Overview

In Kognitos, you can call a subprocess using the start a run syntax. This method allows the parent process to manage the subprocess, including explicitly waiting for it, retrieving the status of the run, and fetching its output. When multiple instances of a subprocess are called this way, they will run in parallel.

            Parent Process
                  │
     ┌------------┴------------┐
     │            │            │
     │            │            │
Subprocess    Subprocess    Subprocess

How Can I Start Parallel Runs?

This section describes how to start parallel runs using the start a run syntax.

📘

Preface Your Automation with learn "internal.run"

To use this functionality, include learn learn "internal.run" at the beginning of your automation.

Syntax

process each {item} as follows:
    start a run where
        the procedure is "{subprocess name}"
        {input}

Parameters

1. item: Object to loop over.
2. subprocess name: The name of the subprocess.
3. input: Objects to pass to the subprocess, if necessary.

Example

the items are "apples", "flour", "eggs", "milk"
process each item as follows
    start a run where
        the procedure is "to make apple pie"
        the ingredient is the item

Click here to walk through an end-to-end example.

🚧

Remember to Publish Your Process

You need to promote your automation from the Playground to a Process in order to start parallel runs. This applies to both the main parent process and the subprocess. Otherwise, you will get an Assertion Error saying procedure not found.

Handling Subprocess Runs

A parent process can explicitly wait for the runs of a subprocess. By waiting for completion, a parent process can also collect the data and outputs from all the subprocess runs.

Adding an Explicit Wait

To add an explicit wait, include the following line into your automation after starting parallel runs:

wait for the runs

Once this is executed, a tab will appear in the user interface, allowing you to view the status of each individual run.

Customizing the Wait Time

You can customize the wait time by adjusting the wait count before starting parallel runs.

• The wait count refers to the number of 5-second intervals.
• The wait time is calculated by multiplying the wait count by 5.
• The default wait count is 60.
• The default wait time is 5 minutes (60 x 5 = 300 seconds, or 5 minutes).

the department's max retry wait count is {intervals}

In the example below, setting intervals to 3 results in a total wait time of 15 seconds (3 x 5 seconds).

the department's max retry wait count is 3

Getting Output From Parallel Runs

To get results from subprocesses that are called with the start a run syntax, you must add an explicit get into the parent process after adding a wait.

Syntax

get the runs's {thing}

Here, the thing is the output you would like to retrieve.

Note: Currently, only one output can be retrieved at a time using this syntax.

Example

get the runs's greeting

🚧

Note: Matching Parent and Subprocess Output

The get call in the parent process must match the name of the data element used in the subprocess when retrieving output.

• Example: If the subprocess data is named status, the parent process should retrieve it using get the runs's status.
• Automatic Pluralization: If there are multiple runs, the system will automatically pluralize the output in the parent process (e.g., statuses).

Walkthrough: Run Parallelization

This is an example walkthrough of invoking multiple instances of a subprocess in parallel runs.

Parent Process

learn "internal.run"
the department's max retry wait count is 2
the names are "Eva", "Ivan", "Leo", "Nia", "Zoe"
process each name as follows
	start a run where
		the procedure is "to greet the user"
		the user is the name 
wait for the runs
get the runs's greeting

Subprocess

Subprocess Output

In line 9, the greetings is referenced as output from the subprocess runs. The results are:

• Hello Eva
• Hello Ian
• Hello Leo
• Hello Nia
• Hello Zoe

Additionally, each individual run can be clicked through in the platform: