AI-Powered Code Reviews & PR Summaries for GitHub
Accelerate your development workflow with intelligent, context-aware code reviews using Claude's advanced reasoning capabilities
π Quick Start β’ π Features β’ βοΈ Configuration β’ π€ Contributing
Transform your code review process with Claude's advanced reasoning. Unlike traditional tools, AIGNE CodeSmith doesn't just check syntaxβit understands context, identifies architectural issues, and provides thoughtful feedback that helps your team grow.
- π§ Context-Aware Reviews: Leverages Claude's superior reasoning to understand code intent and architecture
- β‘ Dual-Model Efficiency: Smart cost optimization with light models for summaries, heavy models for deep analysis
- π― Focus on What Matters: Skips trivial changes, highlights critical issues that impact code quality
- π§ Fully Customizable: Tailor prompts and behavior to match your team's standards and practices
Feature | Description | Benefit |
---|---|---|
π PR Summarization | Auto-generates concise summaries and release notes | Save time understanding changes at a glance |
π Line-by-Line Reviews | Contextual suggestions for code improvements | Catch issues before they reach production |
π° Cost Optimization | Incremental reviews track only changed files | Minimize API costs while maximizing coverage |
π§ Smart Analysis | Uses Claude-3.5-Sonnet for superior reasoning | Get architectural insights, not just syntax checks |
β‘ Intelligent Skipping | Ignores trivial changes (typos, formatting) | Focus reviewer attention on meaningful changes |
π¨ Custom Prompts | Fully customizable review criteria and tone | Adapt to your team's standards and practices |
- Dual-Model Architecture: Light model for summaries, heavy model for deep analysis
- Conversation Mode: Respond to review comments with clarifications
- Flexible Filtering: Include/exclude files with glob patterns
- Multi-Language Support: Reviews in your preferred language
- Debug Mode: Detailed logging for troubleshooting and optimization
Get up and running in under 2 minutes! Just add one file to your repository.
Create .github/workflows/aigne-codesmith.yml
in your repository:
name: π€ AI Code Review
permissions:
contents: read
pull-requests: write
on:
pull_request:
pull_request_review_comment:
types: [created]
concurrency:
group:
${{ github.repository }}-${{ github.event.number || github.head_ref ||
github.sha }}-${{ github.workflow }}-${{ github.event_name ==
'pull_request_review_comment' && 'pr_comment' || 'pr' }}
cancel-in-progress: ${{ github.event_name != 'pull_request_review_comment' }}
jobs:
review:
runs-on: ubuntu-latest
steps:
- uses: aigne-io/aigne-codesmith@latest
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
with:
debug: false
review_simple_changes: false
review_comment_lgtm: false
Add your Anthropic API key to GitHub Secrets:
- Go to your repository Settings β Secrets and variables β Actions
- Click New repository secret
- Name:
ANTHROPIC_API_KEY
- Value: Your API key from Anthropic Console
Open a pull request and watch AIGNE CodeSmith provide intelligent reviews automatically.
- Large Codebases: Maintain consistent standards across multiple repositories
- Security Reviews: Catch security vulnerabilities before they reach production
- Architectural Guidance: Get insights on design patterns and best practices
- Onboarding: Help new team members learn from AI-powered feedback
- Code Quality: Maintain high standards without dedicated reviewers
- Knowledge Sharing: Learn from AI suggestions and improve coding skills
- Time Saving: Focus on building features, not catching basic issues
- 24/7 Reviews: Get feedback immediately, regardless of timezone
- Contributor Guidance: Help newcomers understand project standards
- Consistent Reviews: Maintain quality across diverse contributors
- Documentation: Auto-generate release notes and change summaries
- Accessibility: Support multiple languages for global contributors
"AIGNE CodeSmith caught 3 critical security issues in our authentication system that we missed in manual reviews. The architectural suggestions helped us improve our code structure significantly." β Sarah Chen, Lead Developer at TechFlow
"We reduced our code review time by 60% while actually improving code quality. The AI provides context that junior developers found incredibly valuable." β Marcus Rodriguez, Engineering Manager at DataSync
Parameter | Description | Default |
---|---|---|
anthropic_light_model |
Model for summaries | claude-3-5-sonnet-latest |
anthropic_heavy_model |
Model for detailed reviews | claude-3-5-sonnet-latest |
max_files |
Maximum files to review | 150 |
review_simple_changes |
Review trivial changes | false |
language |
Response language (ISO code) | en-US |
Tailor the AI's personality and focus areas by customizing the system_message
:
π― Security-Focused Reviews
system_message: |
You are a security-focused code reviewer. Prioritize identifying:
- Authentication and authorization flaws
- Input validation issues
- SQL injection vulnerabilities
- XSS prevention
- Data exposure risks
- Cryptographic best practices
ποΈ Architecture & Performance Reviews
system_message: |
You are an architectural reviewer focusing on:
- System design patterns
- Performance bottlenecks
- Scalability concerns
- Code maintainability
- SOLID principles
- Database optimization
Skip AI review for specific PRs by adding this to the PR description:
@codesmith: ignore
Use path_filters
to include/exclude files:
path_filters: |
src/**/*.ts
!**/*.test.ts
!dist/**
For complete configuration options, see action.yml
Team Size | Daily Usage | Estimated Cost* |
---|---|---|
5 developers | ~20 PRs | $3-5/day |
20 developers | ~80 PRs | $12-20/day |
50+ developers | ~200 PRs | $30-50/day |
*Costs vary based on code complexity and review depth. Claude models are significantly more cost-effective than GPT-4.
- Smart Caching: Avoids re-reviewing unchanged code
- Incremental Analysis: Only reviews modified files
- Model Selection: Light models for summaries, heavy models for complex analysis
- Concurrent Processing: Parallel file analysis for faster results
We β€οΈ contributions! Whether you're fixing bugs, adding features, or improving documentation, your help makes AIGNE CodeSmith better for everyone.
# Clone the repository
git clone https://github.com/aigne-io/aigne-codesmith.git
cd aigne-codesmith
# Install dependencies (Node 17+ required)
pnpm install
# Build and test
pnpm run build
pnpm test
# Package for distribution
pnpm run package
- π Found a Bug? Open an issue with reproduction steps
- π‘ Have an Idea? Discuss it in an issue before implementing
- π Improving Docs? Documentation PRs are always welcome
- π§ͺ Adding Tests? Help us improve our test coverage
Command | Purpose |
---|---|
pnpm run build |
Compile TypeScript |
pnpm test |
Run test suite |
pnpm run lint |
Check code style |
pnpm run format |
Format code |
pnpm run all |
Full CI pipeline |
- Code Style: We use Prettier and ESLint - run
pnpm run format
before committing - Tests: Add tests for new features, ensure existing tests pass
- Commits: Use clear, descriptive commit messages
- PRs: Include a clear description of changes and link to related issues
Contributors are recognized in our release notes and CONTRIBUTORS.md. Thank you for making AIGNE CodeSmith amazing! π
- π Documentation: Check our Wiki for detailed guides
- π Bug Reports: Open an issue with details
- π‘ Feature Requests: Suggest improvements
- π¬ Discussions: Join our GitHub Discussions
- β Star this repo to show your support
- π Watch releases to get notified of updates
- π΄ Fork to experiment with your own modifications
- π’ Share with your team and community
This project is licensed under the MIT License - see the LICENSE file for details.
Made with β€οΈ by the AIGNE Team
β Star on GitHub β’ π Report Issues β’ π‘ Request Features