Getting to Know AutoGen(Part2): How AI Agents Work Together

Author(s): Anushka sonawane

Originally published on Towards AI.

In Part 1, we went over the basics — what AI agents are, how they work, and why having multiple agents can really make a difference. That was just an introduction, setting the stage for what’s next. Now, it’s time to take things up a level!

In Part 2, let’s go deeper into AutoGen and how it helps these agents communicate with each other to get things done.

With AutoGen, the agents don’t just work alone. They can actually talk to each other to share information and solve problems together. This makes them much more powerful!

AutoGen’s agents come with two key features:

📍Conversable Agents: Agents that Talk to Each Other. They can share information, ask for help, or update each other, making teamwork easier and faster.

📍Customizable Agents: Agents You Can Customize. Some can write, others can code, and you can even include human help when needed.

Prerequisites

Before diving into the example, let’s make sure you have the following prerequisites covered:

1. AutoGen Setup: Ensure that you have AutoGen installed and ready to use in your environment.

pip install pyautogen

2. API Access: You’ll need API access to Large Language Models (LLMs), like OpenAI’s GPT or Gemini.

Here’s how you can configure OpenAI’s GPT-4 for your agents:

Here’s how you can configure Gemini for your agents:

Now that you’ve set up the LLM configurations, all that’s left is to add this configuration to your AutoGen agents. It’s simple — just pass the llm_config we defined earlier when creating the agents.

guide_gary = ConversableAgent(
 "guide_Gary",
 system_message="Hello, I’m Guide Gary! I specialize in travel tips, destination recommendations, and hidden gems around the world.",
 llm_config={"config_list": [{"model": "gpt-3.5-turbo", "temperature": 0.9, "api_key": "OPENAI_API_KEY"}]},
 human_input_mode="NEVER", 
)

tourist_tina = ConversableAgent(
 "tourist_Tina",
 system_message="Hi there, I’m Tourist Tina! I’m always on the lookout for exciting travel destinations and unique experiences.",
 llm_config={"config_list": [{"model": "gpt-3.5-turbo", "temperature": 0.7, "api_key": "OPENAI_API_KEY"}]},
 human_input_mode="NEVER", 
)

result = tourist_tina.initiate_chat(guide_gary, message="Guide Gary, I'm planning a trip to Norway. Any must-see destinations?",
 max_turns=3)

Here’s what the output looks like:

If you’re excited to see this in details, I’ve put together a GitHub notebook that breaks it all down. Inside, you’ll find:

• A list of LLMs (Large Language Models)
• A code executor
• A function and tool executor
• A component to keep humans in the loop

Language Models (LLMs):

• The agent can use different language models to chat in natural language. This means it can understand and respond to your questions or requests in a friendly way, whether you use simple phrases or more complex sentences.

Code Executor:

• It can run code when necessary. This is great for tasks that need calculations or automating certain processes, making it a handy helper for technical tasks.

Function and Tool Executor:

• The agent can use pre-set functions and tools to perform specific actions, like finding information, doing calculations, or calling up other online services. This makes it really efficient at handling various requests.

Human-in-the-Loop:

• You can set it up to involve people in the conversation. This means the agent can ask for your input or feedback, ensuring that it gets things right and works well with you.

📍AutoGen makes it easy for AI agents to work together, and that’s pretty exciting! These Conversable Agents can chat with each other, sharing information to get tasks done faster.

📍The AssistantAgent helps by creating and improving Python code based on what you need, so you don’t have to start from scratch. On the other hand, the UserProxyAgent keeps you in the loop. It asks for your input and can run code automatically when necessary.

📍Thanks to the auto-reply feature, these agents can chat with each other and handle tasks on their own while still keeping you in the loop. Plus, you can customize them to fit your specific needs, whether it’s for travel advice or coding help.

The image below shows how these agents interact and work together.

Well, that’s the end of Part 2! I hope this gave you a clearer picture of how AutoGen works and how these agents can collaborate to make life easier.

If you’d like to follow along with more insights or discuss any of these topics further, feel free to connect with me:

• LinkedIn
• Medium

Looking forward to chatting and sharing more ideas!

Wait, There’s More!
If you enjoyed this, you’ll love my other blogs! 🎯

Until next time,
Anushka!

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI