Agents with Structured Outputs

When you run your agent or multi-agent framework, you might want it to output the result in a specific format. In this notebook, we will see a simple example of how to apply this to a FunctionAgent!🦙🚀

Let’s first install the needed dependencies

! pip install llama-index

from getpass import getpass
import os

os.environ["OPENAI_API_KEY"] = getpass()

Let’s now define our structured output format

from pydantic import BaseModel, Field


class MathResult(BaseModel):
    operation: str = Field(description="The operation that has been performed")
    result: int = Field(description="Result of the operation")

And a very simple calculator agent

from llama_index.llms.openai import OpenAI
from llama_index.core.agent.workflow import FunctionAgent

llm = OpenAI(model="gpt-4.1")


def add(x: int, y: int):
    """Add two numbers"""
    return x + y


def multiply(x: int, y: int):
    """Multiply two numbers"""
    return x * y


agent = FunctionAgent(
    llm=llm,
    output_cls=MathResult,
    tools=[add, multiply],
    system_prompt="You are a calculator agent that can add or multiply two numbers by calling tools",
    name="calculator",
)

Let’s now run the agent

response = await agent.run("What is the result of 10 multiplied by 4?")

Finally, we can get the structured output

# print the structured output as a plain dictionary
print(response.structured_response)
# print the structured output as a Pydantic model
print(response.get_pydantic_model(MathResult))