Fan-out pipeline (4 nodes)

Run: npm run run:fan-out
File: examples/fan-out-pipeline.json

One source → three parallel leaves.

After source_1 completes, three leaves run in parallel:

Step	What happens
1	source_1 runs first
2	out_a, out_b, out_c run in parallel
3	Each leaf gets the same data.sourceData from source_1

Data flow (what you should see)

The engine sets each leaf’s data.sourceData to the source output before nodeExecutor runs. Demo trace only — src/graph-engine.ts is unchanged.

npm run run:fan-out

Example trace:

  out_a (pipeNode, Branch A):
    receivedFromUpstream (engine sourceData): ["shared upstream payload"]
    emitted (handler return): { "value": ["shared upstream payload"], "label": "Branch A" }

All three leaves show the same upstream string in sourceData.

Payload

{
  "nodes": [
    {
      "id": "source_1",
      "type": "text",
      "position": { "x": 0, "y": 120 },
      "data": { "label": "Source", "nodeData": "shared upstream payload" }
    },
    {
      "id": "out_a",
      "type": "pipeNode",
      "position": { "x": 320, "y": 0 },
      "data": { "label": "Branch A" }
    },
    {
      "id": "out_b",
      "type": "pipeNode",
      "position": { "x": 320, "y": 120 },
      "data": { "label": "Branch B" }
    },
    {
      "id": "out_c",
      "type": "pipeNode",
      "position": { "x": 320, "y": 240 },
      "data": { "label": "Branch C" }
    }
  ],
  "edges": [
    { "id": "e-src-a", "source": "source_1", "target": "out_a" },
    { "id": "e-src-b", "source": "source_1", "target": "out_b" },
    { "id": "e-src-c", "source": "source_1", "target": "out_c" }
  ]
}

Fan-in vs fan-out

Pattern	Shape
Fan-in	Many → one → leaf
Fan-out	One → many leaves

When to use

• One fetch → summarize + translate + tag in parallel
• Broadcast one LLM result to multiple export paths

Fan-in · Docs index