Skip to main content

Build a data model

A model is the shape of a dataset — its fields and their types. Records are validated against it, so your data stays clean.

Plan availability

Starter and above. Free plans have no data store; add-ons raise the dataset cap.

The data page with a dataset open

Define the model​

  1. In Data, create a dataset and open the schema dialog.
  2. Add typed fields — text, number, date, reference, and more.
  3. Save. The model is stored on the dataset, and records are validated against it on the way in.

Display name vs reference ID​

Every field has two names:

  • Display name — the human label shown in the record editor, table headers, and anywhere the field appears. Rename it any time.
  • Reference ID — the stable key used in bindings ({{item.reference_id}}), form field mappings, and CSV import/export.

When you create a field, the Reference ID auto-fills from the display name (for example, Roast preference → roast_preference). Edit it in the same dialog to set your own — handy when you want a short, code-friendly key that differs from the label. Once the field is created the ID is fixed: records, bindings, and forms all point at it, so it can't change without orphaning that data. Pick it deliberately, then rename the display name freely afterward.

Edit records​

Open the typed document editor to add and edit records. Each field renders the input for its type (a date picker for dates, a number field for numbers, and so on), which keeps data consistent.

Tips​

  • Model the data first, then bind it — a clean model makes repeatable components and bindings straightforward.
  • Give fields a description in the schema dialog — it appears as a hint in the record editor, as a tooltip on table headers, and under the field's name in the schema dialog itself. See naming & describing fields.
  • Use a reference field to link records; see relations.