🎈GroupExecutorSender

Everything upstream of this node is just paperwork. GroupExecutorSender is where the group-execution system in Comfyui-LG_GroupExecutor actually does something: it takes the SIGNAL list built by GroupExecutorSingle (and optionally expanded by GroupExecutorRepeater) and triggers the run. It's an output node - it produces no data, it produces execution.

If you've used other workflow-organizing packs you know the shape: rgthree gives you group muting and bypassing, Impact Pack gives you detect-refine loops. This pack's bet is different - it treats a ComfyUI Group as a callable unit and lets you sequence and repeat those units. The Sender is the terminal that commits to that plan. That's a niche power feature, and it's worth saying plainly: this is a small, Chinese-audience pack with zero English reddit footprint, so don't expect a crowd of tutorials. The mechanism is simple enough to reason about on your own, though.

How it works

The Sender receives the signal (a list of execution items), then does one of two things depending on execution_mode:

• 前端执行 (frontend execution) - the node pings the browser, which matches each group_name against the titles of Groups in the loaded graph, builds an API prompt containing only those groups' nodes, and queues the runs one after another. Your cursor can sit in the UI and watch each group fire in sequence.
• 后台执行 (backend execution, the default) - the frontend hands the whole plan to a background thread on the server via a /group_executor/execute_backend route. The UI stays responsive, and there's a queue manager so a second trigger while one is running gets rejected rather than double-firing. Interrupt handling is wired in too: hit cancel and it marks in-flight background tasks for cancellation.

Either way, the work is "grab these groups from the current graph, run them in this order, this many times each." The browser is doing the prompt-building, so this is not a headless/API-only feature.

Inputs

Just two:

• signal (SIGNAL, required) - the execution list from GroupExecutorSingle/Repeater. No signal, no run; the node raises "没有收到执行信号" ("no execution signal received") in the log.
• execution_mode - frontend vs backend. Frontend for small plans where you want to watch; backend for longer batch runs so the UI doesn't lock up.

No outputs. It's the end of the line.

Install

ComfyUI Manager → search "Comfyui-LG_GroupExecutor", or:

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

Restart ComfyUI afterward. No extra dependencies, no model files.

Where people get tripped up

• Nothing happens on trigger. Almost always a group_name mismatch upstream - the Sender can't find the group by title and quietly does nothing useful. Double-check the group titles in the canvas.
• "已有任务在执行中" (a task is already running). Backend mode returns a 409 rather than stacking runs. If you're used to ComfyUI queueing everything, this one refuses to queue a second plan until the first finishes.
• Works only in the browser. The Sender is useless from a pure API call; the frontend JS is doing the orchestration.