Configure bindings

Platform data is not a separate product you click together in the Dashboard. You declare it in wrangler.jsonc next to the Worker (wrangler.json is parsed for listings but not used as the deploy bundle). On deploy, celld creates those bindings on this version. The Dashboard is for inspecting and editing values after the version is ready.

The rule

Lives in wrangler (your repo)	Lives on the version (after deploy)
Binding names (env.DB)	Actual D1 rows, KV keys, R2 objects, queue messages
Resource ids (database_id, KV id, queue name, R2 bucket_name)	Forked copies on child versions
Cron expressions, workflow class names	Cron ticks / workflow instances
vars	Overridable per version in Settings

If it is not in wrangler, env.FOO is undefined. Creating a “database” only in the UI does nothing.

Deploy filename: the orchestrator only treats wrangler.jsonc as the Worker bundle. Use that name.

cellp currently honors one d1_databases entry. A second database in the array fails deploy.

Minimal file

{
  "name": "my-shop",
  "main": "index.js",
  "compatibility_date": "2026-01-01",
  "vars": {
    "STORE_NAME": "Demo shop"
  }
}

vars become env.STORE_NAME. Do not put production secrets here — inject them from CI or version env.

D1

"d1_databases": [
  {
    "binding": "DB",
    "database_name": "commerce",
    "database_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890"
  }
]

In the Worker: env.DB.prepare('select …'), env.DB.exec(schema).

binding is the env key. database_name is what celld uses for import. database_id must stay stable across child versions so the same Worker keeps working — copy it from the parent wrangler; do not mint a new UUID on every PR.

Use D1 in code → · Seed / browse data →

KV

"kv_namespaces": [
  {
    "binding": "CACHE",
    "id": "my-shop-cache"
  }
]

In the Worker: env.CACHE.get(key), env.CACHE.put(key, value). The id is the namespace identity (inherited on branch). KV guide.

R2

"r2_buckets": [
  { "binding": "ASSETS", "bucket_name": "my-shop-assets" }
]

In the Worker: env.ASSETS.put(key, body), get, list. There is no object browser in the Dashboard. R2 guide.

Queues

"queues": {
  "producers": [{ "binding": "FULFILLMENT", "queue": "fulfillment" }]
}

env.FULFILLMENT.send(payload) from a fetch Worker.

Consumer scripts cannot also export fetch

celld: a queue consumer script must not export fetch(). Keep a producer Worker for HTTP and a separate consumer Worker if you need queue(). See Queues and Handlers.

Workflows

"workflows": [
  {
    "binding": "REPORTS",
    "name": "order-report",
    "class_name": "OrderReport"
  }
]

Export the class from the same module (class_name must match). env.REPORTS.create({ params }). Instances do not fork to preview versions. Workflows.

Cron

"triggers": { "crons": ["0 * * * *"] }

Implement scheduled(controller, env, ctx) on the default export. celld fires it; cellp only displays the expression. Cron.

Durable Objects (optional)

"durable_objects": {
  "bindings": [{ "name": "COUNTER", "class_name": "Counter" }]
},
"migrations": [{ "tag": "v1", "new_sqlite_classes": ["Counter"] }]

See the repo example dev/examples/counter. celld DO support is partial. Durable Objects do not show up on GET …/bindings (the control-plane parser has no durable_objects field). Confirm celld compat before you bet the product on them.

Full example (commerce)

The storefront in this repo wires all of the above. Copy wrangler.jsonc, index.js, and storefront.js.

Child versions

When CI sends parent_version_id, cellp copies parent identities onto the child wrangler for D1 database_id, KV id, and R2 bucket_name. You should still ship the same ids in your file so listings match what you expect. Queue names must already match the parent or Start fails.

Changing those ids in a PR does not give you a quietly empty preview for D1/KV/R2 (they get overwritten). An empty preview is usually a missing parent_version_id, a failed branch, or a count mismatch.

The script always comes from this artifact (main). You can change routes and SQL.

What the Dashboard is for

After ready:

• See the binding list (Storage)
• Browse / SQL D1, edit KV keys, peek queues
• Override vars (env)

It will not add a new d1_databases entry. Change wrangler and deploy a new version.