import { Callout } from "zudoku/ui/Callout";
# Deploy to Cloudflare
This page deploys one store: the API server, the dashboard and the storefront,
each a Cloudflare Worker, sharing one D1 database. It assumes you have
[run Wocart locally](/self-hosting/local) from the same checkout.
A Cloudflare account with R2 enabled, a domain on Cloudflare for the API
(see [Domains](#5-domains)), and Wrangler signed in: `npx wrangler login`.
## 1. Create the resources
Choose your own names. Every installation gets new resources: never reuse
another project's.
```bash
npx wrangler d1 create mystore-db
npx wrangler r2 bucket create mystore-media
npx wrangler kv namespace create RATE_LIMIT
npx wrangler kv namespace create OAUTH_KV
npx wrangler kv namespace create DASHBOARD_RATE_LIMIT
npx wrangler queues create mystore-events
```
Keep the IDs each command prints. Then put the names in the root `.env`:
```bash title=".env"
COD_ACCOUNT_ID=
COD_DB_NAME=mystore-db
COD_R2_BUCKET_NAME=mystore-media
COD_SERVER_URL=https://api.yourdomain.com
```
## 2. Configure the API server
In `cod-server/wrangler.toml`, the `[env.production]` section is what a
production deploy reads. Fill in:
- the D1 `database_name` and `database_id`, the bucket name, both KV IDs and
the queue name, in the production bindings;
- the production variables: `WORKER_URL`, `WORKER_SELF_URL`, `BETTER_AUTH_URL`,
`ALLOWED_ORIGINS`, `STOREFRONT_URL`, and the media settings
([Media](/self-hosting/media)).
[Configuration](/self-hosting/configuration#api-server) explains each value.
Check that no placeholder is left:
```bash
grep -n "00000000" cod-server/wrangler.toml cod-client-astro/wrangler.toml
```
It must print nothing.
## 3. Create the database and the store
```bash
cd cod-server
npm run db:migrate:remote
STORE_API_KEY= npm run db:seed:remote
```
`db:migrate:remote` applies every migration to the deployed database: sign-in
fails until it has run. `db:seed:remote` creates the store, its API key (it
keeps only its hash), menus, default collections, legal pages and a shipping
profile; running it again changes nothing the store has. Generate the key with
`openssl rand -hex 32` and keep it for the storefront (step 4).
Then clear uploads that were never prepared, once for the bucket:
```bash
npm run r2:lifecycle
```
## 4. Deploy the three Workers
**The API server:**
```bash
cd cod-server
npm run deploy -- --env production
npx wrangler secret put MCP_LOGIN_TICKET_SECRET --env production
npx wrangler secret put REVIEW_UPLOAD_SECRET --env production
npx wrangler secret put IMAGEKIT_PRIVATE_KEY --env production # with ImageKit
```
**The dashboard.** In `cod-client-astro/wrangler.toml`, set the same D1
database, the `DASHBOARD_RATE_LIMIT` namespace's ID as `RATE_LIMIT_KV`, and
`PUBLIC_APP_URL`, `PUBLIC_API_URL`, `PUBLIC_TRUSTED_ORIGINS`. Then:
```bash
cd cod-client-astro
npm run deploy
npx wrangler secret put BETTER_AUTH_SECRET
npx wrangler secret put MCP_LOGIN_TICKET_SECRET # the same value as the API server's
ADMIN_EMAIL=you@yourdomain.com ADMIN_NAME=You npm run seed:admin:remote
```
`seed:admin:remote` creates your account locally and in the deployed
database, and prints its password and a merchant API key once.
**The storefront.** Theme-00's deploy takes `COD_SERVER_URL` from the root
`.env` and refuses a `localhost` value:
```bash
cd cod-astro/theme-00
npm run deploy
npx wrangler secret put STORE_API_KEY # the key you gave db:seed:remote
```
Theme-00 also builds for Vercel (`npm run build:vercel`) with the same three
values; see [Production and upgrades](/sdk/theme/production) for a theme of
your own.
## 5. Domains
Cloudflare blocks a Worker on `*.workers.dev` from calling another
`*.workers.dev` Worker (error 1042), so a storefront on `workers.dev` cannot
reach an API on `workers.dev`. Put at least the API on your own domain.
In the Cloudflare dashboard, open each Worker, then **Settings → Domains &
Routes → Add → Custom domain**: for example `api.yourdomain.com`,
`dashboard.yourdomain.com` and `shop.yourdomain.com`.
Once the domains are live, make every URL match them, and redeploy what
changed:
| Where | Values |
| --- | --- |
| `cod-server/wrangler.toml` | `WORKER_URL`, `WORKER_SELF_URL`, `BETTER_AUTH_URL`, `ALLOWED_ORIGINS`, `STOREFRONT_URL` |
| `cod-client-astro/wrangler.toml` | `PUBLIC_APP_URL`, `PUBLIC_API_URL`, `PUBLIC_TRUSTED_ORIGINS`. A changed `PUBLIC_API_URL` needs `npm run deploy`, which rebuilds |
| Root `.env` | `COD_SERVER_URL`, then the storefront's `npm run deploy` |
## 6. Check it
```bash
curl -s -o /dev/null -w "%{http_code}\n" https://api.yourdomain.com/api/docs # 200
```
Then sign in to the dashboard, open the storefront, and place a test order:
it appears in the dashboard's orders.
## When something fails
| Symptom | Cause | Fix |
| --- | --- | --- |
| Sign-in answers 500 | The deployed database is not migrated | `npm run db:migrate:remote` in `cod-server` |
| Sign-in answers `403 INVALID_ORIGIN` | The dashboard's address is not in `PUBLIC_TRUSTED_ORIGINS` | Add it, then `npm run deploy` |
| Signed in, but every screen fails with 401 | The API server does not accept the dashboard's tokens | `BETTER_AUTH_URL` must be the dashboard's `/api/auth` address, and `WORKER_SELF_URL` the API's own origin |
| The storefront shows an empty, unavailable store | It cannot read the API | `COD_SERVER_URL` must be the API's public domain, and `STORE_API_KEY` the key given to `db:seed:remote` |
| Image uploads fail | Media is not configured | The error code names what is missing: [Media](/self-hosting/media#when-uploads-fail) |
| Anything else | | `npx wrangler tail ` shows the Worker's errors while you retry |
## Keep it running
- **Logs**: `npx wrangler tail `, or the Worker's **Logs** tab.
- **Roll back** a Worker: `npx wrangler rollback`. A rollback does not undo a
migration, which is why a release keeps working on the database the
previous one left.
- **Back up** the database before an upgrade:
`npx wrangler d1 export --remote --output backup.sql`.
- **Upgrade** with [Upgrading](/self-hosting/upgrading).