What Changed? — AI-Powered Root Cause Analysis

Developers often know that a project worked before, but finding exactly which change caused it to break can require manually reviewing commits, dependencies, configuration, tests, and logs. This process becomes increasingly difficult as repositories grow and multiple changes interact. “What Changed?” is a local AI-powered CLI prototype designed to help developers investigate this problem faster. A developer provides a local Git repository, and the tool analyzes its history and several sources of evidence: Git changes, dependency changes, configuration changes, test results, and available logs. These signals are combined into a structured investigation that identifies the changes most likely related to the failure. The solution produces a self-contained HTML report containing the relevant evidence, ranked causal candidates, explanations, and confidence levels. For the hackathon prototype, the system also uses IBM Bob as the reasoning component: the collected evidence is converted into a structured prompt, Bob analyzes the evidence, and the resulting structured response is incorporated into the final report. The primary target users are software developers, students, and engineering teams who need to diagnose regressions in Git-based projects without manually inspecting the entire history. What makes the prototype distinctive is its evidence-driven workflow. Instead of simply asking an AI model to guess what went wrong, it first gathers concrete repository evidence and then asks Bob to reason over that evidence. The prototype also runs locally without requiring an external AI service. The demo includes a controlled Flask application with four Git commits and an intentional breaking change. The system successfully identifies the breaking commit and explains why it caused the test failure, demonstrating the complete investigation workflow from repository analysis to an actionable report.