DEV Community

Alex Spinov
Alex Spinov

Posted on

AutoGen Has a Free API — Build Multi-Agent AI Conversations

#ai #llm #machinelearning #python

AutoGen: Microsoft's Multi-Agent Conversation Framework

AutoGen by Microsoft Research enables multi-agent conversations where AI agents collaborate, debate, and solve complex tasks through dialogue. Agents can use tools, execute code, and involve humans in the loop.

Why AutoGen

  • Agents converse to solve problems
  • Code execution in sandboxed environment
  • Human-in-the-loop at any step
  • Group chat with multiple agents
  • Tool use and function calling

The Free API 

from autogen import AssistantAgent, UserProxyAgent
import os

config = [{"model": "gpt-4o", "api_key": os.getenv("OPENAI_API_KEY")}]

assistant = AssistantAgent(
    name="coder",
    llm_config={"config_list": config}
)

user = UserProxyAgent(
    name="user",
    human_input_mode="NEVER",
    code_execution_config={"work_dir": "coding", "use_docker": True}
)

# Start conversation — assistant writes code, user executes it
user.initiate_chat(
    assistant,
    message="Write a Python script that downloads Bitcoin price for the last 30 days and plots it"
)
Enter fullscreen mode Exit fullscreen mode

Group Chat (Multiple Agents) 

from autogen import GroupChat, GroupChatManager

researcher = AssistantAgent("researcher", llm_config={"config_list": config},
    system_message="You research topics and provide facts.")

writer = AssistantAgent("writer", llm_config={"config_list": config},
    system_message="You write articles based on research.")

editor = AssistantAgent("editor", llm_config={"config_list": config},
    system_message="You review and improve articles.")

groupchat = GroupChat(
    agents=[researcher, writer, editor, user],
    messages=[],
    max_round=10
)

manager = GroupChatManager(groupchat=groupchat, llm_config={"config_list": config})

user.initiate_chat(manager, message="Write an article about eBPF in networking")
Enter fullscreen mode Exit fullscreen mode

Function Calling 

import requests

def get_weather(city: str) -> str:
    """Get current weather for a city."""
    resp = requests.get(f"https://wttr.in/{city}?format=3")
    return resp.text

assistant = AssistantAgent("assistant", llm_config={"config_list": config})
assistant.register_for_llm(name="get_weather", description="Get weather")(get_weather)
user.register_for_execution(name="get_weather")(get_weather)

user.initiate_chat(assistant, message="What is the weather in Tokyo?")
Enter fullscreen mode Exit fullscreen mode

Real-World Use Case

A research lab needed to automate paper analysis. AutoGen group chat: Paper Reader (extracts key points) -> Critic (identifies weaknesses) -> Summarizer (creates brief). Processed 500 papers in a weekend, human reviewed only flagged ones.

Quick Start 

pip install autogen-agentchat
Enter fullscreen mode Exit fullscreen mode

Resources

Need AI agent data pipelines? Check out my tools on Apify or email spinov001@gmail.com.

Top comments (0)