Float Relay

"Float Relay" is the pass-through node for decimals in this pack's relay family: a FLOAT in, the same FLOAT out, no math performed. It's a graph-organizing tool first and a processing node not at all.

The genuinely useful angle is the same one its integer sibling has: the input is a widget with a default, so the node doubles as a typed constant source. Type 0.65 into the widget and you've got a clean FLOAT you can wire to any node that takes a strength, a weight, a CFG value, whatever. ComfyUI's Primitive node can be flaky about converting between widget types; a relay whose input type is locked to FLOAT never has that problem. If you've ever fought a Primitive that kept feeding a string where a float belonged, you'll appreciate the point immediately.

How it works

One line: return (float_,). The widget holds a float (default 1.0, step 0.01), and that value passes through untouched.

Inputs and output

• float_ (FLOAT, default 1, step 0.01) - feed it or type it.
• FLOAT - unchanged.

When you'd reach for it

Three cases, same as the rest of the family: named anchors for float wires in a crowded graph, a fixed decimal value without Primitive's conversion quirks, and keeping a value pinned while you swap out its upstream source. It's also handy when you're dialing in a parameter by hand and want the current value visible at a glance - a relay parked on the strength wire shows you what's actually flowing.

The honest take

This is a "nice to have" node. Reroute handles most rerouting; this adds typed value display and constant-source convenience on top. Don't install the pack for the relays alone - but if the pack's already in, they cost nothing and occasionally save a headache.

Installing it

It's part of longgui0318/comfyui-common-util:

cd ComfyUI/custom_nodes
git clone https://github.com/longgui0318/comfyui-common-util
cd comfyui-common-util && pip install -r requirements.txt

Restart after, or install via ComfyUI Manager by searching "comfyui-common-util". requirements.txt pins only opencv-python, but the pack imports scipy, scikit-image, and PyWavelets at load - if it fails to load, pip install scipy scikit-image PyWavelets and restart.