Serializing Components

AutoGen provides a {py:class}~autogen_core.Component configuration class that defines behaviours to serialize/deserialize component into declarative specifications. We can accomplish this by calling .dump_component() and .load_component() respectively. This is useful for debugging, visualizing, and even for sharing your work with others. In this notebook, we will demonstrate how to serialize multiple components to a declarative specification like a JSON file.

{warning}


ONLY LOAD COMPONENTS FROM TRUSTED SOURCES.

With serilized components, each component implements the logic for how it is serialized and deserialized - i.e., how the declarative specification is generated and how it is converted back to an object. 

In some cases, creating an object may include executing code (e.g., a serialized function). ONLY LOAD COMPONENTS FROM TRUSTED SOURCES.

{note}

`selector_func` is not serializable and will be ignored during serialization and deserialization process.

Termination Condition Example

In the example below, we will define termination conditions (a part of an agent team) in python, export this to a dictionary/json and also demonstrate how the termination condition object can be loaded from the dictionary/json.

python

from autogen_agentchat.conditions import MaxMessageTermination, StopMessageTermination

max_termination = MaxMessageTermination(5)
stop_termination = StopMessageTermination()

or_termination = max_termination | stop_termination

or_term_config = or_termination.dump_component()
print("Config: ", or_term_config.model_dump_json())

new_or_termination = or_termination.load_component(or_term_config)

Agent Example

In the example below, we will define an agent in python, export this to a dictionary/json and also demonstrate how the agent object can be loaded from the dictionary/json.

python

from autogen_agentchat.agents import AssistantAgent, UserProxyAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient

# Create an agent that uses the OpenAI GPT-4o model.
model_client = OpenAIChatCompletionClient(
    model="gpt-4o",
    # api_key="YOUR_API_KEY",
)
agent = AssistantAgent(
    name="assistant",
    model_client=model_client,
    handoffs=["flights_refunder", "user"],
    # tools=[], # serializing tools is not yet supported
    system_message="Use tools to solve tasks.",
)
user_proxy = UserProxyAgent(name="user")

python

user_proxy_config = user_proxy.dump_component()  # dump component
print(user_proxy_config.model_dump_json())
up_new = user_proxy.load_component(user_proxy_config)  # load component

python

agent_config = agent.dump_component()  # dump component
print(agent_config.model_dump_json())
agent_new = agent.load_component(agent_config)  # load component

A similar approach can be used to serialize the MultiModalWebSurfer agent.

python

from autogen_ext.agents.web_surfer import MultimodalWebSurfer

agent = MultimodalWebSurfer(
    name="web_surfer",
    model_client=model_client,
    headless=False,
)

web_surfer_config = agent.dump_component()  # dump component
print(web_surfer_config.model_dump_json())

Team Example

In the example below, we will define a team in python, export this to a dictionary/json and also demonstrate how the team object can be loaded from the dictionary/json.

python

from autogen_agentchat.agents import AssistantAgent, UserProxyAgent
from autogen_agentchat.conditions import MaxMessageTermination
from autogen_agentchat.teams import RoundRobinGroupChat
from autogen_ext.models.openai import OpenAIChatCompletionClient

# Create an agent that uses the OpenAI GPT-4o model.
model_client = OpenAIChatCompletionClient(
    model="gpt-4o",
    # api_key="YOUR_API_KEY",
)
agent = AssistantAgent(
    name="assistant",
    model_client=model_client,
    handoffs=["flights_refunder", "user"],
    # tools=[], # serializing tools is not yet supported
    system_message="Use tools to solve tasks.",
)

team = RoundRobinGroupChat(participants=[agent], termination_condition=MaxMessageTermination(2))

team_config = team.dump_component()  # dump component
print(team_config.model_dump_json())

await model_client.close()