The same few ideas appear on every page of these docs. Read this once and the rest reads faster.

How your data is organized

Products and ways to use them

You can use both in three ways, against the same workspace and data:

File type, MIME type and AI Schema

  • A file type is the kind of document, such as Invoice or Bank statement. fileAI classifies each document into a file type. The API calls it a blueprint (blueprintId, /v1/file-type).
  • A MIME type is the format of the uploaded bytes, such as application/pdf. You send it as the fileType field on upload. See Supported file types.
  • An AI Schema is the list of fields fileAI extracts for a file type. See AI Schemas.

Approve and Verify

fileAI proposes a new file type and schema when a document matches none you have. You confirm it with approve in the API (PATCH /v1/file-type/approve), which is the same action as Verify in the app. Approved schemas apply to documents you process next.

API keys

Keys work only on the region that issued them. See Authentication and regions.

What is metered

See Billing and credits for the full rules.

API quickstart

Upload a file and read its values from code.

Authentication and regions

Keys, hosts and request conventions.

AI Schemas

Define and read the fields you extract.

Async jobs

Poll or receive callbacks for long-running work.