学習アプリ
Overview
A daily-practice PWA built around a 'study pack' architecture: questions, glossary, and domain weighting are swappable per subject, currently covering the AWS Certified Solutions Architect – Associate exam and WordPress admin/dev practice, with room to add more. Each day's 10 questions are allocated across a pack's domains and picked in deterministic priority order: unseen → missed → related topics → stale. Answers show an explanation and a one-line takeaway immediately, and missed topics get folded into the next day's set for review. Signing in with Google (via Supabase Auth) ties streaks, per-domain accuracy, and the target-date countdown to your account, so progress carries over between phone and desktop. A Learn tab lets you tap underlined acronyms to pop open their full name and meaning inline. Set your own Gemini API key and weak-spot questions get generated on the fly (the AWS pack also factors in the official What's New RSS), with Vercel Cron pre-generating each morning so there's no wait. Questions are original exercises grounded in each subject's official docs, not copies of real exam questions.
Challenges & Solutions
[PROBLEM_1] Deterministic daily question selection
A date-seeded hash drives a pseudo-random shuffle so reopening the app on the same day always yields the same 10 questions. Candidates are ranked unseen → high miss-rate → related topic → stale before drawing, balancing weak-spot review with new material.
[PROBLEM_2] Carrying pre-login progress into an account
Users can start solving immediately before signing in—progress is stashed in sessionStorage until Google login completes, then merged into Supabase via RPC (fetch_progress / push_progress). This preserves a 'try 10 questions, then sign in if you like it' flow while still syncing across devices.
[PROBLEM_3] A study-pack architecture for multiple subjects
Pulled questions, glossary, and domain weighting out of the AWS-specific code into a 'study pack' (StudyPack) abstraction. Selection logic, weak-spot tracking, and the AI generation engine stayed subject-agnostic in a shared core, so adding the WordPress pack meant just writing a new pack definition.
Future Improvements
More study packs
Currently AWS SAA and WordPress—planning to add packs for more certifications and technical subjects over time.
Deeper study analytics
Beyond per-domain accuracy, exploring a dashboard for correlations between missed topics.
Tech Stack
- Framework: Next.js 16 (App Router) / React 19
- Language: TypeScript
- Styling: Tailwind CSS v4（ダークUI）
- Auth: Supabase Auth（Google）
- Database: Supabase（Postgres + RPC）
- AI: Google Gemini API（弱点問題のパーソナライズ生成、ユーザー自身のAPIキー）
- 通知: Web Push（VAPID）+ Vercel Cron（日課リマインダー / AI問題の事前生成）
- 配信: Vercel / PWA（manifest + Service Worker）