Search

Leveraging Agent Actions

The Agent Actions feature is currently in private beta. If you are interested in participating, please reach out to Customer Success to determine whether your use case is a good fit!

While Functionize is a no-code / low-code platform by design, sometimes you may run into a use case that isn’t already supported with an out-of-the-box test action. When you find yourself in that situation there are a few options:

  • Test manually: Just do those specific tests by hand. (Not recommended.)
  • Write custom code: Script out what you need to do using JavaScript.
  • Use Agent Actions: Describe the steps you need performed as a Natural Language Processing (NLP) prompt.

Whenever possible, we recommend using the built-in test action capabilities. These will be more consistent and reliable than any feature dependent on NLP. Choose Agent Actions only when:

  • There is no built-in test action that meets the requirements of the use case,
  • Custom code would be too long or complex to write, and
  • The test case is not a one-off that can be performed manually.

Using Agent Actions

Agent Actions can be added to any existing test case, regardless of whether it was created using Architect or Create Agent. To add an Agent Action:

  • Navigate to the Test Case Details page.
  • Click the + button where you wish to add the Agent Action in the test case.
  • Click Agent.
  • Provide your instruction.
  • Click Insert.

Next time you execute your test case, the instructions for the Agent Action will be processed and performed in sequence.

Additional Notes

  • Agent Actions are designed for highly complex use cases that would be otherwise impossible to accomplish with Functionize.
  • Agent Actions rely on LLMs for NLP, and the LLMs may not process the instructions the same way during every test case execution, so results may vary between test case executions without any other changes.
  • Unlike instructions processed by Create Agent, Agent Actions are not broken out into separate test steps.
  • Agent Actions are able to set variables for use in other test steps.
  • For best results, provide the smallest set of instructions possible per agent action.