Skip to main content

Common Parallel Step Details

29/04/2026

Although each parallel step can be quite different, they all share some common characteristics:

  • Naming “Step - [short desc] - Async”

  • Workflow trigger – a webhook with the following properties:

    • Respond - Immediately

    • Path - internal/[desc] - internal; prefix is required

  • The next node after the webhook must be a prepare input data code node:

image6.png

This node is used to extract input data for the workflow should the Error - Async workflow throw an error.

Please check the stepTag JSON property in prepare input data. It will be used to generate the appropriate message.

  • The third node is usually (although not required to be) the call to the Configuration workflow to get some connection configurations.

  • At the end of the workflow there are two paths, success and failure:

Each path should write the results to PARALLEL_DATA_SYNC data table.

In general, each such step should follow the pattern shown here: the triggering of the webhook leads to the preparation of the input data and the configuration; then, the logic specific to your workflow is executed — for example, connection to external services, the use of AI agents, and the execution of specialized scripts. Once these are done, final cleanup consists of preparing and then storing both output and error data.

As input, the concept expansion sub-workflow takes a text string of user input, along with an optional code indicating the language of the text string. It then extracts relevant concepts and expands them to provide a semantic context for downstream processing such as semantic search and LLM context enhancement. The workflow then returns a JSON object with seed concepts, expanded concepts, and concept URIs in an LLM-friendly format.

The expansion query is currently statically set in the node Build expansion request. It will be made configurable in the future.

In this section: