Input

class Base64Input(base64_string, filename)

Base64-encoded text input.

Parameters:

base64_string (str)
filename (str)

class BaseParameters(model_id, alias=None, webhook_ids=None, polling_options=None, close_file=True)

Base class for parameters accepted by all V2 endpoints.

Parameters:

model_id (str)
alias (str | None)
webhook_ids (List[str] | None)
polling_options (PollingOptions | None)
close_file (bool)

classmethod get_enqueue_slug()

Getter for the enqueue slug.

Return type:: str

get_form_data()

Return the parameters as a config dictionary.

Return type:: Dict[str, Union[str, List[str]]]
Returns:: A dict of parameters.

alias: Optional[str] = None: Use an alias to link the file to your own DB. If empty, no alias will be used.

close_file: bool = True: Whether to close the file after product.

model_id: str: ID of the model, required.

polling_options: Optional[PollingOptions] = None: Options for polling. Set only if having timeout issues.

webhook_ids: Optional[List[str]] = None: IDs of webhooks to propagate the API response to.

class BytesInput(raw_bytes, filename)

Raw bytes input.

Parameters:

raw_bytes (bytes)
filename (str)

class FileInput(file)

A binary file input.

Parameters:: file (BinaryIO)

class InferenceParameters(model_id, alias=None, webhook_ids=None, polling_options=None, close_file=True, _slug='inferences', rag=None, raw_text=None, polygon=None, confidence=None, text_context=None, data_schema=None)

Inference parameters to set when sending a file.

Parameters:

model_id (str)
alias (str | None)
webhook_ids (List[str] | None)
polling_options (PollingOptions | None)
close_file (bool)
_slug (str)
rag (bool | None)
raw_text (bool | None)
polygon (bool | None)
confidence (bool | None)
text_context (str | None)
data_schema (DataSchema | str | dict | None)

get_form_data()

Return the parameters as a config dictionary.

Return type:: Dict[str, Union[str, List[str]]]
Returns:: A dict of parameters.

confidence: Optional[bool] = None: Boost the precision and accuracy of all extractions. Calculate confidence scores for all fields, and fill their confidence attribute.

data_schema: Union[DataSchema, str, dict, None] = None: Dynamic changes to the data schema of the model for this inference. Not recommended, for specific use only.

polygon: Optional[bool] = None: Calculate bounding box polygons for all fields, and fill their locations attribute.

rag: Optional[bool] = None: Enhance extraction accuracy with Retrieval-Augmented Generation.

raw_text: Optional[bool] = None: Extract the full text content from the document as strings, and fill the raw_text attribute.

text_context: Optional[str] = None: Additional text context used by the model during inference. Not recommended, for specific use only.

class InputType(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)

The input type, for internal use.

BASE64 = 'base64'

BYTES = 'bytes'

FILE = 'file'

PATH = 'path'

URL = 'url'

class LocalInputSource(input_type)

Base class for all input sources coming from the local machine.

Parameters:: input_type (InputType)

apply_page_options(page_options)

Apply cut and merge options on multipage documents.

Return type:: None
Parameters:: page_options (PageOptions)

close()

Close the file object.

Return type:: None

compress(quality=85, max_width=None, max_height=None, force_source_text=False, disable_source_text=True)

Compresses the file object, either as a PDF or an image.

Parameters:

quality (int, default: 85) – Quality of the compression. For images, this is the JPEG quality. For PDFs, this affects image quality within the PDF.
max_width (Optional[int], default: None) – Maximum width for image resizing. Ignored for PDFs.
max_height (Optional[int], default: None) – Maximum height for image resizing. Ignored for PDFs.
force_source_text (bool, default: False) – For PDFs, whether to force compression even if source text is present.
disable_source_text (bool, default: True) – For PDFs, whether to disable source text during compression.

Return type:

None

count_doc_pages()

Deprecated. Use page_count instead.

Return type:: int

fix_pdf(maximum_offset=500)

Fix a potentially broken pdf file.

WARNING: this feature alters the data of the enqueued file by removing unnecessary headers.

Reads the bytes of a PDF file until a proper pdf tag is encountered, or until the maximum offset has been reached. If a tag denoting a PDF file is found, deletes all bytes before it.

Parameters:: maximum_offset (int, default: 500) – maximum byte offset where superfluous headers will be removed. Cannot be less than 0.
Return type:: None

has_source_text()

If the file is a PDF, checks if it has source text.

Return type:: bool
Returns:: True if the file is a PDF and has source text. False otherwise.

is_pdf()

Return type:: bool
Returns:: True if the file is a PDF.

is_pdf_empty()

Check if the PDF is empty.

Return type:: bool
Returns:: True if the PDF is empty

merge_pdf_pages(page_numbers)

Create a new PDF from pages and set it to file_object.

Parameters:: page_numbers (set) – List of page numbers to use for merging in the original PDF.
Return type:: None
Returns:: None

process_pdf(behavior, on_min_pages, page_indexes)

Run any required processing on a PDF file.

Return type:

None

Parameters:

behavior (str)
on_min_pages (int)
page_indexes (Sequence)

read_contents(close_file)

Read the contents of the input file.

Parameters:: close_file (bool) – whether to close the file after reading
Return type:: Tuple[str, bytes]
Returns:: a Tuple with the file name and binary data

file_mimetype: str

file_object: BinaryIO

filename: str

filepath: Optional[str]

input_type: InputType

property page_count: int

Count the pages in the document.

Returns:: The number of pages.

class LocalResponse(input_file)

Local response loaded from a file.

Parameters:: input_file (BinaryIO | str | Path | bytes)

deserialize_response(response_class)

Load a local inference.

Typically used when wanting to load a V2 webhook callback.

Return type:: TypeVar(ResponseT, bound= CommonResponse)
Parameters:: response_class (Type[ResponseT])

get_hmac_signature(secret_key)

Returns the hmac signature of the local response, from the secret key provided.

Parameters:: secret_key (Union[str, bytes, bytearray]) – Secret key, either a string or a byte/byte array.
Returns:: The hmac signature of the local response.

is_valid_hmac_signature(secret_key, signature)

Checks if the hmac signature of the local response is valid.

Parameters:

secret_key (Union[str, bytes, bytearray]) – Secret key, given as a string.
signature (str) – HMAC signature, given as a string.

Returns:

True if the HMAC signature is valid.

ResponseT = ~ResponseT

property as_dict: Dict[str, Any]

Returns the dictionary representation of the file.

Returns:: A json-like dictionary.

class PageOptions(page_indexes, operation='KEEP_ONLY', on_min_pages=0)

Options to pass to the parse method for cutting multipage documents.

Parameters:

page_indexes (Sequence[int])
operation (str)
on_min_pages (int)

on_min_pages: int

Apply the operation only if document has at least this many pages.

Default: 0 (apply on all documents)

operation: str

Operation to apply on the document, given the page_indexes specified:

KEEP_ONLY - keep only the specified pages, and remove all others.
REMOVE - remove the specified pages, and keep all others.

page_indexes: Sequence[int]

Zero-based list of page indexes. A negative index can be used, indicating an offset from the end of the document.

[0, -1] represents the fist and last pages of the document.

class PathInput(filepath)

A local path input.

Parameters:: filepath (str | None)

class PollingOptions(initial_delay_sec=2, delay_sec=1.5, max_retries=80)

Options for asynchronous polling.

Parameters:

initial_delay_sec (float)
delay_sec (float)
max_retries (int)

delay_sec: float: Delay between each polling attempt.

initial_delay_sec: float: Initial delay before the first polling attempt.

max_retries: int: Total number of polling attempts.

class UrlInputSource(url)

A local or distant URL input.

Parameters:: url (str)

as_local_input_source(filename=None, username=None, password=None, token=None, headers=None, max_redirects=3)

Convert the URL content to a BytesInput object.

Parameters:

filename (Optional[str], default: None) – Optional filename for the BytesInput.
username (Optional[str], default: None) – Optional username for authentication.
password (Optional[str], default: None) – Optional password for authentication.
token (Optional[str], default: None) – Optional token for authentication.
headers (Optional[dict], default: None) – Optional additional headers for the request.
max_redirects (int, default: 3) – Maximum number of redirects to follow.

Return type:

BytesInput

Returns:

A BytesInput object containing the file content.

save_to_file(filepath, filename=None, username=None, password=None, token=None, headers=None, max_redirects=3)

Save the content of the URL to a file.

Parameters:

filepath (Union[Path, str]) – Path to save the content to.
filename (Optional[str], default: None) – Optional filename to give to the file.
username (Optional[str], default: None) – Optional username for authentication.
password (Optional[str], default: None) – Optional password for authentication.
token (Optional[str], default: None) – Optional token for authentication.
headers (Optional[dict], default: None) – Optional additional headers for the request.
max_redirects (int, default: 3) – Maximum number of redirects to follow.

Return type:

Path

Returns:

The path to the saved file.

url: str: The Uniform Resource Locator.

class WorkflowOptions(alias=None, priority=None, full_text=False, public_url=None, rag=False)

Options to pass to a workflow execution.

Parameters:

alias (str | None)
priority (ExecutionPriority | None)
full_text (bool)
public_url (str | None)
rag (bool)

alias: Optional[str]: Alias for the document.

full_text: bool: Whether to include the full OCR text response in compatible APIs.

priority: Optional[ExecutionPriority]: Priority of the document.

public_url: Optional[str]: A unique, encrypted URL for accessing the document validation interface without requiring authentication.

rag: bool: Whether to enable Retrieval-Augmented Generation.