cai/examples/triage_agent/evals_util.py

24 lines
588 B
Python

from openai import OpenAI
import instructor
from pydantic import BaseModel
from typing import Optional
__client = instructor.from_openai(OpenAI())
class BoolEvalResult(BaseModel):
value: bool
reason: Optional[str]
def evaluate_with_llm_bool(instruction, data) -> BoolEvalResult:
eval_result, _ = __client.chat.completions.create_with_completion(
model="gpt-4o",
messages=[
{"role": "system", "content": instruction},
{"role": "user", "content": data},
],
response_model=BoolEvalResult,
)
return eval_result