Sentinel: The AI Pull Request Reviewer

This link shows a comment made to a PR by our bot: Sentinel-PR/sentinel-sandbox#4 (review)

🤖 An AI-powered GitHub PR review bot using LangGraph multi-agent workflows

Sentinel automatically reviews GitHub pull requests by analyzing code changes and posting structured feedback. It behaves like a senior reviewer, focusing on correctness, security, and maintainability.

🏗️ Architecture

GitHub Webhook → API Service → Redis Queue → Worker → LangGraph → GitHub Review
                     ↓              ↓           ↓
                PostgreSQL ←───────────────────┘

Services:

• API Service (FastAPI) - Receives webhooks, verifies signatures, enqueues jobs
• Worker Service - Consumes jobs, runs AI review workflow, posts reviews
• PostgreSQL - Job tracking, idempotency, audit trail
• Redis - Job queue, locks, SHA coalescing

🚀 Quick Start

Prerequisites

• Docker & Docker Compose
• Python 3.11+
• GitHub App credentials
• OpenAI API key

Setup

1. Clone and configure:

   git clone https://github.com/your-org/sentinel.git
   cd sentinel
   cp .env.example .env
   # Edit .env with your credentials

2. Start services:

   docker compose up -d

3. Verify:

   curl http://localhost:8000/health

Local Development

# Install dependencies
pip install -e ".[dev]"

# Run API service
uvicorn api.main:app --reload

# Run worker (separate terminal)
python -m worker.main

# Run tests
pytest

📁 Project Structure

sentinel/
├── api/                    # FastAPI webhook service
│   ├── routes/             # Endpoint handlers
│   ├── services/           # Business logic
│   └── models/             # Pydantic schemas
├── worker/                 # Background worker
│   ├── github/             # GitHub API client
│   ├── workflow/           # LangGraph review engine
│   └── output/             # Review publishing
├── shared/                 # Shared utilities
│   ├── db/                 # Database models & session
│   └── queue/              # Redis client
└── tests/                  # Test suite

🧠 LangGraph Workflow

Parse Diff → Classify PR → Fetch Context
                              ↓
            ┌─────────────────┼─────────────────┐
            ↓                 ↓                 ↓
   Correctness Agent   Security Agent   Maintainability Agent
            ↓                 ↓                 ↓
            └─────────────────┼─────────────────┘
                              ↓
                         Aggregator → Guardrails → Post Review

📋 Configuration

Key environment variables (see .env.example):

Variable	Description
GITHUB_APP_ID	GitHub App ID
GITHUB_WEBHOOK_SECRET	Webhook signature secret
OPENAI_API_KEY	OpenAI API key for LLM
DATABASE_URL	PostgreSQL connection string
REDIS_URL	Redis connection string

📖 Documentation

• Implementation Plan - Detailed architecture and phases
• API Reference - Coming soon
• Contributing - Coming soon

📜 License

MIT License - see LICENSE