Spaces:
Build error
Build error
Update article_generator.py
Browse files- article_generator.py +59 -50
article_generator.py
CHANGED
|
@@ -58,6 +58,32 @@ def remove_duplicates(text_list):
|
|
| 58 |
return result
|
| 59 |
|
| 60 |
# 記事のセクションをGPT-4で拡張する関数
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
def expand_section_with_gpt4(h2_text, h3_texts, preloaded_data):
|
| 62 |
prompts = []
|
| 63 |
h3_to_text = {}
|
|
@@ -134,7 +160,15 @@ def generate_expanded_article(article_html, h3_to_text, cached_responses):
|
|
| 134 |
if h3.get_text() in h3_to_text:
|
| 135 |
new_paragraph = soup.new_tag('p')
|
| 136 |
new_paragraph.string = h3_to_text[h3.get_text()]
|
| 137 |
-
h3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
|
| 139 |
process_summary_section(soup, cached_responses) # まとめセクションを特別処理し、キャッシュされたレスポンスを渡す
|
| 140 |
|
|
@@ -162,7 +196,7 @@ def setup_plan_and_execute_agent():
|
|
| 162 |
# GPT-4を使用してテキストを生成するヘルパー関数
|
| 163 |
def generate_text_with_gpt4(prompt):
|
| 164 |
response = openai.ChatCompletion.create(
|
| 165 |
-
model="gpt-
|
| 166 |
messages=[{"role": "system", "content": "以下についての詳細な情報をまとめ、適宜箇所書き、もしくは表を使ってオリジナルの内容にしてください。"},
|
| 167 |
{"role": "user", "content": prompt}],
|
| 168 |
temperature=0.7,
|
|
@@ -201,30 +235,6 @@ def process_heading(agent, h2_text, h3_for_this_h2, cached_responses):
|
|
| 201 |
else:
|
| 202 |
return (query, "No cached response found for this heading.")
|
| 203 |
|
| 204 |
-
# IDとセクションを付与する関数
|
| 205 |
-
def assign_ids_and_group_sections(html_content):
|
| 206 |
-
soup = BeautifulSoup(html_content, 'html.parser')
|
| 207 |
-
h2_elements = soup.find_all('h2')
|
| 208 |
-
section_id = 1
|
| 209 |
-
|
| 210 |
-
for h2 in h2_elements:
|
| 211 |
-
section_div = soup.new_tag('div', **{'class': 'section', 'id': f'section-{section_id}'})
|
| 212 |
-
h2['id'] = f'h2-{section_id}'
|
| 213 |
-
section_div.append(h2.extract())
|
| 214 |
-
|
| 215 |
-
next_tag = h2.next_sibling
|
| 216 |
-
while next_tag and (next_tag.name != 'h2'):
|
| 217 |
-
if next_tag.name == 'h3':
|
| 218 |
-
h3_id = f'h3-{section_id}-{next_tag.get_text().split()[0]}'
|
| 219 |
-
next_tag['id'] = h3_id
|
| 220 |
-
section_div.append(next_tag.extract())
|
| 221 |
-
next_tag = next_tag.next_sibling
|
| 222 |
-
|
| 223 |
-
h2.insert_before(section_div)
|
| 224 |
-
section_id += 1
|
| 225 |
-
|
| 226 |
-
return str(soup)
|
| 227 |
-
|
| 228 |
# 記事を生成する関数
|
| 229 |
def generate_article(editable_output2):
|
| 230 |
print("Starting article generation...")
|
|
@@ -268,31 +278,32 @@ def generate_article(editable_output2):
|
|
| 268 |
research_summary = "\n".join([json.dumps(result) for result in research_results])
|
| 269 |
instructions = []
|
| 270 |
|
|
|
|
| 271 |
instructions.append(f"""
|
| 272 |
-
<h1>{h1_text}</h1>
|
| 273 |
-
|
| 274 |
|
| 275 |
sentences = research_summary.split('。')
|
| 276 |
-
|
| 277 |
-
# 質問の数を制限
|
| 278 |
max_questions_per_h3 = 2
|
| 279 |
|
| 280 |
for idx, h2_text in enumerate(h2_texts):
|
| 281 |
h3_for_this_h2 = [h3 for h3 in h3_texts if h3.startswith(f"{idx+1}-")]
|
| 282 |
instructions.append(f"""
|
| 283 |
-
|
| 284 |
-
"{
|
| 285 |
-
|
|
|
|
| 286 |
related_sentences = [sentence for sentence in sentences if h3 in sentence][:max_questions_per_h3]
|
| 287 |
if related_sentences:
|
| 288 |
content_for_h3 = "。".join(related_sentences) + "。"
|
| 289 |
instructions.append(f"""
|
| 290 |
-
<h3>{h3}</h3>
|
| 291 |
-
|
| 292 |
else:
|
| 293 |
instructions.append(f"""
|
| 294 |
-
<h3>{h3}</h3>
|
| 295 |
-
|
|
|
|
| 296 |
|
| 297 |
# トークン数を制限するためにメッセージを分割
|
| 298 |
split_instructions = []
|
|
@@ -335,22 +346,20 @@ def generate_article(editable_output2):
|
|
| 335 |
print("Final generated article content:") # 最終的な記事全体の内容を出力
|
| 336 |
print(final_result)
|
| 337 |
|
| 338 |
-
#
|
| 339 |
-
|
| 340 |
-
print("Grouped HTML content:") # グループ化されたHTML内容を表示
|
| 341 |
-
print(grouped_html)
|
| 342 |
|
| 343 |
-
#
|
| 344 |
-
|
|
|
|
|
|
|
| 345 |
|
| 346 |
-
#
|
| 347 |
-
|
| 348 |
-
expanded_article = generate_expanded_article("\n".join(final_result), h3_to_text, cached_responses)
|
| 349 |
|
|
|
|
| 350 |
with open("output3.txt", "w", encoding="utf-8") as f:
|
| 351 |
-
f.write(
|
| 352 |
|
| 353 |
print("Article generation complete. Output saved to output3.txt.")
|
| 354 |
-
|
| 355 |
-
|
| 356 |
-
return expanded_article
|
|
|
|
| 58 |
return result
|
| 59 |
|
| 60 |
# 記事のセクションをGPT-4で拡張する関数
|
| 61 |
+
def expand_h3_sections(soup, preloaded_data):
|
| 62 |
+
h3_elements = soup.find_all('h3')
|
| 63 |
+
for h3 in h3_elements:
|
| 64 |
+
h3_text = h3.get_text(strip=True)
|
| 65 |
+
section_id = h3['id']
|
| 66 |
+
key = f"{h3_text} {section_id}"
|
| 67 |
+
|
| 68 |
+
if key in preloaded_data:
|
| 69 |
+
context = preloaded_data[key]
|
| 70 |
+
prompt = f"「{h3_text}」について詳しく説明してください。こちらが背景情報です:\n{context}"
|
| 71 |
+
else:
|
| 72 |
+
prompt = f"「{h3_text}」について詳しく説明してください。"
|
| 73 |
+
|
| 74 |
+
expanded_text = generate_text_with_gpt4(prompt)
|
| 75 |
+
new_paragraph = soup.new_tag('p')
|
| 76 |
+
new_paragraph.string = expanded_text
|
| 77 |
+
|
| 78 |
+
# h3タグの次の要素を取得し、その後の要素を探す
|
| 79 |
+
next_sibling = h3.find_next_sibling()
|
| 80 |
+
if next_sibling:
|
| 81 |
+
next_sibling.insert_after(new_paragraph) # 次の要素が存在する場合のみ挿入を行う
|
| 82 |
+
else:
|
| 83 |
+
h3.parent.append(new_paragraph) # h3タグの親が存在する場合、親に直接追加
|
| 84 |
+
|
| 85 |
+
return soup
|
| 86 |
+
|
| 87 |
def expand_section_with_gpt4(h2_text, h3_texts, preloaded_data):
|
| 88 |
prompts = []
|
| 89 |
h3_to_text = {}
|
|
|
|
| 160 |
if h3.get_text() in h3_to_text:
|
| 161 |
new_paragraph = soup.new_tag('p')
|
| 162 |
new_paragraph.string = h3_to_text[h3.get_text()]
|
| 163 |
+
# h3タグの次の要素を取得し、その後に追加する
|
| 164 |
+
next_sibling = h3.find_next_sibling()
|
| 165 |
+
if next_sibling:
|
| 166 |
+
next_sibling.insert_after(new_paragraph)
|
| 167 |
+
else:
|
| 168 |
+
if h3.parent:
|
| 169 |
+
h3.insert_after(new_paragraph)
|
| 170 |
+
else:
|
| 171 |
+
print(f"Error: h3 element '{h3.get_text()}' has no parent.")
|
| 172 |
|
| 173 |
process_summary_section(soup, cached_responses) # まとめセクションを特別処理し、キャッシュされたレスポンスを渡す
|
| 174 |
|
|
|
|
| 196 |
# GPT-4を使用してテキストを生成するヘルパー関数
|
| 197 |
def generate_text_with_gpt4(prompt):
|
| 198 |
response = openai.ChatCompletion.create(
|
| 199 |
+
model="gpt-4-turbo",
|
| 200 |
messages=[{"role": "system", "content": "以下についての詳細な情報をまとめ、適宜箇所書き、もしくは表を使ってオリジナルの内容にしてください。"},
|
| 201 |
{"role": "user", "content": prompt}],
|
| 202 |
temperature=0.7,
|
|
|
|
| 235 |
else:
|
| 236 |
return (query, "No cached response found for this heading.")
|
| 237 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 238 |
# 記事を生成する関数
|
| 239 |
def generate_article(editable_output2):
|
| 240 |
print("Starting article generation...")
|
|
|
|
| 278 |
research_summary = "\n".join([json.dumps(result) for result in research_results])
|
| 279 |
instructions = []
|
| 280 |
|
| 281 |
+
# IDを含むHTMLプロンプトの作成
|
| 282 |
instructions.append(f"""
|
| 283 |
+
<h1 id="title">{h1_text}</h1>
|
| 284 |
+
<p>「{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。</p>""")
|
| 285 |
|
| 286 |
sentences = research_summary.split('。')
|
|
|
|
|
|
|
| 287 |
max_questions_per_h3 = 2
|
| 288 |
|
| 289 |
for idx, h2_text in enumerate(h2_texts):
|
| 290 |
h3_for_this_h2 = [h3 for h3 in h3_texts if h3.startswith(f"{idx+1}-")]
|
| 291 |
instructions.append(f"""
|
| 292 |
+
<div id="section-{idx+1}">
|
| 293 |
+
<h2 id="h2-{idx+1}">{h2_text}</h2>
|
| 294 |
+
<p>「{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)}。</p>""")
|
| 295 |
+
for h3_idx, h3 in enumerate(h3_for_this_h2):
|
| 296 |
related_sentences = [sentence for sentence in sentences if h3 in sentence][:max_questions_per_h3]
|
| 297 |
if related_sentences:
|
| 298 |
content_for_h3 = "。".join(related_sentences) + "。"
|
| 299 |
instructions.append(f"""
|
| 300 |
+
<h3 id="h3-{idx+1}-{h3_idx+1}">{h3}</h3>
|
| 301 |
+
<p>「{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}</p>""")
|
| 302 |
else:
|
| 303 |
instructions.append(f"""
|
| 304 |
+
<h3 id="h3-{idx+1}-{h3_idx+1}">{h3}</h3>
|
| 305 |
+
<p>「{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。</p>""")
|
| 306 |
+
instructions.append("</div>") # 各セクションの終わりにdivタグを閉じる
|
| 307 |
|
| 308 |
# トークン数を制限するためにメッセージを分割
|
| 309 |
split_instructions = []
|
|
|
|
| 346 |
print("Final generated article content:") # 最終的な記事全体の内容を出力
|
| 347 |
print(final_result)
|
| 348 |
|
| 349 |
+
# 更新されたHTMLの解析
|
| 350 |
+
updated_soup = BeautifulSoup(final_result, 'html.parser')
|
|
|
|
|
|
|
| 351 |
|
| 352 |
+
# 初期データをTavily検索で収集する関数
|
| 353 |
+
h3_texts = [h3.get_text(strip=True) for h3 in updated_soup.find_all('h3')]
|
| 354 |
+
cached_responses = perform_initial_tavily_search([], h3_texts)
|
| 355 |
+
save_preloaded_tavily_data(cached_responses)
|
| 356 |
|
| 357 |
+
# h3タグの拡張を行う
|
| 358 |
+
expanded_soup = expand_h3_sections(updated_soup, cached_responses)
|
|
|
|
| 359 |
|
| 360 |
+
final_html = str(expanded_soup)
|
| 361 |
with open("output3.txt", "w", encoding="utf-8") as f:
|
| 362 |
+
f.write(final_html)
|
| 363 |
|
| 364 |
print("Article generation complete. Output saved to output3.txt.")
|
| 365 |
+
return final_html
|
|
|
|
|
|