Skip to content

ag2-tutorial

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

ag2-tutorial

README.md

layout title nav_order has_children
default
AG2 Tutorial
73
true

AG2 Tutorial: Next-Generation Multi-Agent Framework

Build collaborative AI agent systems with AG2, the community-driven successor to AutoGen.

🤖 Open-Source Multi-Agent AI Development

GitHub

🎯 What is AG2?

AG2View Repo is the community-driven successor to Microsoft's AutoGen framework. It provides a powerful, open-source platform for building AI agents that can collaborate to solve complex tasks through natural conversation.

AG2 vs AutoGen

Aspect AG2 Microsoft AutoGen
Governance Open community-driven Microsoft-controlled
Development Community contributions Microsoft team
License Apache 2.0 MIT
Focus Multi-agent collaboration Part of Microsoft Agent Framework
Repository github.com/ag2ai/ag2 github.com/microsoft/autogen

Note: The original AutoGen creators transitioned to AG2 to promote open governance. Microsoft continues developing AutoGen as part of their Agent Framework.

flowchart TD
    A[User Task] --> B[AssistantAgent]
    B <--> C[UserProxyAgent]
    
    B --> D{Needs Tools?}
    D -->|Yes| E[Code Execution]
    D -->|Yes| F[Function Calling]
    D -->|No| G[Direct Response]
    
    E --> H[Results]
    F --> H
    G --> H
    
    H --> I{Task Complete?}
    I -->|No| B
    I -->|Yes| J[Final Output]
    
    C --> K[Human Feedback]
    K --> B
    
    classDef agent fill:#e1f5fe,stroke:#01579b
    classDef decision fill:#f3e5f5,stroke:#4a148c
    classDef action fill:#fff3e0,stroke:#ef6c00
    classDef output fill:#e8f5e8,stroke:#1b5e20
    
    class A,B,C agent
    class D,I decision
    class E,F,G,K action
    class H,J output
Loading

Current Snapshot (auto-updated)

  • repository: ag2ai/ag2
  • stars: about 4.3k
  • latest release: v0.11.3 (published 2026-03-16)

Core Concepts

Agent Types

Agent Purpose Key Features
AssistantAgent AI-powered task solving LLM-based reasoning, tool use
UserProxyAgent Human interaction Code execution, human feedback
GroupChatManager Multi-agent coordination Conversation flow, speaker selection
ConversableAgent Base class for custom agents Flexible conversation patterns

Conversation Patterns

  • Two-Agent Chat - Simple back-and-forth between agents
  • Sequential Chat - Ordered multi-agent workflows
  • Group Chat - Multiple agents collaborating
  • Nested Chat - Hierarchical agent structures

Tutorial Chapters

  1. Chapter 1: Getting Started - Installation, setup, and your first AG2 agents
  2. Chapter 2: Agent Types - Understanding and configuring different agents
  3. Chapter 3: Conversation Patterns - Two-agent, sequential, and group chats
  4. Chapter 4: Code Execution - Secure code execution with Docker
  5. Chapter 5: Tool Integration - Function calling and external tools
  6. Chapter 6: Group Chat - Multi-agent collaboration and speaker selection
  7. Chapter 7: Advanced Patterns - Nested chats, caching, and optimization
  8. Chapter 8: Production Deployment - Scaling, monitoring, and best practices
  9. Chapter 9: Enterprise Operations - Governance, evals, safety, and cost/perf tuning

What You'll Learn

  • Create Conversational Agents that collaborate through natural dialogue
  • Implement Code Execution safely with Docker sandboxing
  • Build Multi-Agent Systems with group chat coordination
  • Integrate External Tools through function calling
  • Design Complex Workflows with nested and sequential patterns
  • Handle Human-in-the-Loop scenarios effectively
  • Deploy Production Systems with proper scaling and monitoring

Prerequisites

  • Python 3.9+
  • OpenAI API key (or other LLM provider)
  • Docker (for code execution)
  • Basic understanding of LLMs

Quick Start

# Install AG2
pip install ag2

# Or with all extras
pip install ag2[all]
from ag2 import AssistantAgent, UserProxyAgent

# Create an assistant agent
assistant = AssistantAgent(
    name="assistant",
    llm_config={
        "model": "gpt-4",
        "api_key": "your-api-key"
    }
)

# Create a user proxy agent
user_proxy = UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",  # Automatic mode
    code_execution_config={
        "work_dir": "coding",
        "use_docker": True
    }
)

# Start a conversation
user_proxy.initiate_chat(
    assistant,
    message="Write a Python function to calculate fibonacci numbers and test it."
)

Example: Research Team

from ag2 import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager

# Create specialized agents
researcher = AssistantAgent(
    name="Researcher",
    system_message="You are a research assistant. Search for and analyze information.",
    llm_config=llm_config
)

writer = AssistantAgent(
    name="Writer", 
    system_message="You are a technical writer. Create clear, well-structured content.",
    llm_config=llm_config
)

critic = AssistantAgent(
    name="Critic",
    system_message="You review content for accuracy and suggest improvements.",
    llm_config=llm_config
)

user_proxy = UserProxyAgent(
    name="User",
    human_input_mode="TERMINATE",
    code_execution_config=False
)

# Create group chat
group_chat = GroupChat(
    agents=[user_proxy, researcher, writer, critic],
    messages=[],
    max_round=10
)

manager = GroupChatManager(groupchat=group_chat, llm_config=llm_config)

# Start research task
user_proxy.initiate_chat(
    manager,
    message="Research and write a brief summary about quantum computing applications."
)

Learning Path

🟢 Beginner Track

  1. Chapters 1-3: Setup, agent types, and basic conversations
  2. Build simple two-agent systems

🟡 Intermediate Track

  1. Chapters 4-6: Code execution, tools, and group chat
  2. Create collaborative multi-agent workflows

🔴 Advanced Track

  1. Chapters 7-8: Advanced patterns and production deployment
  2. Master enterprise-grade agent systems

Ready to build multi-agent AI systems? Let's begin with Chapter 1: Getting Started!

Generated for Awesome Code Docs

Navigation & Backlinks

Full Chapter Map

  1. Chapter 1: Getting Started with AG2
  2. Chapter 2: Agent Types & Configuration
  3. Chapter 3: Conversation Patterns
  4. Chapter 4: Code Execution & Security
  5. Chapter 5: Tool Integration & Function Calling
  6. Chapter 6: Group Chat & Multi-Agent Collaboration
  7. Chapter 7: Advanced Patterns & Optimization
  8. Chapter 8: Production Deployment & Scaling
  9. Chapter 9: Enterprise Operations & Advanced Patterns

Source References

Generated by AI Codebase Knowledge Builder