runNode

workflow.runNode() executes a normal node. It is responsible for interrupt checking, node start/end hooks, execution information logging, and error wrapping.

Signature

workflow.runNode<Input, Output>(
  node: NodeDefinition<Input, Output>,
  input: Input,
  context: NodeContext,
): Promise<Output>

Parameter

Parameter	Type	Description
node	NodeDefinition<Input, Output>	The node object to be executed, usually from a node factory.
input	Input	Node input.
context	NodeContext	runtime context for hooks, provider, KV, files, abortSignal.

NodeDefinition

Field	Description
name	The node name. Required.
standardName	Standard node type name, such as input, llm, output.
metadata	Structure diagram and UI meta information.
run	The node executes the function, receiving input and NodeContext. Required.

Execution process

Steps	Description
1. Check for interruptions	If context.abortSignal is interrupted, throw execution_aborted.
2. Send Start Event	The trigger node starts a hook and records the executionId, nodeName, metadata, and startedAt.
3. Execution node	Call node.run(input, context).
4. Send Completion Event	Record the output, duration, and executionInfo, and trigger the completion of the hook.
5. Wrong packaging	Convert to HarnessError on failure and write report.

Return value

Return	Type	Description
Node output	Promise<Output>	Returns the result of node.run on success.

Error

Situation	Error Type	Description
Break before or during execution	execution_aborted	From context.abortSignal.
Node throw error	node_execution	Unknown errors are wrapped as HarnessError; existing HarnessError is preserved and the context is supplemented.

Example

const parsed = await workflow.runNode(inputNode, input, context);
const answer = await workflow.runNode(llmNode, parsed, context);
const output = await workflow.runNode(outputNode, answer, context);

The difference from runStreamNode

Method	Node type	Return
runNode	NodeDefinition	Single Output.
runStreamNode	StreamNodeDefinition	{ chunks, output }.