🚀 AI Code Reviewer

An Intelligent, Real-Time Code Review Extension for Visual Studio Code

Powered by Local & Cloud LLMs (Codestral, Mistral, DeepSeek, Ollama, OpenAI)

AI Code Reviewer brings AI-powered code analysis, refactoring, and instant explanations directly into your workflow.
Built as a full VS Code extension, it integrates seamlessly with LM Studio, Ollama, or OpenAI-compatible APIs to deliver smart, real-time insights into your code.

⭐ Key Features

🔍 1. Full-File AI Code Review

Get a comprehensive analysis of:

• Bugs & logic errors
• Unhandled edge cases
• Security issues
• Performance problems
• Readability / style issues (optional)

Runs with: Ctrl + Alt + R

✂️ 2. Selection-Based Review (Advanced)

Review only the highlighted part of code: Ctrl + Alt + Shift + R Great for isolating small blocks or debugging tricky logic.

💡 3. Explain Code (Beginner-Friendly)

Select any code → get a clear step-by-step explanation in Markdown. Ctrl + Alt + E Perfect for learning, documentation, or onboarding new developers.

🛠️ 4. AI Refactor (Auto-Fix & Improve Code)

Select code → AI rewrites it with:

• Better readability
• Error handling
• Idiomatic style
• Cleaner logic
• Edge case protection

Run with: Ctrl + Alt + F

📊 5. Issues Panel (Built-In Dashboard)

A dedicated panel inside VS Code showing:

• Issue number
• Severity
• Line number
• Title
• Explanation
• Learning tip

Open anytime with: Ctrl + Alt + I

🔄 6. Continuous Code Review Mode (Real-Time)

Like GitHub Copilot diagnostics — but local AI powered.

When enabled:

• The extension auto-reviews your file after you stop typing
• Fully configurable debounce (default: 4s)

Enable in Settings: AI Code Reviewer → enableContinuousReview

⚙️ 7. Customizable Review Modes

Choose how strict the AI should be:

• Relaxed – only major issues
• Balanced – recommended
• Strict – flags every possible issue

Disable entire categories:

• Style issues
• Performance concerns
• Security warnings

🔄 8. Multi-Model Support

Works with ANY OpenAI-compatible API:

• LM Studio
• Codestral
• Ollama
• Mistral
• DeepSeek
• OpenAI
• Any local GGUF model with an OpenAI endpoint

🧩 9. Fully Offline Capability

Using LM Studio or Ollama, the entire AI functionality can run:

• Without Internet
• Without sending code externally
• Completely free

🧱 Architecture Overview

┌──────────────────────┐ │ VS Code UI │ │ (Editor, Commands) │ └───────────┬──────────┘ │ ▼ ┌──────────────────────┐ │ AI Code Reviewer │ │ (Extension Logic) │ └───────────┬──────────┘ │ Builds prompts ▼ ┌──────────────────────┐ │ Prompt Builder │ └───────────┬──────────┘ │ Sends requests ▼ ┌──────────────────────┐ │ LLM Client │ │ (OpenAI API Schema) │ └───────────┬──────────┘ │ ▼ ┌──────────────────────┐ │ Local / Cloud LLM │ │ Codestral, Mistral… │ └───────────┬──────────┘ │ Returns JSON ▼ ┌──────────────────────────┐ │ Diagnostics + Issue Panel │ └──────────────────────────┘

🖥️ Screenshots (Add Your Own Here!)

Add these after generating screenshots:

• Full-file review
• Selection review
• Explain code panel
• Refactor in action
• Issues panel
• LM Studio server logs
• Continuous mode demo

📦 Installation

1. Clone the repo

git clone https://github.com/your-username/ai-code-reviewer.git
cd ai-code-reviewer

2. Install dependencies

npm install

3. Build the extension

npm run compile

4. Run the extension

In VS Code:

mathematica Run → Run Extension or press F5 This opens the Extension Development Host window.

🤖 LLM Setup (LM Studio Recommended) Download LM Studio Load a model (e.g., codestral-22b-v0.1) Start the server Copy the endpoint (usually: http://127.0.0.1:1234/v1) Update settings in VS Code:

AI Code Reviewer → apiBaseUrl = http://127.0.0.1:1234/v1
AI Code Reviewer → model = codestral-22b-v0.1
AI Code Reviewer → apiKey = test

💡 LM Studio does not require a real API key.

🎛️ Commands & Shortcuts Action Shortcut Command Review Entire File Ctrl+Alt+R codesense-mentor.reviewFile Review Selection Ctrl+Alt+Shift+R codesense-mentor.reviewSelection Explain Code Ctrl+Alt+E codesense-mentor.explainSelection Refactor Code Ctrl+Alt+F codesense-mentor.refactorSelection Open Issues Panel Ctrl+Alt+I codesense-mentor.openIssuesPanel Toggle Continuous Review Ctrl+Alt+C codesense-mentor.toggleContinuousReview

⚙️ Extension Settings Setting Description apiKey API key for OpenAI-compatible endpoints (LM Studio can use anything). apiBaseUrl Endpoint URL (LM Studio, Ollama, etc.) model The model to run (codestral, mistral, llama3, etc.) reviewMode relaxed, balanced, strict includeStyleIssues Show style/readability issues includePerformanceIssues Detect performance problems includeSecurityIssues Detect security issues enableContinuousReview Run automatic reviews while typing continuousReviewDebounceMs Delay before auto-review triggers

🛠️ Tech Stack TypeScript VS Code Extension API LM Studio (local LLMs) Node.js Codestral / Mistral / DeepSeek / Ollama OpenAI API schema

📄 Folder Structure

src/
  extension.ts         → main logic
  llmClient.ts         → handles API calls
  promptBuilder.ts     → builds system/user prompts
  issuesPanel.ts       → dashboard for issues
  diagnostics.ts       → VS Code diagnostics integration

📜 License MIT License — free for personal & commercial use.

💬 Contact If you use this project or want to collaborate, feel free to connect!

Author: Jenil Gohel Location: Waterloo, Canada