app.py

import gradio as gr
import random
import time
import csv
import os
import pandas as pd
from openai import OpenAI
import numpy as np
import ast
import json


# ユーザーの諸々の情報やプロンプトを受け取り返答する
def search_teacher(job_type, job_start_dates, tool_function_arguments):
    user_request = ', '.join(list(tool_function_arguments.values())[:-2])
    user_message_embedded_vector = \
        client.embeddings.create(input=[user_request.replace("\n", " ")], model='text-embedding-3-small').data[
            0].embedding

    data = pd.read_csv(os.environ['DATA_PATH'])

    # job_type の日本語を英語に変換
    if job_type == '常勤':
        job_type = 'full time'
    elif job_type == '非常勤':
        job_type = 'part time'

    # job_start_dates の日本語を英語に変換
    job_start_date_translation = {
        '今年度': 'This Year',
        '来年度': 'Next Year',
        '来来年度': 'Year After Next'
    }
    # job_start_dates が文字列の場合リストに変換(単一選択の場合を考慮)
    if isinstance(job_start_dates, str):
        job_start_dates = [job_start_dates]
    # 英語に変換
    job_start_dates = [job_start_date_translation.get(date, date) for date in job_start_dates]

    # job_type と job_start_dates でフィルタリング
    filtered_data = data[(data['job_type'] == job_type) & (data['job_start_date'].isin(job_start_dates))].copy()

    result = []

    for index, row in filtered_data.iterrows():
        teacher_embedded_vector = ast.literal_eval(row['embedding'])

        similarity = np.dot(user_message_embedded_vector, teacher_embedded_vector) / (
                np.linalg.norm(user_message_embedded_vector) * np.linalg.norm(teacher_embedded_vector))

        if len(result) < 3:
            result.append((index, similarity))
            result.sort(key=lambda x: x[1], reverse=True)
        else:
            if similarity > result[-1][1]:
                result[-1] = (index, similarity)
                result.sort(key=lambda x: x[1], reverse=True)

    formatted_result = []
    for index, similarity in result:
        name = filtered_data.loc[index, 'name']
        temp = f"{name}: {similarity}"
        formatted_result.append(temp)

    return ', '.join(formatted_result)


def openai_api(job_type, job_start_dates, history):
    # GPTにユーザーの入力を送信
    message = client.beta.threads.messages.create(
        thread_id=thread.id,
        role="user",
        content=history[-1][0]
    )

    # 送信した入力を実行
    run = client.beta.threads.runs.create(
        thread_id=thread.id,
        assistant_id=assistant.id,
    )

    while True:
        run = client.beta.threads.runs.retrieve(
            thread_id=thread.id,
            run_id=run.id
        )
        if run.status == 'completed':
            break
        elif run.status == 'requires_action':
            tool_id = run.required_action.submit_tool_outputs.tool_calls[0].id
            tool_function_arguments = json.loads(
                run.required_action.submit_tool_outputs.tool_calls[0].function.arguments)

            tool_function_output = search_teacher(job_type, job_start_dates, tool_function_arguments)

            run = client.beta.threads.runs.submit_tool_outputs(
                thread_id=thread.id,
                run_id=run.id,
                tool_outputs=[
                    {
                        "tool_call_id": tool_id,
                        "output": tool_function_output,
                    }
                ]
            )
            time.sleep(3)

        time.sleep(0.5)

    messages = client.beta.threads.messages.list(
        thread_id=thread.id,
        order="asc"
    )

    return messages.data[-1].content[0].text.value


def user(user_job_type, user_job_start_date, user_message, history):
    return None, history + [[user_message, None]]


def bot(job_type, job_start_date, history):
    prompt = ""
    # apiを叩くためにデータを加工するなりする
    for chat in history[:-1]:
        prompt += '"' + chat[0] + '", "' + chat[1] + '"'

    bot_message = openai_api(job_type, job_start_date, history)

    history[-1][1] = ""
    for character in bot_message:
        history[-1][1] += character
        time.sleep(0.01)

        yield history


with gr.Blocks(theme=gr.themes.Soft()) as demo:
    global assistant
    global thread
    global client

    client = OpenAI(max_retries=5)

    assistant = client.beta.assistants.create(
        name='connpath_demo',
        instructions=os.environ['INSTRUCTIONS'],
        model="gpt-4-0125-preview",
        tools=[{
            "type": "function",
            "function": {
                "name": "connpath_demo_gpt",
                "description": "検索を支援する",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "educational_goals": {"type": "string", "description": "教育の目的"},
                        "student_profile": {"type": "string", "description": "対象のプロフィール"},
                        "required_skills_experience": {"type": "string", "description": "求められるスキル"},
                        "teaching_method_environment": {"type": "string", "description": "教育する環境"},
                        "evaluation_feedback": {"type": "string", "description": "教育の評価方法"}
                    },
                    "required": ["educational_goals", "student_profile", "required_skills_experience",
                                 "teaching_method_environment", "evaluation_feedback"]
                }
            }
        }]
    )

    thread = client.beta.threads.create()

    # これがuser_job_typeを保持 (str型)
    job_type = gr.Radio(["常勤", "非常勤"],
                        label="Job Type",
                        info="探している雇用形態について")

    # これがuser_job_start_dateを保持 (list型)
    job_start_date = gr.CheckboxGroup(["今年度", "来年度", "来来年度"],
                                      label="Start Date",
                                      info="探している就業時期について")

    # これがuser_messageを保持
    msg = gr.Textbox(label="input message",
                     info="課題について教えてください(例:国語の先生を探しています)")

    # これがhistoryを保持
    chatbot = gr.Chatbot(show_copy_button=True)

    clear = gr.Button("clear history")

    msg.submit(user,
               [job_type, job_start_date, msg, chatbot],
               [msg, chatbot],
               queue=True
               ).then(bot, [job_type, job_start_date, chatbot], chatbot)
    clear.click(lambda: None, None, chatbot, queue=False)

if __name__ == "__main__":
    demo.queue()
    demo.launch(auth=(os.environ['USER_NAME'], os.environ['PASSWORD']))