🤖 AI PR Review CLI

An intelligent command-line tool that uses AI to review your code changes and provide instant feedback on issues, typos, and improvements. Perfect for any programming language and project type.

✨ Key Features

• 🔍 Smart Code Analysis - Uses GPT to find bugs, security issues, and improvements
• 🎯 Auto Branch Detection - Automatically detects your base branch (main/master/develop)
• 🔄 Intelligent Batching - Processes ALL files in manageable batches (5 files per batch)
• 📊 No File Limits - Reviews unlimited files without losing any data
• ⚡ Rate Limit Protection - Smart delays between batches prevent API errors
• 🎨 Beautiful Output - Clean, colorful results with progress tracking
• 🌍 Universal Support - Works with any language, framework, or project type
• ⚙️ Flexible Config - Multiple ways to customize behavior
• 🛡️ Error Recovery - Continues processing even if one batch fails

🚀 Quick Start

1. Install

# Install globally
npm install -g ai-pr-review-cli

# Or use without installing
npx ai-pr-review-cli review

2. Get OpenAI API Key

Get your API key from OpenAI

3. Set API Key

Windows:

# PowerShell - current session
$env:OPENAI_API_KEY="your-api-key-here"

# Permanent (via System Properties > Environment Variables)
# Variable name: OPENAI_API_KEY
# Variable value: your-api-key-here

Mac/Linux:

# Current session
export OPENAI_API_KEY="your-api-key-here"

# Permanent (add to ~/.bashrc or ~/.zshrc)
echo 'export OPENAI_API_KEY="your-api-key-here"' >> ~/.bashrc

4. Use in Any Git Repository

# Navigate to your project
cd my-project

# Review your changes
ai-pr-review review

# Short alias also works
aipr review

💻 Usage Examples

Small Projects

# Simple review - auto-detects base branch
ai-pr-review review

# Review against specific branch
ai-pr-review review --base main

# Use GPT-4 for higher quality
ai-pr-review review --model gpt-4o

Large Projects (Multiple Approaches)

Method 1: Automatic Batching (Built-in)

# The standard review now automatically batches files
ai-pr-review review  # Processes ALL files in batches of 5

# With verbose output to see batching progress
ai-pr-review review --verbose

Method 2: Advanced Batch Mode

# Smart batching by file type (React, Vue, Angular, etc.)
ai-pr-review batch --frontend

# Custom batch size for different project needs
ai-pr-review batch --batch-size 6

# Preview what would be reviewed (no API calls)
ai-pr-review batch --dry-run

Method 3: Selective Review

# Review only specific files
git add src/important-file.js src/critical-component.js
ai-pr-review review  # Only reviews staged files

📋 Common Workflows

Feature Development

# Create feature branch
git checkout -b feature/new-feature

# Make changes, then review before committing
ai-pr-review review

# Address feedback, then commit
git add .
git commit -m "Implement new feature"

Large Codebase Updates

# Stage all changes
git add .

# Smart automatic batching handles everything
ai-pr-review review --verbose

# Example output for 23 files:
# 🔄 Processing 5 batches of files (5 files per batch)...
# --- Batch 1/5 ---
# Files: auth.js, login.js, middleware.js, routes.js, config.js
# ✅ Batch 1 completed: 3 issues, 1 typo, 4 improvements
# ⏳ Waiting 3 seconds before next batch...
# [continues for all batches...]

# Final consolidated results:
# 🚨 Issues: bugs, security, performance
# ✏️ Typos: spelling, grammar
# 💡 Improvements: best practices, refactoring

Different Project Types

# Works automatically for any project:
cd react-frontend && ai-pr-review batch --frontend
cd node-api && ai-pr-review batch
cd python-ml && ai-pr-review batch
cd java-enterprise && ai-pr-review batch
cd devops-terraform && ai-pr-review batch

⚙️ Configuration Options

Global Config (One-time Setup)

# Set default preferences
ai-pr-review config --model gpt-4o-mini --base main

# Creates ~/.aiprconfig.json with your defaults

Project Config (.aiprconfig.json)

{
  "provider": "openai",
  "model": "gpt-4o-mini",
  "baseBranch": "develop"
}

Command Line Options

| Option | Description | Example | | ------------------ | ------------------------------- | ----------------------- | | --base <branch> | Compare against specific branch | --base origin/develop | | --model <model> | Use specific AI model | --model gpt-4o | | --verbose | Show detailed analysis info | --verbose | | --batch-size <n> | Files per batch (default: 8) | --batch-size 4 | | --frontend | Optimize for frontend projects | --frontend | | --dry-run | Preview without API calls | --dry-run |

🎯 What Gets Analyzed

Code Changes

• ✅ Committed changes (between branches)
• ✅ Staged changes (git add)
• ✅ Working directory changes (uncommitted)

File Types Supported

• Web: JavaScript, TypeScript, HTML, CSS, SCSS
• Backend: Python, Java, Go, Ruby, PHP, C#
• Mobile: React Native, Flutter, Swift, Kotlin
• DevOps: Terraform, Docker, Kubernetes, YAML
• Data: SQL, R, Jupyter notebooks
• Config: JSON, YAML, TOML, XML
• Documentation: Markdown, reStructuredText

🔄 How Batching Works

Automatic Batching (Default)

The tool automatically processes files in batches to handle any number of files efficiently:

ai-pr-review review  # Automatically batches files

# Example with 23 files:
📁 Found 23 changed files: auth.js, login.js, config.js...

🔄 Processing 5 batches of files (5 files per batch)...

--- Batch 1/5 ---
Files: auth.js, login.js, middleware.js, routes.js, config.js
✅ Batch 1 completed: 3 issues, 1 typo, 4 improvements
⏳ Waiting 3 seconds before next batch...

--- Batch 2/5 ---
Files: utils.js, helpers.js, constants.js, api.js, db.js
✅ Batch 2 completed: 2 issues, 0 typos, 3 improvements
⏳ Waiting 3 seconds before next batch...

[... continues for all files ...]

📋 Final Summary: 12 issues, 4 typos, 18 improvements

Benefits

• ✅ No File Limits - Reviews unlimited files
• ✅ Rate Limit Safe - 3-second delays prevent API errors
• ✅ Progress Tracking - See which batch is processing
• ✅ Error Recovery - If one batch fails, others continue
• ✅ Memory Efficient - Processes files in manageable chunks

🔧 Troubleshooting

Rate Limit Errors (429)

# The tool now automatically handles rate limits with batching
ai-pr-review review  # Built-in 3-second delays between batches

# For very large projects, use advanced batch mode
ai-pr-review batch --batch-size 3  # Smaller batches

# Or wait 1-2 minutes and retry
ai-pr-review review

# Consider upgrading OpenAI plan for higher limits

No Changes Found

# Check git status
git status

# Make sure you're not on the main branch
git checkout -b feature/my-changes

# Or specify different base branch
ai-pr-review review --base origin/master

API Key Issues

# Verify key is set
echo $OPENAI_API_KEY  # Mac/Linux
echo $env:OPENAI_API_KEY  # Windows PowerShell

# Test with simple review
ai-pr-review review --verbose

💡 Pro Tips

Efficient Workflows

# Review before committing
git add changed-files/
ai-pr-review review
# Fix issues, then commit

# Review large changes in batches
ai-pr-review batch --dry-run  # Preview
ai-pr-review batch            # Execute

Cost Optimization

# Use mini model for regular reviews (cheaper)
ai-pr-review review --model gpt-4o-mini

# Use GPT-4 for critical reviews (more thorough)
ai-pr-review review --model gpt-4o

# Set as default
ai-pr-review config --model gpt-4o-mini

Shell Aliases

# Add to ~/.bashrc or ~/.zshrc
alias review="ai-pr-review review"
alias batch-review="ai-pr-review batch"
alias quick-review="ai-pr-review review --model gpt-4o-mini"

# Usage
review --base develop
batch-review --frontend
quick-review

🌍 Platform Support

| Platform | Status | Notes | | ---------------- | ------ | -------------------- | | GitHub | ✅ | Full support | | GitLab | ✅ | Full support | | Bitbucket | ✅ | Full support | | Azure DevOps | ✅ | Full support | | Local Git | ✅ | Works without remote |

📊 Sample Output

Single File Review

git add problematic-file.js
ai-pr-review review

🤖 AI PR Review
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📊 Analyzing changes on branch: feature/auth-fixes
📁 Found 1 changed file: problematic-file.js

🔄 Processing 1 batch of files (5 files per batch)...

--- Batch 1/1 ---
Files: problematic-file.js
✅ Batch 1 completed: 3 issues, 1 typo, 2 improvements
✔ Review completed!

🚨 Issues
  1. Invalid JavaScript syntax: unexpected semicolon and commas
  2. Missing input validation for user input
  3. Use const instead of var for better scoping

✏️ Typos
  1. "recieve" should be "receive" in comment

💡 Improvements
  1. Add proper error handling
  2. Remove unused variables

📋 Review Summary: 3 issues, 1 typo, 2 improvements

Large Project Review

git add .  # 23 files
ai-pr-review review --verbose

🤖 AI PR Review
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📊 Analyzing changes on branch: feature/auth-system
📁 Found 23 changed files: auth.js, login.js, middleware.js, routes.js...

🔄 Processing 5 batches of files (5 files per batch)...

--- Batch 1/5 ---
Files: auth.js, login.js, middleware.js, routes.js, config.js
🔍 Debug - Files with content:
  1. auth.js (1,240 chars): "const jwt = require('jsonwebtoken'); const bcrypt..."
  2. login.js (856 chars): "// Login component with validation..."
✅ Batch 1 completed: 4 issues, 1 typo, 3 improvements
⏳ Waiting 3 seconds before next batch...

--- Batch 2/5 ---
Files: utils.js, helpers.js, constants.js, api.js, db.js
✅ Batch 2 completed: 2 issues, 0 typos, 4 improvements
⏳ Waiting 3 seconds before next batch...

[... continues for all 5 batches ...]

✔ Review completed!

🚨 Issues
  1. Missing input validation in auth.js:15
  2. Potential SQL injection in routes.js:42
  3. Password stored in plain text - use bcrypt
  [... 9 more issues ...]

✏️ Typos
  1. "recieve" should be "receive" in comment (auth.js:8)
  [... 2 more typos ...]

💡 Improvements
  1. Add rate limiting to login endpoint
  2. Use environment variables for JWT secret
  [... 16 more improvements ...]

📋 Review Summary: 12 issues, 3 typos, 19 improvements

🔐 Security & Privacy

• Code is sent to OpenAI's API for analysis
• No permanent storage of your code
• Use .gitignore to exclude sensitive files
• Review OpenAI's data usage policies
• Consider using on non-production codebases first

📊 Performance for Any Repository Size

Smart Handling of Large Codebases

| Repository Size | Processing Method | Time | Success Rate | | ---------------- | ----------------- | -------------- | ------------ | | 1-5 files | Single batch | ~10 seconds | ✅ 100% | | 6-25 files | 2-5 batches | ~30-60 seconds | ✅ 100% | | 26-100 files | 5-20 batches | 2-5 minutes | ✅ 100% | | 100+ files | Smart batching | 5-15 minutes | ✅ 100% |

Before vs After

• Old Approach: Limited to 10 files = Many files ignored ❌
• New Approach: Unlimited files in smart batches = All files reviewed ✅

Real-World Examples

# Small project (5 files)
📁 Found 5 files → 🔄 1 batch → ✅ ~15 seconds

# Medium project (23 files)
📁 Found 23 files → 🔄 5 batches → ✅ ~90 seconds

# Large project (87 files)
📁 Found 87 files → 🔄 18 batches → ✅ ~8 minutes

# Enterprise project (200+ files)
📁 Found 234 files → 🔄 47 batches → ✅ ~15 minutes

Efficiency Features

• ✅ No File Limits - Review unlimited files
• ✅ Smart Rate Limiting (3-second delays between batches)
• ✅ Progress Tracking for long reviews
• ✅ Error Recovery - continues even if one batch fails
• ✅ Memory Efficient - processes in manageable chunks

📄 License

MIT License - see LICENSE file for details.

🤝 Contributing

Contributions welcome! Feel free to:

• Report bugs or request features
• Submit pull requests
• Improve documentation
• Share usage examples

📞 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions

Made with ❤️ for developers who want smarter code reviews

Save time, catch bugs, and improve code quality with AI-powered reviews 🚀