RootTrace — Evidence-Driven Debugging

Debugging with AI often starts with a guess: identify a likely bug, generate a patch, and hope the tests pass. RootTrace takes a different approach by treating debugging as an evidence-driven investigation. RootTrace is designed for developers who need to understand not only what fix works, but why the bug happened and whether the fix is actually justified. The workflow starts with a reproducible bug. RootTrace collects runtime output, stack traces, source-level contracts, relevant test behavior, and data-flow evidence. It then develops competing root-cause hypotheses instead of accepting the first plausible explanation. Each hypothesis is evaluated against actual repository evidence before a root cause is confirmed. Once the root cause is supported by evidence, RootTrace applies the smallest appropriate fix and runs regression tests and an application/runtime check. The investigation is then preserved as a structured Debugging Case File containing the incident, reproduction, evidence, hypotheses, confirmed root cause, fix, validation, and confidence. Our controlled demonstration is a checkout pricing failure. A 20% discount is passed as the value 20 to a function whose documented contract expects a decimal such as 0.20. RootTrace reproduces the failure, traces the value through the code, evaluates competing hypotheses, confirms the contract mismatch, applies the minimal conversion of discountPercent / 100, and verifies that all three tests pass and the application produces the expected $120 result. The target users are developers, engineering teams, and anyone using AI-assisted development who wants debugging decisions to remain explainable and verifiable. What makes RootTrace different is its investigation-first workflow: evidence before diagnosis, competing hypotheses before commitment, and verification after the fix. The result is not simply a generated patch, but a traceable debugging record showing why the fix was made and how it was validated.