Llm

Implementation of API for LLMs.

OpenAIClassifier(k_shot=None, prompt=None, openai_client=None, responses_kwargs=None, classes=None)

Bases: ClassifierMixin, MultiOutputMixin, BaseEstimator

OpenAI classifier.

A classifier that is based on the OpenAI API.

Parameters:

Name	Type	Description	Default
k_shot	int | ArrayLike | None	Number of examples to include into the prompt. If None then no examples are included. If int then it is equal to the number of example that are selected randomly. If an array of indices is provided then it is used to select the examples.	None
prompt	str | None	Prompt for the OpenAI API.	None
openai_client	str | OpenAI | AsyncOpenAI | None	OpenAI client.	None
responses_kwargs	dict | None	Keyword arguments for the OpenAI API.	None
classes	dict | ndarray | list[ndarray] | None	Class labels.	None

Attributes:

Name	Type	Description
k_shot_	int | dict[int, ArrayLike | RandomState] | None	Number of examples per class.
prompt_	str | None	Prompt for the OpenAI API.
openai_client_	OpenAI | AsyncOpenAI	OpenAI client.
responses_kwargs_	dict | None	Keyword arguments for the OpenAI API.
classes_	ndarray | list[ndarray] | None	Class labels.
instructions_	str	Instructions for the OpenAI API.

Source code in src/skai/llm/_openai.py

def __init__(
    self: Self,
    k_shot: int | ArrayLike | None = None,
    prompt: str | None = None,
    openai_client: str | openai.OpenAI | openai.AsyncOpenAI | None = None,
    responses_kwargs: dict | None = None,
    classes: dict | np.ndarray | list[np.ndarray] | None = None,
) -> None:
    self.k_shot = k_shot
    self.prompt = prompt
    self.responses_kwargs = responses_kwargs
    self.openai_client = openai_client
    self.classes = classes

__sklearn_tags__()

Classifier tags.

Source code in src/skai/llm/_openai.py

def __sklearn_tags__(self: Self) -> Tags:
    """Classifier tags."""
    tags = super().__sklearn_tags__()
    tags.classifier_tags = ClassifierTags(
        poor_score=False,
        multi_class=True,
        multi_label=True,
    )
    tags.target_tags = TargetTags(
        required=True,
        multi_output=True,
        single_output=True,
    )
    return tags

fit(X=None, y=None)

Fit the classifier to the training dataset.

Parameters:

Name	Type	Description	Default
X	ArrayLike | None	Input data.	None
y	ArrayLike | None	Target values.	None

Returns:

Name	Type	Description
self	Self	The fitted OpenAI classifier.

Source code in src/skai/llm/_openai.py

@_fit_context(prefer_skip_nested_validation=True)
def fit(self, X: ArrayLike | None = None, y: ArrayLike | None = None) -> Self:
    """Fit the classifier to the training dataset.

    Args:
        X:
            Input data.

        y:
            Target values.

    Returns:
        self:
            The fitted OpenAI classifier.
    """
    return self._fit(X, y)

predict(X)

Predict the class labels for the provided data.

Parameters:

Name	Type	Description	Default
X	ArrayLike	Input data.	required

Returns:

Type	Description
NDArray	The predicted class labels.

Source code in src/skai/llm/_openai.py

def predict(self, X: ArrayLike) -> NDArray:
    """Predict the class labels for the provided data.

    Args:
        X:
            Input data.

    Returns:
        The predicted class labels.
    """
    check_is_fitted(
        self,
        ('k_shot_', 'openai_client_', 'responses_kwargs_', 'prompt_', 'classes_', 'instructions_'),
    )
    X = check_array(X, ensure_2d=False, dtype=str)
    if isinstance(self.openai_client_, openai.OpenAI):
        predictions = self._predict_sync(X)
    else:
        predictions = asyncio.run(self._predict_async(X.tolist()))
    if isinstance(self.classes, dict):
        classes_mapping = {v: k for k, v in self.classes.items()}
        predictions = [classes_mapping.get(pred.strip(), self.classes[0]) for pred in predictions]
    return np.array(predictions, dtype=self.y_dtype_)