Deployment

Deployment

Any host with Node 24+ works (Node 24 is required for the built-in node:sqlite).

Local / first run

git clone https://github.com/dripips/english-trainer.git
cd english-trainer
npm install
cp .env.example server/.env      # fill in, then: chmod 600 server/.env
npm run build:web                # build the PWA
npm start                        # Fastify serves the API + the built PWA

Configuration

All via environment variables in server/.env — see .env.example.

Minimum to boot: AUTH_SECRET, COOKIE_SECRET, SEED_USERS. Everything else is optional and degrades gracefully:

Variable(s)	Enables	Fallback if unset
LLM_TRANSLATE_URL/KEY/MODEL	context-aware translation + AI writing/speaking feedback	LibreTranslate / free MyMemory; feedback disabled
ELEVENLABS_API_KEY (+ MODEL/VOICE/SPEED)	natural TTS narration	browser speechSynthesis voice
VAPID_PUBLIC/PRIVATE/SUBJECT	push reminders	reminders disabled
TEXTBOOK_PDF, LIBRARY_DIR	private PDF reader / graded-reader library	sections hidden
ADMIN_USERNAMES	auto-promote admins on boot	seed roles only

Secrets live only in server/.env (chmod 600) — never commit them.

Reference production setup

git pull → PATH=/opt/node/bin:$PATH npm run build:web → systemctl restart english-trainer, behind a reverse proxy that terminates HTTPS and forwards to 127.0.0.1:$PORT.

• A systemd unit runs node server/src/index.js with the env file.
• Content-only changes (new lesson/vocab .md/.json): just git pull && systemctl restart — no rebuild.
• Code/image changes: rebuild the PWA (npm run build:web) before restart.

Data & backups

Everything stateful is under server/data/:

• app.db — SQLite (users, progress, SRS, settings…)
• tts-cache/ — generated speech (safe to delete; regenerates on demand)

Back up app.db. It survives deploys; keep AUTH_SECRET stable so existing logins keep working.