Prediction¶

Prediction(
    values,
    *,
    turn=None,
    response=None,
    responses=(),
    repairs=(),
    attempts=1,
    probabilities=None,
    measured_by=None,
)

Everything one call produced.

• the outputs, by name (pred.result, pred.reasoning, dict(pred))
• pred.turn: the lmcc turn (inputs, every model and tool step, outputs)
• pred.response / pred.responses: the lm15 responses
• pred.usage: tokens summed over every model call
• pred.repairs: what the reader forgave in the reply
• pred.call_id: the call's id, for functai.rate and the call log

Attributes¶