Using OpenAI Swarm in Python: A Step-by-Step Guide

In this tutorial, you will learn how to set up and use OpenAI Swarm in Python to create a multi-agent system. The tutorial will cover environment setup, and basic usage of the OpenAI Swarm framework.

But what can I do with OpenAI Swarm?

With Firecrawl and OpenAI Swarm, a multi-agent system could crawl the internet and answer the pricing of LaunchFa.st starter kits in JSON format 💥

Prerequisites

You’ll need the following:

Python 3.10 or higher
pip package installer
OpenAI Account

Set up a new virtual environment
Define dependencies
Install the dependencies
Define environment variables
Create a Python script to use OpenAI Swarm
Run the Script

Set up a new virtual environment

# Create a new directory for your project
mkdir openai-swarm-tutorial

# Navigate to the project directory
cd openai-swarm-tutorial

# Create a new virtual environment
python -m venv venv

# Activate the virtual environment

## On Windows:
venv\Scripts\activate

## On macOS and Linux:
source venv/bin/activate

Define dependencies

Create a requirements.txt file in the project directory with the following dependencies:

git+https://github.com/openai/swarm.git
python-dotenv

Install the dependencies

Install the dependencies using pip:

pip install -r requirements.txt

Define environment variables

Create a .env file in the project directory with the following:

OPENAI_API_KEY="your_api_key_here"

Replace your_api_key_here with your actual OpenAI API key.

Create a Python script to use OpenAI Swarm

Create the main script app.py with the following code:

from dotenv import load_dotenv
from swarm import Agent, Swarm

# Load environment variables
load_dotenv()

# Initialize the Swarm client
client = Swarm()

def transfer_to_agent_b():
    return agent_b

# Define Agent A
agent_a = Agent(
    name="Agent A",
    instructions="You are a helpful agent.",
    functions=[transfer_to_agent_b],
)

# Define Agent B
agent_b = Agent(
    name="Agent B",
    instructions="Only speak in Haikus.",
)

# Run the Swarm with Agent A
response = client.run(
    agent=agent_a,
    messages=[{"role": "user", "content": "I want to talk to agent B."}],
)

# Print the last message from the response
print(response.messages[-1]["content"])

This script does the following:

Imports necessary modules and loads the OPENAI_API_KEY environment variable.
Initializes the OpenAI Swarm client.
Defines a transfer_to_agent_b function to switch to Agent B.
Creates two agents: Agent A and Agent B with specific instructions.
Runs the Swarm with Agent A and a user message.
Prints the last message from the response.

Run the Script

Execute the script:

python app.py

This will process the initial message with Agent A, potentially transfer to Agent B, and display the result.

Conclusion

You have now learned how to set up and use OpenAI Swarm to create a basic multi-agent system. This example demonstrates agent definition, inter-agent communication, and basic Swarm execution. To expand on this, consider adding more agents, implementing complex transfer logic, or exploring advanced Swarm features like memory or tool use.

If you have any questions or comments, feel free to reach out to me on Twitter.