Code Review Council - Powered by Claude Haiku 3.5

A multi-agent code review system using Claude Haiku 3.5 with 6 specialized agents.

What You Need

1. Python 3.7 or higher
2. Anthropic API Key - Get one at https://console.anthropic.com/api-keys

Quick Setup

Step 1: Install Dependencies

pip install anthropic

Step 2: Set Up Your API Key

python setup.py

When prompted:

• Enter your Anthropic API key (starts with "sk-ant-api03-...")
• Choose option 1 (environment variable) or 2 (.env file)

Step 3: Run the Code Review

python main.py

When prompted:

• Enter the path to the code file you want to review (e.g., example.py)
• Wait for all 6 agents to analyze your code
• Choose whether to save results to text file

What Each Agent Does

1. Syntax & Logic Agent - Finds bugs, logic errors, and correctness issues
2. Security Agent - Finds vulnerabilities and security issues
3. Performance Agent - Identifies bottlenecks and optimization opportunities
4. Architecture Agent - Reviews design patterns and code structure
5. Testing Agent - Suggests test cases and identifies testing gaps
6. Documentation Agent - Reviews and improves code documentation

Try the Demo (No API Key Needed!)

python DemoMode.py

This shows how the system works without using your API credits.

Review Multiple Files

python BatchReview.py /path/to/your/project

File Structure

├── main.py              # Main entry point
├── BaseAgent.py         # Base class for all agents
├── SyntaxLogicAgent.py  # Syntax and logic error detection
├── SecurityAgent.py     # Security analysis
├── PerformanceAgent.py  # Performance analysis
├── ArchitectureAgent.py # Architecture review
├── TestingAgent.py      # Testing recommendations
├── DocumentationAgent.py # Documentation review
├── CodeReviewCouncil.py # Orchestrates all agents
├── BatchReview.py       # Review multiple files
├── DemoMode.py          # Demo without API
├── setup.py             # API key setup
├── config.json          # Configuration
├── example.py           # Example code to test
└── README.md            # This file

Configuration

Edit config.json to:

• Change the Claude model (default: claude-3-5-haiku-20241022)
• Enable/disable specific agents
• Adjust temperature and max tokens

Example Usage

$ python main.py
Enter the path to the code file to review: example.py

📂 Reading example.py...
📏 File size: 37 lines

🔍 Starting code review for: example.py
============================================================
[████████████] Review complete! (6/6)
============================================================

============================================================
📊 REVIEW SUMMARY
============================================================

📈 Total Issues Found: 17

🔍 Issues by Type:
   🐛 Syntax & Logic: 3 issues
   🔒 Security: 3 issues
   ⚡ Performance: 2 issues
   🏗️ Architecture: 3 issues
   🧪 Testing: 3 issues
   📝 Documentation: 3 issues

⚠️  Severity Breakdown:
   • Critical: 5 🔴
   • Warnings: 8 🟡
   • Suggestions: 4 🟢

💡 Key Findings:
   • Syntax & Logic Agent: List Modification During Iteration (Lines 9-12)
   • Security Agent: SQL Injection (Line 2)
   • Performance Agent: O(n²) Algorithm (Lines 9-12)
   • Architecture Agent: Tight Coupling - Data access mixed with business logic
   • Testing Agent: SQL Injection Testing (Line 2) - Security-critical code untested
   • Documentation Agent: Missing Function Docstrings

============================================================
✅ Review complete! Check the detailed report for full analysis.
============================================================

💾 Save Options:
Save detailed report as .txt? (y/n): y

📝 Detailed report saved to: example_review_20241120_143022.txt

✨ Success! Review saved to: example_review_20241120_143022.txt
   Contains:
   • Summary of issues found
   • Detailed findings by agent
   • Line-specific references
   • Recommended fixes

Also save raw data as JSON? (y/n): n

🎉 Done! Thanks for using Code Review Council.

Output Files

The system generates two types of output:

1. Text Report (Recommended)

   • Human-readable format
   • Includes line numbers and code references
   • Detailed analysis from each agent
   • Code examples and fixes
   • Example: example_review_20241120_143022.txt

2. JSON File (Optional)

   • Raw data for integration with other tools
   • Machine-readable format
   • Example: ReviewResults.json

Cost

Claude Haiku 3.5 is very affordable:

• ~$0.25 per million input tokens
• ~$1.25 per million output tokens
• A typical code review costs less than $0.01

Troubleshooting

"API key not found" error

• Make sure you ran python setup.py and entered your key
• On Windows, restart your terminal after setup

"No module named 'anthropic'" error

• Run: pip install anthropic

"File not found" error

• Make sure the file path is correct
• Try using the full path to your code file