03. 구조화된 출력 체인(with_structered_output)

Chains using structured output (with_structured_output)

# Configuration file for managing API keys as environment variables
from dotenv import load_dotenv

# Load API key information
load_dotenv()

True

# Set up LangSmith tracking. https://smith.langchain.com
# !pip install langchain-teddynote
from langchain_teddynote import logging

# Enter a project name.
logging.langsmith("Structured-Output-Chain")

Start tracking LangSmith.
[Project name]
Structured-Output-Chain

Implement a process to generate a 4-option multiple choice quiz on a specific topic.

The Quiz class defines the quiz's question, difficulty, and four choices.

The ChatOpenAI instance performs natural language processing using the GPT-4o model, and the ChatPromptTemplate defines the conversational prompt for quiz generation.

from langchain.chains.openai_functions import create_structured_output_runnable
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.pydantic_v1 import BaseModel, Field
from typing import List


class Quiz(BaseModel):
    """Extract information from a 4-option multiple choice quiz"""

    question: str = Field(..., description="The best of the best")
    level: str = Field(
        ..., description="Indicates the difficulty of the quiz (easy, medium, hard)"
    )
    options: List[str] = Field(..., description="There are four choices in the quiz.")


llm = ChatOpenAI(model="gpt-4o", temperature=0.1)
prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            "You're a world-famous quizzer and generates quizzes in structured formats.",
        ),
        (
            "human",
            "Please give a 4-option multiple choice quiz related to the content presented in TOPIC.. If there is an actual past exam question, please create a similar question."
            "However, do not include any content or information about the TOPIC in the question. \nTOPIC:\n{topic}",
        ),
        ("human", "Tip: Make sure to answer in the correct format"),
    ]
)

# Create a model for structured output
llm_with_structured_output = llm.with_structured_output(Quiz)

# Create a quiz creation chain
chain = prompt | llm_with_structured_output

# Request the creation of a quiz.
generated_quiz = chain.invoke({"topic": "ADSPData Analysis Associate Professional) Qualification Exam"})

Prints the generated quiz.

# Generated quiz output
print(f"{generated_quiz.question} (Difficulty: {generated_quiz.level})\n")
for i, opt in enumerate(generated_quiz.options):
    print(f"{i+1}) {opt}")

Which of the following is the first step in the data analysis process? (Difficulty: Moderate)

1) Data collection 
2) Data preprocessing 
3) Problem definition 
4) Model evaluation