AI Pull Request Analysis

A powerful tool for automated code review powered by Anthropic's Claude AI. Available both as a GitHub Action and as an NPM package for CLI usage.

Features

• Automated code review using Claude AI
• Detection of potential bugs and vulnerabilities
• Style and consistency checking
• Documentation assessment
• Performance impact analysis
• Security best practices verification
• Cost estimation and tracking
• Flexible file filtering
• Supports multiple analysis levels
• Available as both GitHub Action and CLI tool

Usage Options

1. GitHub Action

Prerequisites

1. Anthropic API key (obtain from Anthropic's website)
2. GitHub repository with pull request workflows enabled

Installation

1. Create .github/workflows/pr-analysis.yml:

name: PR Analysis
on:
  pull_request:
    types: [opened, synchronize, reopened]

jobs:
  analyze:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write

    steps:
      - uses: actions/checkout@v4
      - name: AI Pull Request Analysis
        uses: diekotto/ai-pull-review@v2
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          github_token: ${{ secrets.GITHUB_TOKEN }}

2. Add your Anthropic API key to repository secrets:
   • Repository settings → Secrets and Variables → Actions
   • Create new secret: ANTHROPIC_API_KEY
   • Add your API key as the value

2. CLI Usage

The tool is also available as an NPM package for command-line usage.

Installation

# Global installation
npm install -g @diekotto/ai-pull-review-cli

# Or run directly with npx
npx @diekotto/ai-pull-review-cli

CLI Options

Usage: ai-pull-review-cli [options]

Options:
  -V, --version                       output the version number
  -C, --directory <dir>               Change to directory before performing any operations
  -m, --mode <mode>                   Analysis mode (github or localgit) (default: "localgit")
  -p, --pr <number>                   Pull request number (required for github mode)
  -r, --repo <owner/repo>            Repository (required for github mode)
  -t, --token <token>                GitHub token (default: GITHUB_TOKEN env)
  -k, --key <key>                    Anthropic API key (default: ANTHROPIC_API_KEY env)
  --model <model>                    Claude model to use
  --include <patterns>               File patterns to include (comma-separated)
  --exclude <patterns>               File patterns to exclude (default: node_modules, dist, build, etc.)
  --max-files <number>               Maximum files to analyze (default: "50")
  --max-size <number>                Maximum file size in KB (default: "100")
  --threshold <number>               Comment confidence threshold (default: "0.6")
  -A, --after-context <number>       Number of lines to show after each change (default: "10")
  -B, --before-context <number>      Number of lines to show before each change (default: "10")
  --write-pull-request               Write the analysis to the pull request as a comment
  -o, --output <folder>              Output folder for results
  --base-sha <sha>                   Base SHA or branch for local Git diff
  --head-sha <sha>                   Head SHA for local Git diff (default: "HEAD")
  --dry-run                          Show the prompt without sending to the model
  -h, --help                         display help for command

Example Usage

# Using environment variables
export GITHUB_TOKEN=your_github_token
export ANTHROPIC_API_KEY=your_anthropic_key

# Analyze a GitHub PR with context
ai-pull-review-cli --mode github -p 123 -r owner/repo -A 5 -B 3

# Analyze local git changes with minimal context
ai-pull-review-cli --mode localgit -A 1 -B 1

# Analyze specific files with maximum context
ai-pull-review-cli --include "*.ts,*.tsx" -A 10 -B 10

# Dry run to see what would be analyzed
ai-pull-review-cli --dry-run -A 2 -B 2

# Or with npx
npx ai-pull-review-cli --mode github -p 123 -r owner/repo

Configuration

Available Models

The tool supports various Claude models with different capabilities and pricing:

• claude-3-5-sonnet-20241022 (Default for accuracy)
• claude-3-5-haiku-20241022 (Default for speed)
• claude-3-opus-20240229 (Most capable)

Error Handling and Limitations

• Files exceeding size or complexity limits are automatically skipped
• Warnings are logged for problematic files
• Analysis continues with remaining files when encountering issues
• Detailed error reporting in GitHub Actions log
• Cost tracking and estimation for API usage

Requirements

• Node.js >= 22.0.0
• Yarn >= 1.22.0

License

MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.