Production stack

The engineering behind Anto

Full-stack architecture, conversational AI, and real-time messaging built for reliable everyday use.

React Native (Expo) → Express.js / Socket.IO → MongoDB → OpenAI GPT-5.4 Mini

Production technology stack

Libraries and services that power the Anto mobile app, API, and this website.

Mobile frontend

  • React NativeiOS + Android · shared codebase
  • Expo SDKEAS Build · OTA updates
  • TypeScriptStrict typing across the app
  • React NavigationNative stack + tabs
  • Socket.IO ClientBidirectional events <2.5s
  • Expo NotificationsPush APNs / FCM

Web frontend

  • Next.js 14App Router · SSR/SSG
  • React 18Server + Client Components
  • TypeScript 5tsc --noEmit in CI
  • CSS design tokensVariables · dark mode
  • Service WorkersPWA offline-first
  • socket.io-clientLive sync with backend

Backend

  • Node.jsLTS runtime · JSON APIs
  • Express.jsMiddleware chain · REST
  • MongoDB + MongooseSchemas · chat indexes
  • Socket.IOPer-user rooms · heartbeat
  • WinstonStructured JSON logs
  • SentryProduction error traces

AI and data

  • OpenAI GPT-5.4 MiniChat completions · streaming
  • Prompt pipelineContext + clinical-practical tone
  • Sentiment analysisPre-filter before LLM
  • Crisis detectionPatterns + Twilio / SendGrid
  • RedisSession and response cache
  • Backend i18nES/EN replies by locale

Security

  • JWT + refreshSessions with expiry
  • bcryptCost factor 12 on passwords
  • Helmet.jsCSP · HSTS · XSS
  • JoiAPI payload validation
  • DOMPurifyInput sanitisation
  • Rate limitingAnti-abuse per IP and user

DevOps and integrations

  • GitHub ActionsCI: lint · type-check · build
  • VercelNext.js website · preview branches
  • RenderNode.js API · auto-deploy from Git
  • Mercado PagoIn-app subscriptions
  • SendGridTransactional email
  • TwilioWhatsApp/SMS on crisis
  • ESLint + PrettierUnified style in monorepo

Layered architecture

Five decoupled layers: mobile and web clients, Node.js API with WebSockets, AI pipeline with session context, and MongoDB + Redis persistence. Each layer has its own scaling and security policies.

📱

Mobile client

React Native + Expo (EAS Build). Local state with AsyncStorage, React Navigation, Socket.IO client for real-time events.

React NativeExpo SDKTypeScriptSocket.IO Client
🌐

Web client

Next.js 14 App Router, strict TypeScript, CSS design tokens, PWA with Service Workers, and ES/EN i18n routes.

Next.js 14React 18TypeScriptPWA
⚙️

API and real-time

Node.js + Express.js: JSON REST, security middleware (Helmet, rate limit), Socket.IO WebSockets, and structured logging (Winston).

Node.jsExpress.jsSocket.IOREST + WS
🤖

AI layer

OpenAI GPT-5.4 Mini via REST API: session-context prompts, MongoDB chat history, and sentiment analysis before each reply.

GPT-5.4 MiniOpenAI APINLPPrompt pipeline
💾

Persistence

MongoDB + Mongoose (user, chat, subscription documents). Redis for session cache and frequent-response queues.

MongoDBMongooseRedisCompound indexes

Technical Challenges and Solutions

⚡

Real-Time Performance

Challenge: Process and respond to conversations in real time with low latency.

Solution: WebSockets with Socket.IO, Redis caching, and optimized database queries.

🔒

Security and Privacy

Challenge: Ensure total privacy of sensitive mental health conversations.

Solution: End-to-end encryption, JWT authentication, security headers, and GDPR/HIPAA compliance.

📱

Cross-Platform

Challenge: Maintain consistency across iOS, Android, and Web.

Solution: React Native for mobile, shared code, and responsive design for web.

🤖

AI Integration

Challenge: Integrate GPT efficiently with emotional context.

Solution: Optimized prompt system, common response caching, and prior sentiment analysis.

📊

Scalability

Challenge: Scale to handle millions of users and conversations.

Solution: Modular architecture, distributed databases, CDN, and cloud auto-scaling.

🎨

User Experience

Challenge: Create an intuitive and accessible interface for all users.

Solution: User-centered design, constant testing, WCAG accessibility, and continuous feedback.

Development Metrics

15K+
LOC TypeScript/JS
PWA
Responsive web + iOS/Android apps
97%+
CI test pass rate
<2.5s
Chat latency (p95)
REST+WS
Dual API (HTTP + Socket.IO)
6
Active security layers

Build pipeline

From monorepo to production (Vercel web · Render API · EAS · GitHub Actions)

1

Planning and Design

Requirements analysis, architecture design, wireframes, and prototypes. Technology stack definition and database structure.

  • System architecture design
  • UI/UX prototyping
  • API definition
  • Sprint planning
2

Backend Development

RESTful API implementation, AI service integration, authentication and authorization system, and database configuration.

  • REST API with Express.js
  • GPT-5.4 Mini integration
  • JWT authentication system
  • End-to-end encryption
  • WebSockets for real-time
3

Frontend Development

Mobile app development with React Native and website with Next.js, implementing PWA and performance optimizations.

  • Cross-platform mobile app
  • Responsive website
  • PWA with Service Workers
  • Performance optimizations
  • Animations and micro-interactions
4

Integration and Testing

Integration of all components, unit, integration, and user testing. Performance and security optimization.

  • Unit and integration testing
  • Security testing
  • Performance optimization
  • Load testing
  • Bug fixes
5

Deployment and Monitoring

Production deployment, CI/CD configuration, error and performance monitoring, and continuous updates.

  • Website on Vercel (Next.js)
  • API on Render (Node.js)
  • CI/CD with GitHub Actions
  • Monitoring with Sentry
  • Analytics and metrics
  • Continuous updates
👨‍💻

Lead Developer

Marcelo Ull Marambio

Full-stack engineer: React Native/Expo, Next.js, Node.js, MongoDB, Socket.IO, and OpenAI integrations for production mental-health products.

📧marcelo.ull@antoapps.com

Questions about the stack?

For technical inquiries, integrations, or collaboration on infrastructure, get in touch.

ContactDirect Email