why AgentChat

In earlier chapters, we covered the basics of AutoGen Core.

The official AutoGen documentation suggests starting with AutoGen AgentChat. You might wonder why I focused on AutoGen Core first. And now, especially when we were close to finishing the HR TimeSheet automation use case, only to switch to AgentChat.

The reason is simple: understanding AutoGen Core is essential for building professional AI Agent applications. It’s the foundation you need to become a skilled AI Agent Engineer.

Let’s dive into AgentChat.

AgentChat is a simple, high-level API for creating multi-agent apps, built on top of AutoGen core. It’s great for beginners, while advanced users can use AutoGen-core’s event-driven model for more control. Here’s the thing: AgentChat is largely community-driven and comes with many ready-made Agent and team setups, making it super easy to start with.

One popular design pattern, Magentic-One, which we’ll cover later, was originally built in AutoGen Core 0.2. It helps users quickly create professional use cases. Now, it’s been fully rebuilt in AgentChat.

AgentChat, with its strong community support, has become more than just a basic API. It’s so powerful that even when using AutoGen Core, instead of building complex Agent patterns from scratch, you can just grab a preset Agent and team setup from AgentChat and plug it into your workflow, it works perfectly.

AutoGen AgentChat¶

AutoGen AgentChat is a user-friendly, opinionated, high-level API built on AutoGen Core. When you want to quickly test a feature without writing much code, you can use Autogen Studio, which offers a nice visual interface to try out AI Agents fast.

For more detailed control over your proof of concept, AgentChat steps in. It’s not as beginner-friendly as Autogen Studio’s no-code, drag-and-drop setup, but AgentChat gives you some control over your Agents.

However, Don’t think AgentChat is just for quick, high-level proofs of concept. Lately, it’s grown so popular and strong that it rivals AutoGen-Core itself. In fact, AutoGen’s most famous design pattern, Magentic-One once a highlight of AutoGen 0.2 is now fully built in AgentChat.

Getting started with AgentChat¶

This is an example script from the AutoGen Documentation designed to help you quickly begin using the AutoGen AgentChat API. In the following chapters, we will explore this API in greater depth.

pip install -U "autogen-agentchat" "autogen-ext[openai,azure]"

from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.ui import Console
from autogen_ext.models.openai import OpenAIChatCompletionClient

# Define a model client. You can use other model client that implements
# the `ChatCompletionClient` interface.
model_client = OpenAIChatCompletionClient(
    model="gpt-4o",
    # api_key="YOUR_API_KEY",
)


# Define a simple function tool that the agent can use.
# For this example, we use a fake weather tool for demonstration purposes.
async def get_weather(city: str) -> str:
    """Get the weather for a given city."""
    return f"The weather in {city} is 73 degrees and Sunny."


# Define an AssistantAgent with the model, tool, system message, and reflection enabled.
# The system message instructs the agent via natural language.
agent = AssistantAgent(
    name="weather_agent",
    model_client=model_client,
    tools=[get_weather],
    system_message="You are a helpful assistant.",
    reflect_on_tool_use=True,
    model_client_stream=True,  # Enable streaming tokens from the model client.
)


# Run the agent and stream the messages to the console.
async def main() -> None:
    await Console(agent.run_stream(task="What is the weather in New York?"))


# NOTE: if running this inside a Python script you'll need to use asyncio.run(main()).
await main()