third.py

# -*- coding: utf-8 -*-

import os
import openai
import json
from langchain.chat_models import ChatOpenAI
from langchain.experimental.plan_and_execute import PlanAndExecute, load_agent_executor, load_chat_planner
from langchain.llms import OpenAI
from langchain.utilities import GoogleSearchAPIWrapper
from langchain.agents.tools import Tool
from bs4 import BeautifulSoup
import asyncio
from datetime import timedelta

# APIキーと検索エンジンIDの設定
openai.api_key = os.getenv("OPENAI_API_KEY")
GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
CUSTOM_SEARCH_ENGINE_ID = os.getenv("CUSTOM_SEARCH_ENGINE_ID")

# 実行された指示を追跡するリスト
executed_instructions = []

# 調査結果を保存するリスト
research_results = []

async def main(editable_output2, keyword_id):
    search = GoogleSearchAPIWrapper(google_cse_id=CUSTOM_SEARCH_ENGINE_ID, google_api_key=GOOGLE_API_KEY)

    tools = [
        Tool(
            name = "Search",
            func=search.run,
            description="useful for when you need to answer questions about current events"
        ),
    ]

    # 「Planner」,「Executor」, および Agentの定義
    model_name = "gpt-3.5-turbo-16k"
    llm = ChatOpenAI(model_name=model_name, temperature=0, max_tokens=1000)
    planner = load_chat_planner(llm)
    executor = load_agent_executor(llm, tools, verbose=True)

    # ここでexecutorの内容を出力して確認
    print("Executor:", executor)

    agent = PlanAndExecute(planner=planner, executor=executor, verbose=True, 
    suffix='Answer should be in Japanese.')

    # editable_output2 is the text of the article structure (h1, h2, h3...).
    soup = BeautifulSoup(editable_output2, 'html.parser')

    # Use the text of the h1 tag as the purpose.
    h1_text = soup.find('h1').get_text()

    # Use the text of the h2 tags as part of the purpose.
    h2_texts = [h2.get_text() for h2 in soup.find_all('h2')]

    # Use the text of the h3 tags as part of the purpose.
    h3_texts = [h3.get_text() for h3 in soup.find_all('h3')]

    # Generate the purpose.
    purpose = f"about {h1_text}, focusing particularly on {' and '.join(h2_texts)} and {' and '.join(h3_texts)}, to investigate the latest information and details"

    # Specify the type of information you want to research.
    if "人物" in h1_text or any("人物" in h2_text for h2_text in h2_texts) or any("人物" in h3_text for h3_text in h3_texts):
        # If the topic is about a person, specify that you want to research their name and career.
        purpose += " including the person's name and career"
    elif "商品" in h1_text or any("商品" in h2_text for h2_text in h2_texts) or any("商品" in h3_text for h3_text in h3_texts):
        # If the topic is about a product, specify that you want to research the brand name, product name, and price.
        purpose += " including the brand name, product name, and price"
    elif "イベント" in h1_text or any("イベント" in h2_text for h2_text in h2_texts) or any("イベント" in h3_text for h3_text in h3_texts):
        # If the topic is about an event, specify that you want to research the event's content, schedule, and venue.
        purpose += " including the event's content, schedule, and venue"

    # Convert the purpose into an instruction in the form of a question.
    instruction = f"Can you research {purpose} and include specific details in your response?"

    # Run the instruction with a clear expectation of the output format
    if instruction not in executed_instructions:
        raw_output = agent.run(instruction)
        executed_instructions.append(instruction)
        response_content = raw_output
        research_results.append(response_content)
    else:
        index = executed_instructions.index(instruction)
        response_content = research_results[index]

    # Prepare the system message
    system_message = {
        "role": "system",
        "content": "あなたはプロのライターです。"
    }

    # Prepare the user message
    research_summary = "\n".join(research_results)
    instructions = []

    # 記事タイトルに関する指示
    instructions.append(f'１. "{h1_text}"に関する導入文を作成してください。直接的なコピーまたは近いフレーズを避けて、オリジナルな内容にしてください。')

    # research_summaryを文単位に分割
    sentences = research_summary.split('。')

    # 各見出しと小見出しに関する指示
    for idx, h2_text in enumerate(h2_texts):
        h3_for_this_h2 = [h3 for h3 in h3_texts if h3.startswith(f"{idx+1}-")]
        instructions.append(f'{idx+2}. "{h2_text}"に関する導入文を作成してください。この導入文は、以下の小見出しの内容を考慮してください：{"、".join(h3_for_this_h2)}。直接的なコピーまたは近いフレーズを避けて、オリジナルな内容にしてください。')
        for h3 in h3_for_this_h2:
            # h3のテキストをキーワードとして使用し、関連する文を探す
            related_sentences = [sentence for sentence in sentences if h3 in sentence]
            
            # 関連する文を基に、<h3>の内容に関する記事本文を生成
            if related_sentences:
                content_for_h3 = "。".join(related_sentences) + "。"
                instructions.append(f'次に、"{h3}"に関する詳細な内容として、以下の情報を記述してください：{content_for_h3} ここでも、オリジナルな内容を心がけてください。')
            else:
                instructions.append(f'次に、"{h3}"に関する詳細な内容を記述してください。オリジナルな内容を心がけてください。')


    user_message = {
        "role": "user",
        "content": "\n".join(instructions)
    }

    # Generate a new text using the ChatCompletion API
    response = openai.ChatCompletion.create(
        model="gpt-3.5-turbo-16k",
        messages=[system_message, user_message],
        temperature=0.7,
    )
    result = response.choices[0]["message"]["content"]

    # Save the generated message to output3.txt
    with open('output3.txt', 'w', encoding='utf-8') as f:
        f.write(result)

    # Print the generated message
    print(result)