pr_reviewer/README.md

PR Reviewer

An automated pull request review system using CrewAI and MCP (Model Context Protocol).

Overview

This system provides automated code, security, and infrastructure reviews for pull requests using a multi-agent approach. It leverages CrewAI for orchestrating specialized review agents and MCP (Model Context Protocol) for integrating with various static analysis tools.

Features

• Code Review: Uses Semgrep (via MCP) to check code quality, best practices, and maintainability
• Security Review: Uses Trivy (native MCP) to identify security vulnerabilities
• Infrastructure Review: Uses Hadolint and Checkov (via MCP wrappers) to review Dockerfiles and Kubernetes manifests
• Contextual Review: Incorporates customizable guidelines for code, security, and infrastructure reviews
• Automated Orchestration: Uses CrewAI Flows to manage the review process
• REST API: FastAPI endpoint for triggering reviews
• Containerized: Docker support for easy deployment

Architecture

The system follows a modular architecture with:

• State management using Pydantic models
• LLM factory for flexible provider support (OpenAI, Anthropic, Ollama)
• Context resolution system for incorporating review guidelines
• Crew-based implementation for each review type (code, security, infrastructure)
• MCP server integrations for static analysis tools
• Flow-based orchestration for managing the review process
• RESTful API for integration with CI/CD systems

Installation

Prerequisites

• Python 3.10-3.13
• UV package manager
• Git
• Docker (optional, for containerized deployment)

Local Development

1. Clone the repository
2. Install UV package manager: curl -LsSf https://astral.sh/uv/install.sh | sh
3. Activate UV environment: source $HOME/.local/bin/env
4. Create virtual environment: uv venv .venv
5. Activate virtual environment: source .venv/bin/activate
6. Install dependencies: uv pip install -e .
7. Configure environment variables (see .env.example)

Docker Deployment

1. Build the Docker image: docker build -t pr-reviewer .
2. Run the container: docker run -p 8000:8000 --env-file .env pr-reviewer

Usage

API Endpoints

Health Check

GET /api/v1/health

Returns the health status of the service.

Trigger PR Review

POST /api/v1/review

Initiates a pull request review.

Request Body:

{
  "pr_id": "123",
  "title": "Add new feature",
  "description": "This PR adds a new feature to the application",
  "repo": {
    "name": "my-repo",
    "url": "https://github.com/user/my-repo"
  },
  "source": {
    "branch": "feature/new-feature",
    "commit": "abc123"
  },
  "target": {
    "branch": "main",
    "commit": "def456"
  },
  "files": [
    {
      "path": "src/main.py",
      "content": "print('Hello World')",
      "status": "modified",
      "additions": 1,
      "deletions": 0
    }
  ],
  "context": {
    "code_review": "Follow PEP8 guidelines",
    "security_review": "Check for SQL injection vulnerabilities",
    "infra_review": "Ensure Dockerfile follows best practices"
  }
}

Response:

{
  "review_id": "uuid-string",
  "status": "completed",
  "timestamp": "2023-05-08T10:00:00Z",
  "results": {
    "code_review": "Code review results...",
    "security_review": "Security review results...",
    "infra_review": "Infrastructure review results...",
    "summary": "Synthesized review summary..."
  },
  "metadata": {
    "processing_time_seconds": 45.2,
    "pr_id": "123",
    "repo": {
      "name": "my-repo",
      "url": "https://github.com/user/my-repo"
    }
  }
}

Configuration

Environment Variables

See .env.example for detailed configuration options.

Context Files

Default review guidelines are located in contexts/defaults/:

• code_review.md: Coding practice guidelines
• security_review.md: Security guidelines
• infra_review.md: Infrastructure guidelines

These can be overridden via the API context parameter.

Development

Running Tests

# Run unit tests
pytest

# Run tests with coverage
pytest --cov=src.pr_reviewer

# Run specific test categories
pytest tests/unit/
pytest tests/integration/

Code Style

The project uses Black for code formatting and Flake8 for linting.

Run formatting:

black src/

Run linting:

flake8 src/

Deployment

Kubernetes

Kubernetes manifests are available in the k8s/ directory:

• Secret for LLM configuration
• Deployment for the PR Reviewer service
• Service for exposing the API

Gitea Actions

GitHub Actions workflow for CI/CD is available in .gitea/workflows/deploy.yaml.

License

MIT

Contributing

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Open a pull request