Portal Community
Flow Studio

Resuming the Engine

IWorkflowContinuationOrchestrator re-loads the suspended state, merges the actor's response, and restarts execution from the HIL node's downstream. The engine does not re-execute nodes that have already completed.

Continuation Steps

1

Load suspended execution

HILResumeService loads the row from Process_SuspendedExecutions. Deserialises ExecutionMemory from JSON.

2

Mark token as consumed

Sets ResumedAt on the suspended execution row — prevents double-resume before continuation begins.

3

Merge response data

Actor's responseData is written to ExecutionMemory.nodeOutputs[suspendedNodeId].

4

Resolve output port

Based on response (e.g., Approved → approved port), the engine identifies which downstream edges to follow.

5

Continue execution

IWorkflowContinuationOrchestrator.ContinueAsync() resumes the engine from the identified downstream nodes. Normal execution proceeds.

Nodes Not Re-Executed

The engine uses the restored ExecutionMemory.nodeOutputs map to know which nodes have already run. Any node whose output is already in the map is skipped — only nodes downstream of the HIL node (on the selected port) are executed fresh.

IWorkflowContinuationOrchestrator

public interface IWorkflowContinuationOrchestrator
{
    Task ContinueAsync(
        ExecutionMemory     restoredMemory,
        string              processId,
        string              threadId,
        string              startFromNodeId,  // the HIL node
        string              startFromPort,    // the resolved output port
        CancellationToken   ct);
}
← Injecting Resume Data Next: Failure Modes →