Automates Android application testing by combining UI accessibility trees with multi-modal vision models for human-like interaction and context understanding.
Sponsored
Vis is an innovative autonomous testing agent designed to revolutionize Android app quality assurance. It leverages a unique blend of UIAutomator accessibility trees and advanced multi-modal vision models to interpret screen content semantically, enabling it to interact with apps as a human would. This approach eliminates the reliance on brittle XPath selectors, making tests more robust and self-healing. By utilizing local models via Ollama, Vis ensures data privacy and sovereignty. Its optimized Go core facilitates fast, async screen capture and streaming, offering capabilities like launching apps by human-readable names, performing complex device interactions, generating detailed reports, and orchestrating tests across multiple devices.
主な機能
01Semantic understanding of UI elements
02Self-healing test execution with visual fallbacks
03Local vision model inference via Ollama for data privacy