Auto Generated Agent Chat: GPTAssistant with Code Interpreter

This Jupyter Notebook showcases the integration of the Code Interpreter tool which executes Python code dynamically within applications.

The latest released Assistants API by OpenAI allows users to build AI assistants within their own applications. The Assistants API currently supports three types of tools: Code Interpreter, Retrieval, and Function calling. In this notebook, we demonstrate how to enable GPTAssistantAgent to use code interpreter.

Requirements

AG2 requires Python>=3.9. To run this notebook example, please install:

Install autogen:

pip install autogen

For more information, please refer to the installation guide.

Set your API Endpoint

The config_list_from_json function loads a list of configurations from an environment variable or a json file.

import io

from IPython.display import display
from PIL import Image

import autogen
from autogen.agentchat import UserProxyAgent
from autogen.agentchat.contrib.gpt_assistant_agent import GPTAssistantAgent

config_list = autogen.config_list_from_json(
    "OAI_CONFIG_LIST",
    file_location=".",
    filter_dict={
        "model": ["gpt-3.5-turbo", "gpt-35-turbo", "gpt-4", "gpt4", "gpt-4-32k", "gpt-4-turbo"],
    },
)

Learn more about configuring LLMs for agents here.

Perform Tasks Using Code Interpreter

We demonstrate task solving using GPTAssistantAgent with code interpreter. Pass code_interpreter in tools parameter to enable GPTAssistantAgent with code interpreter. It will write code and automatically execute it in a sandbox. The agent will receive the results from the sandbox environment and act accordingly.

Example 1: Math Problem Solving

In this example, we demonstrate how to use code interpreter to solve math problems.

# Initiate an agent equipped with code interpreter
gpt_assistant = GPTAssistantAgent(
    name="CoderAssistant",
    llm_config={
        "config_list": config_list,
    },
    assistant_config={
        "tools": [{"type": "code_interpreter"}],
    },
    instructions="You are an expert at solving math questions. Write code and run it to solve math problems. Reply TERMINATE when the task is solved and there is no problem.",
)

user_proxy = UserProxyAgent(
    name="user_proxy",
    is_termination_msg=lambda msg: "TERMINATE" in msg["content"],
    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.
    },
    human_input_mode="NEVER",
)

# When all is set, initiate the chat.
user_proxy.initiate_chat(
    gpt_assistant, message="If $725x + 727y = 1500$ and $729x+ 731y = 1508$, what is the value of $x - y$ ?"
)
gpt_assistant.delete_assistant()

Example 2: Plotting with Code Interpreter

Code Interpreter can outputs files, such as generating image diagrams. In this example, we demonstrate how to draw figures and download it.

gpt_assistant = GPTAssistantAgent(
    name="CoderAssistant",
    llm_config={
        "config_list": config_list,
    },
    assistant_config={
        "tools": [{"type": "code_interpreter"}],
    },
    instructions="You are an expert at writing python code to solve problems. Reply TERMINATE when the task is solved and there is no problem.",
)

user_proxy.initiate_chat(
    gpt_assistant,
    message="Draw a line chart to show the population trend in US. Show how you solved it with code.",
    is_termination_msg=lambda msg: "TERMINATE" in msg["content"],
    human_input_mode="NEVER",
    clear_history=True,
    max_consecutive_auto_reply=1,
)

Now we have the file id. We can download and display it.

api_response = gpt_assistant.openai_client.files.with_raw_response.retrieve_content(
    "assistant-tvLtfOn6uAJ9kxmnxgK2OXID"
)

if api_response.status_code == 200:
    content = api_response.content
    image_data_bytes = io.BytesIO(content)
    image = Image.open(image_data_bytes)
    display(image)

gpt_assistant.delete_assistant()

Edit this page

Assistant

Responses are generated using AI and may contain mistakes.