Bug-Detective

Bug Detective — AI-Assisted Debugging & Testing Debugging is where developers lose the most time, and beginners lose the most of all. A typical bug means bouncing between an error message, a stack trace, the actual code, a search engine, an AI chat window, and a test runner — and even after a fix seems to work, there's rarely a real check that it actually solved the problem. That loop is repetitive, fragmented, and easy to get stuck in. Bug Detective replaces that loop with a single guided workflow: Investigate → Diagnose → Fix → Test → Verify Paste the error message, an optional stack trace, and the relevant code. Bug Detective identifies the likely root cause and shows the exact evidence for it — quoted from your actual code, never invented — along with a confidence level (Low, Medium, or High) so you know how much to trust the result. If the pattern is one it recognizes, it explains precisely what's wrong and why; if it isn't, it says so honestly instead of guessing. From there, it suggests a fix shown as a clear before/after diff, explaining what changes and why, along with any assumptions or side effects worth knowing. It then generates a regression test designed to catch the same bug if it ever comes back, and — for JavaScript — actually runs that test in a sandboxed in-browser environment: failing on the original code, then passing on the fix. That's the "Verify" step: proof, not a claim. What it understands today: Bug Detective recognizes six common JavaScript/TypeScript bug patterns — reading a property off undefined/null (directly, chained, or through an array), an API response shaped differently than the code expects, empty-input crashes, ReferenceError, calling something that isn't a function, and destructuring a missing value. Python and Java are supported with real test templates (pytest, JUnit 5) rather than a fake pass/fail, and outside its recognized patterns, it reports Low confidence rather than fabricating a root cause. Built with: Next.js, TypeScript, React, and Tailwind CSS, deployed on Vercel, with 20 automated tests (Vitest) covering the analyzer's patterns and edge cases (empty input, malformed responses, whitespace variance) across JavaScript, TypeScript, Python, and Java. Built for the IBM Bob 2.0 Hackathon: IBM Bob was part of the actual development process, not just the branding. Bob reviewed the project's architecture and flagged two concrete risks — weak Python/Java support and a test generator that silently produced broken output for non-JavaScript languages — then helped fix both, and helped troubleshoot the production deployment until the prototype was live. Who it's for: developers debugging real errors, students learning to read stack traces instead of guessing, and small teams who want a fix that comes with proof it works — not just a plausible-sounding explanation. Don't just fix the bug. Understand it. Test it. Verify it.