Flake Data Split (Select)

FlakeDataSplitSelect does exactly what FlakeDataSplitAll does - it takes a flake_data bundle and unpacks it into the individual values hiding inside - with one difference worth having: the output pins are opt-in. Instead of thirteen wires always cluttering your graph, you use a dropdown on the node and a +/- button to add only the pins you need. Need just model, positive, and cfg to feed your own sampler? Add three pins, done. The graph stays readable, which is the entire point of a pack built around hiding complexity behind a bundle wire in the first place.

The full menu is the same thirteen as the "All" variant: model, clip, vae, positive, negative, latent, filename_prefix, width, height, steps, cfg, sampler_name, scheduler. And the behavior is identical - it's the same split code, so filename_prefix comes out as the fully-built prefix (preset name plus output_stems plus the date/time stamp), and the sampler values are plain strings that plug into any KSampler.

When would you actually bother? The realistic scenarios are: you want to drive your own sampler instead of using Flake Generate; you want to pipe one value out of the bundle into a node that only accepts a plain type; or you're debugging - a bundle from a preset-plus-stack chain does a lot of invisible work, and sometimes you need to look at one number to figure out why the render is 832×1216 instead of what you thought you set. For the "keep the graph tidy" version of all of that, this is the node; for a one-off where you don't care about clutter, the All variant is one less click.

Inputs and outputs. Required: flake_data. Outputs: any subset of the thirteen above, added via the dropdown and +/- controls. It's a pure passthrough/split - no generation happens, no values are modified.

Install. ComfyUI Manager → search ComfyUI Flakes → Install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/JeyzerMC/comfyui-flakes

then restart. PyYAML is the only dependency and it already ships with ComfyUI core.

One small thing to know: because the pins are dynamic, the node is a touch more frontend-heavy than its sibling - the add/remove logic lives in the JS widget. On a brand-new pack that's the seam most likely to feel rough when ComfyUI's UI updates under it. If the dropdown ever looks unresponsive, a refresh usually sorts it.