Environment variables

Where the settings live

Create each app's file from the .env.example beside it, which documents every variable: cp .env.example .env. The examples work as they are for a local run.

Where the settings live

One file, .env in the website's folder, read when the website is built. Every value in it is public, so it never holds a secret: payment and AI keys live in the API's .env. Create it from .env.example, which points at an API on http://localhost:8000: cp .env.example .env.

Where the settings live

One file, .env in the dashboard's folder, read when the dashboard is built. Every value in it is public, so it never holds a secret: the AI keys live in the API's .env. Create it from .env.example, which points at an API on http://localhost:8000: cp .env.example .env.

Where the settings live

One file, .env in the API's folder, read by the API with yarn dev, and by a container when you pass it with --env-file .env. It holds secrets: never commit it. Create it from .env.example, which documents every variable and works as it is for a local run: cp .env.example .env.

The Docker image sets its own defaults, so docker run works without a file: SQLite at /data/database.sqlite, uploads in /data/uploads, CORS_ORIGIN for the two local frontends, PUBLIC_MEDIA_URL=http://localhost:8000/media and SEED_DEMO_DATA=false. Override any of them with -e.

API essentials

The API checks JWT_SECRET and, in production, CORS_ORIGIN before it starts. If one is missing or unusable, it stops with one line saying what to fix.

Generate your own JWT_SECRET with:

node -e "console.log(require('crypto').randomBytes(48).toString('hex'))"

Every optional feature below is off while its lines in .env.example stay commented out.

Use MySQL instead of SQLite

Create an empty database, then set the driver and the connection in the API's .env:

DB_TYPE=mysql
DB_HOST=your-mysql-host
DB_PORT=3306
DB_USERNAME=your-mysql-username
DB_PASSWORD=your-mysql-password
DB_DATABASE=your-database-name

Then run yarn seed for the demo restaurant, or yarn db:sync for the tables with no data. The running API only creates and updates the tables itself when NODE_ENV=development, so with NODE_ENV=production one of those commands runs before the first start (yarn seed:prod or yarn db:sync:prod after yarn build).

Ordering website and staff dashboard

Every NEXT_PUBLIC_* value is compiled into the JavaScript the browser loads, and anyone who opens the page can read it. Never put a secret in these files, and rebuild after changing one.

The website's marketing tags are NEXT_PUBLIC_* values too.

• Analytics and tracking pixels

Docker Compose options

Nothing has to be set for a local run. To change something, put it in a .env file beside docker-compose.yml and run docker compose up --build again: the frontends compile these values in.

Change a port when another program already uses it, such as ADMIN_PORT=3041. The API's CORS_ORIGIN, FRONTEND_URL, STOREFRONT_URL, API_PUBLIC_URL and PUBLIC_MEDIA_URL, and the frontends' API address, are all built from the ports and the three addresses, so they follow by themselves.

The API container also reads back-end/.env when it exists, so payments, mail, media and AI keys are set in one place for both yarn dev and Docker. The compose file wins for the values that differ inside a container: the database path, the upload folder, the port, NODE_ENV=production and the addresses above. The database and the uploads live in the food-studio-data volume.

Media storage

The API serves the seeded pictures itself, from public/media/food-studio, at /media/food-studio/…. Uploads from the dashboard (dish photos, profile pictures, the AI studio's results) are stored on the API's disk under MEDIA_UPLOAD_DIR and served at /media/uploads/…, unless a bucket is set.

PUBLIC_MEDIA_URL=http://localhost:8000/media
MEDIA_UPLOAD_DIR=./public/media/uploads

PUBLIC_MEDIA_URL is the address /media is reached at from a browser. The seed writes it into every picture's address, so set it before the first seed on a server. On a server, point MEDIA_UPLOAD_DIR at persistent storage; with Docker it is in the data volume.

To send uploads to Cloudflare R2, or any S3-compatible bucket, set all five variables, and give both frontends the bucket's public host through NEXT_PUBLIC_MEDIA_HOSTNAME (with Docker Compose, MEDIA_HOSTNAME). Set all five or none.

R2_ACCESS_KEY_ID=your-r2-access-key-id
R2_SECRET_ACCESS_KEY=your-r2-secret-access-key
R2_ENDPOINT=https://your-account-id.r2.cloudflarestorage.com
R2_BUCKET_NAME=your-bucket-name
R2_PUBLIC_URL=https://your-public-url.r2.dev

Payments

Checkout sends the customer to Stripe's or PayPal's own page. Set STRIPE_SECRET_KEY and STRIPE_WEBHOOK_SECRET, or PAYPAL_CLIENT_ID, PAYPAL_CLIENT_SECRET and PAYPAL_WEBHOOK_ID (with PAYPAL_ENV, sandbox by default). With no provider set, checkout offers a built-in demo payment that marks an order paid without moving money.

STRIPE_SECRET_KEY=sk_test_your-stripe-secret-key
STRIPE_WEBHOOK_SECRET=whsec_your-webhook-secret

• Set up payments

Email

Password-reset links are sent through Resend. Both values are needed: a sending-access key, and a From address on a domain verified in Resend. Without them "Forgot password" still answers as usual, no message is sent, and the log says mail is not configured.

RESEND_API_KEY=re_your-sending-access-key
MAIL_FROM=Food Studio <noreply@your-domain.com>
MAIL_MAX_PER_ADDRESS_PER_DAY=5

The link opens the website's reset page at STOREFRONT_URL, in the customer's language, and works once within an hour. One address receives at most MAIL_MAX_PER_ADDRESS_PER_DAY messages a day, 5 by default.

AI features

The MCP server

Demo mode

Going live

Before you deploy anywhere public:

1. In the API's .env, set NODE_ENV=production and a long random JWT_SECRET of your own.
2. Set CORS_ORIGIN to the website's and the dashboard's addresses, FRONTEND_URL to the dashboard's, STOREFRONT_URL to the website's, API_PUBLIC_URL to the API's, and PUBLIC_MEDIA_URL to the API's address followed by /media.
3. On a new database, run yarn build, then yarn seed:prod once (or yarn db:sync:prod for empty tables), then yarn start:prod. There are no migrations: the seeder is the schema tool.
4. Keep the uploads on persistent storage: a bucket, or MEDIA_UPLOAD_DIR on a disk that survives a deploy.
5. Point your host's health check at /api/health, and set TRUST_PROXY when a reverse proxy stands in front of the API.
6. Build each frontend with NEXT_PUBLIC_API_BASE_URL at your deployed API; the dashboard also with NEXT_PUBLIC_WEBSOCKET_BASE_URL, NEXT_PUBLIC_DASHBOARD_URL and NEXT_PUBLIC_STOREFRONT_URL.
7. On the website, set domain.url in src/config/brand.config.ts, and rewrite the privacy policy and terms in messages/legal for your business.
8. In the dashboard, set your kitchens, their hours and delivery postcodes, your fees and promo codes, and switch the service clock to Real: the seed starts it on the demo clock, which keeps every kitchen open.
9. Change the seeded passwords, or delete the accounts, and switch payments to live keys.