Every field in an AI Schema has a type, plus an optional format. The type tells fileAI what to look for and decides how the value comes back from GET /v1/files/{fileId}/values. Pick the narrowest type that fits: an enum beats a free-text string when the answer is one of a known set.

​
Types

Other format values on string fields are latitude and longitude.

​
How values come back

  • number and integer values are JSON numbers. Grouped text such as "199,612.89" is read as 199612.89.
  • Every other scalar value is a string, including boolean ("true" or "false").
  • An empty value is null.
  • A number or table cell whose value is 0 also comes back as null. If your system needs to tell zero from empty, treat null on a number field as “no value or zero”.

​
Tables

A list field is a table. Its items define the columns, and each column has its own type. In the values response, rows are in valueArray: an array of rows, each row an array of cells.
Two other field shapes are stored and returned as list: Tables over 100 rows are truncated unless you pass includeFullTableData=true. To change a cell, use PATCH /v1/files/{fileId}/values with cells. See Correct a value.

​
Enums

An enum field lists its allowed values in enumValues. fileAI returns one of them as a string.

​
Field properties

These properties appear on each field in GET /v1/schemas and can be set with PATCH /v1/files/schema:

AI Schemas

Define fields and read values.

Extraction prompts

Write descriptions that extract reliably.