Inspiration

Reviewers get tired and skip the boring-but-critical stuff — missing tests, security gaps, creeping complexity. We wanted an AI reviewer that catches it first.

What it does

AI PR Reviewer takes a GitHub PR link, pulls the diff, and returns an AI-generated review in seconds — issues sorted by Security, Style, Complexity, and Missing Tests, each with severity and an overall grade. Past reviews are also saved.

How we built it

Next.js 14 (App Router) for frontend + API routes, GitHub's REST API for diffs, Gemini for review generation with strict JSON output, MongoDB for storage, Tailwind for UI.

Challenges we ran into

Getting Gemini to return clean JSON every time took a retry fallback. Large diffs needed truncation to stay in budget. Unauthenticated GitHub API calls hit rate limits faster than expected.

Accomplishments that we're proud of

A full fetch → analyze → store → display pipeline that runs in one smooth flow, and real issues surfacing on real open-source PRs .

What we learned

Structured LLM output takes real prompt iteration. A good loading state changes how fast an AI tool feels, even when latency doesn't change

What's next for AI PR Review

Turn it into a real GitHub App that comments directly on PRs, add per-repo rule configs, and track whether flagged issues actually get fixed on the next push.

Built With

Share this project:

Updates

Submission history