Zap the Flakes! Leveraging AI to Combat Flaky Tests with CANNIER

Explore how to combat flaky tests using AI through the CANNIER approach in this 34-minute conference talk from DevConf.CZ 2025. Learn about the persistent challenge of non-deterministic tests that intermittently fail and pass, particularly in end-to-end testing environments where multiple components interact. Discover how flaky tests can overload CI systems, create lengthy feedback cycles, and erode developer trust in continuous integration processes. Examine the CANNIER methodology designed to reduce test rerun time costs by predicting test flakiness before tests enter the end-to-end suite. Understand how feature sets extracted from existing test code in the KubeVirt project generate training data for AI models that can predict flakiness probability. Gain insights into implementing CANNIER features, generating training data for machine learning models, and analyzing flakiness probability to provide actionable feedback for test authors. Master strategies for improving test reliability, enhancing software quality, and boosting team productivity by identifying potential flakes early in the development lifecycle.