	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > LangChain > LangChain with OpenAI Functions

LangChain with OpenAI Functions

Author: Venkata Sudhakar

LangChain integrates with OpenAI function calling to allow agents to reliably call Python functions with structured arguments. Instead of asking the LLM to return JSON and then parsing it, OpenAI function calling guarantees that the LLM will invoke a specific function with arguments matching a defined schema. LangChain wraps this into a clean tool interface that works the same way across OpenAI, Gemini, and Anthropic models.

The key difference between function calling and regular prompting is determinism. Regular prompts ask the LLM to describe what to do. Function calling tells the LLM to actually do it - selecting the right function and providing structured arguments that your code can execute directly. This is the foundation of reliable agent behavior in production systems.

The below example builds a ShopMax India product assistant using LangChain with OpenAI function calling, showing how the LLM selects and invokes the correct tool based on natural language input.

from langchain_openai import ChatOpenAI
from langchain.tools import tool
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_core.prompts import ChatPromptTemplate

@tool
def search_products(query: str, max_price: int = 100000) -> str:
    """Search ShopMax India product catalogue by query and optional max price in Rs."""
    catalogue = [
        {"name": "Samsung Galaxy S24", "price": 74999, "category": "smartphone"},
        {"name": "MacBook Air M3", "price": 114900, "category": "laptop"},
        {"name": "Sony WH-1000XM5", "price": 29999, "category": "headphones"},
        {"name": "iPad Pro 12.9", "price": 112900, "category": "tablet"},
        {"name": "OnePlus 12", "price": 64999, "category": "smartphone"}
    ]
    results = [p for p in catalogue if query.lower() in p["name"].lower() or query.lower() in p["category"] if p["price"] <= max_price]
    if not results:
        return "No products found matching your criteria."
    return "\n".join([f"{p['name']} - Rs {p['price']}" for p in results])

@tool
def calculate_emi(price: int, months: int = 6) -> str:
    """Calculate 0% EMI monthly instalment for a ShopMax product at given price and tenure."""
    if price < 30000:
        return "EMI not available for orders below Rs 30000."
    monthly = round(price / months)
    return f"Rs {monthly}/month for {months} months (0% interest). Total: Rs {price}"

llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
tools = [search_products, calculate_emi]
prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a ShopMax India shopping assistant. Help customers find products and calculate EMI."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}")
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools)

result = executor.invoke({"input": "Show me smartphones under Rs 70000 and calculate 12-month EMI for the cheapest one."})
print(result["output"])

It gives the following output,

Here are smartphones under Rs 70000 at ShopMax India:
- OnePlus 12 - Rs 64999

For the OnePlus 12 at Rs 64999, the 12-month EMI is:
Rs 5417/month for 12 months (0% interest). Total: Rs 64999

The agent automatically chained two function calls - first searching for smartphones under Rs 70000, then calculating EMI for the cheapest result - without any explicit orchestration code. Function calling ensures arguments are typed and validated before your code runs. For ShopMax production systems, replace the mock data with real API calls to your product catalogue and pricing engine.

Send your comments, suggestions or queries regarding this site to [email protected].