Open Source · Interview Prep
In developmentPrepForge
A structured interview question bank with topic browsing and question-level progress APIs.
My role · Creator
Overview
An interview-preparation platform that organizes questions by technology and topic. Guests can browse the question bank; backend progress APIs support studying and completed states for signed-in users. A shared schema package keeps the web and API contracts aligned.
System design
Client
Service
Data
Shared
Connections
- REST /api/v1 · Fastify reads and writes PostgreSQL · Drizzle
- GraphQL · Mercurius reads and writes PostgreSQL · Drizzle
- Zod schemas package validates contracts for Next.js web app
- Zod schemas package validates contracts for REST /api/v1 · Fastify
01
Web · Next.js App Router
A React 19 frontend composes category and question sections into the landing page. Tailwind CSS 4 and reusable widgets structure the interface.
02
API · Fastify + GraphQL
Fastify registers REST modules for auth, categories, topics, questions and progress under /api/v1. Mercurius exposes schema-first GraphQL queries and a progress mutation.
03
Storage · PostgreSQL + Drizzle
Categories, topics, questions, users and user progress are modeled with Drizzle. Progress is keyed by user and question; structured answer tiers, hints and tags live alongside question content.
04
Contracts · Shared Zod schemas
A pnpm/Turborepo workspace contains apps/api, apps/web and packages/shared. Both applications consume shared validation schemas for questions, topics, users and progress.
Decisions
- A flat learning hierarchy
- Category → topic → tags avoids a separate subtopic layer. Studying and completed are explicit progress states; no progress record means unseen.
- REST and schema-first GraphQL
- REST endpoints expose resource operations while GraphQL lets the frontend request related topic, question and progress data through an explicit SDL contract.
- Batch per-question progress
- A GraphQL DataLoader batches progress lookups rather than resolving every question's progress in isolation.
- Share validation across workspaces
- Zod schemas are packaged separately from the applications, reducing duplicated definitions between the browser and API.
Implementation
- Modular Fastify registration, a health endpoint and centralized error handling for database and authentication failures.
- Drizzle schema and migrations for the content hierarchy, user records and question-level progress, with cascading progress foreign keys.
- GraphQL topic/question queries, user progress reads and a progress mutation backed by service modules and DataLoader.
- A Next.js landing page assembled from hero, category, recent-question and mobile-navigation widgets.
- Local PostgreSQL setup with Docker Compose and workspace scripts coordinated through Turborepo.
Results and learnings
The repository contains backend REST and GraphQL layers, database migrations, shared schemas and frontend landing-page widgets.
The handover records frontend authentication integration as unfinished. Backend auth code does not by itself establish a completed signed-in browser flow.
Spaced repetition, offline/PWA support and deployment/observability goals are not presented here as shipped features.
No production usage, performance or coverage measurements were verified. The main engineering takeaway is keeping shared contracts and per-question progress explicit as the product evolves.