AI-powered GitHub Pull Request reviewer built with Google Gemini, LangGraph, FastAPI, and Tree-sitter.
AI Code Reviewer is an automated code review system that analyzes GitHub Pull Requests using a multi-agent AI architecture.
Instead of relying on a single AI model to review everything, the system uses specialized agents for different categories of code analysis — including bugs, security, performance, code quality, and testing.
The findings are then passed through a Final Reviewer, Critic Agent, and deterministic Validator before verified issues are posted directly to the GitHub Pull Request.
- 🔗 GitHub App Integration
- 🪝 GitHub Webhook-based PR triggering
- 🤖 Google Gemini-powered code analysis
- 🧠 LangGraph multi-agent orchestration
- 🐛 Bug detection
- 🔐 Security vulnerability analysis
- ⚡ Performance analysis
- 🧹 Code quality analysis
- 🧪 Testing analysis
- 🌳 Tree-sitter AST analysis
- 🔍 Code intelligence from AST structure
- 🧐 Final Reviewer for issue consolidation
- 🔎 Critic Agent for false-positive filtering
- ✅ Deterministic review validation
- 💬 Automatic GitHub Pull Request inline comments
- 📁 Multi-file Pull Request analysis
- 🔑 GitHub App JWT and installation-token authentication
- 📊 Structured AI review output using Pydantic
Developer
│
▼
GitHub Pull Request
│
▼
GitHub Webhook
│
▼
FastAPI
│
▼
GitHub App Authentication
│
▼
Fetch PR Information
│
┌─────────┴─────────┐
│ │
▼ ▼
PR Diff / Files Tree-sitter
│
▼
Code Intelligence
│
▼
┌─────────────────────┐
│ LangGraph Pipeline │
└─────────────────────┘
│
┌───────────────┬───────┼────────┬───────────────┐
│ │ │ │ │
▼ ▼ ▼ ▼ ▼
🐛 Bug 🔐 Security ⚡ Performance 🧹 Quality 🧪 Testing
Agent Agent Agent Agent Agent
│ │ │ │ │
└───────────────┴───────────┴───────────┴───────────┘
│
▼
🔍 Final Reviewer
│
▼
🧐 Critic Agent
│
▼
✅ Validator
│
▼
💬 GitHub PR Review
│
▼
Verified Comments