ReaderDocument

Type

A built document as read receives it: its id, source, metadata, and entities, with chunks() and bytes().

The document that read and relinked receive. It carries the document's id, its sourceId (key your rows by it so cleared can drop them), filename, content type, external id on a mapped source, metadata, the entities it asserts, and counts. url is a short-lived link to the file the document came from, valid for the delivery. chunks() returns the document's passages as indexed, each with its entities. bytes() returns the original file, for a reader that needs the drawing itself rather than its text. Each loads on first call and keeps the result.

ReaderDocument.ts

interface ReaderDocument {
  id: string
  sourceId: string
  filename: string | null
  contentType: string | null
  externalId: string | null
  metadata: DocumentMetadata | null
  entities: string[]
  pageCount: number | null
  chunkCount: number | null
  createdAt: string
  url: string | null
  chunks(): Promise<DataSourceChunk[]>
  bytes(): Promise<Uint8Array>
}

Fields

string

The document id. Key your rows by it.

sourceId

string

The data source it belongs to, so cleared can drop its rows.

externalId

string | null

The identity a mapped source replaces by; null on unmapped sources.

entities

string[]

The entities the document asserts, sorted.

url

string | null

A short-lived link to the file it came from, valid for the delivery.

chunks()

() => Promise<DataSourceChunk[]>

Its passages as indexed, each with its entities.

bytes()

() => Promise<Uint8Array>

The original file, for a reader that needs the drawing rather than its text.

Read text or bytes

read: async (document) => {
  const chunks = await document.chunks();
  const image = document.contentType?.startsWith('image/') ? await document.bytes() : null;
}