Action Flow Control
Parallel Dispatcher
Executes a subgraph once per iteration like the For Loop, but runs instances concurrently with a configurable parallelism cap and per-instance retry. All instances run to completion; successes and failures are reported separately.
Node IDParallelDispatchAction
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
Loop Type loop_type | enum | No | ForEach | ForEach iterates over array items, Range iterates a specified number of times. |
Items (ForEach) items | array_object | No | [] | Array of items to dispatch (used when Loop Type is ForEach). |
Start (Range) start | integer | No | 0 | Starting index for Range mode. |
End (Range) end | integer | No | 10 | Ending index for Range mode (exclusive). |
Step (Range) step | integer | No | 1 | Step increment for Range mode. |
Subgraph Path subgraph_path | file_picker | No | "" | Path to the subgraph file to execute for each instance (e.g., './subgraphs/worker.json'). |
Max Parallel max_parallel | integer | No | 4 | Maximum number of subgraph instances running at the same time (minimum 1). |
Max Retries max_retries | integer | No | 0 | How many times a failed instance is re-run before it is reported as failed. |
Retry Delay (ms) retry_delay_ms | integer | No | 1000 | Fixed delay in milliseconds between retry attempts of a failed instance. |
Collect Outputs collect_outputs | boolean | No | true | If true, collects all instance outputs into an array (input order). If false, only returns the output of the highest-index successful instance. |
Aggregate By Key aggregate_by_key | boolean | No | false | If true, merges outputs by their identifier keys into arrays. Only works when collect_outputs is true. |
Inputs
| Input | Type | Required | Description |
|---|---|---|---|
Execute input0 | object | No | Trigger to start the parallel dispatch. |
Outputs
| Output | Type | Identifier | Description |
|---|---|---|---|
Succeeded output0 | object | succeeded_items | Always fires after a completed run. Carries results (per-instance outputs in input order) and stats (total, succeeded, failed). |
Failed output1 | object | failed_items | Always fires after a completed run. Carries failed_items (index, input, error, attempts) and stats (total, succeeded, failed). |