Execution Logic: Learn & Run

tapioka.ai operates in two distinct phases to ensure test reliability and accuracy: Learn and Run.

1. The Learn Phase

In this phase, the AI explores the application to determine how to fulfill your instructions.

• Exploration: The AI performs actions like tap, swipe, long-press, and text input to navigate the app.
• Visual Mapping: At every step, the system captures a screenshot of the current state.
• Validation: Once the instructions are completed, the AI verifies if the final state matches your defined Expected Result.
• Path Persistence: If the final state is correct, the successful path (sequence of actions and screenshots) is saved for future automated execution.

Result: A verified path with visual benchmarks for every step.

2. The Run Phase

This is the automated execution of a previously "learned" test case.

• Repetition: The AI selects the saved path and attempts to repeat the exact sequence of actions.
• Anomaly Detection: After each action, the system captures a new screenshot and compares it to the benchmark from the Learn phase. It checks for:
  • System errors (crashes, overlays).
  • Significant UI discrepancies.
  • Unexpected behavior.
• Final Verification: After the full path is executed, the AI performs a final check of the Expected Result.

Result: A pass/fail status, a list of steps with potential anomalies, and confirmation of requirements fulfillment.