Skip to main content
Version: 2.0.0

Documentation

LLMWhispererClientV2

The LLMWhispererClientV2 class provides a client for interacting with the LLMWhisperer API (v2.x). This client simplifies sending documents for processing and retrieving results from the LLMWhisperer API.

Constructor​

__init__(base_url: str = "", api_key: str = "", logging_level: str = "")

Initializes a new instance of the LLMWhispererClientV2 with optional parameters for the base URL, API key, and logging level.

Parameters:​

  • base_url (str, optional): The base URL for the LLMWhisperer API. If not provided, it defaults to the LLMWHISPERER_BASE_URL_V2 environment variable or a default URL.
  • api_key (str, optional): The API key for authenticating requests. If not provided, it defaults to the LLMWHISPERER_API_KEY environment variable.
  • logging_level (str, optional): The logging level for the client. Accepts values such as:
    • "DEBUG"
    • "INFO"
    • "WARNING"
    • "ERROR"
    Defaults to "DEBUG" or uses the LLMWHISPERER_LOGGING_LEVEL environment variable.

Methods​

get_usage_info() -> dict

Retrieves the usage information of the LLMWhisperer API.

Returns:​

  • dict: A dictionary containing the API usage information.

Raises:​

  • LLMWhispererClientException: If the API request fails.
whisper(
    file_path: str = "", 
    stream: IO[bytes] = None, 
    url: str = "", 
    mode: str = "high_quality", 
    output_mode: str = "layout_preserving", 
    page_seperator: str = "<<<", 
    pages_to_extract: str = "", 
    median_filter_size: int = 0, 
    gaussian_blur_radius: int = 0, 
    line_splitter_tolerance: float = 0.75, 
    horizontal_stretch_factor: float = 1.0, 
    mark_vertical_lines: bool = False, 
    mark_horizontal_lines: bool = False, 
    line_spitter_strategy: str = "left-priority", 
    lang: str = "eng", 
    tag: str = "default", 
    filename: str = "", 
    webhook_metadata: str = "", 
    use_webhook: str = "", 
    wait_for_completion: bool = False, 
    wait_timeout: int = 180
) -> dict
info

Refer to the API documentation for detailed information on the parameters.

Processes a document via the LLMWhisperer API using various input methods (file path, stream, or URL).

Parameters:​

  • file_path (str, optional): The file path of the document to be processed. Defaults to an empty string.
  • stream (IO[bytes], optional): A stream of bytes (e.g., file-like object) to be processed. Defaults to None.
  • url (str, optional): The URL of the document to be processed. Defaults to an empty string.
  • mode (str, optional): The processing mode. Can be:
    • "high_quality" (default)
    • "form"
    • "low_cost"
    • "native_text"
  • output_mode (str, optional): The output format for the processed document. Can be:
    • "layout_preserving" (default)
    • "text"
  • page_seperator (str, optional): A string used to separate pages in the output. Defaults to "<<<".
  • pages_to_extract (str, optional): A string specifying which pages to extract. Defaults to an empty string (processes all pages).
  • median_filter_size (int, optional): The size of the median filter to apply during processing. Defaults to 0.
  • gaussian_blur_radius (int, optional): The radius of the Gaussian blur to apply. Defaults to 0.
  • line_splitter_tolerance (float, optional): Tolerance for line splitting. Defaults to 0.75.
  • horizontal_stretch_factor (float, optional): Factor to stretch the document horizontally. Defaults to 1.0.
  • mark_vertical_lines (bool, optional): Whether to mark vertical lines in the output. Defaults to False.
  • mark_horizontal_lines (bool, optional): Whether to mark horizontal lines in the output. Defaults to False.
  • line_spitter_strategy (str, optional): Strategy for splitting lines. Can be "left-priority" (default).
  • lang (str, optional): The language of the document. Defaults to "eng".
  • tag (str, optional): A custom tag for the document. Defaults to "default".
  • filename (str, optional): The file name for the document in the output. Defaults to an empty string.
  • webhook_metadata (str, optional): Metadata to be sent to the webhook, if used. Defaults to an empty string.
  • use_webhook (str, optional): Name of the webhook to call. Defaults to an empty string.
  • wait_for_completion (bool, optional): Whether to wait for the whisper operation to complete. Defaults to False.
  • wait_timeout (int, optional): Time to wait (in seconds) for the operation to complete if wait_for_completion is True. Defaults to 180.

Returns:​

  • dict: The processed document or status information.

Raises:​

  • LLMWhispererClientException: If the API request fails or document input is invalid.
whisper_status(whisper_hash: str) -> dict

Retrieves the status of a whisper operation.

Parameters:​

  • whisper_hash (str): The unique hash for the whisper operation (returned by the whisper method).

Returns:​

  • dict: A dictionary containing the status of the whisper operation, including:
    • "status_code": The HTTP status code.
    • "status": Status details of the whisper operation.

Raises:​

  • LLMWhispererClientException: If the API request fails.
whisper_retrieve(whisper_hash: str) -> dict

Retrieves the result of a whisper operation.

Parameters:​

  • whisper_hash (str): The unique hash for the whisper operation.

Returns:​

  • dict: A dictionary containing the result of the whisper operation, including:
    • "status_code": The HTTP status code.
    • "extraction": The extracted text from the document.

Raises:​

  • LLMWhispererClientException: If the API request fails.
register_webhook(url: str, auth_token: str, webhook_name: str) -> dict

Registers a webhook to the LLMWhisperer API for event notifications.

Parameters:​

  • url (str): The URL for the webhook.
  • auth_token (str): The authentication token to be sent with webhook events.
  • webhook_name (str): A custom name for the webhook.

Returns:​

  • dict: A dictionary containing the webhook registration status.

Raises:​

  • LLMWhispererClientException: If the API request fails.
get_webhook_details(webhook_name: str) -> dict

Retrieves details about a previously registered webhook.

Parameters:​

  • webhook_name (str): The name of the webhook whose details are to be retrieved.

Returns:​

  • dict: A dictionary containing the details of the webhook, including:
    • "status_code": The HTTP status code.
    • "details": Webhook information.

Raises:​

  • LLMWhispererClientException: If the API request fails.
get_highlight_rect(self,
        line_metadata: list[int],
        target_width: int,
        target_height: int)

Given the line metadata and the line number, this function returns the bounding box of the line in the format (page, x1, y1, x2, y2).

Parameters​

  • line_metadata (list[int]): The line metadata returned by the LLMWhisperer API. The list typically contains information like page number, line position, and line height.
  • target_width (int): The width of the target image or page in the UI.
  • target_height (int): The height of the target image or page in the UI.

Returns​

  • tuple[int, int, int, int, int]: A tuple containing the bounding box of the line in the following format:
    • page (int): The page number.
    • x1 (int): The x-coordinate of the top-left corner of the bounding box.
    • y1 (int): The y-coordinate of the top-left corner of the bounding box, scaled to the target height.
    • x2 (int): The x-coordinate of the bottom-right corner of the bounding box (equal to target_width).
    • y2 (int): The y-coordinate of the bottom-right corner of the bounding box, scaled to the target height.

Exceptions​

LLMWhispererClientException

Exception raised for errors occurring within the LLMWhispererClientV2 class.

Attributes:​

  • message (str): A message describing the error.
  • status_code (int): The HTTP status code returned by the LLMWhisperer API (if applicable).

Methods:​

  • error_message(): Returns the error message.