🔱 Artha Flow Pause

Artha Flow Pause is the rare custom node with zero magic in it: it passes whatever you feed it straight through, and it gives you a button to freeze the whole workflow mid-run. That's the entire job, and it's more useful than it sounds. ComfyUI runs a graph top to bottom and doesn't stop for anyone - you can't park a queue at an expensive KSampler while you go squint at an intermediate preview. Slip this node in before that sampler, hit pause, and the run parks there until you say go.

It's part of the Artha pack's Flow category, which is the pack's "don't overthink it" corner: no API key, no Gemini dependency, nothing to configure beyond a toggle. The one thing you set up is the flow itself.

How it works

The schema is deliberately boring:

• in_slot - an * input, so it accepts any data type. The tooltip says it plainly: "Pass thru in slot."
• out - an * output, "Pass thru out slot."

Execute it and in_slot comes out the other side untouched. The pause is a global state that lives in the node - when it's set, execute sits in a loop (sleeping 0.1s at a time) until you unpause from the button that the pack's front-end adds to the node, or until you hit ComfyUI's stop button, which breaks it out of the loop cleanly and resets the state so your next run isn't stuck. There's a small web endpoint behind it (/artha/flow_toggle_pause_button) but you never touch that directly.

One implementation detail worth knowing: the node returns NaN for its input fingerprint, which stops ComfyUI from caching on it. That's deliberate - a caching pass-through would defeat the point, since you want execution to genuinely stop there, not pretend to.

Why you'd bother

Three genuinely good uses, in my order of preference:

1. Multi-stage pipelines. Chain a cheap first pass, pause, inspect the result, then run the expensive second stage once you're happy. Saves you from burning GPU time on a composition you're going to reject anyway.
2. Debugging. You're fairly sure a node upstream is producing garbage but not sure which. Park the graph right after it and actually look at what's flowing through.
3. Staging batch runs. Pause between items in a queue so you can sanity-check one output before letting the rest fly.

It's not for everyone. If your workflows are single-shot and short, a pause button is a solution looking for a problem. But if you've ever sat watching a 20-minute generation you knew was already off, you know exactly why this node exists.

Install

It ships with the Artha pack, so you get it for free with the rest. ComfyUI Manager, search "Artha," install, restart. Or the manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/Cyrostar/ComfyUI-Artha-Nodes
cd ComfyUI-Artha-Nodes
pip install -r requirements.txt

The only gotcha is temperamental rather than technical: if your paused workflow sits there and seems frozen forever, check the node's button state - the global pause flag survives between runs, so a forgotten pause will happily stall your next queue too. It's not hung; it's waiting on you.