OpenAI Assistants in AG2

Two-agent chat with OpenAI assistants.

This notebook shows a very basic example of the GPTAssistantAgent, which is an experimental AG2 agent class that leverages the OpenAI Assistant API for conversational capabilities, working with UserProxyAgent in AG2.

import logging
import os

from autogen import AssistantAgent, UserProxyAgent, config_list_from_json
from autogen.agentchat.contrib.gpt_assistant_agent import GPTAssistantAgent

logger = logging.getLogger(__name__)
logger.setLevel(logging.WARNING)

assistant_id = os.environ.get("ASSISTANT_ID", None)

config_list = config_list_from_json("OAI_CONFIG_LIST")
llm_config = {"config_list": config_list}

assistant_config = {"assistant_id": assistant_id}

gpt_assistant = GPTAssistantAgent(
    name="assistant",
    instructions=AssistantAgent.DEFAULT_SYSTEM_MESSAGE,
    llm_config=llm_config,
    assistant_config=assistant_config,
)

user_proxy = UserProxyAgent(
    name="user_proxy",
    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.
    is_termination_msg=lambda msg: "TERMINATE" in msg["content"],
    human_input_mode="NEVER",
    max_consecutive_auto_reply=1,
)
user_proxy.initiate_chat(gpt_assistant, message="Print hello world")

user_proxy.initiate_chat(gpt_assistant, message="Write py code to eval 2 + 2", clear_history=True)

gpt_assistant.delete_assistant()

Edit this page