🤖 ReviewPilot — AI Pull Request Reviewer

ReviewPilot is an AI-powered code review agent that analyzes GitHub pull requests, generates structured feedback, and posts semantic inline review comments directly on the changed lines of code.

It transforms LLM reasoning into a real developer workflow.

Untitled.Project.mp4

Live Demo: https://reviewpilot-frontend.onrender.com

✨ Features

• 🔍 Ingests real GitHub Pull Requests
• 🧠 LLM-based multi-point code review
• ⚠️ Risk classification based on change scope
• 💬 Inline comments on the exact lines of code
• 📌 Structured PR summary
• 🧭 Multiple issue detection in a single run
• 🎯 Demo-ready UI for running reviews

---

🧠 Why ReviewPilot?

Most AI code review demos stop at summarizing diffs.

ReviewPilot acts as a real reviewer inside GitHub:

PR → AI analysis → Inline feedback → Developer workflow

This project demonstrates how LLMs can:

• Reason over real code changes
• Make architectural suggestions
• Execute actions in external systems

---

🏗️ Architecture

Flow

1. Fetch PR metadata and diff from GitHub
2. Send diff to LLM for structured analysis
3. Extract multiple review findings
4. Map findings to exact changed lines
5. Post inline review comments via GitHub API
6. Return structured summary to UI

Tech Stack

Backend

• FastAPI
• Groq LLM
• GitHub REST API

AI Layer

• Structured prompting
• JSON parsing & validation
• Multi-finding semantic extraction

Frontend (Demo UI)

• Static HTML
• Vanilla JavaScript
• TailwindCSS

---

📸 Demo Flow

1. Enter repository → owner/repo

2. Enter Pull Request number

3. Click Run AI Review

4. AI:

   • Analyzes code
   • Posts inline feedback on GitHub
   • Returns structured summary

---

⚙️ Setup

1️⃣ Clone the repository

git clone https://github.com/your-username/reviewpilot.git
cd reviewpilot

---

2️⃣ Backend setup

python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

Create a .env file:

GROQ_API_KEY=your_key
GITHUB_TOKEN=your_token

Run the backend:

uvicorn app.main:app --reload

---

3️⃣ Frontend setup

Run the frontend:

python frontend-new/serve.py

The frontend will be available at http://127.0.0.1:5500/ and will call the backend at http://127.0.0.1:8000/review.

---

🔌 API

POST /review

Request

{
  "repo": "owner/repo",
  "pr_number": 1
}

Response

{
  "summary": "...",
  "issues": ["..."],
  "suggestions": ["..."],
  "risk": "low | medium | high"
}

---

🧪 Example Output

• Inline comments on risky code paths
• Architectural improvement suggestions
• Test coverage warnings
• Maintainability feedback

---

🧭 Project Goals

This project is built to demonstrate:

• Real-world AI agent workflows
• LLM orchestration beyond chat interfaces
• Tool integration with external platforms
• AI as a developer productivity system

---

🚀 Future Improvements

• Priority & severity scoring
• Confidence levels per finding
• Review analytics
• Multi-PR batch reviews

---

👨‍💻 Author

Krish Patel

AI / GenAI / LLM Application Engineer

---

⭐ If you found this interesting

Give it a star — it helps the project reach more developers.