🎈GroupExecutorRepeater

If the group-execution trio in this pack is a tiny scripting language for your canvas, GroupExecutorRepeater is the loop construct. It sits between your GroupExecutorSingle chain and the GroupExecutorSender, takes the whole SIGNAL list, and repeats it N times - inserting a configurable pause between each pass. "Run these three groups, wait, run them again, wait, run them a third time." That's the whole job, and it does it without you duplicating nodes.

It's a niche pack, so a quick reality check: Comfyui-LG_GroupExecutor is a small utility pack from LAOGOU-666 aimed mostly at Chinese users, with no English reddit presence to speak of. It has no Python dependencies and downloads no models - the repeater is pure orchestration, and the logic lives in plain Python plus a bit of frontend JS. That makes it pleasant to reason about when something misbehaves.

How it works

The Python side is almost embarrassingly simple, which I mean as a compliment. repeat() takes your signal list and repeat_count, then does:

for i in range(repeat_count):
    repeated_list.extend(execution_list)
    if i < repeat_count - 1:
        repeated_list.append({"group_name": "__delay__", "repeat_count": 1, "delay_seconds": group_delay})

Note the trick: the delay isn't a separate timer. It's inserted as a pseudo execution item named __delay__, which the frontend recognizes and sleeps through before the next pass. So group_delay (0–300 seconds) becomes a pause between repeats - never before the first one. That's worth remembering if you're expecting a lead-in pause.

The expanded list then flows to the Sender, which runs each group the requested number of times, in order.

Inputs

• signal (SIGNAL, required) - the execution list from one or more GroupExecutorSingles.
• repeat_count (1–100) - full passes over the whole list.
• group_delay (0–300 seconds) - pause between passes.

Output

One SIGNAL output carrying the repeated list, wired straight into a GroupExecutorSender.

Install

Same as the rest of the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/LAOGOU-666/Comfyui-LG_GroupExecutor

or via ComfyUI Manager (search "Comfyui-LG_GroupExecutor"), then restart. No models, no pip installs.

Gotchas

• The list can blow up fast. repeat_count is capped at 100, but 100 × a long chain is a lot of queued group runs, and each one is a full sub-prompt execution. Start small.
• Delay is between passes only. First pass starts immediately; there's no __delay__ before it.
• It still needs the browser. Like the rest of the group system, the actual execution is orchestrated by frontend JS, so this won't do anything in a headless API-only setup.
• Backend mode rejects overlap. If a plan is already running in backend mode and you trigger again, you'll get a "task already running" refusal rather than a queue.