Read a document's text
Page through the readable text of one ingested document, in reading order, as the passages answers cite. Use it to show a cited source in your own interface.
GET
Returns one page of a document’s text as it was ingested: the current version, split into the passages that answers cite, in reading order. Each passage’s
id is the source_id a citation carries, so you can show the document and scroll to the passage an answer used.
text is content from your knowledge sources. It is returned as a JSON string for you to display as plain text; do not render it as HTML or Markdown.
limit and offset are clamped rather than refused. A limit below 1 becomes 1, a limit above 50 becomes 50, and a negative offset becomes 0. A value that is not a whole number returns 400. Any role in the organization can read a document’s content. A document that does not exist, belongs to another organization or has a malformed ID returns 404.
string
required
The document ID, from List documents.
integer
default:"20"
Passages per page, from
1 to 50.integer
default:"0"
Number of passages to skip.
Response
object
required
The document, with the same fields List documents
returns for it.
string
required
The version of the document this text belongs to. It changes when a sync finds the document
changed.
object[]
required
The page of passages, in reading order.
integer
required
Number of passages in this version of the document.
integer
required
The page size applied, after clamping.
integer
required
The offset applied, after clamping.
boolean
required
true when more passages follow this page.Last modified on October 13, 2026
