Spaces:
Sleeping
Sleeping
Update src/streamlit_app.py
Browse files- src/streamlit_app.py +189 -209
src/streamlit_app.py
CHANGED
@@ -11,28 +11,13 @@ from collections import Counter
|
|
11 |
import json
|
12 |
import os
|
13 |
from datetime import datetime, timedelta
|
14 |
-
|
15 |
from dotenv import load_dotenv
|
16 |
import traceback
|
17 |
import plotly.graph_objects as go
|
18 |
import schedule
|
19 |
import threading
|
20 |
import matplotlib.pyplot as plt
|
21 |
-
import kss # KoNLPy ๋์ KSS ์ฌ์ฉ
|
22 |
-
from PIL import Image
|
23 |
-
import base64
|
24 |
-
from io import BytesIO
|
25 |
-
import logging
|
26 |
-
|
27 |
-
# ๋ก๊น
์ค์
|
28 |
-
logging.basicConfig(
|
29 |
-
level=logging.INFO,
|
30 |
-
format='%(asctime)s - %(levelname)s - %(message)s',
|
31 |
-
handlers=[
|
32 |
-
logging.StreamHandler(),
|
33 |
-
logging.FileHandler('/tmp/crawler.log')
|
34 |
-
]
|
35 |
-
)
|
36 |
|
37 |
# ์๋ํด๋ผ์ฐ๋ ์ถ๊ฐ
|
38 |
try:
|
@@ -55,52 +40,41 @@ class SchedulerState:
|
|
55 |
global_scheduler_state = SchedulerState()
|
56 |
|
57 |
# API ํค ๊ด๋ฆฌ๋ฅผ ์ํ ์ธ์
์ํ ์ด๊ธฐํ
|
58 |
-
if '
|
59 |
-
st.session_state.
|
60 |
|
61 |
# ์ฌ๋ฌ ๋ฐฉ๋ฒ์ผ๋ก API ํค ๋ก๋ ์๋
|
62 |
load_dotenv() # .env ํ์ผ์์ ๋ก๋ ์๋
|
63 |
|
64 |
-
# OpenAI ํด๋ผ์ด์ธํธ ์ด๊ธฐํ๋ฅผ ์ํ ํจ์
|
65 |
-
def init_openai_client(api_key=None):
|
66 |
-
try:
|
67 |
-
if api_key:
|
68 |
-
client = OpenAI(api_key=api_key)
|
69 |
-
# ๊ฐ๋จํ API ํค ์ ํจ์ฑ ๊ฒ์ฌ
|
70 |
-
client.models.list() # API ํค๊ฐ ์ ํจํ์ง ํ
์คํธ
|
71 |
-
return client
|
72 |
-
return None
|
73 |
-
except Exception as e:
|
74 |
-
st.error(f"API ํค ์ด๊ธฐํ ์ค๋ฅ: {str(e)}")
|
75 |
-
return None
|
76 |
-
|
77 |
# 1. ํ๊ฒฝ ๋ณ์์์ API ํค ํ์ธ
|
78 |
-
|
79 |
-
|
80 |
-
st.session_state.
|
81 |
|
82 |
-
# 2. Streamlit secrets์์ API ํค ํ์ธ
|
83 |
-
if not st.session_state.
|
84 |
try:
|
85 |
if 'OPENAI_API_KEY' in st.secrets:
|
86 |
-
st.session_state.
|
|
|
87 |
except Exception as e:
|
88 |
pass # secrets ํ์ผ์ด ์์ด๋ ์ค๋ฅ ๋ฐ์ํ์ง ์์
|
89 |
|
90 |
-
# NLTK ๋ฐ์ดํฐ ๊ฒฝ๋ก ์ค์
|
91 |
-
|
92 |
-
|
|
|
93 |
|
94 |
-
# ํ์ํ NLTK ๋ฐ์ดํฐ
|
95 |
try:
|
96 |
nltk.data.find('tokenizers/punkt')
|
97 |
except LookupError:
|
98 |
-
nltk.download('punkt', download_dir=
|
99 |
|
100 |
try:
|
101 |
nltk.data.find('corpora/stopwords')
|
102 |
except LookupError:
|
103 |
-
nltk.download('stopwords', download_dir=
|
104 |
|
105 |
# ํ์ด์ง ์ค์
|
106 |
st.set_page_config(page_title="๋ด์ค ๊ธฐ์ฌ ๋๊ตฌ", page_icon="๐ฐ", layout="wide")
|
@@ -116,12 +90,9 @@ with st.sidebar:
|
|
116 |
st.divider()
|
117 |
api_key = st.text_input("OpenAI API ํค ์
๋ ฅ", type="password")
|
118 |
if api_key:
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
st.success("API ํค๊ฐ ์ฑ๊ณต์ ์ผ๋ก ์ค์ ๋์์ต๋๋ค!")
|
123 |
-
else:
|
124 |
-
st.error("์ ํจํ์ง ์์ API ํค์
๋๋ค.")
|
125 |
|
126 |
# ์ ์ฅ๋ ๊ธฐ์ฌ๋ฅผ ๋ถ๋ฌ์ค๋ ํจ์
|
127 |
def load_saved_articles():
|
@@ -141,21 +112,16 @@ def crawl_naver_news(keyword, num_articles=5):
|
|
141 |
"""
|
142 |
๋ค์ด๋ฒ ๋ด์ค ๊ธฐ์ฌ๋ฅผ ์์งํ๋ ํจ์
|
143 |
"""
|
144 |
-
logging.info(f"ํฌ๋กค๋ง ์์: ํค์๋={keyword}, ๊ธฐ์ฌ ์={num_articles}")
|
145 |
url = f"https://search.naver.com/search.naver?where=news&query={keyword}"
|
146 |
results = []
|
147 |
|
148 |
try:
|
149 |
# ํ์ด์ง ์์ฒญ
|
150 |
-
logging.info(f"์์ฒญ URL: {url}")
|
151 |
response = requests.get(url)
|
152 |
-
logging.info(f"์๋ต ์ํ ์ฝ๋: {response.status_code}")
|
153 |
-
|
154 |
soup = BeautifulSoup(response.text, 'html.parser')
|
155 |
|
156 |
# ๋ด์ค ์์ดํ
์ฐพ๊ธฐ
|
157 |
news_items = soup.select('div.sds-comps-base-layout.sds-comps-full-layout')
|
158 |
-
logging.info(f"์ฐพ์ ๋ด์ค ์์ดํ
์: {len(news_items)}")
|
159 |
|
160 |
# ๊ฐ ๋ด์ค ์์ดํ
์์ ์ ๋ณด ์ถ์ถ
|
161 |
for i, item in enumerate(news_items):
|
@@ -190,68 +156,48 @@ def crawl_naver_news(keyword, num_articles=5):
|
|
190 |
'description': description,
|
191 |
'source': source,
|
192 |
'date': date,
|
193 |
-
'content': ""
|
194 |
})
|
195 |
|
196 |
-
logging.info(f"๊ธฐ์ฌ ์ถ์ถ ์ฑ๊ณต: {title}")
|
197 |
-
|
198 |
except Exception as e:
|
199 |
-
|
200 |
continue
|
201 |
|
202 |
except Exception as e:
|
203 |
-
|
204 |
|
205 |
-
logging.info(f"ํฌ๋กค๋ง ์๋ฃ: {len(results)}๊ฐ ๊ธฐ์ฌ ์์ง")
|
206 |
return results
|
207 |
|
208 |
# ๊ธฐ์ฌ ์๋ฌธ ๊ฐ์ ธ์ค๊ธฐ
|
209 |
def get_article_content(url):
|
210 |
-
logging.info(f"๊ธฐ์ฌ ์๋ฌธ ๊ฐ์ ธ์ค๊ธฐ ์์: {url}")
|
211 |
try:
|
212 |
response = requests.get(url, timeout=5)
|
213 |
-
logging.info(f"์๋ฌธ ์์ฒญ ์ํ ์ฝ๋: {response.status_code}")
|
214 |
-
|
215 |
soup = BeautifulSoup(response.text, 'html.parser')
|
216 |
|
217 |
# ๋ค์ด๋ฒ ๋ด์ค ๋ณธ๋ฌธ ์ฐพ๊ธฐ
|
218 |
content = soup.select_one('#dic_area')
|
219 |
if content:
|
220 |
text = content.text.strip()
|
221 |
-
text = re.sub(r'\s+', ' ', text)
|
222 |
-
logging.info("๋ค์ด๋ฒ ๋ด์ค ๋ณธ๋ฌธ ์ถ์ถ ์ฑ๊ณต")
|
223 |
return text
|
224 |
|
225 |
-
# ๋ค๋ฅธ ๋ด์ค ์ฌ์ดํธ ๋ณธ๋ฌธ ์ฐพ๊ธฐ
|
226 |
content = soup.select_one('.article_body, .article-body, .article-content, .news-content-inner')
|
227 |
if content:
|
228 |
text = content.text.strip()
|
229 |
text = re.sub(r'\s+', ' ', text)
|
230 |
-
logging.info("์ผ๋ฐ ๋ด์ค ๋ณธ๋ฌธ ์ถ์ถ ์ฑ๊ณต")
|
231 |
return text
|
232 |
|
233 |
-
logging.warning("๋ณธ๋ฌธ์ ์ฐพ์ ์ ์์")
|
234 |
return "๋ณธ๋ฌธ์ ๊ฐ์ ธ์ฌ ์ ์์ต๋๋ค."
|
235 |
except Exception as e:
|
236 |
-
logging.error(f"์๋ฌธ ๊ฐ์ ธ์ค๊ธฐ ์ค๋ฅ: {str(e)}", exc_info=True)
|
237 |
return f"์ค๋ฅ ๋ฐ์: {str(e)}"
|
238 |
|
239 |
-
# NLTK๋ฅผ ์ด์ฉํ ํค์๋ ๋ถ์
|
240 |
def analyze_keywords(text, top_n=10):
|
241 |
-
# ํ๊ตญ์ด ๋ถ์ฉ์ด ๋ชฉ๋ก
|
242 |
korean_stopwords = ['์ด', '๊ทธ', '์ ', '๊ฒ', '๋ฐ', '๋ฑ', '๋ฅผ', '์', '์', '์์', '์', '์ผ๋ก', '๋ก']
|
243 |
|
244 |
-
|
245 |
-
try:
|
246 |
-
sentences = kss.split_sentences(text)
|
247 |
-
tokens = []
|
248 |
-
for sentence in sentences:
|
249 |
-
# ๊ฐ๋จํ ํ ํฐํ (๊ณต๋ฐฑ ๊ธฐ์ค)
|
250 |
-
tokens.extend(sentence.split())
|
251 |
-
except:
|
252 |
-
# KSS ์คํจ์ ๊ธฐ๋ณธ ํ ํฐํ
|
253 |
-
tokens = text.split()
|
254 |
-
|
255 |
tokens = [word for word in tokens if word.isalnum() and len(word) > 1 and word not in korean_stopwords]
|
256 |
|
257 |
word_count = Counter(tokens)
|
@@ -379,55 +325,36 @@ def analyze_news_content(news_df):
|
|
379 |
results['top_keywords'] = []
|
380 |
return results
|
381 |
|
382 |
-
# OpenAI API๋ฅผ ์ด์ฉํ ์ ๊ธฐ์ฌ ์์ฑ (
|
383 |
def generate_article(original_content, prompt_text):
|
384 |
try:
|
385 |
-
if not st.session_state.
|
386 |
return "OpenAI API ํค๊ฐ ์ค์ ๋์ง ์์์ต๋๋ค."
|
387 |
|
388 |
-
response =
|
389 |
-
model="gpt-4",
|
390 |
messages=[
|
391 |
{"role": "system", "content": "๋น์ ์ ์ ๋ฌธ์ ์ธ ๋ด์ค ๊ธฐ์์
๋๋ค. ์ฃผ์ด์ง ๋ด์ฉ์ ๋ฐํ์ผ๋ก ์๋ก์ด ๊ธฐ์ฌ๋ฅผ ์์ฑํด์ฃผ์ธ์."},
|
392 |
{"role": "user", "content": f"๋ค์ ๋ด์ฉ์ ๋ฐํ์ผ๋ก {prompt_text}\n\n{original_content[:1000]}"}
|
393 |
],
|
394 |
max_tokens=2000
|
395 |
)
|
396 |
-
return response.choices[0].message
|
397 |
except Exception as e:
|
398 |
return f"๊ธฐ์ฌ ์์ฑ ์ค๋ฅ: {str(e)}"
|
399 |
|
400 |
-
# OpenAI API๋ฅผ ์ด์ฉํ ์ด๋ฏธ์ง ์์ฑ (
|
401 |
def generate_image(prompt):
|
402 |
try:
|
403 |
-
if not st.session_state.
|
404 |
return "OpenAI API ํค๊ฐ ์ค์ ๋์ง ์์์ต๋๋ค."
|
405 |
|
406 |
-
|
407 |
-
result = st.session_state.openai_client.images.generate(
|
408 |
-
model="gpt-image-1", # ์๋ก์ด ๋ชจ๋ธ๋ช
์ฌ์ฉ
|
409 |
prompt=prompt,
|
|
|
410 |
size="1024x1024"
|
411 |
)
|
412 |
-
|
413 |
-
# base64 ์ด๋ฏธ์ง ๋ฐ์ดํฐ๋ฅผ ๋์ฝ๋ฉ
|
414 |
-
image_base64 = result.data[0].b64_json
|
415 |
-
image_bytes = base64.b64decode(image_base64)
|
416 |
-
|
417 |
-
# BytesIO ๊ฐ์ฒด๋ก ๋ณํ
|
418 |
-
image = BytesIO(image_bytes)
|
419 |
-
|
420 |
-
# PIL Image๋ก ๋ณํํ์ฌ ํฌ๊ธฐ ์กฐ์ (์ ํ์ฌํญ)
|
421 |
-
pil_image = Image.open(image)
|
422 |
-
pil_image = pil_image.resize((800, 800), Image.LANCZOS) # ํฌ๊ธฐ ์กฐ์
|
423 |
-
|
424 |
-
# ๋ค์ BytesIO๋ก ๋ณํ
|
425 |
-
output = BytesIO()
|
426 |
-
pil_image.save(output, format="JPEG", quality=80, optimize=True)
|
427 |
-
output.seek(0)
|
428 |
-
|
429 |
-
return output
|
430 |
-
|
431 |
except Exception as e:
|
432 |
return f"์ด๋ฏธ์ง ์์ฑ ์ค๋ฅ: {str(e)}"
|
433 |
|
@@ -449,14 +376,11 @@ def run_scheduled_task():
|
|
449 |
traceback.print_exc()
|
450 |
|
451 |
def perform_news_task(task_type, keyword, num_articles, file_prefix):
|
452 |
-
logging.info(f"์ค์ผ์ค๋ฌ ์์
์์: {task_type}, ํค์๋={keyword}")
|
453 |
try:
|
454 |
articles = crawl_naver_news(keyword, num_articles)
|
455 |
-
logging.info(f"์์ง๋ ๊ธฐ์ฌ ์: {len(articles)}")
|
456 |
|
457 |
# ๊ธฐ์ฌ ๋ด์ฉ ๊ฐ์ ธ์ค๊ธฐ
|
458 |
-
for
|
459 |
-
logging.info(f"๊ธฐ์ฌ {i+1}/{len(articles)} ์๋ฌธ ๊ฐ์ ธ์ค๊ธฐ: {article['title']}")
|
460 |
article['content'] = get_article_content(article['link'])
|
461 |
time.sleep(0.5) # ์๋ฒ ๋ถํ ๋ฐฉ์ง
|
462 |
|
@@ -468,12 +392,10 @@ def perform_news_task(task_type, keyword, num_articles, file_prefix):
|
|
468 |
with open(filename, 'w', encoding='utf-8') as f:
|
469 |
json.dump(articles, f, ensure_ascii=False, indent=2)
|
470 |
|
471 |
-
logging.info(f"๊ฒฐ๊ณผ ์ ์ฅ ์๋ฃ: {filename}")
|
472 |
-
|
473 |
global_scheduler_state.last_run = datetime.now()
|
474 |
print(f"{datetime.now()} - {task_type} ๋ด์ค ๊ธฐ์ฌ ์์ง ์๋ฃ: {keyword}")
|
475 |
|
476 |
-
# ์ ์ญ ์ํ์ ์์ง ๊ฒฐ๊ณผ๋ฅผ ์ ์ฅ
|
477 |
result_item = {
|
478 |
'task_type': task_type,
|
479 |
'keyword': keyword,
|
@@ -484,7 +406,7 @@ def perform_news_task(task_type, keyword, num_articles, file_prefix):
|
|
484 |
global_scheduler_state.scheduled_results.append(result_item)
|
485 |
|
486 |
except Exception as e:
|
487 |
-
|
488 |
traceback.print_exc()
|
489 |
|
490 |
def start_scheduler(daily_tasks, interval_tasks):
|
@@ -641,25 +563,9 @@ elif menu == "๊ธฐ์ฌ ๋ถ์ํ๊ธฐ":
|
|
641 |
with keyword_tab1:
|
642 |
keywords = analyze_keywords(selected_article['content'])
|
643 |
|
644 |
-
#
|
645 |
df = pd.DataFrame(keywords, columns=['๋จ์ด', '๋น๋์'])
|
646 |
-
|
647 |
-
go.Bar(
|
648 |
-
x=df['๋จ์ด'],
|
649 |
-
y=df['๋น๋์'],
|
650 |
-
marker_color='rgb(55, 83, 109)'
|
651 |
-
)
|
652 |
-
])
|
653 |
-
|
654 |
-
fig.update_layout(
|
655 |
-
title='ํค์๋ ๋น๋ ๋ถ์',
|
656 |
-
xaxis_title='ํค์๋',
|
657 |
-
yaxis_title='๋น๋์',
|
658 |
-
height=500,
|
659 |
-
margin=dict(l=50, r=50, t=80, b=50)
|
660 |
-
)
|
661 |
-
|
662 |
-
st.plotly_chart(fig, use_container_width=True)
|
663 |
|
664 |
st.write("**์ฃผ์ ํค์๋:**")
|
665 |
for word, count in keywords:
|
@@ -689,14 +595,7 @@ elif menu == "๊ธฐ์ฌ ๋ถ์ํ๊ธฐ":
|
|
689 |
# ํ
์คํธ ํต๊ณ ๊ณ์ฐ
|
690 |
word_count = len(re.findall(r'\b\w+\b', content))
|
691 |
char_count = len(content)
|
692 |
-
|
693 |
-
# KSS๋ก ๋ฌธ์ฅ ๋ถ๋ฆฌ
|
694 |
-
sentences = kss.split_sentences(content)
|
695 |
-
sentence_count = len(sentences)
|
696 |
-
except:
|
697 |
-
# KSS ์คํจ์ ๊ธฐ๋ณธ ๋ฌธ์ฅ ๋ถ๋ฆฌ
|
698 |
-
sentence_count = len(re.split(r'[.!?]+', content))
|
699 |
-
|
700 |
avg_word_length = sum(len(word) for word in re.findall(r'\b\w+\b', content)) / word_count if word_count > 0 else 0
|
701 |
avg_sentence_length = word_count / sentence_count if sentence_count > 0 else 0
|
702 |
|
@@ -716,50 +615,136 @@ elif menu == "๊ธฐ์ฌ ๋ถ์ํ๊ธฐ":
|
|
716 |
with col2:
|
717 |
st.metric("ํ๊ท ๋ฌธ์ฅ ๊ธธ์ด", f"{avg_sentence_length:.1f}๋จ์ด")
|
718 |
|
719 |
-
# ํ
์คํธ ๋ณต์ก์ฑ ์ ์
|
720 |
complexity_score = min(10, (avg_sentence_length / 10) * 5 + (avg_word_length / 5) * 5)
|
721 |
st.progress(complexity_score / 10)
|
722 |
st.write(f"ํ
์คํธ ๋ณต์ก์ฑ ์ ์: {complexity_score:.1f}/10")
|
723 |
-
|
724 |
-
#
|
725 |
-
st.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
726 |
|
727 |
elif analysis_type == "๊ฐ์ ๋ถ์":
|
728 |
if st.button("๊ฐ์ ๋ถ์ํ๊ธฐ"):
|
729 |
-
if st.session_state.
|
730 |
with st.spinner("๊ธฐ์ฌ์ ๊ฐ์ ์ ๋ถ์ ์ค์
๋๋ค..."):
|
731 |
try:
|
732 |
-
|
733 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
734 |
messages=[
|
735 |
-
{"role": "system", "content": "
|
736 |
-
|
737 |
-
{
|
738 |
-
"sentiment": "๊ธ์ ์ /๋ถ์ ์ /์ค๋ฆฝ์ ",
|
739 |
-
"reason": "์ด์ ์ค๋ช
...",
|
740 |
-
"keywords": [
|
741 |
-
{"word": "ํค์๋1", "score": 8},
|
742 |
-
{"word": "ํค์๋2", "score": 7}
|
743 |
-
]
|
744 |
-
}"""},
|
745 |
-
{"role": "user", "content": f"๋ค์ ๋ด์ค ๊ธฐ์ฌ๋ฅผ ๋ถ์ํด ์ฃผ์ธ์:\n\n์ ๋ชฉ: {selected_article['title']}\n\n๋ด์ฉ: {selected_article['content'][:1500]}"}
|
746 |
],
|
747 |
-
max_tokens=800
|
748 |
-
response_format={ "type": "json_object" } # JSON ์๋ต ํ์ ๊ฐ์
|
749 |
)
|
750 |
|
751 |
-
#
|
752 |
-
|
753 |
-
logging.info(f"API ์๋ต: {content}")
|
754 |
-
|
755 |
-
# JSON ํ์ฑ
|
756 |
-
try:
|
757 |
-
analysis_result = json.loads(content)
|
758 |
-
except json.JSONDecodeError as e:
|
759 |
-
logging.error(f"JSON ํ์ฑ ์ค๋ฅ: {str(e)}")
|
760 |
-
logging.error(f"ํ์ฑ ์๋ํ ๋ด์ฉ: {content}")
|
761 |
-
st.error("API ์๋ต์ ํ์ฑํ๋ ์ค ์ค๋ฅ๊ฐ ๋ฐ์ํ์ต๋๋ค. ์๋ต ํ์์ด ์ฌ๋ฐ๋ฅด์ง ์์ต๋๋ค.")
|
762 |
-
st.stop() # return ๋์ st.stop() ์ฌ์ฉ
|
763 |
|
764 |
# ๊ฒฐ๊ณผ ์๊ฐํ
|
765 |
st.subheader("๊ฐ์ ๋ถ์ ๊ฒฐ๊ณผ")
|
@@ -908,9 +893,9 @@ elif menu == "๊ธฐ์ฌ ๋ถ์ํ๊ธฐ":
|
|
908 |
|
909 |
except Exception as e:
|
910 |
st.error(f"๊ฐ์ ๋ถ์ ์ค๋ฅ: {str(e)}")
|
911 |
-
st.
|
912 |
else:
|
913 |
-
st.warning("OpenAI API
|
914 |
|
915 |
elif menu == "์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ":
|
916 |
st.header("์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ")
|
@@ -945,7 +930,7 @@ elif menu == "์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ":
|
|
945 |
generate_image_too = st.checkbox("๊ธฐ์ฌ ์์ฑ ํ ์ด๋ฏธ์ง๋ ํจ๊ป ์์ฑํ๊ธฐ", value=True)
|
946 |
|
947 |
if st.button("์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ"):
|
948 |
-
if st.session_state.
|
949 |
with st.spinner("๊ธฐ์ฌ๋ฅผ ์์ฑ ์ค์
๋๋ค..."):
|
950 |
new_article = generate_article(selected_article['content'], prompt_text)
|
951 |
|
@@ -955,6 +940,7 @@ elif menu == "์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ":
|
|
955 |
# ์ด๋ฏธ์ง ์์ฑํ๊ธฐ (์ต์
์ด ์ ํ๋ ๊ฒฝ์ฐ)
|
956 |
if generate_image_too:
|
957 |
with st.spinner("๊ธฐ์ฌ ๊ด๋ จ ์ด๋ฏธ์ง๋ฅผ ์์ฑ ์ค์
๋๋ค..."):
|
|
|
958 |
image_prompt = f"""์ ๋ฌธ๊ธฐ์ฌ ์ ๋ชฉ "{selected_article['title']}" ์ ๋ณด๊ณ ์ด๋ฏธ์ง๋ฅผ ๋ง๋ค์ด์ค
|
959 |
์ด๋ฏธ์ง์๋ ๋ค์ ์์๊ฐ ํฌํจ๋์ด์ผ ํฉ๋๋ค:
|
960 |
- ๊ธฐ์ฌ๋ฅผ ์ดํดํ ์ ์๋ ๋์
|
@@ -963,13 +949,13 @@ elif menu == "์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ":
|
|
963 |
"""
|
964 |
|
965 |
# ์ด๋ฏธ์ง ์์ฑ
|
966 |
-
|
967 |
|
968 |
-
if
|
969 |
st.subheader("์์ฑ๋ ์ด๋ฏธ์ง:")
|
970 |
-
st.image(
|
971 |
else:
|
972 |
-
st.error(
|
973 |
|
974 |
# ์์ฑ๋ ๊ธฐ์ฌ ์ ์ฅ ์ต์
|
975 |
if st.button("์์ฑ๋ ๊ธฐ์ฌ ์ ์ฅ"):
|
@@ -987,6 +973,8 @@ elif menu == "์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ":
|
|
987 |
else:
|
988 |
st.warning("OpenAI API ํค๋ฅผ ์ฌ์ด๋๋ฐ์์ ์ค์ ํด์ฃผ์ธ์.")
|
989 |
|
|
|
|
|
990 |
elif menu == "๋ด์ค ๊ธฐ์ฌ ์์ฝํ๊ธฐ":
|
991 |
st.header("๋ด์ค ๊ธฐ์ฌ ์์ฝํ๊ธฐ")
|
992 |
|
@@ -1073,30 +1061,6 @@ elif menu == "๋ด์ค ๊ธฐ์ฌ ์์ฝํ๊ธฐ":
|
|
1073 |
with tab3:
|
1074 |
st.subheader("์ค์ผ์ค๋ฌ ์ ์ด ๋ฐ ์ํ")
|
1075 |
|
1076 |
-
# ๋ก๊ทธ ๋ทฐ์ด๋ฅผ ์๋จ์ ๋ฐฐ์น
|
1077 |
-
st.subheader("์ค์๊ฐ ๋ก๊ทธ")
|
1078 |
-
log_container = st.empty()
|
1079 |
-
|
1080 |
-
def update_logs():
|
1081 |
-
try:
|
1082 |
-
with open('/tmp/crawler.log', 'r') as f:
|
1083 |
-
logs = f.readlines()
|
1084 |
-
return ''.join(logs[-100:]) # ์ต๊ทผ 100์ค๋ง ํ์
|
1085 |
-
except Exception as e:
|
1086 |
-
return f"๋ก๊ทธ ํ์ผ์ ์ฝ์ ์ ์์ต๋๋ค: {str(e)}"
|
1087 |
-
|
1088 |
-
# ๋ก๊ทธ ์๋ ์
๋ฐ์ดํธ
|
1089 |
-
if st.checkbox("๋ก๊ทธ ์๋ ์
๋ฐ์ดํธ", value=True):
|
1090 |
-
log_content = update_logs()
|
1091 |
-
log_container.text_area("์ต๊ทผ ๋ก๊ทธ", value=log_content, height=400)
|
1092 |
-
else:
|
1093 |
-
if st.button("๋ก๊ทธ ์๋ก๊ณ ์นจ"):
|
1094 |
-
log_content = update_logs()
|
1095 |
-
log_container.text_area("์ต๊ทผ ๋ก๊ทธ", value=log_content, height=400)
|
1096 |
-
|
1097 |
-
st.divider()
|
1098 |
-
|
1099 |
-
# ์ค์ผ์ค๋ฌ ์ ์ด
|
1100 |
col1, col2 = st.columns(2)
|
1101 |
|
1102 |
with col1:
|
@@ -1180,4 +1144,20 @@ elif menu == "๋ด์ค ๊ธฐ์ฌ ์์ฝํ๊ธฐ":
|
|
1180 |
|
1181 |
# ํธํฐ
|
1182 |
st.markdown("---")
|
1183 |
-
st.markdown("ยฉ ๋ด์ค ๊ธฐ์ฌ ๋๊ตฌ @conanssam")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
import json
|
12 |
import os
|
13 |
from datetime import datetime, timedelta
|
14 |
+
import openai # ๊ตฌ ๋ฒ์ ๋ฐฉ์ ์ฌ์ฉ
|
15 |
from dotenv import load_dotenv
|
16 |
import traceback
|
17 |
import plotly.graph_objects as go
|
18 |
import schedule
|
19 |
import threading
|
20 |
import matplotlib.pyplot as plt
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
# ์๋ํด๋ผ์ฐ๋ ์ถ๊ฐ
|
23 |
try:
|
|
|
40 |
global_scheduler_state = SchedulerState()
|
41 |
|
42 |
# API ํค ๊ด๋ฆฌ๋ฅผ ์ํ ์ธ์
์ํ ์ด๊ธฐํ
|
43 |
+
if 'openai_api_key' not in st.session_state:
|
44 |
+
st.session_state.openai_api_key = None
|
45 |
|
46 |
# ์ฌ๋ฌ ๋ฐฉ๋ฒ์ผ๋ก API ํค ๋ก๋ ์๋
|
47 |
load_dotenv() # .env ํ์ผ์์ ๋ก๋ ์๋
|
48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
# 1. ํ๊ฒฝ ๋ณ์์์ API ํค ํ์ธ
|
50 |
+
if os.environ.get('OPENAI_API_KEY'):
|
51 |
+
st.session_state.openai_api_key = os.environ.get('OPENAI_API_KEY')
|
52 |
+
openai.api_key = st.session_state.openai_api_key
|
53 |
|
54 |
+
# 2. Streamlit secrets์์ API ํค ํ์ธ (try-except๋ก ์ค๋ฅ ๋ฐฉ์ง)
|
55 |
+
if not st.session_state.openai_api_key:
|
56 |
try:
|
57 |
if 'OPENAI_API_KEY' in st.secrets:
|
58 |
+
st.session_state.openai_api_key = st.secrets['OPENAI_API_KEY']
|
59 |
+
openai.api_key = st.session_state.openai_api_key
|
60 |
except Exception as e:
|
61 |
pass # secrets ํ์ผ์ด ์์ด๋ ์ค๋ฅ ๋ฐ์ํ์ง ์์
|
62 |
|
63 |
+
# ์์ ๋๋ ํ ๋ฆฌ๋ฅผ ์ฌ์ฉํ๋๋ก NLTK ๋ฐ์ดํฐ ๊ฒฝ๋ก ์ค์
|
64 |
+
nltk_data_dir = '/tmp/nltk_data'
|
65 |
+
os.makedirs(nltk_data_dir, exist_ok=True)
|
66 |
+
nltk.data.path.insert(0, nltk_data_dir) # ์ด ๊ฒฝ๋ก๋ฅผ ์ฐ์ ๊ฒ์ํ๋๋ก ์ค์
|
67 |
|
68 |
+
# ํ์ํ NLTK ๋ฐ์ดํฐ ๋ค์ด๋ก๋
|
69 |
try:
|
70 |
nltk.data.find('tokenizers/punkt')
|
71 |
except LookupError:
|
72 |
+
nltk.download('punkt', download_dir=nltk_data_dir)
|
73 |
|
74 |
try:
|
75 |
nltk.data.find('corpora/stopwords')
|
76 |
except LookupError:
|
77 |
+
nltk.download('stopwords', download_dir=nltk_data_dir)
|
78 |
|
79 |
# ํ์ด์ง ์ค์
|
80 |
st.set_page_config(page_title="๋ด์ค ๊ธฐ์ฌ ๋๊ตฌ", page_icon="๐ฐ", layout="wide")
|
|
|
90 |
st.divider()
|
91 |
api_key = st.text_input("OpenAI API ํค ์
๋ ฅ", type="password")
|
92 |
if api_key:
|
93 |
+
st.session_state.openai_api_key = api_key
|
94 |
+
openai.api_key = api_key
|
95 |
+
st.success("API ํค๊ฐ ์ค์ ๋์์ต๋๋ค!")
|
|
|
|
|
|
|
96 |
|
97 |
# ์ ์ฅ๋ ๊ธฐ์ฌ๋ฅผ ๋ถ๋ฌ์ค๋ ํจ์
|
98 |
def load_saved_articles():
|
|
|
112 |
"""
|
113 |
๋ค์ด๋ฒ ๋ด์ค ๊ธฐ์ฌ๋ฅผ ์์งํ๋ ํจ์
|
114 |
"""
|
|
|
115 |
url = f"https://search.naver.com/search.naver?where=news&query={keyword}"
|
116 |
results = []
|
117 |
|
118 |
try:
|
119 |
# ํ์ด์ง ์์ฒญ
|
|
|
120 |
response = requests.get(url)
|
|
|
|
|
121 |
soup = BeautifulSoup(response.text, 'html.parser')
|
122 |
|
123 |
# ๋ด์ค ์์ดํ
์ฐพ๊ธฐ
|
124 |
news_items = soup.select('div.sds-comps-base-layout.sds-comps-full-layout')
|
|
|
125 |
|
126 |
# ๊ฐ ๋ด์ค ์์ดํ
์์ ์ ๋ณด ์ถ์ถ
|
127 |
for i, item in enumerate(news_items):
|
|
|
156 |
'description': description,
|
157 |
'source': source,
|
158 |
'date': date,
|
159 |
+
'content': "" # ๋์ค์ ์๋ฌธ ๋ด์ฉ์ ์ ์ฅํ ํ๋
|
160 |
})
|
161 |
|
|
|
|
|
162 |
except Exception as e:
|
163 |
+
st.error(f"๊ธฐ์ฌ ์ ๋ณด ์ถ์ถ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}")
|
164 |
continue
|
165 |
|
166 |
except Exception as e:
|
167 |
+
st.error(f"ํ์ด์ง ์์ฒญ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}")
|
168 |
|
|
|
169 |
return results
|
170 |
|
171 |
# ๊ธฐ์ฌ ์๋ฌธ ๊ฐ์ ธ์ค๊ธฐ
|
172 |
def get_article_content(url):
|
|
|
173 |
try:
|
174 |
response = requests.get(url, timeout=5)
|
|
|
|
|
175 |
soup = BeautifulSoup(response.text, 'html.parser')
|
176 |
|
177 |
# ๋ค์ด๋ฒ ๋ด์ค ๋ณธ๋ฌธ ์ฐพ๊ธฐ
|
178 |
content = soup.select_one('#dic_area')
|
179 |
if content:
|
180 |
text = content.text.strip()
|
181 |
+
text = re.sub(r'\s+', ' ', text) # ์ฌ๋ฌ ๊ณต๋ฐฑ ์ ๊ฑฐ
|
|
|
182 |
return text
|
183 |
|
184 |
+
# ๋ค๋ฅธ ๋ด์ค ์ฌ์ดํธ ๋ณธ๋ฌธ ์ฐพ๊ธฐ (์ฌ๋ฌ ์ฌ์ดํธ ๋์ ํ์)
|
185 |
content = soup.select_one('.article_body, .article-body, .article-content, .news-content-inner')
|
186 |
if content:
|
187 |
text = content.text.strip()
|
188 |
text = re.sub(r'\s+', ' ', text)
|
|
|
189 |
return text
|
190 |
|
|
|
191 |
return "๋ณธ๋ฌธ์ ๊ฐ์ ธ์ฌ ์ ์์ต๋๋ค."
|
192 |
except Exception as e:
|
|
|
193 |
return f"์ค๋ฅ ๋ฐ์: {str(e)}"
|
194 |
|
195 |
+
# NLTK๋ฅผ ์ด์ฉํ ํค์๋ ๋ถ์
|
196 |
def analyze_keywords(text, top_n=10):
|
197 |
+
# ํ๊ตญ์ด ๋ถ์ฉ์ด ๋ชฉ๋ก (์ง์ ์ ์ํด์ผ ํฉ๋๋ค)
|
198 |
korean_stopwords = ['์ด', '๊ทธ', '์ ', '๊ฒ', '๋ฐ', '๋ฑ', '๋ฅผ', '์', '์', '์์', '์', '์ผ๋ก', '๋ก']
|
199 |
|
200 |
+
tokens = word_tokenize(text)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
201 |
tokens = [word for word in tokens if word.isalnum() and len(word) > 1 and word not in korean_stopwords]
|
202 |
|
203 |
word_count = Counter(tokens)
|
|
|
325 |
results['top_keywords'] = []
|
326 |
return results
|
327 |
|
328 |
+
# OpenAI API๋ฅผ ์ด์ฉํ ์ ๊ธฐ์ฌ ์์ฑ (๊ตฌ ๋ฒ์ ๋ฐฉ์)
|
329 |
def generate_article(original_content, prompt_text):
|
330 |
try:
|
331 |
+
if not st.session_state.openai_api_key:
|
332 |
return "OpenAI API ํค๊ฐ ์ค์ ๋์ง ์์์ต๋๋ค."
|
333 |
|
334 |
+
response = openai.ChatCompletion.create(
|
335 |
+
model="gpt-4.1-mini",
|
336 |
messages=[
|
337 |
{"role": "system", "content": "๋น์ ์ ์ ๋ฌธ์ ์ธ ๋ด์ค ๊ธฐ์์
๋๋ค. ์ฃผ์ด์ง ๋ด์ฉ์ ๋ฐํ์ผ๋ก ์๋ก์ด ๊ธฐ์ฌ๋ฅผ ์์ฑํด์ฃผ์ธ์."},
|
338 |
{"role": "user", "content": f"๋ค์ ๋ด์ฉ์ ๋ฐํ์ผ๋ก {prompt_text}\n\n{original_content[:1000]}"}
|
339 |
],
|
340 |
max_tokens=2000
|
341 |
)
|
342 |
+
return response.choices[0].message['content']
|
343 |
except Exception as e:
|
344 |
return f"๊ธฐ์ฌ ์์ฑ ์ค๋ฅ: {str(e)}"
|
345 |
|
346 |
+
# OpenAI API๋ฅผ ์ด์ฉํ ์ด๋ฏธ์ง ์์ฑ (๊ตฌ ๋ฒ์ ๋ฐฉ์)
|
347 |
def generate_image(prompt):
|
348 |
try:
|
349 |
+
if not st.session_state.openai_api_key:
|
350 |
return "OpenAI API ํค๊ฐ ์ค์ ๋์ง ์์์ต๋๋ค."
|
351 |
|
352 |
+
response = openai.Image.create(
|
|
|
|
|
353 |
prompt=prompt,
|
354 |
+
n=1,
|
355 |
size="1024x1024"
|
356 |
)
|
357 |
+
return response['data'][0]['url']
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
358 |
except Exception as e:
|
359 |
return f"์ด๋ฏธ์ง ์์ฑ ์ค๋ฅ: {str(e)}"
|
360 |
|
|
|
376 |
traceback.print_exc()
|
377 |
|
378 |
def perform_news_task(task_type, keyword, num_articles, file_prefix):
|
|
|
379 |
try:
|
380 |
articles = crawl_naver_news(keyword, num_articles)
|
|
|
381 |
|
382 |
# ๊ธฐ์ฌ ๋ด์ฉ ๊ฐ์ ธ์ค๊ธฐ
|
383 |
+
for article in articles:
|
|
|
384 |
article['content'] = get_article_content(article['link'])
|
385 |
time.sleep(0.5) # ์๋ฒ ๋ถํ ๋ฐฉ์ง
|
386 |
|
|
|
392 |
with open(filename, 'w', encoding='utf-8') as f:
|
393 |
json.dump(articles, f, ensure_ascii=False, indent=2)
|
394 |
|
|
|
|
|
395 |
global_scheduler_state.last_run = datetime.now()
|
396 |
print(f"{datetime.now()} - {task_type} ๋ด์ค ๊ธฐ์ฌ ์์ง ์๋ฃ: {keyword}")
|
397 |
|
398 |
+
# ์ ์ญ ์ํ์ ์์ง ๊ฒฐ๊ณผ๋ฅผ ์ ์ฅ (UI ์
๋ฐ์ดํธ์ฉ)
|
399 |
result_item = {
|
400 |
'task_type': task_type,
|
401 |
'keyword': keyword,
|
|
|
406 |
global_scheduler_state.scheduled_results.append(result_item)
|
407 |
|
408 |
except Exception as e:
|
409 |
+
print(f"์์
์คํ ์ค ์ค๋ฅ ๋ฐ์: {e}")
|
410 |
traceback.print_exc()
|
411 |
|
412 |
def start_scheduler(daily_tasks, interval_tasks):
|
|
|
563 |
with keyword_tab1:
|
564 |
keywords = analyze_keywords(selected_article['content'])
|
565 |
|
566 |
+
# ์๊ฐํ
|
567 |
df = pd.DataFrame(keywords, columns=['๋จ์ด', '๋น๋์'])
|
568 |
+
st.bar_chart(df.set_index('๋จ์ด'))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
569 |
|
570 |
st.write("**์ฃผ์ ํค์๋:**")
|
571 |
for word, count in keywords:
|
|
|
595 |
# ํ
์คํธ ํต๊ณ ๊ณ์ฐ
|
596 |
word_count = len(re.findall(r'\b\w+\b', content))
|
597 |
char_count = len(content)
|
598 |
+
sentence_count = len(re.split(r'[.!?]+', content))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
599 |
avg_word_length = sum(len(word) for word in re.findall(r'\b\w+\b', content)) / word_count if word_count > 0 else 0
|
600 |
avg_sentence_length = word_count / sentence_count if sentence_count > 0 else 0
|
601 |
|
|
|
615 |
with col2:
|
616 |
st.metric("ํ๊ท ๋ฌธ์ฅ ๊ธธ์ด", f"{avg_sentence_length:.1f}๋จ์ด")
|
617 |
|
618 |
+
# ํ
์คํธ ๋ณต์ก์ฑ ์ ์ (๊ฐ๋จํ ์์)
|
619 |
complexity_score = min(10, (avg_sentence_length / 10) * 5 + (avg_word_length / 5) * 5)
|
620 |
st.progress(complexity_score / 10)
|
621 |
st.write(f"ํ
์คํธ ๋ณต์ก์ฑ ์ ์: {complexity_score:.1f}/10")
|
622 |
+
|
623 |
+
# ์ถํ ๋น๋ ๋ง๋ ๊ทธ๋ํ
|
624 |
+
st.subheader("ํ์ฌ๋ณ ๋ถํฌ (ํ๊ตญ์ด/์์ด ์ง์)")
|
625 |
+
try:
|
626 |
+
# KoNLPy ์ค์น ํ์ธ
|
627 |
+
try:
|
628 |
+
from konlpy.tag import Okt
|
629 |
+
konlpy_installed = True
|
630 |
+
except ImportError:
|
631 |
+
konlpy_installed = False
|
632 |
+
st.warning("ํ๊ตญ์ด ํํ์ ๋ถ์์ ์ํด KoNLPy๋ฅผ ์ค์นํด์ฃผ์ธ์: pip install konlpy")
|
633 |
+
|
634 |
+
# ์์ด POS tagger ์ค๋น
|
635 |
+
from nltk import pos_tag
|
636 |
+
try:
|
637 |
+
nltk.data.find('taggers/averaged_perceptron_tagger')
|
638 |
+
except LookupError:
|
639 |
+
nltk.download('averaged_perceptron_tagger', download_dir=nltk_data_dir)
|
640 |
+
|
641 |
+
# ์ธ์ด ๊ฐ์ง (๊ฐ๋จํ ๋ฐฉ์)
|
642 |
+
is_korean = bool(re.search(r'[๊ฐ-ํฃ]', content))
|
643 |
+
|
644 |
+
if is_korean and konlpy_installed:
|
645 |
+
# ํ๊ตญ์ด ํํ์ ๋ถ์
|
646 |
+
okt = Okt()
|
647 |
+
tagged = okt.pos(content)
|
648 |
+
|
649 |
+
# ํ๊ตญ์ด ํ์ฌ ๋งคํ
|
650 |
+
pos_dict = {
|
651 |
+
'Noun': '๋ช
์ฌ', 'NNG': '๋ช
์ฌ', 'NNP': '๊ณ ์ ๋ช
์ฌ',
|
652 |
+
'Verb': '๋์ฌ', 'VV': '๋์ฌ', 'VA': 'ํ์ฉ์ฌ',
|
653 |
+
'Adjective': 'ํ์ฉ์ฌ',
|
654 |
+
'Adverb': '๋ถ์ฌ',
|
655 |
+
'Josa': '์กฐ์ฌ', 'Punctuation': '๊ตฌ๋์ ',
|
656 |
+
'Determiner': '๊ดํ์ฌ', 'Exclamation': '๊ฐํ์ฌ'
|
657 |
+
}
|
658 |
+
|
659 |
+
pos_counts = {'๋ช
์ฌ': 0, '๋์ฌ': 0, 'ํ์ฉ์ฌ': 0, '๋ถ์ฌ': 0, '์กฐ์ฌ': 0, '๊ตฌ๋์ ': 0, '๊ดํ์ฌ': 0, '๊ฐํ์ฌ': 0, '๊ธฐํ': 0}
|
660 |
+
|
661 |
+
for _, pos in tagged:
|
662 |
+
if pos in pos_dict:
|
663 |
+
pos_counts[pos_dict[pos]] += 1
|
664 |
+
elif pos.startswith('N'): # ๊ธฐํ ๋ช
์ฌ๋ฅ
|
665 |
+
pos_counts['๋ช
์ฌ'] += 1
|
666 |
+
elif pos.startswith('V'): # ๊ธฐํ ๋์ฌ๋ฅ
|
667 |
+
pos_counts['๋์ฌ'] += 1
|
668 |
+
else:
|
669 |
+
pos_counts['๊ธฐํ'] += 1
|
670 |
+
|
671 |
+
else:
|
672 |
+
# ์์ด POS ํ๊น
|
673 |
+
tokens = word_tokenize(content.lower())
|
674 |
+
tagged = pos_tag(tokens)
|
675 |
+
|
676 |
+
# ์์ด ํ์ฌ ๋งคํ
|
677 |
+
pos_dict = {
|
678 |
+
'NN': '๋ช
์ฌ', 'NNS': '๋ช
์ฌ', 'NNP': '๊ณ ์ ๋ช
์ฌ', 'NNPS': '๊ณ ์ ๋ช
์ฌ',
|
679 |
+
'VB': '๋์ฌ', 'VBD': '๋์ฌ', 'VBG': '๋์ฌ', 'VBN': '๋์ฌ', 'VBP': '๋์ฌ', 'VBZ': '๋์ฌ',
|
680 |
+
'JJ': 'ํ์ฉ์ฌ', 'JJR': 'ํ์ฉ์ฌ', 'JJS': 'ํ์ฉ์ฌ',
|
681 |
+
'RB': '๋ถ์ฌ', 'RBR': '๋ถ์ฌ', 'RBS': '๋ถ์ฌ'
|
682 |
+
}
|
683 |
+
|
684 |
+
pos_counts = {'๋ช
์ฌ': 0, '๋์ฌ': 0, 'ํ์ฉ์ฌ': 0, '๋ถ์ฌ': 0, '๊ธฐํ': 0}
|
685 |
+
|
686 |
+
for _, pos in tagged:
|
687 |
+
if pos in pos_dict:
|
688 |
+
pos_counts[pos_dict[pos]] += 1
|
689 |
+
else:
|
690 |
+
pos_counts['๊ธฐํ'] += 1
|
691 |
+
|
692 |
+
# ๊ฒฐ๊ณผ ์๊ฐํ
|
693 |
+
pos_df = pd.DataFrame({
|
694 |
+
'ํ์ฌ': list(pos_counts.keys()),
|
695 |
+
'๋น๋': list(pos_counts.values())
|
696 |
+
})
|
697 |
+
|
698 |
+
st.bar_chart(pos_df.set_index('ํ์ฌ'))
|
699 |
+
|
700 |
+
if is_korean:
|
701 |
+
st.info("ํ๊ตญ์ด ํ
์คํธ๊ฐ ๊ฐ์ง๋์์ต๋๋ค.")
|
702 |
+
else:
|
703 |
+
st.info("์์ด ํ
์คํธ๊ฐ ๊ฐ์ง๋์์ต๋๋ค.")
|
704 |
+
except Exception as e:
|
705 |
+
st.error(f"ํ์ฌ ๋ถ์ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}")
|
706 |
+
st.error(traceback.format_exc())
|
707 |
|
708 |
elif analysis_type == "๊ฐ์ ๋ถ์":
|
709 |
if st.button("๊ฐ์ ๋ถ์ํ๊ธฐ"):
|
710 |
+
if st.session_state.openai_api_key:
|
711 |
with st.spinner("๊ธฐ์ฌ์ ๊ฐ์ ์ ๋ถ์ ์ค์
๋๋ค..."):
|
712 |
try:
|
713 |
+
# ๊ฐ์ ๋ถ์ ํ๋กฌํํธ ์ค์ (๊ตฌ ๋ฒ์ ๋ฐฉ์)
|
714 |
+
prompt = """
|
715 |
+
๋ค์ ๋ด์ค ๊ธฐ์ฌ์ ๊ฐ์ ๊ณผ ๋
ผ์กฐ๋ฅผ ๋ถ์ํ๊ณ , ์๋ ์์์ฒ๋ผ JSON๋ง ๋ฐํํ์ธ์.
|
716 |
+
๋ถํ์ํ ์ค๋ช
, ์ธ์ฌ๋ง, ๊ธฐํ ํ
์คํธ ์์ด ๋ฐ๋์ JSON๋ง ์ถ๋ ฅํ์ธ์.
|
717 |
+
|
718 |
+
์์:
|
719 |
+
{
|
720 |
+
"sentiment": "๊ธ์ ์ ",
|
721 |
+
"reason": "์ด ๊ธฐ์ฌ์์๋ ๊ธ์ ์ ์ธ ๋จ์ด์ ํํ์ด ๋ง์ด ์ฌ์ฉ๋์์ต๋๋ค.",
|
722 |
+
"keywords": [
|
723 |
+
{"word": "ํฌ๋ง", "score": 8},
|
724 |
+
{"word": "์ฑ๊ณต", "score": 7},
|
725 |
+
{"word": "๊ธฐ๋", "score": 6},
|
726 |
+
{"word": "์ฑ์ฅ", "score": 7},
|
727 |
+
{"word": "ํ์ ", "score": 8}
|
728 |
+
]
|
729 |
+
}
|
730 |
+
|
731 |
+
๋ถ์ํ ๊ธฐ์ฌ:
|
732 |
+
์ ๋ชฉ: {title}
|
733 |
+
๋ด์ฉ: {content}
|
734 |
+
"""
|
735 |
+
|
736 |
+
# ๊ฐ์ ๋ถ์
|
737 |
+
response = openai.ChatCompletion.create(
|
738 |
+
model="gpt-4.1-mini",
|
739 |
messages=[
|
740 |
+
{"role": "system", "content": "๋น์ ์ ํ
์คํธ์ ๊ฐ์ ๊ณผ ๋
ผ์กฐ๋ฅผ ๋ถ์ํ๋ ์ ๋ฌธ๊ฐ์
๋๋ค. ๋ค์ ๋ด์ค ๊ธฐ์ฌ์ ๊ฐ์ ๊ณผ ๋
ผ์กฐ๋ฅผ ๋ถ์ํ๊ณ , '๊ธ์ ์ ', '๋ถ์ ์ ', '์ค๋ฆฝ์ ' ์ค ํ๋๋ก ๋ถ๋ฅํด ์ฃผ์ธ์. ๋ํ ๊ธฐ์ฌ์์ ๋๋ฌ๋๋ ํต์ฌ ๊ฐ์ ํค์๋๋ฅผ 5๊ฐ ์ถ์ถํ๊ณ , ๊ฐ ํค์๋๋ณ๋ก 1-10 ์ฌ์ด์ ๊ฐ๋ ์ ์๋ฅผ ๋งค๊ฒจ์ฃผ์ธ์. JSON ํ์์ผ๋ก ๋ค์๊ณผ ๊ฐ์ด ์๋ตํด์ฃผ์ธ์: {'sentiment': '๊ธ์ ์ /๋ถ์ ์ /์ค๋ฆฝ์ ', 'reason': '์ด์ ์ค๋ช
...', 'keywords': [{'word': 'ํค์๋1', 'score': 8}, {'word': 'ํค์๋2', 'score': 7}, ...]}"},
|
741 |
+
{"role": "user", "content": prompt.format(title=selected_article['title'], content=selected_article['content'][:1500])}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
742 |
],
|
743 |
+
max_tokens=800
|
|
|
744 |
)
|
745 |
|
746 |
+
# JSON ํ์ฑ (๊ตฌ ๋ฒ์ ๋ฐฉ์)
|
747 |
+
analysis_result = json.loads(response.choices[0].message['content'])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
748 |
|
749 |
# ๊ฒฐ๊ณผ ์๊ฐํ
|
750 |
st.subheader("๊ฐ์ ๋ถ์ ๊ฒฐ๊ณผ")
|
|
|
893 |
|
894 |
except Exception as e:
|
895 |
st.error(f"๊ฐ์ ๋ถ์ ์ค๋ฅ: {str(e)}")
|
896 |
+
st.code(traceback.format_exc())
|
897 |
else:
|
898 |
+
st.warning("OpenAI API ํค๊ฐ ์ค์ ๋์ด ์์ง ์์ต๋๋ค. ์ฌ์ด๋๋ฐ์์ API ํค๋ฅผ ์ค์ ํด์ฃผ์ธ์.")
|
899 |
|
900 |
elif menu == "์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ":
|
901 |
st.header("์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ")
|
|
|
930 |
generate_image_too = st.checkbox("๊ธฐ์ฌ ์์ฑ ํ ์ด๋ฏธ์ง๋ ํจ๊ป ์์ฑํ๊ธฐ", value=True)
|
931 |
|
932 |
if st.button("์ ๊ธฐ์ฌ ์์ฑํ๊ธฐ"):
|
933 |
+
if st.session_state.openai_api_key:
|
934 |
with st.spinner("๊ธฐ์ฌ๋ฅผ ์์ฑ ์ค์
๋๋ค..."):
|
935 |
new_article = generate_article(selected_article['content'], prompt_text)
|
936 |
|
|
|
940 |
# ์ด๋ฏธ์ง ์์ฑํ๊ธฐ (์ต์
์ด ์ ํ๋ ๊ฒฝ์ฐ)
|
941 |
if generate_image_too:
|
942 |
with st.spinner("๊ธฐ์ฌ ๊ด๋ จ ์ด๋ฏธ์ง๋ฅผ ์์ฑ ์ค์
๋๋ค..."):
|
943 |
+
# ์ด๋ฏธ์ง ์์ฑ ํ๋กฌํํธ ์ค๋น
|
944 |
image_prompt = f"""์ ๋ฌธ๊ธฐ์ฌ ์ ๋ชฉ "{selected_article['title']}" ์ ๋ณด๊ณ ์ด๋ฏธ์ง๋ฅผ ๋ง๋ค์ด์ค
|
945 |
์ด๋ฏธ์ง์๋ ๋ค์ ์์๊ฐ ํฌํจ๋์ด์ผ ํฉ๋๋ค:
|
946 |
- ๊ธฐ์ฌ๋ฅผ ์ดํดํ ์ ์๋ ๋์
|
|
|
949 |
"""
|
950 |
|
951 |
# ์ด๋ฏธ์ง ์์ฑ
|
952 |
+
image_url = generate_image(image_prompt)
|
953 |
|
954 |
+
if image_url and not image_url.startswith("์ด๋ฏธ์ง ์์ฑ ์ค๋ฅ"):
|
955 |
st.subheader("์์ฑ๋ ์ด๋ฏธ์ง:")
|
956 |
+
st.image(image_url)
|
957 |
else:
|
958 |
+
st.error(image_url)
|
959 |
|
960 |
# ์์ฑ๋ ๊ธฐ์ฌ ์ ์ฅ ์ต์
|
961 |
if st.button("์์ฑ๋ ๊ธฐ์ฌ ์ ์ฅ"):
|
|
|
973 |
else:
|
974 |
st.warning("OpenAI API ํค๋ฅผ ์ฌ์ด๋๋ฐ์์ ์ค์ ํด์ฃผ์ธ์.")
|
975 |
|
976 |
+
|
977 |
+
|
978 |
elif menu == "๋ด์ค ๊ธฐ์ฌ ์์ฝํ๊ธฐ":
|
979 |
st.header("๋ด์ค ๊ธฐ์ฌ ์์ฝํ๊ธฐ")
|
980 |
|
|
|
1061 |
with tab3:
|
1062 |
st.subheader("์ค์ผ์ค๋ฌ ์ ์ด ๋ฐ ์ํ")
|
1063 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1064 |
col1, col2 = st.columns(2)
|
1065 |
|
1066 |
with col1:
|
|
|
1144 |
|
1145 |
# ํธํฐ
|
1146 |
st.markdown("---")
|
1147 |
+
st.markdown("ยฉ ๋ด์ค ๊ธฐ์ฌ ๋๊ตฌ @conanssam")
|
1148 |
+
|
1149 |
+
def extract_json_from_response(response_text):
|
1150 |
+
# JSON ๊ฐ์ฒด ๋ถ๋ถ๋ง ์ถ์ถ (๊ฐ์ฅ ๋จผ์ ๋์ค๋ ์ค๊ดํธ ์)
|
1151 |
+
match = re.search(r'\{.*\}', response_text, re.DOTALL)
|
1152 |
+
if match:
|
1153 |
+
json_str = match.group(0)
|
1154 |
+
try:
|
1155 |
+
return json.loads(json_str)
|
1156 |
+
except Exception as e:
|
1157 |
+
st.error(f"JSON ํ์ฑ ์ค๋ฅ: {str(e)}")
|
1158 |
+
st.code(json_str)
|
1159 |
+
return None
|
1160 |
+
else:
|
1161 |
+
st.error("์๋ต์์ JSON์ ์ฐพ์ ์ ์์ต๋๋ค.")
|
1162 |
+
st.code(response_text)
|
1163 |
+
return None
|