#Deploy Your Own

StoryLark is built to be deployed once per brand from the same codebase: one Cloudflare Worker, one D1 database, one R2 bucket per brand. This guide stands up a new branded site end to end.

Everything below uses placeholders — <your-...>, example.com, 00000000-.... Never commit real account IDs, database IDs, domains, or secrets. The base brand's wrangler.jsonc env ships with placeholder IDs on purpose; fill in your own only in your deployment, and keep secrets out of the repo (wrangler secret put, below).

#1. Create your brand

Copy the base brand and give it a new id:

cp -r brands/storylark brands/<your-id>

Edit brands/<your-id>/brand.json (identity, origins, nouns, TTS) and brands/<your-id>/theme.css (colors + fonts). The full field-by-field reference is in build-your-own-theme.md. At minimum, set:

• id — must match the folder name and the --mode you build with.
• appName, name, shortName, tagline, author.
• appOrigin — where the app is served (e.g. https://app.example.com).
• contentOrigin — where R2 content is served (e.g. https://content.example.com).
• themeColor / backgroundColor — must match your theme's paper color.
• layout and nouns — see build-your-own-presentation.md.

#Icons

The manifest references three PNGs under brands/<your-id>/assets/icons/:

Drop in your own artwork, or generate neutral placeholder icons in your brand colors:

node tools/gen-icons.mjs --brand <your-id>

(The base brand also carries favicon.svg, favicon-32.png, favicon-180.png, and a logo.svg — supply your own equivalents if your HTML references them.)

#2. Add a Wrangler env for the brand

wrangler.jsonc defines one env per brand, selected with --env. Copy the storylark block, rename it to <your-id>, and fill in your own resources:

"env": {
  "<your-id>": {
    "name": "<your-id>",
    "routes": [{ "pattern": "app.example.com", "custom_domain": true }],
    "d1_databases": [
      {
        "binding": "DB",
        "database_name": "<your-id>",
        "database_id": "<your-d1-database-id>",
        "migrations_dir": "worker/migrations"
      }
    ],
    "r2_buckets": [
      { "binding": "CONTENT", "bucket_name": "<your-id>-content" }
    ],
    "vars": {
      "BRAND": "<your-id>",
      "APP_ORIGIN": "https://app.example.com",
      "CONTENT_ORIGIN": "https://content.example.com",
      "MAIL_FROM": "Your App <noreply@example.com>",
      "APP_NAME": "Your App"
    }
  }
}

The Worker serves app/dist as static assets and runs code only for /api/* (run_worker_first: ["/api/*"]), with SPA fallback for everything else.

#3. Provision Cloudflare resources

Authenticate once (npx wrangler login), then create the per-brand resources.

D1 database — create it, copy the returned database_id into your env's database_id, then apply the migrations in worker/migrations/:

npx wrangler d1 create <your-id>
npx wrangler d1 migrations apply <your-id> --env <your-id> --remote

R2 bucket — the content bucket is named <your-id>-content (the publish pipeline derives this name from the brand id):

npx wrangler r2 bucket create <your-id>-content

Attach an R2 custom domain so the bucket serves at your contentOrigin. The pipeline uploads objects at the bucket root, and an R2 custom domain serves the bucket root at the domain root — so content.example.com/manifest.json maps to the manifest.json object. (See architecture.md for why content bypasses the Worker.)

#4. Set secrets

These are Worker secrets, never committed. Set the ones you need per env:

npx wrangler secret put VAPID_PUBLIC_KEY   --env <your-id>
npx wrangler secret put VAPID_PRIVATE_KEY  --env <your-id>
npx wrangler secret put ADMIN_KEY          --env <your-id>
npx wrangler secret put RESEND_API_KEY     --env <your-id>
npx wrangler secret put GOOGLE_CLIENT_ID   --env <your-id>
npx wrangler secret put GOOGLE_CLIENT_SECRET --env <your-id>

Password + passkey sign-in need no secrets.

#5. Build and deploy

npm run build -w app -- --mode <your-id>
npx wrangler deploy --env <your-id>

The root npm run deploy is hardcoded to the storylark brand/env; for your own brand, run the two commands above (or add a matching script to package.json).

#6. Publish content

Your site boots as an empty shelf until you publish. Point the pipeline at your content source and a parser you own:

node tools/publish.mjs --brand <your-id> \
  --source <path-to-your-content> \
  --parser <path-to-your-parser.mjs>

Full pipeline reference, flags, and the parser contract: content-pipeline.md.

#Free-tier note

The default architecture is tuned to fit Cloudflare + Azure Speech free tiers (see the budget table in architecture.md). The publish pipeline enforces a monthly TTS character budget with a hard stop; heavier usage means moving off the free tiers.

Found a gap? StoryLark is open source — improve these docs on GitHub.