🔍 AI Code Review Platform

A high-scale microservices backend providing automated AI-powered code reviews, utilizing secure RESTful APIs and LangChain for intelligent feedback loops integrated via Git hooks.

Architecture

graph TD
    A[Git Hooks / CLI] -->|POST /api/v1/reviews| B(Gateway: Node.js)
    C[GitHub Webhooks] -->|POST /api/v1/reviews| B
    B -->|Publish| D{Kafka: code-review-requests}
    D -->|Consume| E[Review Worker: LangChain]
    E -->|Publish Results| F{Kafka: review-results}
    F -->|Consume| G[Gate Check Service: Spring Boot]
    E -->|API| H[GitHub PR Comments]
    G -->|API| I[GitHub Status Checks]

Loading

Services

Service	Technology	Port	Description
gateway	Node.js + Express	3000	REST API ingestion gateway
review-worker	Node.js + LangChain	-	Kafka consumer, AI review engine
gate-service	Java + Spring Boot	8080	Gate checks & metadata extraction
cli	Node.js	-	Developer CLI tool

Quick Start

Prerequisites

• Node.js >= 18
• Java >= 17
• Docker & Docker Compose (for Kafka)

1. Start Infrastructure

docker-compose up -d

2. Install & Run Services

# Gateway
cd services/gateway && npm install && npm run dev

# Review Worker
cd services/review-worker && npm install && npm run dev

# Gate Service
cd services/gate-service && ./gradlew bootRun

3. Install CLI

cd cli && npm install && npm link

4. Run a Review

# Review staged changes
ai-review review --staged

# Review a specific commit
ai-review review --commit HEAD~1

# Review a PR
ai-review review --pr https://github.com/org/repo/pull/1

Configuration

Environment Variables

Each service requires specific environment variables to operate. Create a .env file in the root of each service.

Gateway (services/gateway/.env)

PORT=3000
KAFKA_BROKERS=localhost:19092
CORS_ORIGIN=*
RATE_LIMIT_WINDOW_MS=900000
RATE_LIMIT_MAX=100

Local Config File

Create a .ai-review.yaml in your project root:

llm:
  provider: openai       # openai | anthropic | gemini
  model: gpt-4o
  temperature: 0.2

review:
  categories:
    - security
    - performance
    - code-quality
    - best-practices
  severity_threshold: WARNING  # CRITICAL | WARNING | INFO

gate_checks:
  max_critical_issues: 0
  max_warnings: 5
  require_tests: true

License

MIT