OpenAI functions agent in LangChain - Mini Project: Build & Apply

Build a Simple OpenAI Functions Agent with LangChain

📖 Scenario: You want to create a small assistant that can answer questions by calling a function you define. This is useful when you want your AI to do specific tasks like math or fetching data.

🎯 Goal: Build a LangChain OpenAI Functions agent that uses a custom function to answer questions about simple math operations.

📋 What You'll Learn

Create a function that adds two numbers

Set up a LangChain OpenAI Functions agent

Configure the agent to use the custom addition function

Run the agent to answer a math question using the function

💡 Why This Matters

🌍 Real World

OpenAI Functions agents let you connect AI chat models to your own code, so the AI can perform tasks like calculations, database queries, or API calls automatically.

💼 Career

Understanding how to build OpenAI Functions agents is useful for AI developers, chatbot creators, and software engineers who want to integrate AI with custom business logic.

1

Create the addition function

Write a Python function called add_numbers that takes two parameters a and b and returns their sum.

LangChain

# Your code here
def add_numbers(a, b):
    # Your code here

Need a hint?

Think of a function that takes two inputs and returns their sum using the return keyword.

2

Set up the OpenAI Functions agent

Import OpenAI and create_openai_functions_agent from langchain. Then create an OpenAI instance called llm with temperature=0.

LangChain

def add_numbers(a, b):
    return a + b

# Your code here
from langchain.chat_models import OpenAI
from langchain.agents import create_openai_functions_agent

llm = OpenAI(temperature=0)

Need a hint?

Use from langchain.chat_models import OpenAI and from langchain.agents import create_openai_functions_agent. Then create llm = OpenAI(temperature=0).

3

Configure the agent with the addition function

Create a list called functions with one dictionary describing the add_numbers function. The dictionary must have keys: name with value 'add_numbers', description with value 'Add two numbers', and parameters describing two required integer properties a and b. Then create the agent by calling create_openai_functions_agent(llm, functions) and assign it to agent.

LangChain

def add_numbers(a, b):
    return a + b

from langchain.chat_models import OpenAI
from langchain.agents import create_openai_functions_agent

llm = OpenAI(temperature=0)

# Your code here
functions = [
    {
        "name": "add_numbers",
        "description": "Add two numbers",
        "parameters": {
            "type": "object",
            "properties": {
                "a": {"type": "integer"},
                "b": {"type": "integer"}
            },
            "required": ["a", "b"]
        }
    }
]

agent = create_openai_functions_agent(llm, functions)

Need a hint?

Define functions as a list with one dictionary describing the function name, description, and parameters. Then create agent using create_openai_functions_agent.

4

Run the agent to answer a math question

Call agent.run with the string 'What is the sum of 5 and 7?' and assign the result to answer.

LangChain

def add_numbers(a, b):
    return a + b

from langchain.chat_models import OpenAI
from langchain.agents import create_openai_functions_agent

llm = OpenAI(temperature=0)

functions = [
    {
        "name": "add_numbers",
        "description": "Add two numbers",
        "parameters": {
            "type": "object",
            "properties": {
                "a": {"type": "integer"},
                "b": {"type": "integer"}
            },
            "required": ["a", "b"]
        }
    }
]

agent = create_openai_functions_agent(llm, functions)

# Your code here
answer = agent.run("What is the sum of 5 and 7?")

Need a hint?

Use answer = agent.run("What is the sum of 5 and 7?") to get the agent's response.