trulens.feedback.computer¶

trulens.feedback.computer ¶

Classes¶

RecordGraphNode ¶

Graph form of a record (i.e. a list of spans).

Functions¶

compute_feedback_by_span_group ¶

compute_feedback_by_span_group(
    events: DataFrame,
    feedback_name: str,
    feedback_function: Callable[
        [Any], Union[float, Tuple[float, Dict[str, Any]]]
    ],
    higher_is_better: bool,
    kwarg_to_selector: Dict[str, Selector],
    feedback_aggregator: Optional[
        Callable[[List[float]], float]
    ] = None,
    raise_error_on_no_feedbacks_computed: bool = True,
) -> None

Compute feedback based on span groups in events.

PARAMETER	DESCRIPTION
`events`	DataFrame containing trace events. TYPE: `DataFrame`
`feedback_name`	Name of the feedback function. TYPE: `str`
`feedback_function`	Function to compute feedback. TYPE: `Callable[[Any], Union[float, Tuple[float, Dict[str, Any]]]]`
`higher_is_better`	Whether higher values are better. TYPE: `bool`
`kwarg_to_selector`	Mapping from function kwargs to span selectors TYPE: `Dict[str, Selector]`
`feedback_aggregator`	Aggregator function to combine feedback scores. TYPE: `Optional[Callable[[List[float]], float]]` DEFAULT: `None`
`raise_error_on_no_feedbacks_computed`	Raise an error if no feedbacks were computed. Default is True. TYPE: `bool` DEFAULT: `True`