Middleware API

The Middleware interface allows you to inject cross-cutting concerns into workflow execution. Middleware can wrap node executions, run before or after nodes, and is essential for features like observability, transactions, and logging.

Interface Definition

export interface Middleware<TContext extends Record<string, any> = Record<string, any>> {
	beforeNode?: (ctx: ContextImplementation<TContext>, nodeId: string) => void | Promise<void>
	afterNode?: (
		ctx: ContextImplementation<TContext>,
		nodeId: string,
		result: NodeResult | undefined,
		error: Error | undefined,
	) => void | Promise<void>
	aroundNode?: (
		ctx: ContextImplementation<TContext>,
		nodeId: string,
		next: () => Promise<NodeResult>,
	) => Promise<NodeResult>
}

Hooks

beforeNode

Runs before a node executes. Useful for setup or logging.

• Parameters:
  • ctx: The workflow context
  • nodeId: The ID of the node about to execute

afterNode

Runs after a node executes, regardless of success or failure.

• Parameters:
  • ctx: The workflow context
  • nodeId: The ID of the executed node
  • result: The result of the node execution (undefined if error)
  • error: The error thrown (undefined if success)

aroundNode

Wraps the entire node execution. This is the most powerful hook as it can control the execution flow.

• Parameters:
  • ctx: The workflow context
  • nodeId: The ID of the node being executed
  • next: Function to call to proceed with execution
• Returns: The NodeResult from the node or modified result

Usage with FlowRuntime

Middleware is provided to the FlowRuntime constructor:

const runtime = new FlowRuntime({
  middleware: [myMiddleware1, myMiddleware2],
})

Middleware executes in LIFO order (last in, first out).

Built-in Middleware

• @flowcraft/opentelemetry-middleware: Provides distributed tracing using OpenTelemetry standards.

For examples, see the Middleware Guide.