Change Investigator

Modern software repositories contain enormous amounts of information about what changed, but much less about why a change was made. When developers investigate an unfamiliar commit, they often have to manually inspect diffs, previous implementations, tests, comments, documentation, and related commits before they can reconstruct the reasoning behind the change. This becomes especially difficult in large, mature codebases. Change Investigator is a developer-focused investigation tool designed to address this problem. Given a repository and a commit, it automatically collects and analyzes repository evidence surrounding that change. It examines the changed files and symbols, before-and-after code, tests, comments and documentation, Git history, parent and sibling commits, and structural relationships with related implementations. Instead of simply summarizing the diff, Change Investigator reconstructs the most plausible rationale supported by the available evidence. Its output separates documented facts from conclusions inferred from code and patterns, identifies uncertainty, and provides a confidence assessment. This is important because the tool does not claim to know what the original developer was thinking when the repository does not explicitly document it. The target users are developers, maintainers, reviewers, researchers, and engineers onboarding to unfamiliar codebases. A user provides a repository path and commit SHA, selects “Investigate,” and receives a structured explanation of what changed, what problem the change appears to address, which evidence supports that interpretation, and where uncertainty remains. The project's distinctive approach is treating the repository itself as the primary source of evidence. Rather than relying on a generic coding assistant to generate an explanation from a diff alone, Change Investigator performs a structured investigation across multiple forms of repository evidence and makes the boundary between evidence and reconstruction explicit. The result is a practical way to turn difficult historical code archaeology into a repeatable investigation process while preserving uncertainty instead of presenting speculation as fact.