SimplyDrive.gr: Architecture Notes from the Developer’s Desk

SimplyDrive.gr exists for a single reason: to give Greek learner-drivers the quickest, smartest way to master Κ.Ο.Κ. road-sign theory. Below is the tech-stack rundown of how I engineered high-speed loads, AI explanations, and progress-safe offline mode using Next.js 14, Prisma, and PostgreSQL.

1. Runtime Backbone — Next.js 14 App Router

1.1 Partial Prerendering (PPR)

All public pages stream a static shell in well under a second, then hydrate quiz data on the fly. Users feel native-app speed even on 3 G.

1.2 Server Actions

Mutations like submitAnswer() run as Server Actions—no separate REST layer. The action writes via Prisma and returns the updated score bar in a single round-trip.

2. Data Layer — Prisma ORM + PostgreSQL

Schema highlights:

User — UUID primary key, unique email, password hash, relation to answers.
Question — auto-increment ID, text prompt, correct answer reference.
Answer — tracks user-selected choice, correctness flag, timestamp; indexed by userId + questionId.

The database sits behind PgBouncer in transaction mode to survive the parallelism of Server Actions, while prisma migrate deploy handles schema changes inside the Vercel build.

3. User Accounts & Security

NextAuth v5 (Credentials provider) with Argon2id-hashed passwords.
HTTP-only, SameSite=Lax cookies store JWT sessions—never exposed to client JS.
All Prisma queries filtered by the current user’s ID to prevent horizontal privilege escalation.
middleware.ts injects CSP, X-Frame-Options, and Referrer-Policy headers via Helmet.

4. Real-Time & Offline UX

4.1 Optimistic SWR Store

A lightweight custom SWR hook shows instant UI feedback; server diffs reconcile automatically.

4.2 Service Worker

Cache-first strategy for static assets (JS, CSS, SVG road signs).
Background Sync queues any failed submissions and replays them when connectivity returns.
Last 50 answers kept in localStorage for crash recovery.

5. AI-Powered Explanations

Server-side calls to OpenAI create 75-word Greek explanations for each question. Results cache in Postgres for 24 hours to stay within rate limits.

6. Observability & CI/CD

GitHub Actions: ESLint, Vitest, and Playwright run on every pull request.
Automatic preview deployments on Vercel complete in < 90 seconds.
Structured logs shipped to Logtail; Web Vitals tracked via Vercel Analytics (target ≥ 90).

7. Roadmap

Edge-cache pre-warming once Partial Prerendering leaves beta.
Progress import/export via QR codes (PWA share-target).
Redis-backed leaderboard for friendly competition.

8. Final Thoughts

SimplyDrive.gr demonstrates how a lean stack—Next.js streaming, Prisma + Postgres, and targeted AI—can turn rote exam prep into an engaging experience that never loses a learner’s progress.

Enter a URL above to start browsing

No URL loadedStatus: ReadyReady

A Technical Overview of SimplyDrive.gr: Architecture and Functionality