日本語 Vocab — Japanese Study App

A personal Next.js app for studying Japanese vocabulary. Each word has 5 attributes: Kanji, Romaji, English meaning, Tips (Marathi), Category (noun/verb/adjective/…).

Features

• Accounts — Supabase email/password auth. Each user has their own private vocab list (isolated in the database via Row Level Security). Open sign-up.
• Add Vocab — form to add words.
• Flashcards (spaced repetition) — front shows Kanji + Romaji; I remember records a confident recall, or flip to reveal the meaning and grade Got it / Forgot. A science-backed SM-2 scheduler resurfaces forgotten cards frequently and pushes remembered ones to growing intervals. Category filter + Marathi hint. Design notes: docs/spaced-repetition.md.
• Vocab List — searchable table of all columns with inline edit & delete.
• Export — download the whole list as .docx or PDF (kanji + Marathi render correctly via embedded Noto fonts).
• Import — upload CSV / Excel (.xlsx, .xls) / Word (.docx) / PDF with a 5-column table; review/edit the parsed preview, then confirm to bulk-insert.

Tech

Next.js (App Router, TS) · Tailwind CSS v4 · Supabase (Postgres) · docx, pdf-lib (export) · papaparse, xlsx, mammoth, pdf-parse (import).

Setup

1. Create the Supabase project

1. Create a free project at https://supabase.com.
2. Open SQL Editor → New query, paste the contents of supabase/schema.sql, and run it to create the vocab table.
3. Go to Project Settings → API keys and copy the Project URL and the publishable (anon) key.

2. Configure auth (Supabase dashboard)

• Authentication → Providers → Email: enable it, and turn "Confirm email" OFF for the smoothest sign-up (the free-tier email sender is rate-limited; the login UI handles both modes).
• Authentication → Sign In / Providers: keep "Allow new users to sign up" ON (open sign-up).
• Authentication → URL Configuration: set the Site URL to your production origin (not localhost), and add Redirect URLs for every origin you use, e.g. https://your-domain.com/**, http://localhost:3000/** (and your ngrok HTTPS origin if you tunnel the dev server). Password-reset links use the deployed site URL; if the production origin is missing here, Supabase falls back to Site URL (often localhost).

3. Configure environment

cp .env.local.example .env.local

Fill in:

NEXT_PUBLIC_SUPABASE_URL=https://<your-project>.supabase.co
NEXT_PUBLIC_SUPABASE_PUBLISHABLE_KEY=<your-publishable-key>
NEXT_PUBLIC_SITE_URL=https://your-domain.com

NEXT_PUBLIC_SITE_URL is optional locally (the app derives the origin from the request). Set it in production so password-reset emails always link to your deployed site. Per-user data isolation is enforced by Row Level Security, so the publishable key is all the app needs — no service-role key is stored in the app.

4. Run, sign up, then enable per-user lists

npm install
npm run dev

Open http://localhost:3000 → you'll be redirected to /login. Create your owner account. Then enable auth scoping + migrate your existing words: open SQL Editor, set the owner email at the top of supabase/0002_auth_rls.sql, and run it. (It adds user_id, backfills your existing rows to your account, and turns on RLS + per-user policies.)

5. Enable study decks and scheduling

In SQL Editor, run the remaining migrations in numeric order: 0003_srs.sql through 0010_kanji_prompt_hints.sql. They add the independent word/Kanji schedules, per-character Smart Kanji cards, curated selections, lossless reconciliation, atomic review commits and history, idempotent review retries, and safe All Kanjis reading hints. Existing words enter the flashcard rotation as new cards. How the scheduler works: docs/spaced-repetition.md.

Import file format

The first row may be a header. Recognized column names (case-insensitive, any order): kanji/japanese/word, romaji/reading, english/meaning, tips/marathi/hint, category/type. If no header is recognized, columns are read positionally in the order above. CSV and Excel parse most reliably; PDF table parsing is best-effort — always review the preview.

Fonts

PDF export embeds public/fonts/NotoSansJP-Regular.ttf (kanji) and public/fonts/NotoSansDevanagari-Regular.ttf (Marathi). Default PDF fonts cannot render these scripts.

Auth & data model

• Email/password via Supabase Auth, using @supabase/ssr (cookie sessions). Session refresh and route gating live in proxy.ts (Next.js 16 renamed middleware → proxy).
• Every API route runs as the signed-in user through a per-request, RLS-scoped client (lib/supabase/server.ts); vocab.user_id + RLS policies isolate data.
• Exports (.docx/PDF) contain only the signed-in user's words.

Notes

• xlsx (SheetJS) from the npm registry has known advisories; acceptable for personal/local use.