Orylo - Real-Time Fraud Detection for Stripe

Orylo protège votre compte Stripe des fraudes que Radar laisse passer, grâce à une IA collective qui devient plus forte à chaque marchand protégé.

Orylo is an AI-powered fraud detection platform for Stripe merchants. Get real-time alerts on suspicious transactions and take action with one click.

Overview

Orylo provides real-time fraud detection for Stripe merchants, helping you identify and block fraudulent transactions before they become chargebacks. Our AI-powered system analyzes multiple risk factors in parallel, delivering detection results in under 350ms.

Features

• 7 AI Detectors: Velocity, geolocation, trust score, amount anomaly, time-of-day, device fingerprint, BIN analysis
• Real-Time Updates: Server-Sent Events (SSE) push new detections instantly to your dashboard
• Action-First: Block or whitelist customers with one click
• Trust Score System: Adaptive scoring based on transaction history and chargebacks
• Custom Rules: Define your own fraud rules (Epic 4)
• Multi-Tenant: Supports multiple organizations with complete data isolation
• Production-Ready: Idempotent webhooks, retry logic, dead letter queue, observability stack

Quick Start (Beta Users)

1. Sign up at orylo.com (beta access required)
2. Connect your Stripe account via OAuth
3. Configure webhook endpoint (automatic)
4. Start monitoring detections in real-time

For detailed setup instructions, see Beta Onboarding Guide (coming soon).

🏗️ Architecture

Monorepo Structure

Monorepo Turborepo avec Bun comme gestionnaire de packages.

orylo/
├── apps/
│   └── web/              # Frontend Next.js 16
├── packages/
│   ├── fraud-engine/     # 🧠 Fraud Detection Engine
│   ├── database/         # 💾 Drizzle ORM schemas
│   └── ui/               # 🎨 Shadcn/ui components
└── docs/                 # 📚 Documentation

Tech Stack

• Frontend: Next.js 16 (App Router), React 19, Tailwind CSS v4, Shadcn/ui
• Backend: Next.js API Routes, Better Auth, Drizzle ORM
• Database: Neon PostgreSQL (serverless)
• Cache: Upstash Redis
• Observability: Vercel Analytics, PostHog, tslog, Sentry (optional)
• Testing: Vitest (unit/integration), Playwright (E2E)
• Deployment: Vercel
• Monorepo: Turborepo
• Runtime: Bun

System Architecture

User Browser
    ↓
Next.js App (Vercel)
    ↓
┌─────────────────────────────────────┐
│  API Routes                         │
│  ├─ /api/detections (GET)          │
│  ├─ /api/customers/[id]/block      │
│  ├─ /api/webhooks/stripe (POST)    │
│  └─ /api/events (SSE stream)       │
└─────────────────────────────────────┘
    ↓                    ↓
Neon PostgreSQL      Redis Cache
    ↓
Stripe API (webhooks, payments)

Fraud Detection Pipeline

1. Webhook Ingestion: Stripe sends payment_intent.created event
2. Idempotency Check: Verify event not already processed (Story 3.1)
3. Detection: 7 detectors run in parallel (<350ms target)
4. Decision: Aggregate scores → Allow/Review/Block
5. Persistence: Save detection to PostgreSQL
6. Real-Time Push: SSE broadcasts detection to connected clients
7. Action: Merchant can block/whitelist via dashboard

Performance

• P95 Latency: <350ms (webhook → detection saved)
• Detector Speed: Each <100ms (parallel execution)
• SSE Latency: <500ms (detection created → UI update)
• Bundle Size: <500KB (frontend, optimized)
• Database: Indexed queries, selective column fetching
• Cache: Redis with 1h TTL for trust scores, 5min for velocity data

🚀 Quick Start

Prérequis

• Bun >= 1.2.3
• Node.js >= 20.0.0
• PostgreSQL database (ou Neon serverless)

Installation

# Installer les dépendances
bun install

# Copier le fichier d'environnement
cp apps/web/.env.example apps/web/.env

# Configurer votre DATABASE_URL dans apps/web/.env

Développement

# Démarrer le dev server (tous les packages)
bun dev

# Démarrer uniquement le web app
bun --filter @orylo/web dev

# Build tous les packages
bun run build

# Linter
bun run lint

# Type-checking
bun run type-check

Database

# Générer les migrations Drizzle
bun --filter @orylo/database db:generate

# Appliquer les migrations
bun --filter @orylo/database db:push

# Ouvrir Drizzle Studio
bun --filter @orylo/database db:studio

📦 Packages

@orylo/fraud-engine

Engine de détection de fraude modulaire avec injection de dépendances.

Features :

• ✅ Interface IDetector pour extensibilité
• ✅ Branded types pour type-safety
• ✅ Stratégies de scoring pluggables
• ✅ Exécution parallèle des détecteurs
• ✅ Performance < 250ms garantie

@orylo/database

Schémas Drizzle ORM pour PostgreSQL.

Tables :

• organizations - Multi-tenancy
• fraud_detections - Résultats de détection
• customer_trust_scores - Trust score par client
• custom_rules - Règles personnalisées

🧪 POC Validation

Pour valider que tout fonctionne :

# Démarrer le dev server
bun dev

# Ouvrir http://localhost:3000/demo

La page de démo montre :

• ✅ Import du package @orylo/fraud-engine dans Next.js
• ✅ Exécution du Fraud Detection Engine
• ✅ Résultats en temps réel

📚 Documentation

• Architecture ADRs
• Brainstorming Results
• GTM Strategy

🛠️ Stack Technique

• Framework: Next.js 16 (App Router)
• Runtime: Bun
• Database: PostgreSQL (Neon Serverless)
• ORM: Drizzle ORM
• Auth: Better Auth (avec Organizations)
• UI: Shadcn/ui + Tailwind CSS v4
• Payments: Stripe
• State: Zustand + React Query
• Background Jobs: Trigger.dev
• Monorepo: Turborepo

🎯 Roadmap MVP (6 semaines)

• Sprint 0 (S1): Foundation ✅ POC validé !
• Sprint 1-2 (S2-3): 6 Detectors + Scoring
• Sprint 3-4 (S4-5): Dashboard UI
• Sprint 5-6 (S6): Beta testing

Security & Compliance

• PCI Compliant: No full card numbers stored (only Stripe tokens: pi_xxx, cus_xxx)
• GDPR Compliant: 90-day data retention, right to deletion, data export API
• Multi-Tenancy: Row-level security, organizationId isolation
• HTTPS Only: Enforced via Vercel
• Secure Sessions: HttpOnly, Secure, SameSite cookies (Better Auth)
• Data Encryption: Database encryption at rest (Neon PostgreSQL)

Contributing (Beta Phase)

We're currently in private beta. If you encounter bugs or have feature requests:

1. Check existing issues
2. Use issue templates for bug reports or feature requests
3. Join our Discord community for support (beta invite required)

Support

• Discord: Join Beta Community (invite-only during beta)
• Email: support@orylo.com
• Documentation: docs.orylo.com (coming soon)

📝 License

Proprietary - © 2026 Orylo (Beta - internal use only)