---

• Overview
• Architecture Diagram
• Components
• Data Flow
• Deployment & Scaling

---

This page describes the end-to-end architecture for Autio’s MVP:

• Frontend: React SPA for login & chat
• Backend: FastAPI service exposing /auth/* and /chat/*
• Database: SQLite (dev) → PostgreSQL (prod)
• AI Service: OpenAI Chat Completions API

---

---

• Framework: React (Create React App)
• Routes:
  • /login → email/password form, stores JWT in localStorage
  • /chat → chat UI, sends messages with Bearer token
• Responsibilities: UI state, input validation, message rendering.

• Purpose: TLS termination, routing to backend instances.
• Examples: Nginx, AWS ALB, Cloudflare.

• Routers:
  • Auth: /auth/signup, /auth/login
  • Chat: /chat/ (protected)
• Security: OAuth2 password flow + JWT (python-jose, PassLib)
• DB Access: SQLAlchemy sessions via dependency injection
• AI Integration: calls OpenAI ChatCompletion in services/openai_client.py

• Dev: SQLite (file-based, no config)
• Prod: PostgreSQL (managed service)
• Tables:
  • users (id, email, hashed_pw, is_active)
  • (future) messages (id, user_id, role, text, timestamp)

• Service: OpenAI Chat Completions API
• Model: gpt-3.5-turbo (or newer)
• Config: temperature 0.7, system prompt “You are Autio, recommending cars.”

---

1. User Login

   1. Frontend POST /auth/login with {email,password}
   2. Backend verifies or creates user → issues JWT
   3. Frontend stores JWT

2. Chat Interaction

   1. User sends message via React → POST /chat/ + Authorization: Bearer <JWT>
   2. Backend validates JWT → extracts user
   3. Backend calls OpenAI → receives AI reply
   4. (Optional) Store user message & AI reply in DB
   5. Backend returns { reply: "..." }
   6. Frontend appends messages to chat window

---

• Containerization: Dockerize both frontend & backend
• Hosting:
  • Frontend → Vercel, Netlify, S3+CloudFront
  • Backend → AWS ECS/Fargate, Heroku, DigitalOcean App Platform
• Database: Migrate to managed Postgres (AWS RDS, Cloud SQL)
• Scaling Strategies:
  • Auto-scale backend instances behind load balancer
  • Introduce Redis for rate-limiting or caching
  • Use a queue (e.g. RabbitMQ, AWS SQS) for heavy processing or batching AI calls

---

This Wiki page captures Autio’s core system architecture for the MVP. Expand with network diagrams, security considerations, and operational runbooks as the project evolves.