Page details

result.pages returns parse output page by page. Each page object includes metadata, a structured block list, and a lightweight content list. Additionally, result.detail provides a flat paragraph-level view across all pages.

Page object fields

Field	Type	Meaning
page_id	int	Page number (1-based)
angle	float	Page rotation angle (after correction)
height	int	Page pixel height
width	int	Page pixel width
image_id	string	Original image identifier (usually empty)
durations	float	Processing time for this page (seconds)
status	string	Processing status, e.g. Success
structured	array	Structured block list for fine-grained rendering and positioning
content	array	Lightweight content list for quick text access

structured — structured blocks

Best for detailed rendering, highlight positioning, and content reconstruction. Each block:

Field	Type	Meaning
id	int	Block sequence number
type	string	Block type, e.g. doc_title, paragraph_title, text, table, image
text	string	Extracted text
pos	float[]	Quadrilateral coordinates [x1,y1, x2,y2, x3,y3, x4,y4] (top-left, top-right, bottom-right, bottom-left)
outline_level	int	Heading level (-1 for non-heading)
content	int[]	Referenced text block IDs linking to the content array

{
  "id": 0,
  "type": "doc_title",
  "text": "# Sample PDF",
  "pos": [141, 154, 509, 154, 509, 220, 141, 220],
  "outline_level": -1,
  "content": [0]
}

Common block types:

Type	Meaning
doc_title	Document title
paragraph_title	Paragraph heading
text	Body text
table	Table
image	Image
figure	Figure / chart
header / footer	Page header / footer
footnote	Footnote
formula	Formula

content — lightweight content

Better for quick consumption — concatenate into page-level text or build search indexes. Each item:

Field	Type	Meaning
id	int	Content block ID, linked from structured[].content
type	string	Block type
text	string	Text content
pos	float[]	Quadrilateral coordinates
score	float	Recognition confidence
angle	float	Text angle

{
  "id": 0,
  "type": "doc_title",
  "text": "# Sample PDF",
  "pos": [141, 154, 509, 154, 509, 220, 141, 220],
  "score": 0.5958,
  "angle": 0
}

result.detail — cross-page paragraph view

result.detail aggregates all paragraphs in reading order into a single flat array, eliminating the need to iterate through pages manually. Each record:

Field	Type	Meaning
paragraph_id	int	Paragraph sequence number
page_id	int	Source page number
type	string	Fixed as paragraph
sub_type	string	Paragraph sub-type, e.g. doc_title, paragraph_title, text, table
text	string	Paragraph text
position	float[]	Quadrilateral coordinates
outline_level	int	Heading level
tags	string[]	Custom tags

{
  "paragraph_id": 1,
  "page_id": 1,
  "type": "paragraph",
  "sub_type": "doc_title",
  "text": "# Sample PDF",
  "position": [141, 154, 509, 154, 509, 220, 141, 220],
  "outline_level": -1,
  "tags": []
}

Typical usage

• For UI highlighting, read structured[].pos or content[].pos
• For block-level filtering, select by structured[].type
• For reading views, use result.markdown or iterate through content
• For structured downstream processing, iterate result.pages and consume structured
• For cross-page paragraph processing, iterate result.detail directly

Related pages

• See Response overview for the top-level object layout
• See Metrics for processing-quality and performance fields