The e2e framework allows users to describe test goals in natural language, where an AI agent executes the necessary steps to reach the desired state. It combines these agent actions with standard locators and assertions to verify results. A key feature is the ability to record agent steps and replay them deterministically without invoking model calls, provided the application UI remains unchanged.
- Write tests using natural language goals instead of brittle selector chains.
- AI agent actions are recorded and replayed deterministically for speed.
- Model inference is skipped during replay unless the app changes.
- Integrates standard assertions with AI-driven interaction steps.
