🚀 CodeInsights AI – Detailed Project Documentation

🧠 1. Project Overview

CodeInsights AI is a full-stack web application that analyzes user-submitted code and provides AI-powered optimization suggestions along with detailed code quality metrics and performance insights.

The goal of this project is to help developers:

• Improve code quality
• Understand performance implications
• Learn best coding practices interactively

---

🎯 2. Problem Statement

Most beginners write code that works but is:

• inefficient
• hard to maintain
• poorly structured

Existing tools:

• focus only on syntax errors OR
• provide static analysis without explanation

👉 There is no unified system that:

• explains improvements
• compares before vs after
• tracks user progress

---

💡 3. Solution

I built CodeInsights AI, which combines:

• AI-based code optimization
• Static code analysis
• Performance metrics
• User accuracy tracking

This creates a learning + evaluation system, not just a tool.

---

🏗️ 4. System Architecture

Frontend (React)
   ↓
Node.js Backend (API layer)
   ↓
Python AI Service (Analysis engine)
   ↓
PostgreSQL Database

---

🔹 5. Frontend (React)

Responsibilities:

• Code editor input
• Display optimized code
• Show metrics and charts
• User authentication (JWT)
• Dashboard (accuracy tracking)

Key Features:

• Monaco Editor for code input
• Recharts for visualization
• Axios with interceptors for API calls
• Context API for global state (user, token)

---

🔹 6. Backend (Node.js + Express)

Responsibilities:

• API handling
• Authentication (JWT)
• Database interaction
• Communication with Python service

Key Modules:

1. Review System

• Accepts user code
• Sends to Python service
• Stores results in DB

2. Authentication

• Signup/Login using bcrypt + JWT
• Middleware for route protection

3. Dashboard System

• Calculates:

  • average accuracy
  • best score
  • review history

---

🔹 7. Python AI Service

This is the core intelligence layer.

Components:

1. LLM Optimization

• Uses Groq API

• Generates:

  • optimized code
  • explanation
  • improvement points

2. Metrics Engine

• LOC
• Cyclomatic complexity (Radon)
• Nesting depth (AST)
• Maintainability score
• Time & space complexity (heuristics)

3. Comparison Engine

• Compares original vs optimized code

• Generates:

  • summary
  • verdict
  • accuracy score

---

📊 8. Accuracy Scoring System (Key Innovation)

Purpose:

To quantify how much the code improved.

Logic:

• Base score = 50

• Add points for:

  • improved maintainability
  • reduced complexity
  • reduced nesting
  • reduced code smells

• Penalize regressions

Special Cases:

• Same code → 100
• Same logic → ~90
• Syntax fix → ~95

👉 This makes the system interactive and gamified

---

🗄️ 9. Database Design (PostgreSQL)

Reviews Table:

• original_code
• optimized_code
• original_metrics (JSON)
• optimized_metrics (JSON)
• improvement_summary (JSON)
• accuracy_score (INTEGER)

Users Table:

• id
• name
• email
• password (hashed)
• provider

---

🔐 10. Authentication System

Manual Auth:

• Password hashing using bcrypt
• JWT token generation
• Middleware for protected routes

Flow:

1. User logs in
2. Token stored in frontend
3. Axios interceptor attaches token
4. Backend verifies token

---

📈 11. Dashboard System

Displays:

• Average accuracy
• Best score
• History of reviews

👉 Helps users track improvement over time

---

⚙️ 12. API Flow

User submits code
   ↓
Frontend → /code/analyze
   ↓
Backend → Python service
   ↓
AI + Metrics + Comparison
   ↓
Backend stores result
   ↓
Response sent to frontend

---

🧪 13. Example Use Case

Input:

print("hello")

Output:

• Optimized code (structured version)
• Metrics comparison
• Summary
• Accuracy score (~90 if same logic)

---

🚀 14. Key Features Summary

• AI-powered code improvement
• Real-time metrics analysis
• Visual comparison dashboard
• Accuracy scoring system
• JWT authentication
• Full-stack architecture

---

🧠 15. Challenges Faced

1. Comparing code correctness

   • Solved using metric-based logic instead of strict AST

2. Handling invalid code

   • Used AST parsing with fallback logic

3. LLM JSON parsing issues

   • Implemented safe JSON extraction

4. Token-based authentication issues

   • Solved using Axios interceptors

---

🌟 16. Unique Selling Points

• Combines AI + static analysis
• Accuracy scoring system (gamification)
• Full-stack + microservice architecture
• Real-world engineering practices

---

🔮 17. Future Improvements

• Multi-language support
• Better complexity detection
• Code execution sandbox
• Team collaboration
• Leaderboards & streaks

---

🎤 18. Conclusion

CodeInsights AI is not just a code optimizer but a learning platform that helps developers understand and improve their coding practices using AI and data-driven insights.

---