benjamin5607 commited on
Commit
7001ffd
ยท
verified ยท
1 Parent(s): ac5219b

Update streamlit_app.py

Browse files
Files changed (1) hide show
  1. streamlit_app.py +76 -96
streamlit_app.py CHANGED
@@ -22,7 +22,7 @@ else:
22
  st.stop()
23
 
24
  # ==============================================================================
25
- # ๐ŸŒ [UI] ๋‹ค๊ตญ์–ด ๋”•์…”๋„ˆ๋ฆฌ (์‹œ์žฅ ๋‚ ์”จ ์ถ”๊ฐ€)
26
  # ==============================================================================
27
  UI_TEXT = {
28
  "KR": {
@@ -33,21 +33,22 @@ UI_TEXT = {
33
  "search_title": "๐Ÿง  ๋‹ค๊ตญ์–ด AI ์ฃผ์‹ ๋น„์„œ (๋ฉ€ํ‹ฐ ํƒ€๊ฒŸ)",
34
  "search_placeholder": "์˜ˆ: ์‚ผ์„ฑ์ „์ž๋ž‘ TSMC ๋น„๊ตํ•ด์ค˜, ๋ชฝ๊ณจ ๊ณ ๋น„ ์บ์‹œ๋ฏธ์–ด...",
35
  "btn_analyze": "๋ถ„์„ ์‹œ์ž‘",
36
- "status_thinking": "๐Ÿง  AI๊ฐ€ ์‹œ์žฅ ์ƒํ™ฉ๊ณผ ๊ธฐ์—…์„ ๋ถ„์„ ์ค‘์ž…๋‹ˆ๋‹ค...",
37
  "status_news": "๐Ÿ“ฐ ๋‰ด์Šค ์ˆ˜์ง‘ ์ค‘...",
38
  "metric_price": "ํ˜„์žฌ๊ฐ€",
39
  "metric_source": "์ถœ์ฒ˜",
40
  "expander_news": "๐Ÿ“ฐ ๋‰ด์Šค ์›๋ฌธ ๋ณด๊ธฐ",
41
  "insight_title": "๐Ÿง  AI ์ข…ํ•ฉ ๋ถ„์„ ๋ฆฌํฌํŠธ",
42
  "scanner_title": "๐Ÿ“ก ๊ธ€๋กœ๋ฒŒ ์‹œ์žฅ ์ „๊ด‘ํŒ (Frontier Included)",
43
- "scanner_caption": "โ€ป ๋ชฝ๊ณจ, ๋ฐฉ๊ธ€๋ผ๋ฐ์‹œ ๋“ฑ ๋ฐ์ดํ„ฐ ์ทจ์•ฝ ๊ตญ๊ฐ€๋Š” '์›น ๊ฒ€์ƒ‰'์œผ๋กœ ๊ฐ€๊ฒฉ์„ ์ถ”์ ํ•ฉ๋‹ˆ๋‹ค.",
44
- "btn_scan": "์ „์ฒด ์ข…๋ชฉ ์Šค์บ” ์‹œ์ž‘",
45
  "tab_gainers": "๐Ÿ”ฅ ๊ธ‰๋“ฑ Top 10",
46
  "tab_losers": "๐Ÿ’ง ๊ธ‰๋ฝ Top 10",
47
  "col_name": "๊ธฐ์—…๋ช…",
48
  "col_price": "๊ฐ€๊ฒฉ",
49
  "col_change": "๋“ฑ๋ฝ๋ฅ ",
50
  "col_source": "์ถœ์ฒ˜",
 
51
  "msg_fail": "๋ฐ์ดํ„ฐ ์ˆ˜์ง‘ ์‹คํŒจ",
52
  "llm_lang_instruction": "Korean"
53
  },
@@ -59,32 +60,33 @@ UI_TEXT = {
59
  "search_title": "๐Ÿง  AI Stock Assistant (Multi-Target)",
60
  "search_placeholder": "e.g. Compare Samsung vs TSMC...",
61
  "btn_analyze": "Analyze",
62
- "status_thinking": "๐Ÿง  AI is analyzing market context & targets...",
63
  "status_news": "๐Ÿ“ฐ Fetching news...",
64
  "metric_price": "Price",
65
  "metric_source": "Source",
66
  "expander_news": "๐Ÿ“ฐ View Source News",
67
  "insight_title": "๐Ÿง  AI Comprehensive Report",
68
  "scanner_title": "๐Ÿ“ก Global Market Scanner (Frontier Included)",
69
- "scanner_caption": "โ€ป Frontier markets (Mongolia, BD) rely on Web Search.",
70
- "btn_scan": "Start Full Scan",
71
  "tab_gainers": "๐Ÿ”ฅ Top 10 Gainers",
72
  "tab_losers": "๐Ÿ’ง Top 10 Losers",
73
  "col_name": "Company",
74
  "col_price": "Price",
75
  "col_change": "Change(%)",
76
  "col_source": "Source",
 
77
  "msg_fail": "Data collection failed",
78
  "llm_lang_instruction": "English"
79
  }
80
  }
81
 
82
  # ==============================================================================
83
- # ๐Ÿ’พ [๋ฐ์ดํ„ฐ] ํ‹ฐ์ปค ๋งคํ•‘ (ํ”„๋ก ํ‹ฐ์–ด ๋งˆ์ผ“ ํฌํ•จ Full List)
84
  # ==============================================================================
85
  TICKER_NAMES = {
86
  # ๐Ÿ‡ญ๐Ÿ‡ฐ ํ™์ฝฉ (HK)
87
- "1299.HK": "AIA Group", "0388.HK": "HKEX (ํ™์ฝฉ๊ฑฐ๋ž˜์†Œ)", "0005.HK": "HSBC Holdings",
88
  "0016.HK": "SHK Properties", "1113.HK": "CK Asset", "0002.HK": "CLP Holdings",
89
  "0003.HK": "HK & China Gas", "0066.HK": "MTR Corp", "0011.HK": "Hang Seng Bank",
90
  "0027.HK": "Galaxy Ent", "0019.HK": "Swire Pacific", "9988.HK": "Alibaba",
@@ -171,37 +173,22 @@ MARKET_SAMPLES = {
171
  }
172
 
173
  # ==============================================================================
174
- # ๐Ÿ› ๏ธ [์—”์ง„] ์ฃผ์š” ์ง€์ˆ˜(S&P 500 ๋“ฑ) ์ˆ˜์ง‘ ํ•จ์ˆ˜ (NEW)
175
  # ==============================================================================
176
  def get_market_indices():
177
- """
178
- ์‚ฌ์ด๋“œ๋ฐ” ๋ฐ AI ๋ถ„์„์šฉ ์‹œ์žฅ ์ง€์ˆ˜ ์ˆ˜์ง‘
179
- """
180
- indices = {
181
- "S&P 500": "^GSPC",
182
- "NASDAQ": "^IXIC",
183
- "KOSPI": "^KS11",
184
- "VN-INDEX": "^VNINDEX"
185
- }
186
  data = {}
187
  try:
188
- tickers = list(indices.values())
189
- df = yf.download(tickers, period="5d", progress=False)['Close']
190
-
191
  for name, ticker in indices.items():
192
  if ticker in df.columns:
193
  series = df[ticker].dropna()
194
  if len(series) >= 2:
195
- curr = series.iloc[-1]
196
- prev = series.iloc[-2]
197
- pct = ((curr - prev)/prev)*100
198
- data[name] = (curr, pct)
199
  except: pass
200
  return data
201
 
202
- # ==============================================================================
203
- # ๐Ÿ› ๏ธ [์—”์ง„] ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘ ํ•จ์ˆ˜๋“ค (Robust)
204
- # ==============================================================================
205
  FAKE_HEADERS = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36'}
206
 
207
  def fetch_news_robust(keyword):
@@ -210,18 +197,12 @@ def fetch_news_robust(keyword):
210
  try:
211
  with DDGS() as ddgs:
212
  results = list(ddgs.news(keyword, timelimit="m", max_results=3))
213
- if not results:
214
- time.sleep(random.uniform(0.5, 1.0))
215
- results = list(ddgs.news(keyword, max_results=3))
216
- if not results:
217
- time.sleep(random.uniform(0.5, 1.0))
218
- results = list(ddgs.text(f"{keyword} latest news", max_results=3))
219
  for r in results:
220
  url = r.get('url') or r.get('href')
221
  if url not in seen_urls:
222
- date = r.get('date', '?')
223
- title = r.get('title', '')
224
- summary += f"[{date}] {title}\n"
225
  seen_urls.add(url)
226
  except: return "No news."
227
  return summary if summary else "No news."
@@ -246,11 +227,8 @@ def get_price_data_robust(ticker, name):
246
  elif ".KS" in ticker or ".KQ" in ticker: fdr_symbol = ticker.split('.')[0]
247
  hist = fdr.DataReader(fdr_symbol, start=(datetime.datetime.now() - datetime.timedelta(days=7)))
248
  if not hist.empty:
249
- curr = hist['Close'].iloc[-1]
250
- prev = hist['Close'].iloc[-2] if len(hist) > 1 else curr
251
- pct = 0.0
252
- if prev != 0: pct = ((curr - prev)/prev)*100
253
- return curr, pct, "FDR"
254
  except: pass
255
  try:
256
  stock = yf.Ticker(ticker)
@@ -268,8 +246,7 @@ def get_chart_data(ticker):
268
  if ".VN" in ticker: fdr_symbol = ticker.split('.')[0]
269
  elif ".JK" in ticker: fdr_symbol = f"IDX:{ticker.split('.')[0]}"
270
  elif ".T" in ticker: fdr_symbol = f"TSE:{ticker.split('.')[0]}"
271
- start = (datetime.datetime.now() - datetime.timedelta(days=90)).strftime('%Y-%m-%d')
272
- hist = fdr.DataReader(fdr_symbol, start)
273
  if not hist.empty: return hist.rename(columns={'Close':'Close', 'Open':'Open', 'High':'High', 'Low':'Low'}), "FDR"
274
  except: pass
275
  try:
@@ -318,39 +295,35 @@ with st.sidebar:
318
  T = UI_TEXT[lang_code]
319
  st.title(T['sidebar_title'])
320
 
321
- # ๐Ÿ“Š [NEW] ์‹œ์žฅ ๋‚ ์”จ (Market Weather)
322
  st.divider()
323
  st.caption(T['sidebar_weather'])
324
-
325
  market_data = get_market_indices()
326
- market_context_str = "" # AI์—๊ฒŒ ๋˜์ ธ์ค„ ์‹œ์žฅ ์ƒํ™ฉ ์š”์•ฝ๊ธ€
327
 
328
  if market_data:
329
- # UI ํ‘œ์‹œ (2x2 Grid)
330
- col_w1, col_w2 = st.columns(2)
331
  if "S&P 500" in market_data:
332
  p, c = market_data["S&P 500"]
333
- col_w1.metric("๐Ÿ‡บ๐Ÿ‡ธ S&P 500", f"{p:,.0f}", f"{c:+.2f}%")
334
- market_context_str += f"S&P 500 is {c:+.2f}%. "
335
  if "NASDAQ" in market_data:
336
  p, c = market_data["NASDAQ"]
337
- col_w2.metric("๐Ÿ‡บ๐Ÿ‡ธ NASDAQ", f"{p:,.0f}", f"{c:+.2f}%")
338
- market_context_str += f"NASDAQ is {c:+.2f}%. "
339
-
340
- col_w3, col_w4 = st.columns(2)
341
  if "KOSPI" in market_data:
342
  p, c = market_data["KOSPI"]
343
- col_w3.metric("๐Ÿ‡ฐ๐Ÿ‡ท KOSPI", f"{p:,.0f}", f"{c:+.2f}%")
344
  if "VN-INDEX" in market_data:
345
  p, c = market_data["VN-INDEX"]
346
- col_w4.metric("๐Ÿ‡ป๐Ÿ‡ณ VN-IDX", f"{p:,.0f}", f"{c:+.2f}%")
347
- market_context_str += f"Vietnam Index is {c:+.2f}%. "
348
- else:
349
- st.caption("Load failed.")
350
 
351
  st.divider()
352
  menu = st.radio("MENU", [T['menu_search'], T['menu_scanner']], index=0)
353
 
 
354
  if menu == T['menu_search']:
355
  st.subheader(T['search_title'])
356
  c1, c2 = st.columns([3, 1])
@@ -373,12 +346,9 @@ if menu == T['menu_search']:
373
  h, src = get_chart_data(ticker)
374
  news_data = get_polyglot_news(eng_key, native_key)
375
  curr, pct = "N/A", 0
376
-
377
  if not h.empty:
378
- curr = h['Close'].iloc[-1]
379
- if len(h) >= 2:
380
- prev = h['Close'].iloc[-2]
381
- if prev != 0: pct = ((curr - prev)/prev)*100
382
  else:
383
  c_web, p_web, s_web = get_price_data_robust(ticker, name)
384
  if c_web: curr, pct, src = c_web, p_web, s_web
@@ -389,7 +359,6 @@ if menu == T['menu_search']:
389
 
390
  if not h.empty: st.plotly_chart(plot_candle_chart(h, name, src), use_container_width=True)
391
  else: st.warning(f"Chart N/A")
392
-
393
  with st.expander(T['expander_news']): st.text(news_data)
394
 
395
  collected_data.append(f"[TARGET {i+1}] {name} ({ticker})\nPrice: {curr}, Change: {pct:.2f}%\nNews: {news_data[:500]}...")
@@ -398,27 +367,17 @@ if menu == T['menu_search']:
398
  st.divider()
399
  st.subheader(T['insight_title'])
400
 
401
- target_lang = T['llm_lang_instruction']
402
-
403
- # [NEW] ํ”„๋กฌํ”„ํŠธ์— ์‹œ์žฅ ์ƒํ™ฉ ์ฃผ์ž… (AI Context Injection)
404
  prompt = f"""
405
- [ROLE] Global Investment Analyst.
406
- [CURRENT MARKET CONTEXT] {market_context_str} (Consider this global sentiment).
407
- [USER QUERY] "{query}"
408
- [DATA OF TARGETS] {chr(10).join(collected_data)}
409
-
410
- [TASK]
411
- 1. Compare targets based on data and news.
412
- 2. Evaluate their 'Relative Strength' against the market context (e.g., Is the stock strong while S&P 500 is weak?).
413
- 3. Provide a verdict.
414
-
415
- [OUTPUT LANGUAGE] **{target_lang}** (Strictly output in {target_lang}).
416
  """
417
  msg = [{"role": "user", "content": prompt}]
418
  stream = client.chat.completions.create(model="Qwen/Qwen2.5-72B-Instruct", messages=msg, stream=True)
419
  st.write_stream(parse_stream(stream))
420
  else: st.error("AI Error")
421
 
 
422
  elif menu == T['menu_scanner']:
423
  st.subheader(T['scanner_title'])
424
  st.caption(T['scanner_caption'])
@@ -438,24 +397,14 @@ elif menu == T['menu_scanner']:
438
  for i, t in enumerate(tickers):
439
  name_display = TICKER_NAMES.get(t, t)
440
  p, c, s = None, None, None
441
-
442
  if not batch_data.empty and t in batch_data.columns:
443
  series = batch_data[t].dropna()
444
  if len(series) >= 2:
445
- p = series.iloc[-1]
446
- prev = series.iloc[-2]
447
- if prev != 0:
448
- c = ((p - prev)/prev)*100
449
- s = "Yahoo (Batch)"
450
-
451
- if p is None:
452
- p, c, s = get_price_data_robust(t, name_display)
453
-
454
  if p is not None:
455
- results.append({
456
- T['col_name']: name_display, 'Ticker': t,
457
- T['col_price']: p, 'Change(%)': c, T['col_source']: s
458
- })
459
  bar.progress((i+1)/len(tickers))
460
 
461
  if results:
@@ -463,10 +412,41 @@ elif menu == T['menu_scanner']:
463
  st.plotly_chart(plot_bar_chart(df, T), use_container_width=True)
464
 
465
  disp_cols = [T['col_name'], 'Ticker', T['col_price'], 'Change(%)', T['col_source']]
466
-
467
  df_gainers = df[df['Change(%)'] > 0].head(10)
468
  df_losers = df[df['Change(%)'] < 0].sort_values('Change(%)', ascending=True).head(10)
469
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
470
  c_up, c_down = st.columns(2)
471
  with c_up:
472
  st.success(T['tab_gainers'])
 
22
  st.stop()
23
 
24
  # ==============================================================================
25
+ # ๐ŸŒ [UI] ๋‹ค๊ตญ์–ด ๋”•์…”๋„ˆ๋ฆฌ
26
  # ==============================================================================
27
  UI_TEXT = {
28
  "KR": {
 
33
  "search_title": "๐Ÿง  ๋‹ค๊ตญ์–ด AI ์ฃผ์‹ ๋น„์„œ (๋ฉ€ํ‹ฐ ํƒ€๊ฒŸ)",
34
  "search_placeholder": "์˜ˆ: ์‚ผ์„ฑ์ „์ž๋ž‘ TSMC ๋น„๊ตํ•ด์ค˜, ๋ชฝ๊ณจ ๊ณ ๋น„ ์บ์‹œ๋ฏธ์–ด...",
35
  "btn_analyze": "๋ถ„์„ ์‹œ์ž‘",
36
+ "status_thinking": "๐Ÿง  AI๊ฐ€ ๋ถ„์„ ์ค‘์ž…๋‹ˆ๋‹ค...",
37
  "status_news": "๐Ÿ“ฐ ๋‰ด์Šค ์ˆ˜์ง‘ ์ค‘...",
38
  "metric_price": "ํ˜„์žฌ๊ฐ€",
39
  "metric_source": "์ถœ์ฒ˜",
40
  "expander_news": "๐Ÿ“ฐ ๋‰ด์Šค ์›๋ฌธ ๋ณด๊ธฐ",
41
  "insight_title": "๐Ÿง  AI ์ข…ํ•ฉ ๋ถ„์„ ๋ฆฌํฌํŠธ",
42
  "scanner_title": "๐Ÿ“ก ๊ธ€๋กœ๋ฒŒ ์‹œ์žฅ ์ „๊ด‘ํŒ (Frontier Included)",
43
+ "scanner_caption": "โ€ป 100๊ฐœ ๊ธฐ์—… ์Šค์บ” ํ›„, AI๊ฐ€ ์‹œ์žฅ ์ „์ฒด๋ฅผ ์š”์•ฝํ•˜๊ณ  ์œ ๋ง ์ข…๋ชฉ์„ ๋ธŒ๋ฆฌํ•‘ํ•ฉ๋‹ˆ๋‹ค.",
44
+ "btn_scan": "์ „์ฒด ์ข…๋ชฉ ์Šค์บ” & AI ๋ฆฌํฌํŠธ ์ƒ์„ฑ",
45
  "tab_gainers": "๐Ÿ”ฅ ๊ธ‰๋“ฑ Top 10",
46
  "tab_losers": "๐Ÿ’ง ๊ธ‰๋ฝ Top 10",
47
  "col_name": "๊ธฐ์—…๋ช…",
48
  "col_price": "๊ฐ€๊ฒฉ",
49
  "col_change": "๋“ฑ๋ฝ๋ฅ ",
50
  "col_source": "์ถœ์ฒ˜",
51
+ "briefing_title": "๐Ÿง  AI Market Briefing (์˜ค๋Š˜์˜ ์‹œ์žฅ ๋ถ„์„)",
52
  "msg_fail": "๋ฐ์ดํ„ฐ ์ˆ˜์ง‘ ์‹คํŒจ",
53
  "llm_lang_instruction": "Korean"
54
  },
 
60
  "search_title": "๐Ÿง  AI Stock Assistant (Multi-Target)",
61
  "search_placeholder": "e.g. Compare Samsung vs TSMC...",
62
  "btn_analyze": "Analyze",
63
+ "status_thinking": "๐Ÿง  AI is analyzing...",
64
  "status_news": "๐Ÿ“ฐ Fetching news...",
65
  "metric_price": "Price",
66
  "metric_source": "Source",
67
  "expander_news": "๐Ÿ“ฐ View Source News",
68
  "insight_title": "๐Ÿง  AI Comprehensive Report",
69
  "scanner_title": "๐Ÿ“ก Global Market Scanner (Frontier Included)",
70
+ "scanner_caption": "โ€ป Scans 100+ stocks and generates an AI market briefing.",
71
+ "btn_scan": "Scan & Generate AI Report",
72
  "tab_gainers": "๐Ÿ”ฅ Top 10 Gainers",
73
  "tab_losers": "๐Ÿ’ง Top 10 Losers",
74
  "col_name": "Company",
75
  "col_price": "Price",
76
  "col_change": "Change(%)",
77
  "col_source": "Source",
78
+ "briefing_title": "๐Ÿง  AI Market Briefing",
79
  "msg_fail": "Data collection failed",
80
  "llm_lang_instruction": "English"
81
  }
82
  }
83
 
84
  # ==============================================================================
85
+ # ๐Ÿ’พ [๋ฐ์ดํ„ฐ] ํ‹ฐ์ปค ๋งคํ•‘ (Full List)
86
  # ==============================================================================
87
  TICKER_NAMES = {
88
  # ๐Ÿ‡ญ๐Ÿ‡ฐ ํ™์ฝฉ (HK)
89
+ "1299.HK": "AIA Group", "0388.HK": "HKEX", "0005.HK": "HSBC Holdings",
90
  "0016.HK": "SHK Properties", "1113.HK": "CK Asset", "0002.HK": "CLP Holdings",
91
  "0003.HK": "HK & China Gas", "0066.HK": "MTR Corp", "0011.HK": "Hang Seng Bank",
92
  "0027.HK": "Galaxy Ent", "0019.HK": "Swire Pacific", "9988.HK": "Alibaba",
 
173
  }
174
 
175
  # ==============================================================================
176
+ # ๐Ÿ› ๏ธ [์—”์ง„] ์ง€์ˆ˜ & ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘
177
  # ==============================================================================
178
  def get_market_indices():
179
+ indices = {"S&P 500": "^GSPC", "NASDAQ": "^IXIC", "KOSPI": "^KS11", "VN-INDEX": "^VNINDEX"}
 
 
 
 
 
 
 
 
180
  data = {}
181
  try:
182
+ df = yf.download(list(indices.values()), period="5d", progress=False)['Close']
 
 
183
  for name, ticker in indices.items():
184
  if ticker in df.columns:
185
  series = df[ticker].dropna()
186
  if len(series) >= 2:
187
+ curr, prev = series.iloc[-1], series.iloc[-2]
188
+ data[name] = (curr, ((curr - prev)/prev)*100)
 
 
189
  except: pass
190
  return data
191
 
 
 
 
192
  FAKE_HEADERS = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36'}
193
 
194
  def fetch_news_robust(keyword):
 
197
  try:
198
  with DDGS() as ddgs:
199
  results = list(ddgs.news(keyword, timelimit="m", max_results=3))
200
+ if not results: results = list(ddgs.news(keyword, max_results=3))
201
+ if not results: results = list(ddgs.text(f"{keyword} latest news", max_results=3))
 
 
 
 
202
  for r in results:
203
  url = r.get('url') or r.get('href')
204
  if url not in seen_urls:
205
+ summary += f"[{r.get('date', '?')}] {r.get('title', '')}\n"
 
 
206
  seen_urls.add(url)
207
  except: return "No news."
208
  return summary if summary else "No news."
 
227
  elif ".KS" in ticker or ".KQ" in ticker: fdr_symbol = ticker.split('.')[0]
228
  hist = fdr.DataReader(fdr_symbol, start=(datetime.datetime.now() - datetime.timedelta(days=7)))
229
  if not hist.empty:
230
+ curr, prev = hist['Close'].iloc[-1], hist['Close'].iloc[-2]
231
+ return curr, ((curr - prev)/prev)*100, "FDR"
 
 
 
232
  except: pass
233
  try:
234
  stock = yf.Ticker(ticker)
 
246
  if ".VN" in ticker: fdr_symbol = ticker.split('.')[0]
247
  elif ".JK" in ticker: fdr_symbol = f"IDX:{ticker.split('.')[0]}"
248
  elif ".T" in ticker: fdr_symbol = f"TSE:{ticker.split('.')[0]}"
249
+ hist = fdr.DataReader(fdr_symbol, start=(datetime.datetime.now() - datetime.timedelta(days=90)).strftime('%Y-%m-%d'))
 
250
  if not hist.empty: return hist.rename(columns={'Close':'Close', 'Open':'Open', 'High':'High', 'Low':'Low'}), "FDR"
251
  except: pass
252
  try:
 
295
  T = UI_TEXT[lang_code]
296
  st.title(T['sidebar_title'])
297
 
 
298
  st.divider()
299
  st.caption(T['sidebar_weather'])
 
300
  market_data = get_market_indices()
301
+ market_context_str = ""
302
 
303
  if market_data:
304
+ c1, c2 = st.columns(2)
 
305
  if "S&P 500" in market_data:
306
  p, c = market_data["S&P 500"]
307
+ c1.metric("๐Ÿ‡บ๐Ÿ‡ธ S&P 500", f"{p:,.0f}", f"{c:+.2f}%")
308
+ market_context_str += f"S&P 500 {c:+.2f}%, "
309
  if "NASDAQ" in market_data:
310
  p, c = market_data["NASDAQ"]
311
+ c2.metric("๐Ÿ‡บ๐Ÿ‡ธ NASDAQ", f"{p:,.0f}", f"{c:+.2f}%")
312
+ market_context_str += f"NASDAQ {c:+.2f}%, "
313
+ c3, c4 = st.columns(2)
 
314
  if "KOSPI" in market_data:
315
  p, c = market_data["KOSPI"]
316
+ c3.metric("๐Ÿ‡ฐ๐Ÿ‡ท KOSPI", f"{p:,.0f}", f"{c:+.2f}%")
317
  if "VN-INDEX" in market_data:
318
  p, c = market_data["VN-INDEX"]
319
+ c4.metric("๐Ÿ‡ป๐Ÿ‡ณ VN-IDX", f"{p:,.0f}", f"{c:+.2f}%")
320
+ market_context_str += f"Vietnam Index {c:+.2f}%."
321
+ else: st.caption("Loading failed.")
 
322
 
323
  st.divider()
324
  menu = st.radio("MENU", [T['menu_search'], T['menu_scanner']], index=0)
325
 
326
+ # --- AI ๊ฒ€์ƒ‰ ---
327
  if menu == T['menu_search']:
328
  st.subheader(T['search_title'])
329
  c1, c2 = st.columns([3, 1])
 
346
  h, src = get_chart_data(ticker)
347
  news_data = get_polyglot_news(eng_key, native_key)
348
  curr, pct = "N/A", 0
 
349
  if not h.empty:
350
+ curr, prev = h['Close'].iloc[-1], h['Close'].iloc[-2]
351
+ pct = ((curr - prev)/prev)*100
 
 
352
  else:
353
  c_web, p_web, s_web = get_price_data_robust(ticker, name)
354
  if c_web: curr, pct, src = c_web, p_web, s_web
 
359
 
360
  if not h.empty: st.plotly_chart(plot_candle_chart(h, name, src), use_container_width=True)
361
  else: st.warning(f"Chart N/A")
 
362
  with st.expander(T['expander_news']): st.text(news_data)
363
 
364
  collected_data.append(f"[TARGET {i+1}] {name} ({ticker})\nPrice: {curr}, Change: {pct:.2f}%\nNews: {news_data[:500]}...")
 
367
  st.divider()
368
  st.subheader(T['insight_title'])
369
 
 
 
 
370
  prompt = f"""
371
+ [ROLE] Global Analyst. [MARKET CONTEXT] {market_context_str}
372
+ [USER QUERY] "{query}" [DATA] {chr(10).join(collected_data)}
373
+ [TASK] Comparative Analysis & Verdict (Consider Relative Strength). [LANG] {T['llm_lang_instruction']}.
 
 
 
 
 
 
 
 
374
  """
375
  msg = [{"role": "user", "content": prompt}]
376
  stream = client.chat.completions.create(model="Qwen/Qwen2.5-72B-Instruct", messages=msg, stream=True)
377
  st.write_stream(parse_stream(stream))
378
  else: st.error("AI Error")
379
 
380
+ # --- ์‹œ์žฅ ์Šค์บ๋„ˆ & ๋ฆฌํฌํŠธ ---
381
  elif menu == T['menu_scanner']:
382
  st.subheader(T['scanner_title'])
383
  st.caption(T['scanner_caption'])
 
397
  for i, t in enumerate(tickers):
398
  name_display = TICKER_NAMES.get(t, t)
399
  p, c, s = None, None, None
 
400
  if not batch_data.empty and t in batch_data.columns:
401
  series = batch_data[t].dropna()
402
  if len(series) >= 2:
403
+ p, prev = series.iloc[-1], series.iloc[-2]
404
+ c, s = ((p - prev)/prev)*100, "Yahoo (Batch)"
405
+ if p is None: p, c, s = get_price_data_robust(t, name_display)
 
 
 
 
 
 
406
  if p is not None:
407
+ results.append({T['col_name']: name_display, 'Ticker': t, T['col_price']: p, 'Change(%)': c, T['col_source']: s})
 
 
 
408
  bar.progress((i+1)/len(tickers))
409
 
410
  if results:
 
412
  st.plotly_chart(plot_bar_chart(df, T), use_container_width=True)
413
 
414
  disp_cols = [T['col_name'], 'Ticker', T['col_price'], 'Change(%)', T['col_source']]
 
415
  df_gainers = df[df['Change(%)'] > 0].head(10)
416
  df_losers = df[df['Change(%)'] < 0].sort_values('Change(%)', ascending=True).head(10)
417
 
418
+ # [NEW] AI Market Briefing (์‹œ์žฅ ์š”์•ฝ ๋ฆฌํฌํŠธ)
419
+ st.divider()
420
+ st.subheader(T['briefing_title'])
421
+
422
+ # AI์—๊ฒŒ ๋˜์งˆ ๋ฐ์ดํ„ฐ ์š”์•ฝ
423
+ gainers_str = ", ".join([f"{r[T['col_name']]} ({r['Change(%)']:.2f}%)" for _, r in df_gainers.head(5).iterrows()])
424
+ losers_str = ", ".join([f"{r[T['col_name']]} ({r['Change(%)']:.2f}%)" for _, r in df_losers.head(5).iterrows()])
425
+
426
+ prompt = f"""
427
+ [ROLE] Chief Market Strategist.
428
+ [TARGET MARKET] {target}
429
+ [GLOBAL CONTEXT] {market_context_str}
430
+ [SCAN DATA]
431
+ - Top Gainers: {gainers_str}
432
+ - Top Losers: {losers_str}
433
+
434
+ [TASK]
435
+ 1. **Market Summary**: Briefly describe today's market sentiment based on gainers/losers.
436
+ 2. **Sector Insight**: Why are these stocks moving? (e.g., Tech rally, Energy dip).
437
+ 3. **Top Pick**: Recommend 1 stock from the gainers with strong momentum.
438
+ 4. **Caution**: Warn about 1 stock/sector from the losers.
439
+
440
+ [OUTPUT LANGUAGE] **{T['llm_lang_instruction']}** (Use professional financial tone).
441
+ """
442
+
443
+ # ๋ฆฌํฌํŠธ ์ŠคํŠธ๋ฆฌ๋ฐ
444
+ msg = [{"role": "user", "content": prompt}]
445
+ stream = client.chat.completions.create(model="Qwen/Qwen2.5-72B-Instruct", messages=msg, stream=True)
446
+ st.write_stream(parse_stream(stream))
447
+ st.divider()
448
+
449
+ # ๊ธฐ์กด ๋ฐ์ดํ„ฐํ”„๋ ˆ์ž„ ํ‘œ์‹œ
450
  c_up, c_down = st.columns(2)
451
  with c_up:
452
  st.success(T['tab_gainers'])