🤖 @prajitzala/ai-code-reviewer

Add AI-powered PR code reviews to any GitHub repo in one command.

⚡ Install in 60 Seconds

npx @prajitzala/ai-code-reviewer init

That's it. Run this from the root of any GitHub repo — it sets up everything automatically.

Package: npm/@prajitzala/ai-code-reviewer · Source: github.com/Prajitzala/AI-Code-Reviewer

🎥 Demo

Live Action output on a real PR: Prajitzala/Prajitzala#1 (summary comment + inline comments on the diff).

🎬 What Happens

$ npx @prajitzala/ai-code-reviewer init

  ╔══════════════════════════════════╗
  ║     🤖  AI Code Reviewer         ║
  ║     Powered by GPT-4o            ║
  ╚══════════════════════════════════╝

? Which OpenAI model should the reviewer use?
  ❯ GPT-4o  (best quality, recommended)
    GPT-4o-mini  (faster, cheaper)

? Max diff size to review per PR?
  ❯ 12,000 chars  (recommended)

  ✔  .github/workflows/ai-review.yml
  ✔  scripts/review.py
  ✔  scripts/requirements.txt

  🚀 One step left:
  Add OPENAI_API_KEY to your repo secrets.
  Repo → Settings → Secrets → New secret
  Done! Open a PR and watch the magic 🎉

Then add your OpenAI API key as a GitHub secret and open any PR.

🔍 What the Review Looks Like

Every PR automatically gets a comment like this:

✅ AI Code Review — GPT-4o

Overall clean implementation. One security issue and two
suggestions were found.

Verdict: REQUEST_CHANGES

---
🔍 Issues Found

🔴 🔐 Hardcoded Secret Key
- Severity: critical   Category: security
- File: middleware/auth.js  line: 14
- JWT secret is hardcoded. Use process.env.JWT_SECRET instead.

🟡 🐛 Missing Error Handling
- Severity: warning   Category: bug
- File: routes/user.js  line: 42
- Async function has no try/catch — will crash in production.

---
👍 What's Good
- Clean separation of concerns
- Consistent use of async/await

Plus inline comments posted directly on the diff lines.

📋 What Gets Reviewed

| Category | What It Checks | |---|---| | 🐛 Bug | Logic errors, null refs, unhandled promises | | 🔐 Security | Hardcoded secrets, injection risks, auth issues | | ⚡ Performance | N+1 queries, unnecessary re-renders, blocking ops | | ✨ Quality | Naming, duplication, complexity, missing tests |

Lock files and binary assets are automatically skipped.

🛠️ Commands

npx @prajitzala/ai-code-reviewer init     # Set up in current repo
npx @prajitzala/ai-code-reviewer remove   # Remove from current repo
npx @prajitzala/ai-code-reviewer help     # Show help

🗂️ Files Created in Your Repo

your-repo/
├── .github/
│   └── workflows/
│       └── ai-review.yml      # Triggers on every PR
└── scripts/
    ├── review.py              # GPT-4o review logic
    └── requirements.txt       # Python deps

⚙️ How It Works

PR opened / updated
       │
       ▼
GitHub Action triggers
       │
       ▼
Python fetches PR diff (GitHub API)
       │
       ▼
Diff sent to GPT-4o (JSON mode)
       │
       ├──▶ Summary comment posted on PR
       └──▶ Inline comments on diff lines

🔮 Roadmap

• [ ] Support Anthropic Claude as alternative LLM
• [ ] .aireview config file per repo
• [ ] Severity threshold flag — only post if critical issues found
• [ ] Auto-approve clean PRs
• [ ] Re-review on new commits (deduplication)

🧑‍💻 Author

Prajit Zala — linkedin.com/in/prajitzala · github.com/Prajitzala

📄 License

MIT — use it, fork it, improve it.