🤝 Top Collaborators
🤓 Latest Submissions
Git Actions Flaky Agent Fixer
In the modern software engineering ecosystem, Continuous Integration and Continuous Deployment (CI/CD) pipelines serve as the central nervous system for product delivery, providing an immediate signal regarding the stability of newly committed code. However, this critical feedback loop is increasingly compromised by the proliferation of "flaky tests" automated software tests that are inconsistent failure which can be caused by: Asynchronous Waits (Async Wait), Concurrency and Race Conditions, Test Order Dependency (TOD), Unordered Collections & Environmental Constraints. The Autonomous CI Flaky Test Debugging Agent is an agentic workflow integrated natively into the CI pipeline and powered exclusively by the IBM Bob 2.0 platform. The product can be installed on the repo through Github Apps. This system's main functionality is the multi-language handling for rerunning tests on Git Actions .yml file for different languages; MVP includes Java, JS, Python, GoLang, and .Net. The backend uses a deterministic LangGraph state machine orchestrates the context—carrying the source code, logs, and git diffs—and routes non-deterministic (flaky) failures to an investigator persona or deterministic failures directly to a fixer. Intelligence is fully offloaded to the IBM Bob 2.0 CLI executing as a subprocess, where a unified subagent persona sequentially acts as a statistical analyst, TOCTOU delay analyst, and test order dependency analyst to inspect and reason over the repository's source code and persist structured findings into state. Standard CI/CD failure fixing agents typically fail to address flaky, non-deterministic errors. This system fills that gap by pinpointing and directly patching the underlying source code responsible for intermittent test failures across Java, JavaScript, Python, Go, and .NET environments.
27 Sep 2026
