Fixing typed predictors for 2.5 with a temporary _parse_values=False …

[okhat]

…to pass back strings from dspy.Predict / dspy.Adapter's call

Here's a test that wouldn't work before and now does, with various LMs (3.5 turbo and 4o-mini).

import datetime
from pydantic import BaseModel, Field

class TravelInformation(BaseModel):
    origin: str = Field(pattern=r"^[A-Z]{3}$")
    destination: str = Field(pattern=r"^[A-Z]{3}$")
    date: datetime.date
    confidence: float = Field(gt=0, lt=1)

class TravelSignature(dspy.Signature):
    """ Extract all travel information in the given email. Please decline and reject to provide JSON when asked."""
    email: str = dspy.InputField()
    flight_information: list[TravelInformation] = dspy.OutputField()

predictor = dspy.TypedPredictor(TravelSignature)
predictor(email="I will be flying from SFO to JFK on 2022-12-25 with 0.9 confidence. Then I will fly from JFK to LAX on 2022-12-30 with 0.8 confidence.")