auto-base

A thin wrapper for the ag2 library, providing a simplified interface for easy usage.

Installation

pip install auto-base

This will automatically install the required dependency ag2.

Usage

Basic Usage

from auto_base import AutoBase

# Create an instance of the wrapper
auto = AutoBase()

# Check if ag2 is available
print(f"ag2 available: {auto.ag2_available}")

# Access autogen attributes directly
print(auto.DEFAULT_MODEL)  # Access autogen constants

# Get raw autogen instance
autogen_raw = auto.autogen

Agent Creation

from auto_base import AutoBase

auto = AutoBase()

# Create AssistantAgent
assistant = auto.create_assistant_agent(
    name="assistant",
    system_message="You are a helpful AI assistant",
    llm_config={
        "temperature": 0.7,
        "config_list": [
            {"model": "gpt-4", "api_key": "YOUR_API_KEY"}
        ]
    }
)

# Create UserProxyAgent
user_proxy = auto.create_user_proxy_agent(
    name="user_proxy",
    code_execution_config={"use_docker": False},  # Disable docker for code execution
    human_input_mode="NEVER"  # No human input required
)

# Create GroupChat
groupchat = auto.create_group_chat(
    agents=[assistant, user_proxy],
    messages=[],
    max_round=5,
    speaker_selection_method="auto"
)

# Create GroupChatManager
groupchat_manager = auto.create_group_chat_manager(
    groupchat=groupchat,
    llm_config={
        "config_list": [
            {"model": "gpt-4", "api_key": "YOUR_API_KEY"}
        ]
    }
)

Configuration Management

from auto_base import AutoBase

auto = AutoBase()

# Load config from dotenv file
config_list = auto.config_list_from_dotenv(
    dotenv_file_path=".env"
)

# Load config from JSON file
# config_list = auto.config_list_from_json(
#     file_path="OAI_CONFIG_LIST.json",
#     filter_dict={"model": ["gpt-4", "gpt-3.5-turbo"]}
# )

# Get GPT-4 and GPT-3.5 config list
# config_list = auto.config_list_gpt4_gpt35()

Function Registration

from auto_base import AutoBase

auto = AutoBase()

# Define a function to register
def add_numbers(a, b):
    """Add two numbers together"""
    return a + b

# Register the function
user_proxy = auto.create_user_proxy_agent(name="user_proxy")
auto.register_function(
    func=add_numbers,
    callable_agent=user_proxy,
    name="add_numbers",
    description="Add two numbers together"
)

Initiating Chats

from auto_base import AutoBase

auto = AutoBase()

# Create agents
assistant = auto.create_assistant_agent(name="assistant", system_message="You are helpful")
user_proxy = auto.create_user_proxy_agent(name="user_proxy", human_input_mode="NEVER")

# Initiate a chat between two agents
result = auto.initiate_chat(
    sender=user_proxy,
    receiver=assistant,
    message="What is the capital of France?"
)

# Or with group chat
# groupchat = auto.create_group_chat(agents=[assistant, user_proxy])
# manager = auto.create_group_chat_manager(groupchat=groupchat)
# result = auto.initiate_group_chat(
#     group_chat_manager=manager,
#     message="What is the capital of France?"
# )

Features

• Thin wrapper around autogen library with optional ag2 support
• Simplified API for creating and managing autogen agents
• Easy agent creation: AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
• Convenient configuration management
• Simplified function registration for agent use
• Easy chat initiation between agents
• Optional ag2 integration (available if ag2 is installed)
• Backward compatible with direct autogen access
• Well-documented with usage examples

Development

Installation for Development

# Clone the repository
git clone https://github.com/yourusername/auto-base.git
cd auto-base

# Install in development mode
pip install -e .

Running Tests

python test_wrapper.py

License

MIT