AI Code Reviewer Agent

A comprehensive Java-based AI code reviewer agent that analyzes Java source code for quality issues, best practices violations, and potential bugs.

Features

• Automatic Code Analysis: Scans Java code for common issues

• Multiple Review Categories:

  • Exception Handling (empty catch blocks, generic exceptions)
  • Code Quality (magic numbers, naming conventions)
  • Thread Safety (public static mutable fields)
  • Code Complexity (method length, nested loops)
  • Logging Issues (System.out/err usage)
  • Code Duplication Detection

• Severity Levels: CRITICAL, MAJOR, MINOR, INFO

• Detailed Reporting: Line numbers, code snippets, suggestions

• Performance Metrics: Review execution time tracking

• Configurable Analysis: Customize review rules via ConfigManager

Project Structure

src/main/java/com/ai/reviewer/
├── agent/
│   └── CodeReviewerAgent.java           # Main reviewer agent
├── analyzer/
│   ├── CodeAnalyzer.java               # Interface for analyzers
│   └── JavaCodeAnalyzer.java           # Java code analyzer implementation
├── model/
│   ├── ReviewComment.java              # Individual code review comment
│   ├── ReviewResult.java               # Complete review result
│   └── ReviewSeverity.java             # Severity levels
├── util/
│   ├── ConfigManager.java              # Configuration management
│   └── FileReader.java                 # File I/O utilities
└── CodeReviewerMain.java               # Main entry point

Build & Run

Prerequisites

• Java 11 or higher
• Maven 3.6 or higher

Build

mvn clean package

Run

mvn exec:java -Dexec.mainClass="com.ai.reviewer.CodeReviewerMain"

Or run the JAR:

java -jar target/code-reviewer-agent-1.0.0.jar

Usage Example

// Create the agent
CodeReviewerAgent agent = new CodeReviewerAgent("AI Code Reviewer", "1.0.0");

// Review code
String code = """
    public class Example {
        public void process() {
            try {
                // code here
            } catch (Exception e) {
            }
        }
    }
    """;

ReviewResult result = agent.reviewCode("Example.java", code);

// Display results
System.out.println(result);
System.out.println("Overall Score: " + result.getOverallScore());
System.out.println("Issues Found: " + result.getTotalIssues());

Review Categories

1. Exception Handling

• Empty catch blocks
• Catching generic Exception
• Missing exception logging

2. Code Quality

• Magic numbers (constants without names)
• Naming conventions violations
• Code duplication

3. Thread Safety

• Public static mutable fields
• Shared state without synchronization

4. Code Complexity

• Long methods (>30 lines)
• Nested loops
• High cyclomatic complexity

5. Logging

• System.out/err usage instead of logging framework
• Missing error handling

Extending the Agent

Adding a Custom Analyzer

public class CustomAnalyzer implements CodeAnalyzer {
    @Override
    public ReviewResult analyze(String fileName, String code) {
        ReviewResult result = new ReviewResult(fileName, code);
        // Add your analysis logic
        return result;
    }
}

// Use custom analyzer
CodeReviewerAgent agent = new CodeReviewerAgent("Agent", "1.0.0");
agent.setAnalyzer(new CustomAnalyzer());

Configuration

Configure the agent via ConfigManager:

ConfigManager config = new ConfigManager();
config.setProperty("agent.name", "My Reviewer");
config.setProperty("review.max.method.lines", "50");
config.setProperty("review.check.magic.numbers", "true");

Severity Levels

• CRITICAL (Score -10): Issues that must be fixed (empty catch blocks, etc.)
• MAJOR (Score -5): Significant issues affecting code quality or safety
• MINOR (Score -2): Code quality improvements
• INFO (Score -0.5): Informational items (TODOs, style issues)

Score Calculation

• Base score: 100
• Score decreases based on issue severity
• Minimum score: 0
• Final score = 100 - (sum of severity deductions)

Logging

The agent uses SLF4J with Logback for logging. Configuration is in src/main/resources/logback.xml.

Logs are written to:

• Console (INFO level and above)
• File: code-reviewer.log (DEBUG level for com.ai.reviewer)

Future Enhancements

• Integration with LLM APIs for AI-powered analysis
• Support for multiple programming languages
• HTML/JSON report generation
• IDE plugin support
• CI/CD pipeline integration
• Performance metrics analysis
• Security vulnerability detection
• Architecture pattern recognition

License

MIT License

Author

AI Code Reviewer Team