⌚Smart Caching

Closed Beta: Smart Caching is currently under development and only available to a selected group of companies on paid plans.

Baseline Creation

After a successful run where GPT Driver interacts with the application, a baseline (essentially a cache) is automatically created. This baseline stores information about the screens encountered during the run, along with the actions GPT Driver took for each screen.

Subsequent Runs

During future test editor and cloud runs, GPT Driver checks two things:

1. Prompt Match: Is the prompt used exactly the same as the previous successful run?

2. Screen Match: Does the current sequence of screens exactly match what was encountered in the baseline?

Action Reuse: If both the prompt and screens match the baseline, GPT Driver efficiently replays the stored actions for each screen. This saves time and ensures consistency in repetitive tasks.

Handling Discrepancies: If there are any differences between the current screens and the baseline (e.g., a change in button text), GPT Driver recognizes the variation and relies on its reasoning capabilities to determine the best course of action.

Minor Screen Differences Don't Affect Cache

To improve efficiency, the cache is tolerant of minor discrepancies in the screen appearance. For example, slight variations in display time on the emulator won't cause GPT Driver to abandon the cached actions.