Pipe Out Smart Folder

What it is

Every node in ComfyUI_Eclipse's pipe ecosystem is built on the same bargain: bundle a dozen settings into one dict, push them down a single wire, and pull them out only where you need them. Pipe Out Smart Folder is the pull-out end of that deal. You feed it the pipe that the Smart Folder node produces, and it hands you back the individual values as ordinary sockets - the output path, the dimensions, the video timing stuff, the seed.

If you've ever opened a shared Eclipse workflow and seen a node with twelve outputs spilling off the side, this is probably it. It exists so you don't have to dig through the pipe dict by hand, and so the rest of your graph stays built from plain old wires you can read.

How it works

The source is refreshingly boring. It takes the incoming pipe (it insists on a dict-style pipe - the legacy tuple format was deleted in v4.0.0, so feed it a modern one), reads a set of known keys, and pushes each out as a typed output. The neat trick is the latent output: if the pipe carries a width, height, and batch size, the node builds a fresh empty latent at exactly those dimensions on the spot. That means you can wire the latent straight from this node into a sampler and skip the Empty Latent node entirely - one less place for the resolution to drift out of sync.

The inputs and outputs that matter

Only one required input, and it's the only one you'll ever touch:

• pipe - the output of a Smart Folder node, carrying the generation mode (image or video) and every parameter that mode uses.

Outputs worth knowing:

• path - the resolved output folder; falls back to ComfyUI's output directory if the pipe has none.
• width, height, batch_size - the generation dimensions.
• latent - the auto-generated empty latent at that size.
• frame_rate, frame_load_cap, context_length, overlap, skip_first_frames, select_every_nth - the video-mode parameters, for wiring into a video sampler or loader.
• seed, loop_count - the random seed and how many times a loop should iterate.

The outputs you don't use just stay unconnected. That's fine; the node only computes the fields that are present in the pipe.

When you'd actually use it

The honest use case: Smart Folder is a configuration node, and config nodes are only useful if something downstream reads them. If you want to keep the path, seed, and dimensions defined in one place and shared across a whole workflow, Pipe Out Smart Folder is how the rest of the graph reads them. It pairs naturally with the other Pipe Out nodes and the Pipe 12/24/36CH buses - pull what you need, leave the pipe intact for the next consumer.

Installing it

It ships with the whole ComfyUI_Eclipse pack. ComfyUI Manager (search "Eclipse") handles it, or:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI_Eclipse
pip install -r ComfyUI_Eclipse/requirements.txt

then restart ComfyUI. One gotcha worth knowing before you get bitten: this pack used to be RvTools_v2, and v4.0.0 removed all legacy nodes. Old workflows from that era won't load the same way - the pack ships a Workflow Migration Tool node and a python tools/migrate_workflow.py CLI to upgrade them automatically.

Common issues

The classic failure is feeding it a non-dict pipe (an old-style tuple) and watching it throw. If you see a "expects dict-style pipes only" error, the pipe you're feeding is from a legacy node - migrate or replace it. Other than that, this node is pure extraction; there's very little to go wrong. If the latent comes out at the wrong size, check that width, height, and batch_size are all present in the pipe, because the node only generates the latent when all three exist.