Layout, blocks, geometry

Block

Reading-order-grouped paragraph with bounding box, layout class, and content. Emitted on the response when include_blocks=True.

Block

Bases: _Frozen

ATTRIBUTE	DESCRIPTION
id	TYPE: int
layout_id	TYPE: int
class_name	TYPE: Annotated[str, Field(alias='class')]
bounding_box	TYPE: BoundingBox
content	TYPE: str
order_index	TYPE: int

id instance-attribute

id: int

layout_id instance-attribute

layout_id: int

class_name instance-attribute

class_name: Annotated[str, Field(alias='class')]

bounding_box instance-attribute

bounding_box: BoundingBox

content instance-attribute

content: str

order_index instance-attribute

order_index: int

TextItem

A single recognised word/line with confidence and bounding box.

TextItem

Bases: _Frozen

ATTRIBUTE	DESCRIPTION
text	TYPE: str
confidence	TYPE: float
bounding_box	TYPE: BoundingBox
id	TYPE: int | None
layout_id	TYPE: int | None
source	TYPE: Literal['ocr', 'pdf', 'geometric', 'auto', 'auto_verified'] | None

text instance-attribute

text: str

confidence instance-attribute

confidence: float

bounding_box instance-attribute

bounding_box: BoundingBox

id class-attribute instance-attribute

id: int | None = None

layout_id class-attribute instance-attribute

layout_id: int | None = None

source class-attribute instance-attribute

source: Literal['ocr', 'pdf', 'geometric', 'auto', 'auto_verified'] | None = None

BoundingBox

BoundingBox

Bases: _Frozen

ATTRIBUTE	DESCRIPTION
points	TYPE: Annotated[Quad, Field(description='4-corner polygon')]
aabb	TYPE: tuple[int, int, int, int]
center	TYPE: Point
width	TYPE: int
height	TYPE: int

points instance-attribute

points: Annotated[Quad, Field(description='4-corner polygon')]

aabb cached property

aabb: tuple[int, int, int, int]

center cached property

center: Point

width cached property

width: int

height cached property

height: int

LayoutBox

LayoutBox

Bases: _Frozen

ATTRIBUTE	DESCRIPTION
class_name	TYPE: Annotated[str, Field(alias='class')]
class_id	TYPE: int
confidence	TYPE: float
bounding_box	TYPE: BoundingBox
id	TYPE: int | None

class_name instance-attribute

class_name: Annotated[str, Field(alias='class')]

class_id instance-attribute

class_id: int

confidence instance-attribute

confidence: float

bounding_box instance-attribute

bounding_box: BoundingBox

id class-attribute instance-attribute

id: int | None = None

LayoutLabel

LayoutLabel

Bases: StrEnum

ATTRIBUTE	DESCRIPTION
abstract
algorithm
aside_text
chart
content
display_formula
doc_title
figure_title
footer
footer_image
footnote
formula_number
header
header_image
image
inline_formula
number
paragraph_title
reference
reference_content
seal
table
text
vertical_text
vision_footnote
supplementary_region

abstract class-attribute instance-attribute

abstract = 'abstract'

algorithm class-attribute instance-attribute

algorithm = 'algorithm'

aside_text class-attribute instance-attribute

aside_text = 'aside_text'

chart class-attribute instance-attribute

chart = 'chart'

content class-attribute instance-attribute

content = 'content'

display_formula class-attribute instance-attribute

display_formula = 'display_formula'

doc_title class-attribute instance-attribute

doc_title = 'doc_title'

figure_title class-attribute instance-attribute

figure_title = 'figure_title'

footer class-attribute instance-attribute

footer = 'footer'

footer_image class-attribute instance-attribute

footer_image = 'footer_image'

footnote class-attribute instance-attribute

footnote = 'footnote'

formula_number class-attribute instance-attribute

formula_number = 'formula_number'

header class-attribute instance-attribute

header = 'header'

header_image class-attribute instance-attribute

header_image = 'header_image'

image class-attribute instance-attribute

image = 'image'

inline_formula class-attribute instance-attribute

inline_formula = 'inline_formula'

number class-attribute instance-attribute

number = 'number'

paragraph_title class-attribute instance-attribute

paragraph_title = 'paragraph_title'

reference class-attribute instance-attribute

reference = 'reference'

reference_content class-attribute instance-attribute

reference_content = 'reference_content'

seal class-attribute instance-attribute

seal = 'seal'

table class-attribute instance-attribute

table = 'table'

text class-attribute instance-attribute

text = 'text'

vertical_text class-attribute instance-attribute

vertical_text = 'vertical_text'

vision_footnote class-attribute instance-attribute

vision_footnote = 'vision_footnote'

supplementary_region class-attribute instance-attribute

supplementary_region = 'SupplementaryRegion'

Table

Table

Bases: _Frozen

A recognized table.

Server-recognized tables (tables=True, SLANet-Plus) carry html (the full <table> markup), layout_id, and the recognizer confidence. Tables synthesized from layout blocks (no table backend requested) carry only the region text and id.

ATTRIBUTE	DESCRIPTION
bounding_box	TYPE: BoundingBox
html	TYPE: str | None
layout_id	TYPE: int | None
confidence	TYPE: float | None
id	TYPE: int | None
text	TYPE: str

bounding_box instance-attribute

bounding_box: BoundingBox

html class-attribute instance-attribute

html: str | None = None

layout_id class-attribute instance-attribute

layout_id: int | None = None

confidence class-attribute instance-attribute

confidence: float | None = None

id class-attribute instance-attribute

id: int | None = None

text class-attribute instance-attribute

text: str = ''

Formula

Formula

Bases: _Frozen

A recognized formula.

Server-recognized formulas (formulas=True, PP-FormulaNet) carry latex, layout_id, and the recognizer confidence; text mirrors latex so downstream code can treat both sources uniformly. Formulas synthesized from layout blocks carry only the region text and id.

METHOD	DESCRIPTION
model_post_init

ATTRIBUTE	DESCRIPTION
bounding_box	TYPE: BoundingBox
latex	TYPE: str | None
layout_id	TYPE: int | None
confidence	TYPE: float | None
id	TYPE: int | None
text	TYPE: str
is_inline	TYPE: bool

bounding_box instance-attribute

bounding_box: BoundingBox

latex class-attribute instance-attribute

latex: str | None = None

layout_id class-attribute instance-attribute

layout_id: int | None = None

confidence class-attribute instance-attribute

confidence: float | None = None

id class-attribute instance-attribute

id: int | None = None

text class-attribute instance-attribute

text: str = ''

is_inline class-attribute instance-attribute

is_inline: bool = False

model_post_init

model_post_init(__context: object) -> None

Tables and formulas are strict opt-ins

Pass tables=True / formulas=True (server v3.1+, with TABLE_BACKEND=slanext / FORMULA_BACKEND=ppformulanet_s loaded) and the server recognizes structure: Table.html carries full <table> markup and Formula.latex the LaTeX source, each with a recognizer confidence. Without the opt-in the SDK synthesizes region-level entries from layout blocks (text only; html / latex stay None). Requesting a stage the server doesn't have raises BackendDisabled — check client.capabilities() first.