Apply Values to Remote Nodes

Once you're dispatching a whole workflow to a remote machine - via RemoteQueueSimple-Nux or RemoteChainStart-Nux - you'll eventually want that remote run to differ from your local one in some small way: a different checkpoint, a different step count, a different seed. Rebuilding a second workflow file just to change one value is annoying. This node lets you skip that: tell it which node, which parameter, and what value, and it packages that override to be applied to the remote graph right before the job is dispatched.

Inputs and outputs

Four fields, all typed in directly:

• nodeid - the target node's internal ID inside the remote workflow, as a string.
• param - the parameter name on that node you want to change, e.g. seed, steps, ckpt_name.
• value - what to set it to, typed as plain text regardless of the underlying field's real type.
• type - defaults to "STRING", and here it's a free-typed field rather than a locked dropdown (its sibling, RemoteApplyValues-Nux, restricts this to a STRING/INT/FLOAT/BOOL selector instead).

One output: remote_apply, a REMOTEAPPLY type. Wire it into any one of the ten remoteapply1–remoteapply10 optional slots on RemoteQueueSimple-Nux or RemoteChainStart-Nux, and the override gets applied to the remote graph before that job runs. Ten slots means you can stack up to ten overrides per queue call.

How this differs from its sibling

The pack ships two nodes with the identical display name "Apply Values to Remote Nodes" - this one and RemoteApplyValues-Nux. The fields are nearly the same; the only schema difference visible is that type here is a free string, while the non-Multi version pins it to a dropdown. In practice, use whichever one is already sitting in your workflow - they wire into the same slots and do the same job.

Installing it

Search ComfyUI_NetDist_Plus in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/nux1111/ComfyUI_NetDist_Plus
pip install requests

Restart ComfyUI. No model downloads involved.

Where people get burned

nodeid has to match the actual internal ID in the remote workflow's saved JSON - not necessarily the number you see labeled in your local canvas, especially if the remote workflow has been edited since it was last saved. If your override silently does nothing, open the remote workflow directly and check the node's real ID (dev mode / node properties) rather than assuming.

Because value is always entered as plain text here, getting type wrong - leaving it at STRING for a field the target node actually expects as an INT or FLOAT - is an easy way to have the override fail quietly or throw an error on the remote side, somewhere you may not be watching the console.

This is squarely an advanced-tier node (its category is literally remote/advanced). If you're just doing the basic two-GPU setup from the README, you likely don't need it at all - it's for the moment you want per-batch control over what runs remotely without maintaining a second workflow file by hand.