Executing Tests
Last updated
Last updated
Important: Remember to save your test prompt frequently while editing. Exiting the test editor without saving will result in the loss of any recent changes.
Step 1: Initiate the test execution by selecting Execute prompt to start GPT Driver.
Step 2: Allow GPT Driver to work its magic. The tool will analyze the screen with a 'Thinking...' text displayed and execute the relevant action command accordingly.
Step 3: While the steps are being executed, scroll down to the execution log to observe how GPT Driver is interacting with the app. Each interaction is represented by a yellow circle with a red dot in the middle. The execution log also provides details about the interaction and the reasoning behind it. This information becomes crucial if the expected behavior of GPT Driver deviates, prompting the need for editing the test prompt.
Step 4: Once GPT Driver has completed the steps, you will see either a "Succeeded" or "Failed" message along with a reason for the status. Feel free to review the execution logs to verify that the intended actions are occurring as expected.
Much like the Recorded Actions, you'll find the Execution Log now visible within the Prompt Editor. This log serves a comparable purpose, but with a detailed tracking of every interaction the GPT Driver has with the app, along with its underlying reasoning throughout the Test Prompt steps. Utilize the execution log to follow each step closely and pinpoint areas where enhancements to the prompt may be required
Once your test has succeeded in the a sufficient number of times (we recommend x3 as a minimum threshold), it's ready to be executed from the . However, if the test prompt still encounters failures, consider further revision within the to increase its accuracy.
Anything unclear? Check out our section for additional guidance. Alternatively feel free to reach out to us directly via Slack or Email.
