Learn Japanese kanji the fun way — built with Next.js, tRPC, and Supabase.

• Authentication — sign up / sign in via Supabase Auth; all study routes are protected.
• Learn — guided sessions that introduce new kanji (5 at a time).
• Review — spaced repetition (SRS) sessions that surface kanji based on your performance. See docs/srs.md for a full breakdown of the 10-stage system.
• Kanji browser — browse all JLPT N1–N5 kanji with meanings and readings.
• Play — two games to reinforce memory:
  • Memo — match kanji pairs (8 kanji per round).
  • Flashcards — quiz yourself on meanings and readings (10 kanji per round).

• Bun (package manager and runtime)
• Docker (required for local Supabase)
• Supabase CLI — available via bunx supabase (no global install needed)

bun install

cp .env.example .env

Then fill in .env:

DATABASE_URL=   # Supabase project URL
DATABASE_KEY=   # Supabase anon key

For local development, start Supabase first and get the values from bunx supabase status.

Start the local Supabase stack:

bunx supabase start

Seed kanji data, reset the database schema, and regenerate TypeScript types:

bun run supabase:reload

This command generates supabase/seed.sql from scripts/data.json, resets the local DB, and regenerates src/types/supabase.ts.

bun run dev

Open http://localhost:3000. Sign up for an account and start learning.

src/
  app/          # Next.js App Router pages and tRPC HTTP handler
  components/   # UI components (auth, kanji, learn, review, games, layout)
  server/       # tRPC router definitions (learn, review, kanji, flashcards, memo-game)
  actions/      # Next.js Server Actions (auth: sign-in, sign-up, sign-out)
  database/     # Supabase client, session middleware, seed script
  constants/    # App-wide constants (routes, game config, SRS config)
  hooks/        # Custom React hooks
  utils/        # Pure utility functions
  types/        # TypeScript types including generated Supabase types
supabase/
  migrations/   # SQL migration files
  seed.sql      # Generated kanji seed data (do not edit by hand)
scripts/
  scrape_kanji.py   # Python scraper — fetches JLPT kanji from jisho.org
  main.py           # Entry point — runs the scraper and writes data.json
  data.json         # Scraped kanji data (source of truth for the seed)
docs/
  srs.md        # Spaced repetition system specification

Kanji data (JLPT N1–N5) is scraped from jisho.org using the Python scripts in scripts/. The output (scripts/data.json) is the source for the database seed. Running bun run supabase:reload converts it to supabase/seed.sql via src/database/seed.ts and loads it into the local database. You only need to re-run the scraper if you want to refresh the kanji data.

To run the scraper (requires Python 3 + requests + beautifulsoup4):

cd scripts && python main.py