CodeReviewAgent 🔍

An automated code review system that generates review comments and estimates code quality using transformer-based models.

Project Structure

• codeReviewer/
  FastAPI service for automated code review classification (e.g., "Review Needed" vs "No Review Needed").
• commentGeneration/
  FastAPI service for generating review comments for code diffs.
• streamlit-ui/
  Streamlit web UI for interacting with the code review and comment generation services.
• trainCodeReviewModel.py
  Python file for model training, evaluation, and analysis.
• models/
  Contains trained model weights, tokenizer files, and configs.

Quick Start

Option 1: Try Online Demo

Visit our deployed web interface: https://streamlit-ui-586748801796.us-central1.run.app/

Option 2: Run Locally

Prerequisites

• Python 3.10+
• Git

Installation

# Clone the repository
git clone https://github.com/Dhanshreeb05/CodeReviewAgent.git
cd CodeReviewAgent

# Install dependencies
pip install -r requirements.txt

Usage

# Run the demo
python demo.py

# Run tests
pytest test_demo.py -v

📁 Project Structure

CodeReviewAgent/
├── demo.py                    # Main demo application
├── requirements.txt           # Python dependencies
├── commentGeneration/         # Comment generation models
│   └── models/
├── codeReviewer/             # Code quality models
│   └── models/
└── test_demo.py              # Unit tests

🔧 Configuration

Verify model paths in demo.py:

• Line 274: COMMENT_GENERATION_MODEL_PATH = "commentGeneration/models"
• Line 334: CODE_REVIEWER_MODEL_PATH = "codeReviewer/models"

💻 How to Use

Web Interface

1. Check API Status: Ensure both Quality API and Comment API show "Online"
2. Input Code Diff:
   • Select from pre-configured samples, OR
   • Paste your own git diff (follow UI instructions)
3. Analyze: Click "Analyze Code" button
4. View Results:
   • Quality estimation with confidence score
   • Generated review comment (if review needed)
   • Processing times and technical details

Local Demo

Run python demo.py and follow the interactive prompts to:

• Input code diffs
• Get quality predictions
• Generate review comments

🎯 Features

• Code Quality Estimation: Binary prediction of whether code needs review
• Comment Generation: Automated review comment generation
• Multi-language Support: Works with various programming languages
• Real-time Analysis: Fast processing with confidence scores

📊 Models Used

• CodeT5-Base: For comment generation
• Transformer-based: For code quality estimation
• Beam Search: Decoding with beam size 5 for quality outputs

🧪 Testing

pytest test_demo.py -v

---

Repository: https://github.com/Dhanshreeb05/CodeReviewAgent