Structured Output Chain

• Author: JeongHo Shin

• Peer Review:

• Proofread : Juni Lee

• This is a part of LangChain Open Tutorial

Overview

This tutorial demonstrates how to implement structured output generation using LangChain and OpenAI's language models.

We'll build a quiz generation system that creates multiple-choice questions with consistent formatting and structure.

Table of Contents

• Overview

• Environment Setup

• Implementing Structured Output Chain

• Invoking the Chain

Environment Setup

Setting up your environment is the first step. See the Environment Setup guide for more details.

[Note]

The langchain-opentutorial is a package of easy-to-use environment setup guidance, useful functions and utilities for tutorials. Check out the langchain-opentutorial for more details.

%%capture --no-stderr
%pip install langchain-opentutorial

# Install required packages
from langchain_opentutorial import package

package.install(
    [
        "langsmith",
        "langchain",
        "langchain_core",
        "langchain-anthropic",
        "langchain_community",
        "langchain_text_splitters",
        "langchain_openai",
    ],
    verbose=False,
    upgrade=False,
)

You can set API keys in a .env file or set them manually.

[Note] If you’re not using the .env file, no worries! Just enter the keys directly in the cell below, and you’re good to go.

from dotenv import load_dotenv
from langchain_opentutorial import set_env

# Attempt to load environment variables from a .env file; if unsuccessful, set them manually.
if not load_dotenv():
    set_env(
        {
            "OPENAI_API_KEY": "",
            "LANGCHAIN_API_KEY": "",
            "LANGCHAIN_TRACING_V2": "true",
            "LANGCHAIN_ENDPOINT": "https://api.smith.langchain.com",
            "LANGCHAIN_PROJECT": "Structured-Output-Chain", 
        }
    )

Implementing Structured Output Chain

This tutorial walks you through the process of generating 4-option multiple-choice quizzes for a given topic.

The Quiz class defines the structure of the quiz, including the question, difficulty level, and four answer options.

A ChatOpenAI instance leverages the GPT-4o model for natural language processing, while a ChatPromptTemplate specifies the conversational instructions for generating the quizzes dynamically.

# Import required modules and libraries
from langchain.chains.openai_functions import create_structured_output_runnable
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from pydantic import BaseModel, Field
from typing import List

# Define the Quiz class - Represents the structure of a 4-option multiple-choice quiz
class Quiz(BaseModel):
    """Extracts information for a 4-option multiple-choice quiz"""

    question: str = Field(..., description="The quiz question")  # Quiz question
    level: str = Field(
        ..., description="The difficulty level of the quiz (easy, medium, hard)"
    )
    options: List[str] = Field(..., description="The 4 answer options for the quiz")  # Answer options


# Set up the GPT-4o model with appropriate parameters
llm = ChatOpenAI(model="gpt-4o", temperature=0.1)

# Define a prompt template to guide the model in generating quizzes
prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            "You're a world-famous quizzer and generate quizzes in structured formats.",
        ),
        (
            "human",
            "Please create a 4-option multiple-choice quiz related to the topic provided below. "
            "If possible, base the question on existing trivia but do not directly include details from the topic in the question. "
            "\nTOPIC:\n{topic}",
        ),
        ("human", "Tip: Make sure to answer in the correct format"),
    ]
)

# Create a structured output model to match the Quiz class structure
llm_with_structured_output = llm.with_structured_output(Quiz)

# Combine the prompt and the structured output model into a single chain
chain = prompt | llm_with_structured_output

Invoking the Chain

In this section, we demonstrate how to invoke the structured output chain to generate quizzes dynamically. The chain combines a prompt template and a structured output model to ensure the output adheres to the desired Quiz structure.

# Request the generation of a quiz based on a given topic
generated_quiz = chain.invoke({"topic": "Korean Food"})

# Print the generated quiz
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 a traditional Korean dish made by fermenting vegetables, primarily napa cabbage and Korean radishes, with a variety of seasonings? (Difficulty: medium)
    
    1) Kimchi
    2) Sushi
    3) Tacos
    4) Paella