fileScout turns an uploaded file into structured data. It identifies each element in the file (text blocks, tables, charts, images and more) and processes every one on the path suited to it, then fills the fields you define in an AI Schema. Financial reports, claim forms and contracts come back in the same shape every time.

​
How fileScout processes a file

1

Map the elements

fileScout reads the file and records each element and where it sits: headings, paragraphs, tables, lists, form fields, charts and graphs, images, logos, signatures, handwriting, stamps and QR codes.
2

Choose a path for each element

Tables keep their rows and columns, text is read with its layout, and visual elements such as charts, photos and stamps are read visually. Scanned pages are read with OCR first. One file can use several models and tools, one per kind of element.
3

Skip what carries no data

Running headers and footers are left out, so they cost no processing.
4

Fill your schema

The results from every path come together in the fields of your AI Schema, returned by GET /v1/files/{fileId}/values.
Why it matters: each element gets the method that reads it best, instead of one model for the whole file, and processing is spent only where an element needs it. That keeps accuracy high on mixed files and token use down. The element breakdown is internal; the API returns the structured values. See it on a real-world file in Multi-page financial report.

​
What’s included

​
Turn it on

Add FILE_SCOUT to enabledProducts when you request an upload:
fileScout cannot be combined with storageOnly: true. Choose products covers the other combinations, including adding fileForensics.

​
How usage is counted

fileScout is metered per page, counted once processing completes:

​
Next steps

Quickstart

Upload a file and read its values in five minutes.

Extract fields

Define the fields fileScout fills.