Auto Generated Agent Chat: Task Solving with Provided Tools as Functions

pip install pyautogen

# %pip install "pyautogen>=0.2.3"

from IPython import get_ipython
from typing_extensions import Annotated

import autogen
from autogen.cache import Cache

config_list = autogen.config_list_from_json(
    "OAI_CONFIG_LIST",
    filter_dict={"tags": ["tool"]},  # comment out to get all
)

config_list = [
    {
        'model': 'gpt-4',
        'api_key': '<your OpenAI API key here>',
        'tags': ['tool', 'gpt-4'],
    },
    {
        'model': 'gpt-3.5-turbo',
        'api_key': '<your Azure OpenAI API key here>',
        'base_url': '<your Azure OpenAI API base here>',
        'api_type': 'azure',
        'api_version': '2024-02-01
        'tags': ['tool', 'gpt-3.5-turbo'],
    },
    {
        'model': 'gpt-3.5-turbo-16k',
        'api_key': '<your Azure OpenAI API key here>',
        'base_url': '<your Azure OpenAI API base here>',
        'api_type': 'azure',
        'api_version': '2024-02-01
        'tags': ['tool', 'gpt-3.5-turbo-16k'],
    },
]

llm_config = {
    "config_list": config_list,
    "timeout": 120,
}
chatbot = autogen.AssistantAgent(
    name="chatbot",
    system_message="For coding tasks, only use the functions you have been provided with. Reply TERMINATE when the task is done.",
    llm_config=llm_config,
)

# create a UserProxyAgent instance named "user_proxy"
user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    is_termination_msg=lambda x: x.get("content", "") and x.get("content", "").rstrip().endswith("TERMINATE"),
    human_input_mode="NEVER",
    max_consecutive_auto_reply=10,
    code_execution_config={
        "work_dir": "coding",
        "use_docker": False,
    },  # Please set use_docker=True if docker is available to run the generated code. Using docker is safer than running the generated code directly.
)


# define functions according to the function description


# one way of registering functions is to use the register_for_llm and register_for_execution decorators
@user_proxy.register_for_execution()
@chatbot.register_for_llm(name="python", description="run cell in ipython and return the execution result.")
def exec_python(cell: Annotated[str, "Valid Python cell to execute."]) -> str:
    ipython = get_ipython()
    result = ipython.run_cell(cell)
    log = str(result.result)
    if result.error_before_exec is not None:
        log += f"\n{result.error_before_exec}"
    if result.error_in_exec is not None:
        log += f"\n{result.error_in_exec}"
    return log


# another way of registering functions is to use the register_function
def exec_sh(script: Annotated[str, "Valid Python cell to execute."]) -> str:
    return user_proxy.execute_code_blocks([("sh", script)])


autogen.agentchat.register_function(
    exec_python,
    caller=chatbot,
    executor=user_proxy,
    name="sh",
    description="run a shell script and return the execution result.",
)

with Cache.disk() as cache:
    # start the conversation
    user_proxy.initiate_chat(
        chatbot,
        message="Draw two agents chatting with each other with an example dialog. Don't add plt.show().",
        cache=cache,
        max_turns=3,
    )

user_proxy (to chatbot):

Draw two agents chatting with each other with an example dialog. Don't add plt.show().

--------------------------------------------------------------------------------
chatbot (to user_proxy):

***** Suggested tool call (call_BiLzujDvfB7WMZ0hqcgBdjN2): python *****
Arguments: 
{"cell":"import matplotlib.pyplot as plt\nimport numpy as np\n\n# Create a simple representation of two agents\nagent1_x, agent1_y = [1, 2], [1, 1]\nagent2_x, agent2_y = [4, 3], [1, 1]\n\n# Create dialog bubbles\nbubble1_x = np.linspace(1.5, 2.5, 100)\nbubble1_y = np.sin(np.pi * (bubble1_x - 1.5)) + 1.2\nbubble2_x = np.linspace(3.5, 2.5, 100)\nbubble2_y = np.sin(np.pi * (bubble2_x - 2.5)) + 1.2\n\n# Drawing agents and dialog bubbles\nplt.figure(figsize=(6, 3))\nplt.plot(agent1_x, agent1_y, 'ko-', markersize=20)\nplt.plot(agent2_x, agent2_y, 'ko-', markersize=20)\nplt.plot(bubble1_x, bubble1_y, 'k')\nplt.plot(bubble2_x, bubble2_y, 'k')\nplt.fill_between(bubble1_x, 1, bubble1_y, color = 'grey', alpha = 0.5)\nplt.fill_between(bubble2_x, 1, bubble2_y, color = 'grey', alpha = 0.5)\n\n# Example Dialog\nplt.text(1.5, 1.5, 'Hi!', fontsize=12)\nplt.text(3.5, 1.5, 'Hello!', fontsize=12)\n\nplt.xlim(0, 5)\nplt.ylim(0, 2.5)\nplt.axis('off')\n"}
***********************************************************************

--------------------------------------------------------------------------------

>>>>>>>> EXECUTING FUNCTION python...

user_proxy (to chatbot):

user_proxy (to chatbot):

***** Response from calling tool (call_BiLzujDvfB7WMZ0hqcgBdjN2) *****
(0.0, 5.0, 0.0, 2.5)
**********************************************************************

--------------------------------------------------------------------------------
chatbot (to user_proxy):

I've drawn two agents chatting with each other, including example dialogues saying "Hi!" and "Hello!" within their speech bubbles. This visualization is created using matplotlib, and the conversation is represented graphically with the agents and their dialogue bubbles placed on a 2D plot. The drawing does not include `plt.show()`, adhering to your instruction.

--------------------------------------------------------------------------------
user_proxy (to chatbot):



--------------------------------------------------------------------------------
chatbot (to user_proxy):

It seems your last message was empty. How can I assist you further?

--------------------------------------------------------------------------------