Spaces:
Sleeping
Sleeping
Update article_generator.py
Browse files- article_generator.py +62 -20
article_generator.py
CHANGED
|
@@ -26,43 +26,85 @@ class EnhancedTavilySearchTool:
|
|
| 26 |
else:
|
| 27 |
raise Exception("Failed to fetch data from Tavily API")
|
| 28 |
|
| 29 |
-
#
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
planner = load_chat_planner(llm)
|
| 35 |
executor = load_agent_executor(llm, tools, verbose=True)
|
| 36 |
-
agent = PlanAndExecute(planner=planner, executor=executor, verbose=True)
|
| 37 |
-
return agent
|
| 38 |
|
| 39 |
-
agent =
|
| 40 |
|
| 41 |
-
#
|
| 42 |
-
def generate_article(editable_output2):
|
| 43 |
soup = BeautifulSoup(editable_output2, 'html.parser')
|
| 44 |
h1_text = soup.find('h1').get_text()
|
| 45 |
h2_texts = [h2.get_text() for h2 in soup.find_all('h2')]
|
| 46 |
h3_texts = [h3.get_text() for h3 in soup.find_all('h3')]
|
| 47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
system_message = {
|
| 49 |
"role": "system",
|
| 50 |
"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。"
|
| 51 |
}
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
|
| 57 |
for idx, h2_text in enumerate(h2_texts):
|
| 58 |
h3_for_this_h2 = [h3 for h3 in h3_texts if h3.startswith(f"{idx+1}-")]
|
| 59 |
-
instructions.append(
|
| 60 |
-
|
| 61 |
-
|
| 62 |
for h3 in h3_for_this_h2:
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
|
| 67 |
user_message = {
|
| 68 |
"role": "user",
|
|
@@ -72,7 +114,7 @@ def generate_article(editable_output2):
|
|
| 72 |
response = openai.ChatCompletion.create(
|
| 73 |
model="gpt-4-0125-preview",
|
| 74 |
messages=[system_message, user_message],
|
| 75 |
-
temperature=0.7
|
| 76 |
)
|
| 77 |
result = response.choices[0]["message"]["content"]
|
| 78 |
|
|
|
|
| 26 |
else:
|
| 27 |
raise Exception("Failed to fetch data from Tavily API")
|
| 28 |
|
| 29 |
+
# 実行された指示を追跡するリスト
|
| 30 |
+
executed_instructions = []
|
| 31 |
+
# 調査結果を保存するリスト
|
| 32 |
+
research_results = []
|
| 33 |
+
|
| 34 |
+
async def main(editable_output2, keyword_id):
|
| 35 |
+
tavily_search_tool = Tool(
|
| 36 |
+
name="TavilySearch",
|
| 37 |
+
func=EnhancedTavilySearchTool().search,
|
| 38 |
+
description="Enhanced search tool using Tavily API"
|
| 39 |
+
)
|
| 40 |
+
|
| 41 |
+
tools = [tavily_search_tool]
|
| 42 |
+
|
| 43 |
+
# PlannerとExecutorの拡張定義
|
| 44 |
+
model_name = "gpt-3.5-turbo-1106"
|
| 45 |
+
llm = ChatOpenAI(model_name=model_name, temperature=0, max_tokens=1000)
|
| 46 |
planner = load_chat_planner(llm)
|
| 47 |
executor = load_agent_executor(llm, tools, verbose=True)
|
|
|
|
|
|
|
| 48 |
|
| 49 |
+
agent = PlanAndExecute(planner=planner, executor=executor, verbose=True)
|
| 50 |
|
| 51 |
+
# HTML解析
|
|
|
|
| 52 |
soup = BeautifulSoup(editable_output2, 'html.parser')
|
| 53 |
h1_text = soup.find('h1').get_text()
|
| 54 |
h2_texts = [h2.get_text() for h2 in soup.find_all('h2')]
|
| 55 |
h3_texts = [h3.get_text() for h3 in soup.find_all('h3')]
|
| 56 |
|
| 57 |
+
purpose = f"about {h1_text}, focusing particularly on {' and '.join(h2_texts)} and {' and '.join(h3_texts)}, to investigate the latest information and details"
|
| 58 |
+
|
| 59 |
+
# 特定情報の指定
|
| 60 |
+
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):
|
| 61 |
+
purpose += " including the person's name and career"
|
| 62 |
+
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):
|
| 63 |
+
purpose += " including the brand name, product name, and price"
|
| 64 |
+
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):
|
| 65 |
+
purpose += " including the event's content, schedule, and venue"
|
| 66 |
+
|
| 67 |
+
instruction = f"Can you research {purpose} and include specific details in your response? Please provide the information in Japanese."
|
| 68 |
+
|
| 69 |
+
if instruction not in executed_instructions:
|
| 70 |
+
raw_output = agent.run(instruction)
|
| 71 |
+
executed_instructions.append(instruction)
|
| 72 |
+
response_content = raw_output
|
| 73 |
+
research_results.append(response_content)
|
| 74 |
+
else:
|
| 75 |
+
index = executed_instructions.index(instruction)
|
| 76 |
+
response_content = research_results[index]
|
| 77 |
+
|
| 78 |
system_message = {
|
| 79 |
"role": "system",
|
| 80 |
"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。"
|
| 81 |
}
|
| 82 |
|
| 83 |
+
research_summary = "\n".join(research_results)
|
| 84 |
+
instructions = []
|
| 85 |
+
|
| 86 |
+
instructions.append(f"""
|
| 87 |
+
<h1>{h1_text}</h1>
|
| 88 |
+
"{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。""")
|
| 89 |
+
|
| 90 |
+
sentences = research_summary.split('。')
|
| 91 |
|
| 92 |
for idx, h2_text in enumerate(h2_texts):
|
| 93 |
h3_for_this_h2 = [h3 for h3 in h3_texts if h3.startswith(f"{idx+1}-")]
|
| 94 |
+
instructions.append(f"""
|
| 95 |
+
<h2>{h2_text}</h2>
|
| 96 |
+
"{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。""")
|
| 97 |
for h3 in h3_for_this_h2:
|
| 98 |
+
related_sentences = [sentence for sentence in sentences if h3 in sentence]
|
| 99 |
+
if related_sentences:
|
| 100 |
+
content_for_h3 = "。".join(related_sentences) + "。"
|
| 101 |
+
instructions.append(f"""
|
| 102 |
+
<h3>{h3}</h3>
|
| 103 |
+
"{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。""")
|
| 104 |
+
else:
|
| 105 |
+
instructions.append(f"""
|
| 106 |
+
<h3>{h3}</h3>
|
| 107 |
+
"{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。""")
|
| 108 |
|
| 109 |
user_message = {
|
| 110 |
"role": "user",
|
|
|
|
| 114 |
response = openai.ChatCompletion.create(
|
| 115 |
model="gpt-4-0125-preview",
|
| 116 |
messages=[system_message, user_message],
|
| 117 |
+
temperature=0.7,
|
| 118 |
)
|
| 119 |
result = response.choices[0]["message"]["content"]
|
| 120 |
|