daft.functions.jev_choice#

def jev_choice(
    row: Expression,
    question: str,
    options: list[str],
    *,
    model: str | None = None,
    api_key: str | None = None,
    base_url: str | None = None,
    timeout: float = 60.0,
) -> Expression:
    """The most probable of `options` as the answer to `question` for each row.

    Args:
        row (Expression): The value to classify. Pass `to_struct(...)` to classify on several columns.
        question (str): A plain-English question, e.g. "which team should handle this?".
        options (list[str]): Between 1 and 255 distinct candidate answers; duplicates are dropped.
        model (str | None): The Jev model. Defaults to `TYPESAFE_DEFAULT_MODEL`, else `jev-latest`.
        api_key (str | None): TypeSafe API key. Defaults to `TYPESAFE_API_KEY`.
        base_url (str | None): API root, e.g. to route through a proxy. Defaults to `TYPESAFE_BASE_URL`, else `https://api.typesafe.ai`.
        timeout (float): Seconds allowed for each request, which carries up to 25 rows. Defaults to 60.

    Returns:
        Expression: A String drawn from `options`, or null where `row` is null.

    Examples:
        >>> import daft
        >>> from daft.functions import jev_choice
        >>> df = daft.from_pydict({"ticket": ["I was charged twice", "The app crashes on login"]})
        >>> teams = ["billing", "technical", "security", "sales"]
        >>> df.with_column(
        ...     "team", jev_choice(df["ticket"], "which team should handle this?", teams)
        ... ).show()  # doctest: +SKIP
    """
    options = list(dict.fromkeys(options))
    if not 1 <= len(options) <= 255:
        raise ValueError(f"jev_choice requires between 1 and 255 options, got {len(options)}")
    return _client(api_key, model, base_url, timeout).choice(serialize(row, format="json"), question, options)