path (STRING), lora_text (STRING) - the LoRA metadata slots
• positive (CONDITIONING), negative (CONDITIONING)
• latent_image (LATENT)
• seed (INT), steps (INT), cfg (FLOAT), denoise (FLOAT)

Every one of those outputs is explicitly allowed to be None. SetPipe defaults every input to None too, so a pipe with only a model and a seed packed is perfectly valid - GetPipe will hand back a None for all the empty slots. That's a feature when you're building partial pipes, and a footgun when you forget to pack something and don't realize it until a downstream node chokes.

The only input that matters is pipe, and it must come from a Prepack SetPipe (or anything else that emits the same dict shape - but don't count on cross-pack compatibility; pipe shapes are not standardized).

Where it fits in a real workflow

The typical shape: SetPipe collects model, clip, vae, conditioning, and your sampling settings from the loaders; you drag that one wire through whatever organizational nodes you like; GetPipe at the sampler end splits it back out into the inputs of a KSampler (or this pack's own 💀Prepack Ksampler). Because the pipe carries steps, cfg, seed, and denoise as first-class fields, you can also pack a new pipe with different sampling values and swap it in without rewiring the whole sampler - the "settings presets" use case, done with a single wire change.

One thing to keep straight: GetPipe gives you the raw values. The seed, steps, and cfg outputs are plain numbers; if you want them back inside a packed pipe for another leg of the workflow, you route them into another SetPipe rather than assuming they re-pack themselves.

Installing it

It's in the ComfyUI Prepack pack - install the pack once and you get all eighteen nodes, including SetPipe and GetPipe together:

cd ComfyUI/custom_nodes
git clone https://github.com/S4MUEL-404/ComfyUI-Prepack.git
pip install -r ComfyUI-Prepack/requirements.txt

Or search "Prepack" in ComfyUI Manager. Dependencies are just PyTorch, NumPy, and Pillow - you already have them. Restart, look under 💀Prepack, no model downloads required.

If GetPipe errors on load with "pipe is None", you've wired it to the wrong node - check that the pipe wire really originates from a Prepack SetPipe and that the workflow isn't feeding it something else. If an output comes back None mid-run, the fix is upstream: the field simply wasn't packed.