Utils

Defines utilities for gllm_inference.

`SizeUnit`

Bases: IntEnum

Defines the valid size units.

`base64_to_bytes(value, *, allowed_mimetypes=DEFAULT_BASE64_ALLOWED_MIMETYPES)`

Decode a base64 string to bytes based on allowed MIME type.

The conversion steps are as follows

The function first attempts to decode the given string from base64.
If decoding succeeds, it checks the MIME type of the decoded content.
When the MIME type matches one of the allowed patterns (e.g., "image/*"), the raw bytes are returned. Otherwise, the original string is returned unchanged.

Parameters:

Name	Type	Description	Default
`value`	`str`	Input data to decode.	required
`allowed_mimetypes`	`tuple[str, ...]`	MIME type prefixes that are allowed to be decoded into bytes. Defaults to ("image/", "audio/", "video/*").	`DEFAULT_BASE64_ALLOWED_MIMETYPES`

Returns:

Type	Description
`str \| bytes`	str \| bytes: Base64-encoded string or raw bytes if MIME type is allowed; otherwise returns original string.

Raises:

Type	Description
`ValueError`	If the input is not a string.

`build_google_auth_params(api_key, credentials_path, credentials_info, project_id, location)`

Build Google client authentication parameters based on the provided credentials.

Parameters:

Name	Type	Description	Default
`api_key`	`str \| None`	Required for Google Gen AI authentication. Cannot be used with `credentials_path` or `credentials_info`.	required
`credentials_path`	`str \| None`	For Google Vertex AI authentication. Path to the service account credentials JSON file. Cannot be used together with `api_key` or `credentials_info`.	required
`credentials_info`	`dict[str, Any] \| None`	For Google Vertex AI authentication. Service account credentials JSON contents. Cannot be used with `api_key` or `credentials_path`.	required
`project_id`	`str \| None`	The Google Vertex AI project ID.	required
`location`	`str \| None`	The Google Vertex AI location.	required

Returns:

Type	Description
`dict[str, Any]`	dict[str, Any]: The Google Gen AI and Vertex AI authentication parameters.

Raises:

Type	Description
`ValueError`	If more than one of `api_key`, `credentials_path`, or `credentials_info` is provided.

`build_openai_embedded_error(message, code, status_code=None)`

Build a raisable error from an OpenAI-compatible embedded provider error payload.

Parameters:

Name	Type	Description	Default
`message`	`str \| None`	The provider-supplied error message, if any.	required
`code`	`Any`	The provider-supplied error code.	required
`status_code`	`Any`	The provider-supplied HTTP status code, when reported separately from `code`. A numeric `code` is used as the fallback.	`None`

Returns:

Name	Type	Description
`Exception`	`Exception`	The error to raise so the translator can classify it.

`coerce_openai_status_code(value)`

Coerce a provider-supplied code or status code to an int HTTP status code.

Parameters:

Name	Type	Description	Default
`value`	`Any`	An int, a digit string, or anything else.	required

Returns:

Type	Description
`int \| None`	int \| None: The int form of a numeric value, None otherwise.

`get_openai_model_id_path(base_url)`

Get ModelId path for OpenAI-compatible endpoints.

Returns None when base_url points to the default OpenAI endpoint, otherwise returns the provided base_url value.

Parameters:

Name	Type	Description	Default
`base_url`	`str`	The base URL of the OpenAI-compatible endpoint.	required

Returns:

Type	Description
`str \| None`	str \| None: The ModelId path for OpenAI-compatible endpoints.

`get_size(obj, unit=SizeUnit.BYTES)`

Get the pickle-serialized size of an object in the given unit.

Note

This measures the size of the object when serialized using pickle, NOT its in-memory footprint.

Parameters:

Name	Type	Description	Default
`obj`	`Any`	The object to get the size of.	required
`unit`	`SizeUnit`	The unit to get the size in. Defaults to SizeUnit.BYTES.	`BYTES`

Returns:

Name	Type	Description
`float`	`float`	The size of the object in the given mode.

`is_valid_url(text)`

Check if a URL is valid.

This method validate URL format and rejects dangerous schemes and private/local addresses.

Parameters:

Name	Type	Description	Default
`text`	`str`	The URL string to validate.	required

Returns:

Name	Type	Description
`bool`	`bool`	True if URL is valid, False otherwise.

`load_langchain_model(model_class_path, model_name, model_kwargs)`

Loads the LangChain's model instance.

Parameters:

Name	Type	Description	Default
`model_class_path`	`str`	The path to the LangChain's class, e.g. "langchain_openai.ChatOpenAI".	required
`model_name`	`str`	The model name.	required
`model_kwargs`	`dict[str, Any]`	The additional keyword arguments.	required

Returns:

Type	Description
`BaseChatModel \| Embeddings`	BaseChatModel \| Embeddings: The LangChain's model instance.

`parse_model_data(model)`

Parses the model data from LangChain's BaseChatModel or Embeddings instance.

Parameters:

Name	Type	Description	Default
`model`	`BaseChatModel \| Embeddings`	The LangChain's BaseChatModel or Embeddings instance.	required

Returns:

Type	Description
`dict[str, str]`	dict[str, str]: The dictionary containing the model name and path.

Raises:

Type	Description
`ValueError`	If the model name is not found in the model data.

`validate_os(supported_os, module_name)`

Validate if the current system is supported.

Parameters:

Name	Type	Description	Default
`supported_os`	`str \| tuple[str, ...]`	The supported operating system(s).	required
`module_name`	`str`	The name of the module to validate.	required

Raises:

Type	Description
`OSError`	If the current operating system is not supported.

`validate_trace_attributes(trace_attributes, logger)`

Return OpenTelemetry-compatible scalar trace attributes.

Parameters:

Name	Type	Description	Default
`trace_attributes`	`dict[str, Any] \| None`	User-provided trace attributes.	required
`logger`	`Logger`	Logger used to report skipped attributes.	required

Returns:

Type	Description
`dict[str, TraceValue]`	dict[str, TraceValue]: The supported trace attributes.