Manage workflows

A NL Flow workflow, once published, is a program that, when run, receives JSON input, typically—but not necessarily!—processes it and always produces a JSON output.

One of the simplest and most used workflow can analyze text written in a given language using a predictive model: the input is a JSON containing the text to analyze and the output is a JSON containing the result of the analysis.
NL Flow, however, allows building much more complex workflows, as mentioned in the introduction to this manual, and nothing prevents you from building workflows that do not use predictive models at all, if you deem it useful.

NL Flow users can build as many workflows as they need using the graphical editor, but until a workflow is published in a runtime, it is only a blueprint, like the source code of a program.
With publication, the workflow becomes a program loaded in the memory of a computer and ready to run on demand, that is, when an input to be processed is submitted to it via API or test UI.
When a workflow is not needed anymore, it can be unpublished, so freeing computing resources for other programs.

To manage workflows, use the Workflows view of the dashboard. However, some workflow operations are also possible in other views.
The following articles describe all the workflow management procedures.