LLM Code Audit - Comprehensive Documentation

---

📌 Internship Project

This project was part of my Summer Internship at Clover IT Services Pvt Ltd.
I was responsible for end-to-end design and development, including frontend (React + Vite), backend (Firebase + Firestore), and AI integrations (ChatGPT, Gemini, DeepSeek, LLaMA, Mistral).

Link for the Live Website

https://multi-model-llm-code-audit.vercel.app/

LLM Code Audit is an intelligent, developer-focused tool designed to generate, evaluate, and compare code outputs from multiple Large Language Models (LLMs), including ChatGPT, Gemini, DeepSeek, LLaMA, and Mistral.

Whether you're prototyping, benchmarking, or auditing AI-generated code, this platform helps you make data-driven decisions by providing detailed metrics around code complexity, maintainability, and readability.

With real-time analysis, visual dashboards, and history tracking, LLM Code Audit enables a transparent and measurable way to choose the best code output for your needs — all in a clean, user-friendly interface.

---

Table of Contents

1. Introduction
2. Problem Statement
3. Solution Overview
4. Features
5. Technical Architecture
6. File Structure
7. Detailed Page Descriptions
8. Metrics Explanation
9. Conclusion
10. Contributions

---

Introduction

LLM Code Audit helps developers analyze and compare code generated by multiple AI tools to identify the best-performing and most maintainable solution.

---

Problem Statement

The Problem

• AI-generated code is often unoptimized or poorly structured.
• Manually comparing outputs is tedious and error-prone.
• Lack of tools that provide actionable feedback and ranking across LLMs.

Why Current Tools Fall Short

• Tools like SonarQube evaluate only post-development.
• No side-by-side LLM comparisons.
• No automatic quality metrics or recommendations.

---

Solution Overview

LLM Code Audit provides:

• Multi-LLM integration
• AI code comparison
• 5 quality metrics
• Scoring

---

Features

1. Multi-LLM Code Generation

• One prompt → multiple AI outputs
• Side-by-side comparison
• Add user code for evaluation

2. Advanced Code Analysis

• Detects code issues
• Auto-scores and recommends best output

3. Smart Metrics System

• ACI (Complexity)
• AMR (Maintainability Risk)
• ARS (Readability)
• ADR (Dependency Risk)
• ARF (Redundancy)

4. History Tracking

• Stores prompts and outputs
• Tracks best LLM per prompt
• Allows re-analysis

5. User-Friendly Interface

• Minimal, modern UI
• Interactive visualizations
• Fully responsive

---

Technical Architecture

Frontend

• React.js with Vite
• Modular components
• Styled using custom CSS

Backend

• Firebase Authentication
• Firestore for prompt/output storage

AI Integrations

• ChatGPT
• Gemini
• DeepSeek
• LLaMA
• Mistral

---

File Structure

node_modules/
public/
src/
├── api/
├── assets/
├── backend/
├── components/
│   ├── AuthModal.jsx
│   ├── Header.jsx
│   ├── HistoryItem.jsx
│   ├── QueryInput.jsx
│   └── ResultsDisplay.jsx
├── hooks/
├── pages/
│   ├── AboutPage.jsx
│   ├── AuthPage.jsx
│   ├── ComparisonDetail.jsx
│   ├── Dashboard.jsx
│   ├── HistoryPage.jsx
│   └── HomePage.jsx
├── styles/
├── App.css
├── App.jsx
├── firebaseConfig.js
├── firestoreService.js
├── index.css
├── main.jsx
.env
.gitignore
eslint.config.js
index.html
package-lock.json
package.json
vite.config.js

---

Detailed Page Descriptions

Home Page

• Enter prompt
• Choose LLMs
• Add custom code
• Click "Generate" for outputs

Dashboard Page

• View AI-generated code
• Start code evaluation

Results Page

• Scores and charts for each output
• Final recommendation

History Page

• View and revisit previous prompts
• Compare LLM performance over time

About Page

• Project vision
• Real-world applications

---

Metrics Explanation

AI Complexity Index (ACI)

ACI = (0.5 × Cognitive Complexity) + (0.3 × Method Length) + (0.2 × Nesting Level)

AI Maintainability Risk (AMR)

AMR = (0.4 × Code Smells per 100 LOC) + (0.4 × Duplication %) + (0.2 × % Uncommented Methods)

AI Readability Score (ARS)

ARS = (0.4 × Avg Line Length) + (0.4 × % Comments) + (0.2 × Naming Consistency)

Conclusion

LLM Code Audit empowers developers to take full control over AI-generated code, ensuring quality, maintainability, and clarity before integrating into production environments.

By offering clear comparisons, actionable insights, and in-depth metrics, this tool bridges the gap between rapid AI code generation and robust software engineering standards.

Ready to reduce tech debt and make AI code work for you?
Try LLM Code Audit now!