AI Review Engine

Mesrai’s AI Review Engine is the core of our code review platform. It uses advanced language models to understand your code at a deep level and provide actionable feedback.

How It Works

The AI Review Engine processes your pull requests in three phases:

1. Code Understanding

When you open a PR, Mesrai:

• Analyzes the changed files and understands the code structure
• Maps relationships and dependencies across your codebase
• Builds a contextual understanding of what your code does

2. Analysis

The engine performs multiple analysis passes:

• Security Analysis: Identifies potential vulnerabilities and security anti-patterns
• Performance Analysis: Detects inefficient code patterns and optimization opportunities
• Architecture Analysis: Evaluates code organization and adherence to best practices
• Style Analysis: Checks for consistency with your team’s coding standards

3. Feedback Generation

Finally, the engine generates human-readable feedback:

• Specific line-by-line comments with suggested fixes
• Summary of overall PR quality
• Prioritized list of issues by severity

Key Features

Context-Aware Reviews

Unlike simple linters, Mesrai understands the broader context of your changes:

// Mesrai understands this function is called from an async context
// and suggests appropriate error handling
function fetchUserData(userId) {
  return database.query(`SELECT * FROM users WHERE id = ${userId}`);
  // ⚠️ Mesrai: SQL injection vulnerability detected
  // Suggestion: Use parameterized queries
}

Language Support

The AI Review Engine supports all major programming languages:

Configuration via Platform

Repository-Level Rules

Configure review rules for each repository directly from the Mesrai platform:

1. Go to app.mesrai.com and sign in
2. Select your repository from the dashboard
3. Navigate to Settings → Review Rules
4. Configure your preferences:
   • Enable/disable specific analysis types
   • Set severity levels for different issue types
   • Add custom instructions for your team

Custom Instructions

You can provide custom instructions to guide how Mesrai reviews your code:

• Coding standards — Tell Mesrai about your team’s conventions
• Focus areas — Prioritize security, performance, or other aspects
• Ignore patterns — Specify files or patterns to skip
• Team preferences — Add context about your project requirements

How to Configure

1. Open app.mesrai.com
2. Select your Organization or Repository
3. Go to Settings
4. Configure:
   • Review Settings — Enable/disable review types
   • Instructions — Add custom guidance for reviews
   • Exclusions — Specify files to ignore

Review Settings

Analysis Types

Control which types of analysis are performed:

Severity Levels

Configure how strictly Mesrai reviews your code:

Best Practices

1. Start with default settings — The engine is pre-configured for common use cases
2. Add custom instructions gradually — Begin with key guidelines and expand over time
3. Review AI suggestions — While highly accurate, always verify suggestions match your intent
4. Provide feedback — Use the 👍/👎 buttons on comments to help improve accuracy

Feedback Loop

Help improve Mesrai’s reviews for your codebase:

• 👍 Helpful — Mark suggestions that were useful
• 👎 Not Helpful — Flag suggestions that missed the mark
• Comment — Add context to help Mesrai learn your preferences

Your feedback helps Mesrai provide better, more relevant reviews over time.

Next Steps

• Configure Context Awareness — Enhance review accuracy
• Set up Architectural Analysis — Deep design analysis
• Integrate with GitHub — Connect your repositories
• View Dashboard — Configure your settings