article_generator.py

import os
import openai
import json
import requests
from bs4 import BeautifulSoup
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.agents.tools import Tool

# APIキーの設定
openai.api_key = os.getenv("OPENAI_API_KEY")
tavily_api_key = os.getenv('TAVILY_API_KEY')

# Tavily APIのカスタムツールを定義
class EnhancedTavilySearchTool:
    def search(self, query):
        if len(query) < 5:
            query += " details"  # クエリを拡張して必要な文字数にする
        
        params = {
            'api_key': tavily_api_key,
            'query': query,
            'max_results': 5,
            'detail_level': 'high'
        }
        response = requests.post('https://api.tavily.com/search', json=params)
        if response.status_code == 200:
            return response.json()['results']
        else:
            raise Exception(f"Failed to fetch data from Tavily API: {response.status_code}, {response.text}")

# 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 = []

# 生成状態を保存するファイル
state_file = "state.json"

def save_state(state):
    with open(state_file, "w", encoding="utf-8") as f:
        json.dump(state, f, ensure_ascii=False, indent=4)

def load_state():
    if os.path.exists(state_file):
        with open(state_file, "r", encoding="utf-8") as f:
            return json.load(f)
    return None

def clear_state():
    if os.path.exists(state_file):
        os.remove(state_file)
    global executed_instructions, research_results
    executed_instructions = []
    research_results = []
    return "状態がクリアされました"

def generate_article(editable_output2):
    # 途中から再開する場合のために状態を読み込み
    state = load_state() or {'executed_instructions': [], 'research_results': [], 'current_index': 0}
    executed_instructions = state['executed_instructions']
    research_results = state['research_results']
    current_index = state['current_index']

    tavily_search_tool = Tool(
        name="TavilySearch",
        func=EnhancedTavilySearchTool().search,
        description="Enhanced search tool using Tavily API"
    )

    tools = [tavily_search_tool]

    # PlannerとExecutorの拡張定義
    model_name = "gpt-3.5-turbo-0125"
    llm = ChatOpenAI(model_name=model_name, temperature=0, max_tokens=1000)
    planner = load_chat_planner(llm)
    executor = load_agent_executor(llm, tools, verbose=True)

    agent = PlanAndExecute(planner=planner, executor=executor, verbose=True)

    # HTML解析
    soup = BeautifulSoup(editable_output2, 'html.parser')
    h1_text = soup.find('h1').get_text()
    h2_texts = [h2.get_text() for h2 in soup.find_all('h2')]
    h3_texts = [h3.get_text() for h3 in soup.find_all('h3')]

    for h2_text in h2_texts:
        h3_for_this_h2 = [h3 for h3 in h3_texts if h3.startswith(f"{h2_texts.index(h2_text)+1}-")]

        # 特定情報の指定
        purpose = f"Research details about {h2_text} including {' and '.join(h3_for_this_h2)}"
        if "人物" in h1_text or any("人物" in h2 for h2 in h2_texts) or any("人物" in h3 for h3 in h3_texts):
            purpose += " including the person's name and career"
        elif "商品" in h1_text or any("商品" in h2 for h2 in h2_texts) or any("商品" in h3 for h3 in h3_texts):
            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 for h2 in h2_texts) or any("イベント" in h3 for h3 in h3_texts):
            purpose += " including the event's content, schedule, and venue"

        if purpose not in executed_instructions:
            try:
                response = agent.run(purpose)
                executed_instructions.append(purpose)
                research_results.append(response)
                save_state({'executed_instructions': executed_instructions, 'research_results': research_results, 'current_index': h2_texts.index(h2_text) + 1})
            except Exception as e:
                print(f"Error occurred: {str(e)}")
                # エラー時に状態を保存して中断
                save_state({'executed_instructions': executed_instructions, 'research_results': research_results, 'current_index': h2_texts.index(h2_text)})
                return str(e)
        else:
            index = executed_instructions.index(purpose)
            response = research_results[index]

    system_message = {
        "role": "system",
        "content": "あなたはプロのライターです。すべての回答を日本語でお願いします。"
    }

    research_summary = "\n".join(research_results)
    instructions = []

    instructions.append(f"""
    <h1>{h1_text}</h1> 
    "{h1_text}"に関する導入文を日本語で作成してください。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。""")

    sentences = research_summary.split('。')

    # 質問の数を制限
    max_questions_per_h3 = 2

    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"""
            <h2>{h2_text}</h2>
            "{h2_text}"に関する導入文を日本語で作成してください。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:
            related_sentences = [sentence for sentence in sentences if h3 in sentence][:max_questions_per_h3]
            if related_sentences:
                content_for_h3 = "。".join(related_sentences) + "。"
                instructions.append(f"""
                    <h3>{h3}</h3>
                    "{h3}"に関する詳細な内容として、以下の情報を日本語で記述してください：{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"""
                    <h3>{h3}</h3>
                    "{h3}"に関する詳細な内容を日本語で記述してください。オリジナルな内容を心がけてください。""")

    # トークン数を制限するためにメッセージを分割
    split_instructions = []
    current_chunk = ""
    max_tokens_per_chunk = 8000  # トークン数の上限を設定

    for instruction in instructions:
        if len(current_chunk + instruction) > max_tokens_per_chunk:
            split_instructions.append(current_chunk)
            current_chunk = instruction
        else:
            current_chunk += instruction

    if current_chunk:
        split_instructions.append(current_chunk)

    results = []
    for i, split_instruction in enumerate(split_instructions):
        user_message = {
            "role": "user",
            "content": f"{i+1}/{len(split_instructions)}: {split_instruction}"
        }
        try:
            response = openai.ChatCompletion.create(
                model="gpt-4o",
                messages=[system_message, user_message],
                temperature=0.7,
            )
            results.append(response.choices[0]["message"]["content"])
        except Exception as e:
            error_message = f"Error occurred during ChatCompletion: {str(e)}"
            print(error_message)  # ログにエラーメッセージを出力
            results.append(error_message)
            # 途中で止まった場合の状態を保存
            save_state({
                "executed_instructions": executed_instructions,
                "research_results": research_results,
                "split_instructions": split_instructions,
                "results": results,
                "current_index": i + 1
            })
            return error_message

    final_result = "\n".join(results)

    with open("output3.txt", "w", encoding="utf-8") as f:
        f.write(final_result)

    print(final_result)  # ログに最終結果を出力

    # 生成が完了したら状態ファイルを削除
    if os.path.exists(state_file):
        os.remove(state_file)

    return final_result

def continue_generate_article():
    state = load_state()
    if not state:
        return "再開する状態がありません。"

    executed_instructions = state.get("executed_instructions", [])
    research_results = state.get("research_results", [])
    split_instructions = state.get("split_instructions", [])
    results = state.get("results", [])
    current_index = state.get("current_index", 0)

    system_message = {
        "role": "system",
        "content": "あなたはプロのライターです。すべての回答を日本語でお願いします。"
    }

    for i in range(current_index, len(split_instructions)):
        user_message = {
            "role": "user",
            "content": f"{i+1}/{len(split_instructions)}: {split_instructions[i]}"
        }
        try:
            response = openai.ChatCompletion.create(
                model="gpt-4-turbo",
                messages=[system_message, user_message],
                temperature=0.7,
            )
            results.append(response.choices[0]["message"]["content"])
        except Exception as e:
            error_message = f"Error occurred during ChatCompletion: {str(e)}"
            print(error_message)  # ログにエラーメッセージを出力
            results.append(error_message)
            # 途中で止まった場合の状態を保存
            save_state({
                "executed_instructions": executed_instructions,
                "research_results": research_results,
                "split_instructions": split_instructions,
                "results": results,
                "current_index": i + 1
            })
            return error_message

    final_result = "\n".join(results)

    with open("output3.txt", "w", encoding="utf-8") as f:
        f.write(final_result)

    print(final_result)  # ログに最終結果を出力

    # 生成が完了したら状態ファイルを削除
    if os.path.exists(state_file):
        os.remove(state_file)

    return final_result