03_Code_Intepreter_Analysis.py

from dotenv import load_dotenv
load_dotenv()
from openai import OpenAI
import os
import requests
import time
import json

client = OpenAI()

# V1
# def create_assistant(instructions, file_path):
#     with open(file_path, "rb") as file:
#         file_obj = client.files.create(file=file, purpose='assistants')
#         file_id = file_obj.id

#     assistant = client.beta.assistants.create(
#         instructions=instructions,
#         model="gpt-4",
#         tools=[{"type": "code_interpreter"}],
#         file_ids=[file_id]
#     )
#     return assistant

# V2
def create_assistant(instructions, file_id):
    assistant = client.beta.assistants.create(
        instructions=instructions,
        model="gpt-4-turbo",
        tools=[{"type": "code_interpreter"}],
        tool_resources={
            "code_interpreter": {
                "file_ids": [file_id]
            }
        }
    )
    return assistant

# def create_thread(user_message, file_id):
#     thread = client.beta.threads.create(
#         messages=[
#             {
#                 "role": "user",
#                 "content": user_message,
#                 "file_ids": [file_id]
#             }
#         ]
#     )
#     return thread

def create_thread(user_message, file_id):
    thread = client.beta.threads.create(
        messages=[
            {
                "role": "user",
                "content": user_message,
                "attachments": [
                    {
                        "file_id": file_id,
                        "tools": [{"type": "code_interpreter"}]
                    }
                ]
            }
        ]
    )
    return thread

def run_assistant(thread_id, assistant_id):
    run = client.beta.threads.runs.create(
        thread_id=thread_id,
        assistant_id=assistant_id
    )
    return run

def poll_run_status(client, thread_id, run_id, interval=20):
    while True:
        run = client.beta.threads.runs.retrieve(thread_id=thread_id, run_id=run_id)
        print(f"Run的轮询信息:\n{run}\n")
        if run.status in ['requires_action', 'completed']:
            return run
        time.sleep(interval)

def get_assistant_reply(thread_id):
    response = client.beta.threads.messages.list(thread_id=thread_id)
    message = response.data[-1]
    message_content = message.content[0].text
    annotations = message_content.annotations
    citations = []

    for index, annotation in enumerate(annotations):
        message_content.value = message_content.value.replace(annotation.text, f' [{index}]')

        if (file_citation := getattr(annotation, 'file_citation', None)):
            cited_file = client.files.retrieve(file_citation.file_id)
            citations.append(f'[{index}] {file_citation.quote} from {cited_file.filename}')
        elif (file_path := getattr(annotation, 'file_path', None)):
            cited_file = client.files.retrieve(file_path.file_id)
            citations.append(f'[{index}] Click <here> to download {cited_file.filename}')
            download_file(cited_file.filename, file_path.file_id)

    message_content.value += '\n' + '\n'.join(citations)
    return message_content.value

def download_file(filename, file_id):
    file_content = client.files.content(file_id)
    with open(filename, 'wb') as file:
        file.write(file_content)

def main():
    instructions = """
    Please use the flower_sales.csv data to complete the following analysis tasks:
    1. Group the data by region and calculate the total revenue for each region. Visualize the results using a bar chart.
    2. Group the data by date and calculate the daily total revenue. Create a line chart to show the revenue trend over time.
    3. Calculate the sales proportion of each flower type and display the results in a pie chart.
    4. Find the top 5 most profitable flowers based on the total profit.
    5. Using the historical sales data, forecast the total revenue for the next 7 days using a Random Forest Regressor model.
    """

    file_path = "80_LLM实战课代码/03_Asst_Code_Intepreter/flower_sales.csv"

    with open(file_path, "rb") as file:
        file_obj = client.files.create(file=file, purpose='assistants')
        file_id = file_obj.id

    assistant = create_assistant(instructions, file_id)

    user_message = "Please perform the data analysis tasks as instructed."
    thread = create_thread(user_message, file_id)

    run = run_assistant(thread.id, assistant.id)
    print(f"Run的初始信息: {run.status}")

    # 轮询Run直到完成或需要操作
    run = poll_run_status(client, thread.id, run.id)

    # 获取Assistant在Thread中的回应
    messages = client.beta.threads.messages.list(thread_id=thread.id)
    print("全部的message", messages)

    # 输出Assistant的最终回应
    print('下面打印最终的Assistant回应:')
    for message in messages.data:
        if message.role == "assistant":
            print(f"{message.content}\n")

if __name__ == "__main__":
    main()