Separator Node
TLDR: Separator nodes duplicate work or unbatch grouped agents. They do not choose routes. For routing, use Path Control on a Source or Processor with multiple outgoing connections.
What A Separator Does
| Mode | Behavior | Use When |
|---|---|---|
| Duplicate | Create multiple copies of the same agent. | One request triggers parallel review, testing, or notification work. |
| Split | Unbatch a group that came from a Batcher. | A pallet, batch, or bundle is broken back into individual items. |
What A Separator Does Not Do
A Separator is not the place to model 70/30 routing, shortest-queue routing, or conditional branching. Those decisions live on the upstream node's Path Control when that node has multiple outgoing connections.
Common mistake
If you want some agents to go left and some to go right, do not use a Separator. Connect the Source or Processor to both destinations, then configure its Path Control.
Modeling Notes
Duplication increases the amount of downstream work, so it can create bottlenecks quickly. After duplicating agents, watch downstream queues and utilization.