Upload app.py

app.py

@@ -0,0 +1,1078 @@
+# HUGING_FACE_V3.1.2.py (整合 Bert_predict 和 XGBoost 版本)
+
+# 系統套件
+import os
+from datetime import datetime, timedelta
+import google.generativeai as genai 
+import pandas as pd
+import numpy as np
+import yfinance as yf
+from dash import Dash, dcc, html, callback
+import dash
+import plotly.express as px
+import plotly.graph_objects as go
+from plotly.subplots import make_subplots
+import re
+from bs4 import BeautifulSoup
+import requests
+import time # 引用 time 模組以處理時間戳
+
+# ========================= 引用外部模組 START =========================
+# 引用您組員的預測器程式
+from Bert_predict import BertPredictor
+
+# 引用新的模型預測器
+from model_predictor import XGBoostModel
+# ========================== 引用外部模組 END ==========================
+
+# ========================= 全域設定 START =========================
+# 【【【模型切換開關】】】
+# False: 使用簡易統計模型 (預設)
+# True:  使用 model_predictor.py 中的進階 XGBoost 模型
+USE_ADVANCED_MODEL = False
+
+# ========================= CACHE 設定 START =========================
+# 分析結果的快取字典
+ANALYSIS_CACHE = {}
+# 快取有效時間（秒），例如：4 小時 = 4 * 60 * 60 = 14400 秒
+CACHE_DURATION_SECONDS = 8 * 60 * 60
+# ========================== CACHE 設定 END ==========================
+# ========================== 全域設定 END ==========================
+
+# 台股代號對應表 (移除台指期，因為它現在是獨立個塊)
+TAIWAN_STOCKS = {
+    '元大台灣50': '0050.TW',
+    '台積電': '2330.TW',
+    '聯發科': '2454.TW',
+    '鴻海': '2317.TW',
+    '台達電': '2308.TW',
+    '廣達': '2382.TW',
+    '富邦金': '2881.TW',
+    '中信金': '2891.TW',
+    '國泰金': '2882.TW',
+    '聯電': '2303.TW',
+    '中華電': '2412.TW',
+    '玉山金': '2884.TW',
+    '兆豐金': '2886.TW',
+    '日月光投控': '3711.TW',
+    '華碩': '2357.TW',
+    '統一': '1216.TW',
+    '元大金': '2885.TW',
+    '智邦': '2345.TW',
+    '緯創': '3231.TW',
+    '聯詠': '3034.TW',
+    '第一金': '2892.TW',
+    '瑞昱': '2379.TW',
+    '緯穎': '6669.TWO',
+    '永豐金': '2890.TW',
+    '合庫金': '5880.TW',
+    '華南金': '2880.TW',
+    '台光電': '2383.TW',
+    '世芯-KY': '3661.TWO',
+    '奇鋐': '3017.TW',
+    '凱基金': '2883.TW',
+    '大立光': '3008.TW',
+    '長榮': '2603.TW',
+    '光寶科': '2301.TW',
+    '中鋼': '2002.TW',
+    '中租-KY': '5871.TW',
+    '國巨': '2327.TW',
+    '台新金': '2887.TW',
+    '上海商銀': '5876.TW',
+    '台泥': '1101.TW',
+    '台灣大': '3045.TW',
+    '和碩': '4938.TW',
+    '遠傳': '4904.TW',
+    '和泰車': '2207.TW',
+    '研華': '2395.TW',
+    '台塑': '1301.TW',
+    '統一超': '2912.TW',
+    '藥華藥': '6446.TWO',
+    '南亞': '1303.TW',
+    '陽明': '2609.TW',
+    '萬海': '2615.TW',
+    '台塑化': '6505.TW',
+    '慧洋-KY': '2637.TW',
+    '上銀': '2049.TW',
+    '南亞科': '2408.TW',
+    '旺宏': '2337.TW',
+    '譜瑞-KY': '4966.TWO',
+    '貿聯-KY': '3665.TW',
+    '驊訊': '6870.TWO',
+    '穩懋': '3105.TWO'
+}
+
+# 產業分類
+INDUSTRY_MAPPING = {
+    '0050.TW': 'ETF',
+    '2330.TW': '半導體',
+    '2454.TW': '半導體',
+    '2317.TW': '電子組件',
+    '2308.TW': '電子',
+    '2382.TW': '電子',
+    '2881.TW': '金融',
+    '2891.TW': '金融',
+    '2882.TW': '金融',
+    '2303.TW': '半導體',
+    '2412.TW': '電信',
+    '2884.TW': '金融',
+    '2886.TW': '金融',
+    '3711.TW': '半導體',
+    '2357.TW': '電子',
+    '1216.TW': '食品',
+    '2885.TW': '金融',
+    '2345.TW': '網通設備',
+    '3231.TW': '電子',
+    '3034.TW': '半導體',
+    '2892.TW': '金融',
+    '2379.TW': '半導體',
+    '6669.TWO': '電子',
+    '2890.TW': '金融',
+    '5880.TW': '金融',
+    '2880.TW': '金融',
+    '2383.TW': '電子',
+    '3661.TWO': '半導體',
+    '3017.TW': '電子',
+    '2883.TW': '金融',
+    '3008.TW': '光學',
+    '2603.TW': '航運',
+    '2301.TW': '電子',
+    '2002.TW': '鋼鐵',
+    '5871.TW': '金融',
+    '2327.TW': '電子被動元件',
+    '2887.TW': '金融',
+    '5876.TW': '金融',
+    '1101.TW': '營建',
+    '3045.TW': '電信',
+    '4938.TW': '電子',
+    '4904.TW': '電信',
+    '2207.TW': '汽車',
+    '2395.TW': '電腦周邊',
+    '1301.TW': '塑膠',
+    '2912.TW': '百貨',
+    '6446.TWO': '生技',
+    '1303.TW': '塑膠',
+    '2609.TW': '航運',
+    '2615.TW': '航運',
+    '6505.TW': '塑膠',
+    '2637.TW': '散裝航運',
+    '2049.TW': '工具機',
+    '2408.TW': 'DRAM',
+    '2337.TW': 'NFLSH',
+    '4966.TWO': '高速傳輸',
+    '3665.TW': '連接器',
+    '6870.TWO': '軟體整合',
+    '3105.TWO': 'PA功率'
+}
+
+def get_stock_data(symbol, period='1y'):
+    """獲取股票資料"""
+    try:
+        stock = yf.Ticker(symbol)
+        data = stock.history(period=period)
+        if data.empty and symbol == 'TXF=F':
+            stock = yf.Ticker('0050.TW')
+            data = stock.history(period=period)
+            if data.empty:
+                stock = yf.Ticker('^TWII')
+                data = stock.history(period=period)
+        return data
+    except:
+        return pd.DataFrame()
+
+def get_us_market_data():
+    """獲取美股指數數據"""
+    try:
+        indices = {
+            'DJI': '^DJI',      # 道瓊指數
+            'NAS': '^IXIC',     # 那斯達克
+            'SOX': '^SOX',      # 費城半導體
+            'S&P_500': '^GSPC', # S&P 500
+            'TSM_ADR': 'TSM'    # 台積電ADR
+        }
+        
+        market_data = {}
+        for name, symbol in indices.items():
+            try:
+                data = yf.Ticker(symbol).history(period='5d')
+                if not data.empty:
+                    market_data[name] = data['Close'].iloc[-1]
+                else:
+                    market_data[name] = 0
+            except:
+                market_data[name] = 0
+        
+        return market_data
+    except Exception as e:
+        print(f"獲取美股數據時發生錯誤: {e}")
+        return {'DJI': 0, 'NAS': 0, 'SOX': 0, 'S&P_500': 0, 'TSM_ADR': 0}
+
+def get_exchange_rate():
+    """獲取台幣匯率 (USD/TWD)"""
+    try:
+        data = yf.Ticker('USDTWD=X').history(period='5d')
+        if not data.empty:
+            return data['Close'].iloc[-1]
+        else:
+            return 31.5  # 預設值
+    except:
+        return 31.5
+
+def simple_statistical_predict(data, predict_days=5):
+    """【備用模型】簡化的統計預測模型。"""
+    if len(data) < 60: 
+        return None
+        
+    prices = data['Close'].values
+    ma_short = np.mean(prices[-5:])
+    ma_medium = np.mean(prices[-20:])
+    ma_long = np.mean(prices[-60:])
+    recent_trend = np.polyfit(range(20), prices[-20:], 1)[0]
+    volatility = np.std(prices[-20:]) / np.mean(prices[-20:])
+    base_change = recent_trend * predict_days
+    trend_factor = 1.0 + (0.02 if ma_short > ma_medium > ma_long else -0.02 if ma_short < ma_medium < ma_long else 0)
+    noise_factor = np.random.normal(1, volatility * 0.1)
+    predicted_price = prices[-1] * trend_factor + base_change + (prices[-1] * noise_factor * 0.01)
+    change_pct = ((predicted_price - prices[-1]) / prices[-1]) * 100
+    return {
+        'predicted_price': predicted_price, 
+        'change_pct': change_pct, 
+        'confidence': max(0.6, 1 - volatility * 2)
+    }
+
+def advanced_xgboost_predict(predict_days=5):
+    """
+    【進階模型】使用 XGBoost 模型進行預測
+    """
+    try:
+        print(f"開始使用 XGBoost 模型進行 {predict_days} 天預測...")
+        
+        # 初始化 XGBoost 模型
+        xgb_model = XGBoostModel()
+        
+        # 獲取台指期數據 (作為主要標的)
+        taiex_data = get_stock_data('^TWII', '2y')
+        if taiex_data.empty or len(taiex_data) < 60:
+            print("台指期數據不足，無法進行XGBoost預測")
+            return None
+        
+        # 計算技術指標
+        taiex_data = calculate_technical_indicators(taiex_data)
+        
+        # 獲取美股市場數據
+        us_market = get_us_market_data()
+        
+        # 獲取匯率數據
+        exchange_rate = get_exchange_rate()
+        
+        # 獲取新聞情緒分數
+        try:
+            if predictor is not None:
+                sentiment_score_raw = predictor.get_news_index()
+                if sentiment_score_raw is None:
+                    sentiment_score_raw = 0
+            else:
+                sentiment_score_raw = 0
+        except:
+            sentiment_score_raw = 0
+        
+        # 準備特徵數據 (使用最新的數據點)
+        latest_data = taiex_data.iloc[-1]
+        
+        # 建立特徵向量 (按照訓練數據記錄的順序)
+        features_list = [
+            latest_data['Close'],  # close
+            latest_data['Volume'],  # volume
+            exchange_rate,  # rate
+            us_market['DJI'],  # DJI
+            us_market['NAS'],  # NAS
+            us_market['SOX'],  # SOX
+            us_market['S&P_500'],  # S&P_500
+            us_market['TSM_ADR'],  # TSM_ADR
+            sentiment_score_raw,  # NEWS (使用原始 sentiment_score_raw)
+            latest_data['RSI'] if not pd.isna(latest_data['RSI']) else 50,  # RSI
+            latest_data['MACD'] if not pd.isna(latest_data['MACD']) else 0,  # MACD
+            latest_data['MACD_Signal'] if not pd.isna(latest_data['MACD_Signal']) else 0,  # MACDsign
+            latest_data['MACD_Histogram'] if not pd.isna(latest_data['MACD_Histogram']) else 0,  # MACDvol
+            latest_data['K'] if not pd.isna(latest_data['K']) else 50,  # K
+            latest_data['D'] if not pd.isna(latest_data['D']) else 50,  # D
+            latest_data['+DI'] if not pd.isna(latest_data['+DI']) else 25,  # +DI
+            latest_data['-DI'] if not pd.isna(latest_data['-DI']) else 25,  # -DI
+            latest_data['ADX'] if not pd.isna(latest_data['ADX']) else 25,  # ADX
+            15,    # business_climate (手動填入值)
+            46.7   # PMI (手動填入值)
+        ]
+        
+        # 轉換為 DataFrame (XGBoost 模型期望的格式)
+        column_names = [
+            'close', 'volume', 'rate', 'DJI', 'NAS', 'SOX', 'S&P_500', 'TSM_ADR',
+            'NEWS', 'RSI', 'MACD', 'MACDsign', 'MACDvol', 'K', 'D', 
+            '+DI', '-DI', 'ADX', 'business_climate', 'PMI'
+        ]
+        
+        input_df = pd.DataFrame([features_list], columns=column_names)
+        
+        print("特徵數據準備完成:")
+        print(f"- 收盤價: {features_list[0]:.2f}")
+        print(f"- 成交量: {features_list[1]:,.0f}")
+        print(f"- 匯率: {features_list[2]:.2f}")
+        print(f"- 道瓊指數: {features_list[3]:.2f}")
+        print(f"- 新聞情緒: {features_list[8]:.4f}")
+        print(f"- RSI: {features_list[9]:.2f}")
+        
+        # 進行預測
+        predictions = xgb_model.predict('xgboost_model', input_df)
+        
+        # 根據預測天數選擇對應的預測值
+        pred_mapping = {
+            1: 'Close_t0_pred',
+            5: 'Close_t5_pred', 
+            10: 'Close_t10_pred',
+            20: 'Close_t20_pred'
+        }
+        
+        # 找到最接近的預測天數
+        available_days = [1, 5, 10, 20]
+        closest_day = min(available_days, key=lambda x: abs(x - predict_days))
+        pred_key = pred_mapping[closest_day]
+        
+        predicted_price = predictions[pred_key]
+        current_price = features_list[0]  # close price
+        change_pct = ((predicted_price - current_price) / current_price) * 100
+        
+        print(f"XGBoost 預測完成:")
+        print(f"- 預測天數: {predict_days} (使用 {closest_day} 天模型)")
+        print(f"- 當前價格: {current_price:.2f}")
+        print(f"- 預測價格: {predicted_price:.2f}")
+        print(f"- 預測變化: {change_pct:+.2f}%")
+        
+        return {
+            'predicted_price': predicted_price,
+            'change_pct': change_pct,
+            'confidence': 0.75  # XGBoost 模型的信心度
+        }
+        
+    except Exception as e:
+        print(f"XGBoost 預測時發生錯誤: {e}")
+        import traceback
+        traceback.print_exc()
+        return None
+
+def get_prediction(data, predict_days=5):
+    """
+    【【模型預測控制器】】
+    根據 USE_ADVANCED_MODEL 的設定，呼叫對應的預測模型。
+    """
+    if USE_ADVANCED_MODEL:
+        print(f"模式: 進階XGBoost模型 | 預測天期: {predict_days}天")
+        prediction = advanced_xgboost_predict(predict_days)
+        # 如果進階模型預測失敗，則自動降級使用簡易模型
+        if prediction is not None:
+            return prediction
+        else:
+            print("進階模型預測失敗，自動降級為簡易統計模型。")
+    
+    # 預設或降級時執行簡易模型
+    print(f"模式: 簡易統計模型 | 預測天期: {predict_days}天")
+    return simple_statistical_predict(data, predict_days)
+
+def calculate_technical_indicators(df):
+    """計算技術指標"""
+    if df.empty: return df
+    
+    # 移動平均線
+    df['MA5'] = df['Close'].rolling(window=5).mean()
+    df['MA20'] = df['Close'].rolling(window=20).mean()
+    
+    # RSI
+    delta = df['Close'].diff()
+    gain = (delta.where(delta > 0, 0)).rolling(window=14).mean()
+    loss = (-delta.where(delta < 0, 0)).rolling(window=14).mean()
+    rs = gain / loss
+    df['RSI'] = 100 - (100 / (1 + rs))
+    
+    # MACD
+    exp1 = df['Close'].ewm(span=12).mean()
+    exp2 = df['Close'].ewm(span=26).mean()
+    df['MACD'] = exp1 - exp2
+    df['MACD_Signal'] = df['MACD'].ewm(span=9).mean()
+    df['MACD_Histogram'] = df['MACD'] - df['MACD_Signal']
+    
+    # 布林通道
+    df['BB_Middle'] = df['Close'].rolling(window=20).mean()
+    bb_std = df['Close'].rolling(window=20).std()
+    df['BB_Upper'] = df['BB_Middle'] + (bb_std * 2)
+    df['BB_Lower'] = df['BB_Middle'] - (bb_std * 2)
+    
+    # KD 指標
+    low_min = df['Low'].rolling(window=9).min()
+    high_max = df['High'].rolling(window=9).max()
+    rsv = (df['Close'] - low_min) / (high_max - low_min) * 100
+    df['K'] = rsv.ewm(com=2).mean()
+    df['D'] = df['K'].ewm(com=2).mean()
+    
+    # 威廉指標
+    low_min_14 = df['Low'].rolling(window=14).min()
+    high_max_14 = df['High'].rolling(window=14).max()
+    df['Williams_R'] = -100 * (high_max_14 - df['Close']) / (high_max_14 - low_min_14)
+    
+    # DMI 指標
+    df['up_move'] = df['High'] - df['High'].shift(1)
+    df['down_move'] = df['Low'].shift(1) - df['Low']
+    df['+DM'] = np.where((df['up_move'] > df['down_move']) & (df['up_move'] > 0), df['up_move'], 0)
+    df['-DM'] = np.where((df['down_move'] > df['up_move']) & (df['down_move'] > 0), df['down_move'], 0)
+    df['TR'] = np.max([df['High'] - df['Low'], abs(df['High'] - df['Close'].shift(1)), abs(df['Low'] - df['Close'].shift(1))], axis=0)
+    df['+DI'] = (df['+DM'].ewm(com=13, adjust=False).mean() / df['TR'].ewm(com=13, adjust=False).mean()) * 100
+    df['-DI'] = (df['-DM'].ewm(com=13, adjust=False).mean() / df['TR'].ewm(com=13, adjust=False).mean()) * 100
+    df['DX'] = abs(df['+DI'] - df['-DI']) / (df['+DI'] + df['-DI']) * 100
+    df['ADX'] = df['DX'].ewm(com=13, adjust=False).mean()
+    
+    return df
+
+def calculate_volume_profile(df, num_bins=50):
+    if df.empty or 'High' not in df.columns or 'Low' not in df.columns or 'Volume' not in df.columns: 
+        return None, None, None
+    all_prices = np.concatenate([df['High'].values, df['Low'].values])
+    min_price, max_price = all_prices.min(), all_prices.max()
+    price_for_volume = (df['High'] + df['Low'] + df['Close']) / 3
+    df_vol_profile = df.copy()
+    df_vol_profile['Price_Indicator'] = price_for_volume
+    hist, bin_edges = np.histogram(df_vol_profile['Price_Indicator'], bins=num_bins, range=(min_price, max_price), weights=df_vol_profile['Volume'])
+    price_centers = (bin_edges[:-1] + bin_edges[1:]) / 2
+    return bin_edges, hist, price_centers
+
+def get_business_climate_data():
+    try:
+        if not os.path.exists('business_climate.csv'): return pd.DataFrame()
+        df = pd.read_csv('business_climate.csv')
+        if 'Date' not in df.columns: df.columns = ['Date', 'Index'] if len(df.columns) == 2 else df.columns
+        if 'Date' in df.columns:
+            try: df['Date'] = pd.to_datetime(df['Date'] + '-01', format='%Y-%m-%d', errors='coerce')
+            except: df['Date'] = pd.to_datetime(df['Date'], errors='coerce')
+        df = df.dropna(subset=['Date'])
+        return df
+    except Exception as e:
+        print(f"無法獲取景氣燈號資料: {str(e)}")
+        return pd.DataFrame()
+
+def get_pmi_data():
+    try:
+        if not os.path.exists('taiwan_pmi.csv'): return pd.DataFrame()
+        df = pd.read_csv('taiwan_pmi.csv')
+        if 'DATE' in df.columns: df = df.rename(columns={'DATE': 'Date', 'INDEX': 'Index'})
+        elif len(df.columns) == 2: df.columns = ['Date', 'Index']
+        if 'Date' in df.columns:
+            try: df['Date'] = pd.to_datetime(df['Date'] + '-01', format='%Y-%m-%d', errors='coerce')
+            except: df['Date'] = pd.to_datetime(df['Date'], errors='coerce')
+        df = df.dropna(subset=['Date'])
+        return df
+    except Exception as e:
+        print(f"無法獲取 PMI 資料: {str(e)}")
+        return pd.DataFrame()
+
+def generate_gemini_analysis(stock_name, stock_symbol, period, data):
+    """
+    使用 Gemini API 生成基本面和市場展望分析。
+    """
+    api_key = os.getenv("GEMINI_API_KEY")
+    if not api_key:
+        return "無法讀取 GEMINI API 金鑰", "請在系統環境變數中設定您的金鑰"
+
+    try:
+        genai.configure(api_key=api_key)
+        model = genai.GenerativeModel('gemini-1.5-flash')
+
+        price_change = ((data['Close'].iloc[-1] - data['Close'].iloc[0]) / data['Close'].iloc[0]) * 100
+        rsi_current = data['RSI'].iloc[-1]
+        macd_current = data['MACD'].iloc[-1]
+        macd_signal_current = data['MACD_Signal'].iloc[-1]
+        industry = INDUSTRY_MAPPING.get(stock_symbol, '綜合')
+
+        prompt = f"""
+        請扮演一位專業、資深的台灣股市金融分析師。
+        我將提供一檔台股的即時技術指標數據，請你基於這些數據，結合你對這家公司、其所在產業以及當前市場趨勢的理解，為我生成一段專業的「基本面分析」和一段「市場展望與投資建議」。
+
+        **股票資訊:**
+        - **公司名稱:** {stock_name} ({stock_symbol})
+        - **分析期間:** 最近 {period}
+        - **所屬產業:** {industry}
+        - **期間價格變動:** {price_change:+.2f}%
+        - **目前 RSI 指標:** {rsi_current:.2f}
+        - **目前 MACD 指標:** MACD線為 {macd_current:.3f}, 信號線為 {macd_signal_current:.3f}
+
+        **你的任務:**
+        1.  **基本面分析 (約 150 字):**
+            -   評論這家公司的產業地位、近期營運亮點或挑戰。
+            -   提及任何可能影響其基本面的關鍵因素 (例如：財報、法說會、政策、供應鏈變化等)。
+            -   請用專業、客觀的語氣撰寫。
+
+        2.  **市場展望與投資建議 (約 150 字):**
+            -   基於上述所有資訊，提供對該股票的短期和中期市場展望。
+            -   提出具體的投資建議，例如：適合何種類型的投資人、潛在的風險點。
+            -   請直接提供分析內容，不要包含任何問候語。
+
+        **輸出格式:**
+        請嚴格按照以下格式回傳，使用"$$"作為兩個段落之間的分隔符：
+        [基本面分析內容]$$[市場展望與投資建議內容]
+        """
+
+        response = model.generate_content(prompt)
+        parts = response.text.split('$$')
+        if len(parts) == 2:
+            fundamental_analysis = parts[0].strip()
+            market_outlook = parts[1].strip()
+            return dcc.Markdown(fundamental_analysis), dcc.Markdown(market_outlook)
+        else:
+            # Fallback for unexpected response format
+            return dcc.Markdown("無法解析 Gemini 回應，請稍後再試。"), dcc.Markdown(response.text)
+
+    except Exception as e:
+        error_message = f"呼叫 Gemini API 時發生錯誤: {str(e)}"
+        print(error_message)
+        return dcc.Markdown(error_message), dcc.Markdown("請檢查後台日誌或 API 金鑰設定")
+
+# 建立 Dash 應用程式
+app = dash.Dash(__name__, suppress_callback_exceptions=True)
+
+# 初始化模型
+try:
+    print("正在初始化新聞情緒分析模型...")
+    predictor = BertPredictor(max_news_per_keyword=5)
+    print("新聞情緒分析模型初始化成功。")
+except Exception as e:
+    print(f"錯誤：新聞情緒分析模型初始化失敗 - {e}")
+    predictor = None
+
+# 應用程式佈局
+app.layout = html.Div([
+    html.H1("台股分析儀表板", style={'text-align': 'center', 'margin-bottom': '30px'}),
+    html.Div([
+        html.H2("🤖 AI深度學習預測 - 台指期指數", style={'text-align': 'center','color': '#FFCC22','margin-bottom': '25px'}),
+        html.Div([
+            html.Div([
+                html.Label("預測期間:", style={'font-weight': 'bold', 'color': '#FFCC22'}),
+                dcc.Dropdown(id='taiex-prediction-period',
+                    options=[
+                        {'label': '1日後預測', 'value': 1},{'label': '5日後預測', 'value': 5},
+                        {'label': '10日後預測', 'value': 10},{'label': '20日後預測', 'value': 20},
+                        {'label': '60日後預測', 'value': 60}], value=5,
+                    style={'margin-bottom': '10px', 'color': '#272727'})
+            ], style={'width': '30%', 'display': 'inline-block'}),
+            html.Div(id='taiex-prediction-results', style={'width': '65%', 'display': 'inline-block', 'margin-left': '5%'})
+        ]),
+        html.Div([dcc.Graph(id='taiex-prediction-chart')], style={'margin-top': '20px'})
+    ], style={'background': 'linear-gradient(135deg, #667eea 0%, #764ba2 100%)','padding': '25px','border-radius': '15px','box-shadow': '0 8px 25px rgba(0,0,0,0.15)','color': 'white','margin-bottom': '40px'}),
+    
+    html.Div([
+        html.H3("📰 市場情緒與新聞分析", style={'color': '#E74C3C', 'margin-bottom': '20px'}),
+        html.Div([
+            html.Div([
+                html.H4("市場情緒指標", style={'color': '#8E44AD'}),
+                html.Div(id='sentiment-gauge')
+            ], style={'width': '48%', 'display': 'inline-block'}),
+            html.Div([
+                html.H4("關鍵新聞摘要", style={'color': '#27AE60'}),
+                html.Div(id='news-summary', style={'background': '#f8f9fa','padding': '15px','border-radius': '8px','max-height': '200px','overflow-y': 'auto'})
+            ], style={'width': '48%', 'display': 'inline-block', 'margin-left': '4%'})
+        ])
+    ], style={'margin-top': '30px','padding': '20px','background': 'white','border-radius': '10px','box-shadow': '0 2px 10px rgba(0,0,0,0.1)'}),
+
+    html.Div([
+        html.H3("景氣燈號與 PMI 分析"),
+        html.Div([
+            html.Div([dcc.Graph(id='business-climate-chart')], style={'width': '48%', 'display': 'inline-block'}),
+            html.Div([dcc.Graph(id='pmi-chart')], style={'width': '48%', 'display': 'inline-block', 'margin-left': '2%'})
+        ])
+    ], style={'margin-top': '30px'}),
+    
+    html.Div([
+        html.Div([
+            html.Label("選擇股票:"),
+            dcc.Dropdown(id='stock-dropdown', options=[{'label': name, 'value': symbol} for name, symbol in TAIWAN_STOCKS.items()], value='2330.TW', style={'margin-bottom': '10px'})
+        ], style={'width': '30%', 'display': 'inline-block', 'vertical-align': 'top'}),
+        html.Div([
+            html.Label("時間範圍:"),
+            dcc.Dropdown(id='period-dropdown',
+                options=[{'label': '1個月', 'value': '1mo'},{'label': '3個月', 'value': '3mo'},{'label': '6個月', 'value': '6mo'},{'label': '1年', 'value': '1y'},{'label': '2年', 'value': '2y'}],
+                value='1mo', style={'margin-bottom': '10px'})
+        ], style={'width': '30%', 'display': 'inline-block', 'margin-left': '5%', 'vertical-align': 'top'}),
+        html.Div([
+            html.Label("圖表類型:"),
+            dcc.Dropdown(id='chart-type', options=[{'label': '線圖', 'value': 'line'},{'label': '蠟燭圖', 'value': 'candlestick'}], value='candlestick', style={'margin-bottom': '10px'})
+        ], style={'width': '30%', 'display': 'inline-block', 'margin-left': '5%', 'vertical-align': 'top'})
+    ], style={'margin-bottom': '30px'}),
+
+    html.Div(id='stock-info-cards', style={'margin-bottom': '30px'}),
+    html.Div([html.Div([dcc.Graph(id='price-chart')], style={'width': '100%', 'display': 'inline-block', 'vertical-align': 'top'})]),
+    html.Div([
+        html.H3("📊 進階技術指標分析", style={'margin-bottom': '20px'}),
+        html.Div([
+            html.Label("選擇技術指標:", style={'font-weight': 'bold', 'margin-right': '10px'}),
+            dcc.Dropdown(id='technical-indicator-selector',
+                options=[{'label': 'RSI 相對強弱指標', 'value': 'RSI'},{'label': 'MACD 指數平滑異同移動平均線', 'value': 'MACD'},{'label': '布林通道 Bollinger Bands', 'value': 'BB'},
+                         {'label': 'KD 隨機指標', 'value': 'KD'},{'label': '威廉指標 %R', 'value': 'WR'},{'label': 'DMI 動向指標', 'value': 'DMI'}],
+                value='RSI', style={'width': '100%'})
+        ], style={'margin-bottom': '20px'}),
+        html.Div([dcc.Graph(id='advanced-technical-chart')])
+    ], style={'margin-top': '20px','padding': '20px','background': 'white','border-radius': '10px','box-shadow': '0 2px 10px rgba(0,0,0,0.1)'}),
+    html.Div([dcc.Graph(id='volume-chart')], style={'margin-top': '20px'}),
+    html.Div([html.H3("產業表現分析"), dcc.Graph(id='industry-analysis')], style={'margin-top': '30px'}),
+    html.Div([
+        html.H3("📊 分析師觀點與市場解讀", style={'color': '#2E86AB', 'margin-bottom': '20px'}),
+        html.Div([
+            html.Div([
+                html.H4("📝 技術面分析", style={'color': '#A23B72', 'margin-bottom': '15px'}),
+                html.Div(id='technical-analysis-text', style={'background': '#f8f9fa','padding': '15px','border-radius': '8px','border-left': '4px solid #A23B72','min-height': '150px','font-size': '14px','line-height': '1.6'})
+            ], style={'width': '48%', 'display': 'inline-block', 'vertical-align': 'top'}),
+            html.Div([
+                html.H4("📈 基本面分析 (AI 生成)", style={'color': '#F18F01', 'margin-bottom': '15px'}),
+                html.Div(id='fundamental-analysis-text', style={'background': '#f8f9fa','padding': '15px','border-radius': '8px','border-left': '4px solid #F18F01','min-height': '150px','font-size': '14px','line-height': '1.6'})
+            ], style={'width': '48%', 'display': 'inline-block', 'margin-left': '4%', 'vertical-align': 'top'})
+        ]),
+        html.Div([
+            html.H4("🎯 市場展望與投資建議 (AI 生成)", style={'color': '#C73E1D', 'margin-bottom': '15px', 'margin-top': '25px'}),
+            html.Div(id='market-outlook-text', style={'background': 'linear-gradient(135deg, #667eea 0%, #764ba2 100%)','color': 'white','padding': '20px','border-radius': '10px','min-height': '100px','font-size': '15px','line-height': '1.7','box-shadow': '0 4px 15px rgba(0,0,0,0.1)'})
+        ])
+    ], style={'margin-top': '30px','padding': '25px','background': 'white','border-radius': '12px','box-shadow': '0 4px 20px rgba(0,0,0,0.08)','border': '1px solid #e9ecef'}),
+    html.Div([
+        html.H3("📊 多檔股票比較分析", style={'margin-bottom': '20px'}),
+        html.Div([
+            html.Div([
+                html.Label("選擇比較股票（最多5檔）:", style={'font-weight': 'bold'}),
+                dcc.Dropdown(id='comparison-stocks', options=[{'label': name, 'value': symbol} for name, symbol in TAIWAN_STOCKS.items()], value=['0050.TW', '2330.TW', '2454.TW'], multi=True, style={'margin-bottom': '5px'}),
+                html.Small('（元大台灣50 (0050.TW) 為固定比較基準，不可移除）', style={'display': 'block', 'font-style': 'italic', 'color': 'gray'})
+            ], style={'width': '60%', 'display': 'inline-block'}),
+            html.Div([
+                html.Label("比較期間:", style={'font-weight': 'bold'}),
+                dcc.Dropdown(id='comparison-period', options=[{'label': '1個月', 'value': '1mo'},{'label': '3個月', 'value': '3mo'},{'label': '6個月', 'value': '6mo'},{'label': '1年', 'value': '1y'}], value='3mo')
+            ], style={'width': '35%', 'display': 'inline-block', 'margin-left': '5%', 'vertical-align': 'top'})
+        ]),
+        html.Div([
+            html.Div([dcc.Graph(id='comparison-chart')], style={'width': '65%', 'display': 'inline-block'}),
+            html.Div([html.H4("比較結果", style={'color': '#2E86AB'}), html.Div(id='comparison-table')], style={'width': '33%', 'display': 'inline-block', 'margin-left': '2%', 'vertical-align': 'top'})
+        ])
+    ], style={'margin-top': '30px','padding': '20px','background': 'white','border-radius': '10px','box-shadow': '0 2px 10px rgba(0,0,0,0.1)'}),
+])
+
+# 回調函數區域
+@app.callback(
+    [dash.dependencies.Output('taiex-prediction-results', 'children'),
+     dash.dependencies.Output('taiex-prediction-chart', 'figure')],
+    [dash.dependencies.Input('taiex-prediction-period', 'value')]
+)
+def update_taiex_prediction(predict_days):
+    data = get_stock_data('^TWII', '2y')
+    if data.empty: return html.Div("無法獲取台指期資料"), {}
+    
+    # === 修改點：統一呼叫 get_prediction 控制器 ===
+    final_prediction = get_prediction(data, predict_days)
+    
+    if final_prediction is None: return html.Div("資料不足，無法進行預測"), {}
+    current_price, last_date = data['Close'].iloc[-1], data.index[-1]
+    predicted_price, change_pct, confidence = final_prediction['predicted_price'], final_prediction['change_pct'], final_prediction['confidence']
+    
+    prediction_paths = {1: [1], 5: [1, 5], 10: [1, 5, 10], 20: [1, 10, 20], 60: [1, 10, 20, 60]}
+    intervals_to_predict = prediction_paths.get(predict_days, [predict_days])
+    prediction_dates, prediction_prices = [last_date], [current_price]
+    
+    for days in intervals_to_predict:
+        # === 修改點：迴圈內也使用統一的預測控制器 ===
+        interim_prediction = get_prediction(data, days)
+        if interim_prediction:
+            prediction_dates.append(last_date + timedelta(days=days))
+            prediction_prices.append(interim_prediction['predicted_price'])
+            
+    # 後續繪圖邏輯不變
+    color, arrow = ('red', '📈') if change_pct >= 0 else ('green', '📉')
+    result_card = html.Div([
+        html.H4(f"{predict_days}日後預測結果", style={'margin': '0 0 15px 0', 'color': 'white'}),
+        html.Div([html.Span(f"{arrow} ", style={'font-size': '24px'}), html.Span(f"{change_pct:+.2f}%", style={'font-size': '28px','font-weight': 'bold','color': color})], style={'margin': '10px 0'}),
+        html.P(f"目前價格: {current_price:.2f}", style={'margin': '5px 0'}), html.P(f"預測價格: {predicted_price:.2f}", style={'margin': '5px 0'}),
+        html.P(f"信心度: {confidence:.1%}", style={'margin': '5px 0', 'font-size': '14px'})
+    ], style={'background': 'rgba(255,255,255,0.1)','padding': '20px','border-radius': '10px','border': '1px solid rgba(255,255,255,0.2)'})
+    fig = go.Figure()
+    recent_data = data.tail(30)
+    fig.add_trace(go.Scatter(x=recent_data.index, y=recent_data['Close'], mode='lines', name='歷史價格', line=dict(color='#FFA726', width=2)))
+    fig.add_trace(go.Scatter(x=prediction_dates, y=prediction_prices, mode='lines+markers', name=f'{predict_days}日預測路徑', line=dict(color=color, width=3, dash='dash'), marker=dict(size=8)))
+    fig.update_layout(title=f'台指期 {predict_days}日預測走勢', xaxis_title='日期', yaxis_title='指數點位', height=350, plot_bgcolor='rgba(0,0,0,0)', paper_bgcolor='rgba(0,0,0,0)', font=dict(color='white'))
+    return result_card, fig
+
+@app.callback(
+    dash.dependencies.Output('stock-info-cards', 'children'),
+    [dash.dependencies.Input('stock-dropdown', 'value')]
+)
+def update_stock_info(selected_stock):
+    data = get_stock_data(selected_stock, '5d')
+    if data.empty: return html.Div("無法獲取股票資料")
+    current_price = data['Close'].iloc[-1]
+    prev_price = data['Close'].iloc[-2] if len(data) > 1 else current_price
+    change = current_price - prev_price
+    change_pct = (change / prev_price) * 100
+    stock_name = [name for name, symbol in TAIWAN_STOCKS.items() if symbol == selected_stock][0]
+    color, arrow = ('red', '▲') if change >= 0 else ('green', '▼')
+    return html.Div([
+        html.Div([
+            html.H3(f"{stock_name} ({selected_stock})", style={'margin': '0'}),
+            html.H2(f"${current_price:.2f}", style={'margin': '5px 0', 'color': color}),
+            html.P(f"{arrow} {change:+.2f} ({change_pct:+.2f}%)", style={'margin': '0', 'color': color, 'font-weight': 'bold'})
+        ], style={'background': 'white','padding': '20px','border-radius': '10px','box-shadow': '0 2px 10px rgba(0,0,0,0.1)','display': 'inline-block','margin-right': '20px'}),
+        html.Div([
+            html.H4("今日統計", style={'margin': '0 0 10px 0'}),
+            html.P(f"最高: ${data['High'].iloc[-1]:.2f}", style={'margin': '5px 0'}),
+            html.P(f"最低: ${data['Low'].iloc[-1]:.2f}", style={'margin': '5px 0'}),
+            html.P(f"成交量: {data['Volume'].iloc[-1]:,.0f}", style={'margin': '5px 0'})
+        ], style={'background': 'white','padding': '20px','border-radius': '10px','box-shadow': '0 2px 10px rgba(0,0,0,0.1)','display': 'inline-block'})
+    ])
+
+@app.callback(
+    dash.dependencies.Output('price-chart', 'figure'),
+    [dash.dependencies.Input('stock-dropdown', 'value'),
+     dash.dependencies.Input('period-dropdown', 'value'),
+     dash.dependencies.Input('chart-type', 'value')]
+)
+def update_price_chart(selected_stock, period, chart_type):
+    data = get_stock_data(selected_stock, period)
+    if data.empty: return {}
+    data = calculate_technical_indicators(data)
+    stock_name = [name for name, symbol in TAIWAN_STOCKS.items() if symbol == selected_stock][0]
+    fig = make_subplots(rows=1, cols=2, shared_yaxes=True, column_widths=[0.8, 0.2], horizontal_spacing=0.01)
+    if chart_type == 'candlestick':
+        fig.add_trace(go.Candlestick(x=data.index, open=data['Open'], high=data['High'], low=data['Low'], close=data['Close'], name=stock_name, increasing_line_color='red', decreasing_line_color='green'), row=1, col=1)
+    else:
+        fig.add_trace(px.line(data, y='Close').data[0], row=1, col=1)
+    fig.add_trace(go.Scatter(x=data.index, y=data['MA5'], mode='lines', name='MA5', line=dict(color='orange')), row=1, col=1)
+    fig.add_trace(go.Scatter(x=data.index, y=data['MA20'], mode='lines', name='MA20', line=dict(color='blue')), row=1, col=1)
+    bin_edges, volume_per_bin, price_centers = calculate_volume_profile(data, num_bins=50)
+    if volume_per_bin is not None:
+        fig.add_trace(go.Bar(orientation='h', y=price_centers, x=volume_per_bin, name='Volume Profile', text=[f'{vol/1000:.0f}k' for vol in volume_per_bin], textposition='auto', marker=dict(color='rgba(173, 216, 230, 0.6)', line=dict(color='rgba(30, 144, 255, 0.8)', width=1))), row=1, col=2)
+    fig.update_layout(title_text=f'{stock_name} 股價走勢與成交量分佈', height=500, showlegend=True, xaxis1=dict(title='日期', type='date', rangeslider_visible=False), yaxis1=dict(title='價格 (TWD)'), xaxis2=dict(title='成交量', showticklabels=True), yaxis2=dict(showticklabels=False), bargap=0.05)
+    return fig
+
+@app.callback(
+    dash.dependencies.Output('advanced-technical-chart', 'figure'),
+    [dash.dependencies.Input('technical-indicator-selector', 'value'),
+     dash.dependencies.Input('stock-dropdown', 'value'),
+     dash.dependencies.Input('period-dropdown', 'value')]
+)
+def update_advanced_technical_chart(indicator, selected_stock, period):
+    data = get_stock_data(selected_stock, period)
+    if data.empty: return {}
+    data = calculate_technical_indicators(data)
+    stock_name = [name for name, symbol in TAIWAN_STOCKS.items() if symbol == selected_stock][0]
+    fig = go.Figure()
+    if indicator == 'RSI':
+        fig = go.Figure()
+        fig.add_trace(go.Scatter(x=data.index, y=data['RSI'], mode='lines', name='RSI', line=dict(color='purple', width=2)))
+        fig.add_hline(y=70, line_dash="dash", line_color="green", annotation_text="超買線(70)")
+        fig.add_hline(y=30, line_dash="dash", line_color="red", annotation_text="超賣線(30)")
+        fig.add_hline(y=50, line_dash="dot", line_color="gray", annotation_text="中線(50)")
+        fig.update_layout(title=f'{stock_name} - RSI 相對強弱指標', xaxis_title='日期', yaxis_title='RSI', height=450, yaxis=dict(range=[0, 100]))
+    elif indicator == 'MACD':
+        fig = make_subplots(rows=2, cols=1, shared_xaxes=True, vertical_spacing=0.1, row_heights=[0.7, 0.3], subplot_titles=('價格走勢', 'MACD 指標'))
+        fig.add_trace(go.Scatter(x=data.index, y=data['Close'], mode='lines', name='收盤價', line=dict(color='black', width=1.5)), row=1, col=1)
+        fig.add_trace(go.Scatter(x=data.index, y=data['MACD'], mode='lines', name='MACD (快線)', line=dict(color='blue', width=2)), row=2, col=1)
+        fig.add_trace(go.Scatter(x=data.index, y=data['MACD_Signal'], mode='lines', name='Signal (慢線)', line=dict(color='red', width=2)), row=2, col=1)
+        colors = ['red' if x >= 0 else 'green' for x in data['MACD_Histogram']]
+        fig.add_trace(go.Bar(x=data.index, y=data['MACD_Histogram'], name='MACD柱狀圖', marker_color=colors), row=2, col=1)
+        fig.update_layout(title_text=f'{stock_name} - MACD 指數平滑異同移動平均線', height=550)
+    elif indicator == 'BB':
+        fig = go.Figure()
+        fig.add_trace(go.Scatter(x=data.index, y=data['Close'], mode='lines', name='收盤價', line=dict(color='black', width=2)))
+        fig.add_trace(go.Scatter(x=data.index, y=data['BB_Upper'], mode='lines', name='上軌', line=dict(color='red', width=1, dash='dash')))
+        fig.add_trace(go.Scatter(x=data.index, y=data['BB_Middle'], mode='lines', name='中軌(MA20)', line=dict(color='blue', width=1)))
+        fig.add_trace(go.Scatter(x=data.index, y=data['BB_Lower'], mode='lines', name='下軌', line=dict(color='green', width=1, dash='dash')))
+        fig.update_layout(title=f'{stock_name} - 布林通道 (20日, 2σ)', xaxis_title='日期', yaxis_title='價格 (TWD)', height=450)
+    elif indicator == 'KD':
+        fig = make_subplots(rows=2, cols=1, shared_xaxes=True, vertical_spacing=0.1, row_heights=[0.6, 0.4], subplot_titles=('價格走勢', 'KD指標'))
+        fig.add_trace(go.Scatter(x=data.index, y=data['Close'], mode='lines', name='收盤價', line=dict(color='black', width=1)), row=1, col=1)
+        fig.add_trace(go.Scatter(x=data.index, y=data['K'], mode='lines', name='K線', line=dict(color='blue', width=2)), row=2, col=1)
+        fig.add_trace(go.Scatter(x=data.index, y=data['D'], mode='lines', name='D線', line=dict(color='red', width=2)), row=2, col=1)
+        fig.add_hline(y=80, line_dash="dash", line_color="green", annotation_text="超買線(80)", row=2, col=1)
+        fig.add_hline(y=20, line_dash="dash", line_color="red", annotation_text="超賣線(20)", row=2, col=1)
+        fig.update_layout(title=f'{stock_name} - KD 隨機指標 (9,3,3)', height=500, yaxis2_range=[0, 100])
+    elif indicator == 'WR':
+        fig = make_subplots(rows=2, cols=1, shared_xaxes=True, vertical_spacing=0.1, row_heights=[0.6, 0.4], subplot_titles=('價格走勢', '威廉指標 %R'))
+        fig.add_trace(go.Scatter(x=data.index, y=data['Close'], mode='lines', name='收盤價', line=dict(color='black', width=1)), row=1, col=1)
+        fig.add_trace(go.Scatter(x=data.index, y=data['Williams_R'], mode='lines', name='威廉%R', line=dict(color='purple', width=2)), row=2, col=1)
+        fig.add_hline(y=-20, line_dash="dash", line_color="green", annotation_text="超買線(-20)", row=2, col=1)
+        fig.add_hline(y=-80, line_dash="dash", line_color="red", annotation_text="超賣線(-80)", row=2, col=1)
+        fig.update_layout(title=f'{stock_name} - 威廉指標 %R (14日)', height=500, yaxis2_range=[-100, 0])
+    elif indicator == 'DMI':
+        fig = make_subplots(rows=2, cols=1, shared_xaxes=True, vertical_spacing=0.1, row_heights=[0.6, 0.4], subplot_titles=('價格走勢', 'DMI 指標'))
+        data_filtered = data.iloc[14:]
+        fig.add_trace(go.Scatter(x=data_filtered.index, y=data_filtered['Close'], mode='lines', name='收盤價', line=dict(color='black', width=1)), row=1, col=1)
+        fig.add_trace(go.Scatter(x=data_filtered.index, y=data_filtered['+DI'], mode='lines', name='+DI', line=dict(color='red', width=2)), row=2, col=1)
+        fig.add_trace(go.Scatter(x=data_filtered.index, y=data_filtered['-DI'], mode='lines', name='-DI', line=dict(color='green', width=2)), row=2, col=1)
+        fig.add_trace(go.Scatter(x=data_filtered.index, y=data_filtered['ADX'], mode='lines', name='ADX', line=dict(color='blue', width=2, dash='dot')), row=2, col=1)
+        fig.update_layout(title=f'{stock_name} - DMI 動向指標 (14日)', height=500, showlegend=True, yaxis2_range=[0, 100])
+    return fig
+
+@app.callback(
+    dash.dependencies.Output('volume-chart', 'figure'),
+    [dash.dependencies.Input('stock-dropdown', 'value'),
+     dash.dependencies.Input('period-dropdown', 'value')]
+)
+def update_volume_chart(selected_stock, period):
+    data = get_stock_data(selected_stock, period)
+    if data.empty: return {}
+    stock_name = [name for name, symbol in TAIWAN_STOCKS.items() if symbol == selected_stock][0]
+    colors = ['red' if data['Close'].iloc[i] > data['Open'].iloc[i] else 'green' for i in range(len(data))]
+    fig = go.Figure(go.Bar(x=data.index, y=data['Volume'], marker_color=colors, name='成交量'))
+    fig.update_layout(title=f'{stock_name} 成交量', xaxis_title='日期', yaxis_title='成交量', height=300)
+    return fig
+
+@app.callback(
+    dash.dependencies.Output('industry-analysis', 'figure'),
+    [dash.dependencies.Input('stock-dropdown', 'value')]
+)
+def update_industry_analysis(selected_stock):
+    performance_data = []
+    for name, symbol in TAIWAN_STOCKS.items():
+        data = get_stock_data(symbol, '1mo')
+        if not data.empty and len(data) > 1:
+            return_pct = ((data['Close'].iloc[-1] - data['Close'].iloc[0]) / data['Close'].iloc[0]) * 100
+            performance_data.append({
+                '股票': name,
+                '代碼': symbol,
+                '月報酬率(%)': return_pct,
+                '絕對波動': abs(return_pct)
+            })
+    if not performance_data:
+        fig = go.Figure().add_annotation(text="無法計算產業資料", showarrow=False)
+        fig.update_layout(title="近一月市場波動最大標的", height=400)
+        return fig
+    df_performance = pd.DataFrame(performance_data)
+    df_top_movers = df_performance.sort_values(by='絕對波動', ascending=False).head(10)
+    fig = px.pie(
+        df_top_movers,
+        values='絕對波動',
+        names='股票',
+        title='近一月市場波動最大 Top 10 標的',
+        hover_data={'月報酬率(%)': ':.2f'}
+    )
+    fig.update_traces(
+        textposition='inside',
+        textinfo='percent+label',
+        hovertemplate="<b>%{label}</b><br>月報酬率: %{customdata[0]:.2f}%<extra></extra>"
+    )
+    fig.update_layout(height=400, showlegend=False)
+    return fig
+
+@app.callback(
+    dash.dependencies.Output('business-climate-chart', 'figure'),
+    [dash.dependencies.Input('stock-dropdown', 'value')]
+)
+def update_business_climate_chart(selected_stock):
+    df = get_business_climate_data()
+    if df.empty:
+        fig = go.Figure().add_annotation(text="無法載入景氣燈號資料", showarrow=False)
+        fig.update_layout(title="台灣景氣燈號", height=300)
+        return fig
+    def get_light_color(score):
+        if score >= 32: return 'red'
+        elif score >= 24: return 'orange'
+        elif score >= 17: return 'yellow'
+        elif score >= 10: return 'lightgreen'
+        else: return 'blue'
+    colors = [get_light_color(score) for score in df['Index']]
+    fig = go.Figure()
+    fig.add_trace(go.Scatter(x=df['Date'], y=df['Index'], mode='lines+markers', name='景氣燈號', line=dict(color='darkblue', width=2), marker=dict(size=8, color=colors, line=dict(width=2, color='darkblue'))))
+    fig.add_hline(y=32, line_dash="dash", line_color="red", annotation_text="紅燈(32)")
+    fig.add_hline(y=17, line_dash="dash", line_color="yellow", annotation_text="黃燈(17)")
+    fig.update_layout(title="台灣景氣燈號走勢", xaxis_title='日期', yaxis_title='燈號分數', height=300, yaxis=dict(range=[0, 40]))
+    return fig
+
+# ========================= MODIFIED SECTION START (CACHE INTEGRATED) =========================
+@app.callback(
+    [dash.dependencies.Output('technical-analysis-text', 'children'),
+     dash.dependencies.Output('fundamental-analysis-text', 'children'),
+     dash.dependencies.Output('market-outlook-text', 'children')],
+    [dash.dependencies.Input('stock-dropdown', 'value'),
+     dash.dependencies.Input('period-dropdown', 'value')]
+)
+def update_analysis_text(selected_stock, period):
+    # 建立快取的唯一鍵值
+    cache_key = f"{selected_stock}-{period}"
+    current_time = time.time()
+
+    # 1. 檢查快取
+    if cache_key in ANALYSIS_CACHE:
+        cached_data = ANALYSIS_CACHE[cache_key]
+        if current_time - cached_data['timestamp'] < CACHE_DURATION_SECONDS:
+            print(f"從快取載入分析: {cache_key}")
+            # 直接回傳快取的內容
+            return cached_data['technical'], cached_data['fundamental'], cached_data['outlook']
+
+    print(f"重新生成分析: {cache_key}")
+    # --- 如果快取沒有，才繼續執行以下程式 ---
+
+    data = get_stock_data(selected_stock, period)
+    stock_name = [name for name, symbol in TAIWAN_STOCKS.items() if symbol == selected_stock][0]
+    if data.empty or len(data) < 20:
+        return "資料不足，無法分析", "資料不足，無法分析", "資料不足，無法分析"
+    
+    data = calculate_technical_indicators(data)
+    
+    # 2. 技術面分析
+    price_change = ((data['Close'].iloc[-1] - data['Close'].iloc[0]) / data['Close'].iloc[0]) * 100
+    rsi_current = data['RSI'].iloc[-1] if not pd.isna(data['RSI'].iloc[-1]) else 50
+    macd_current = data['MACD'].iloc[-1] if not pd.isna(data['MACD'].iloc[-1]) else 0
+    macd_signal_current = data['MACD_Signal'].iloc[-1] if not pd.isna(data['MACD_Signal'].iloc[-1]) else 0
+    
+    technical_text = html.Div([
+        html.P([html.Strong("價格趨勢："), f"在最近 {period} 期間內，{stock_name} 股價呈現", html.Span(f"{'上漲' if price_change > 5 else '下跌' if price_change < -5 else '盤整'}", style={'color': 'red' if price_change > 5 else 'green' if price_change < -5 else 'orange', 'font-weight': 'bold'}), f"走勢，累計變動 {price_change:+.1f}%。"]),
+        html.P([html.Strong("RSI 指標："), f"目前的 RSI 值為 {rsi_current:.1f}，", html.Span("處於超買區(>70)" if rsi_current > 70 else "處於超賣區(<30)" if rsi_current < 30 else "在正常範圍內", style={'color': 'green' if rsi_current > 70 else 'red' if rsi_current < 30 else 'blue', 'font-weight': 'bold'}), "。"]),
+        html.P([html.Strong("MACD 指標："), f"MACD 快線 ({macd_current:.3f}) 目前", html.Span("高於" if macd_current > macd_signal_current else "低於", style={'color': 'red' if macd_current > macd_signal_current else 'green', 'font-weight': 'bold'}), f" Signal 慢線 ({macd_signal_current:.3f})，", f"顯示市場動能偏向{'多頭' if macd_current > macd_signal_current else '空頭'}。"]),
+    ])
+    
+    # 3. 基本面與展望分析 (呼叫 Gemini)
+    fundamental_text, market_outlook_text = generate_gemini_analysis(stock_name, selected_stock, period, data)
+    
+    # 4. 將新產生的結果存入快取
+    ANALYSIS_CACHE[cache_key] = {
+        'technical': technical_text,
+        'fundamental': fundamental_text,
+        'outlook': market_outlook_text,
+        'timestamp': current_time
+    }
+
+    return technical_text, fundamental_text, market_outlook_text
+# ========================== MODIFIED SECTION END ==========================
+
+@app.callback(
+    dash.dependencies.Output('pmi-chart', 'figure'),
+    [dash.dependencies.Input('stock-dropdown', 'value')]
+)
+def update_pmi_chart(selected_stock):
+    df = get_pmi_data()
+    if df.empty:
+        fig = go.Figure().add_annotation(text="無法載入PMI資料", showarrow=False)
+        fig.update_layout(title="台灣PMI指數", height=300)
+        return fig
+    colors = ['red' if value >= 50 else 'green' for value in df['Index']]
+    fig = go.Figure()
+    fig.add_trace(go.Scatter(x=df['Date'], y=df['Index'], mode='lines+markers', name='PMI指數', line=dict(color='darkblue', width=2), marker=dict(size=8, color=colors, line=dict(width=2, color='darkblue'))))
+    fig.add_hline(y=50, line_dash="dash", line_color="black", annotation_text="榮枯線(50)")
+    fig.update_layout(title="台灣PMI指數走勢", xaxis_title='日期', yaxis_title='PMI指數', height=300, yaxis=dict(range=[35, 60]))
+    return fig
+
+def summarize_news_with_gemini(news_list: list) -> str:
+    """
+    使用 Gemini API 將英文新聞標題列表摘要成一段繁體中文。
+    """
+    api_key = os.getenv("GEMINI_API_KEY")
+    if not api_key:
+        return "錯誤：找不到 GEMINI_API_KEY。請在 Hugging Face Secrets 中設定。"
+
+    try:
+        genai.configure(api_key=api_key)
+        model = genai.GenerativeModel('gemini-1.5-flash')
+
+        formatted_news = "\n".join([f"- {news}" for news in news_list])
+
+        prompt = f"""
+        請扮演一位專業的金融市場分析師。
+        以下是幾則最新的英文財經新聞標題，請將它們整合成一段簡潔、流���、約 200 字的繁體中文市場動態摘要，與利多哪些產業，利空哪些產業。
+        提供3段重點，
+        請專注於可能影響市場情緒和股價的關鍵資訊，並直接提供摘要內容，不要包含任何額外的問候語或說明。
+
+        英文新聞標題如下：
+        {formatted_news}
+        """
+
+        response = model.generate_content(prompt)
+        return response.text
+
+    except Exception as e:
+        print(f"呼叫 Gemini API 時發生錯誤: {e}")
+        return f"無法生成新聞摘要，請稍後再試。錯誤訊息：{e}"
+
+@app.callback(
+    [dash.dependencies.Output('comparison-chart', 'figure'),
+     dash.dependencies.Output('comparison-table', 'children')],
+    [dash.dependencies.Input('comparison-stocks', 'value'),
+     dash.dependencies.Input('comparison-period', 'value')]
+)
+def update_comparison_analysis(selected_stocks, period):
+    fixed_stock = '0050.TW'
+    if not selected_stocks: selected_stocks = [fixed_stock]
+    elif fixed_stock not in selected_stocks: selected_stocks.insert(0, fixed_stock)
+    selected_stocks = selected_stocks[:5]
+    fig = go.Figure()
+    comparison_data = []
+    for stock in selected_stocks:
+        data = get_stock_data(stock, period)
+        if not data.empty:
+            stock_name = next((name for name, symbol in TAIWAN_STOCKS.items() if symbol == stock), stock)
+            normalized_prices = (data['Close'] / data['Close'].iloc[0]) * 100
+            fig.add_trace(go.Scatter(x=data.index, y=normalized_prices, mode='lines', name=stock_name, line=dict(width=2)))
+            total_return = ((data['Close'].iloc[-1] / data['Close'].iloc[0]) - 1) * 100
+            volatility = data['Close'].pct_change().std() * np.sqrt(252) * 100
+            comparison_data.append({'name': stock_name, 'return': total_return, 'volatility': volatility, 'current_price': data['Close'].iloc[-1]})
+    fig.update_layout(title=f'股票績效比較 - {period}', xaxis_title='日期', yaxis_title='相對績效 (基期=100)', height=400, hovermode='x unified')
+    if comparison_data:
+        table_rows = []
+        for item in sorted(comparison_data, key=lambda x: x['return'], reverse=True):
+            color = 'red' if item['return'] > 0 else 'green'
+            table_rows.append(html.Tr([html.Td(item['name'], style={'font-weight': 'bold'}), html.Td(f"{item['return']:+.1f}%", style={'color': color, 'font-weight': 'bold'}), html.Td(f"{item['volatility']:.1f}%"), html.Td(f"${item['current_price']:.2f}")]))
+        table = html.Table([html.Thead(html.Tr([html.Th("股票"), html.Th("報酬率"), html.Th("波動率"), html.Th("現價")])), html.Tbody(table_rows)], style={'width': '100%'})
+        return fig, table
+    return fig, html.Div("無可比較資料")
+
+@app.callback(
+    [dash.dependencies.Output('sentiment-gauge', 'children'),
+     dash.dependencies.Output('news-summary', 'children')],
+    [dash.dependencies.Input('stock-dropdown', 'value')]
+)
+def update_sentiment_analysis(selected_stock):
+    if predictor is None:
+        error_fig = go.Figure().add_annotation(text="情緒指標模型載入失敗", showarrow=False)
+        error_fig.update_layout(height=200)
+        return dcc.Graph(figure=error_fig), html.P("新聞分析模型載入失敗，請檢查後台日誌。")
+
+    sentiment_score_raw = predictor.get_news_index()
+
+    if sentiment_score_raw is not None:
+        sentiment_score_normalized = (sentiment_score_raw + 1) * 50
+        sentiment_score_normalized = max(0, min(100, sentiment_score_normalized))
+        if sentiment_score_normalized >= 65:
+            bar_color, level_text = "#5cb85c", "樂觀"
+        elif sentiment_score_normalized >= 35:
+            bar_color, level_text = "#f0ad4e", "中性"
+        else:
+            bar_color, level_text = "#d9534f", "悲觀"
+        gauge_fig = go.Figure(go.Indicator(
+            mode = "gauge+number", value = sentiment_score_normalized,
+            domain = {'x': [0, 1], 'y': [0, 1]},
+            title = {'text': f"昨日市場情緒: {level_text}", 'font': {'size': 18}},
+            gauge = {'axis': {'range': [0, 100]}, 'bar': {'color': bar_color},
+                     'steps': [{'range': [0, 35], 'color': "rgba(217, 83, 79, 0.2)"},
+                               {'range': [35, 65], 'color': "rgba(240, 173, 78, 0.2)"},
+                               {'range': [65, 100], 'color': "rgba(92, 184, 92, 0.2)"}]}
+        ))
+        gauge_fig.update_layout(height=200, margin=dict(l=30, r=30, t=50, b=20))
+        gauge_content = dcc.Graph(figure=gauge_fig)
+    else:
+        error_fig = go.Figure().add_annotation(text="今日尚無情緒分數", showarrow=False)
+        error_fig.update_layout(height=200)
+        gauge_content = dcc.Graph(figure=error_fig)
+
+    top_news_list = predictor.get_news()
+    news_content = None
+
+    if top_news_list and isinstance(top_news_list, list):
+        summary_text = summarize_news_with_gemini(top_news_list)
+        news_content = dcc.Markdown(summary_text, style={
+            'margin': '8px 0', 'padding-left': '5px',
+            'font-size': '15px', 'line-height': '1.7'
+        })
+    elif top_news_list == []:
+         news_content = html.P("昨日無重大相關新聞。", style={'text-align': 'center', 'padding-top': '50px'})
+    else:
+        news_content = html.P("讀取新聞時發生錯誤。", style={'text-align': 'center', 'padding-top': '50px'})
+
+    return gauge_content, news_content
+
+# 主程式執行
+if __name__ == '__main__':
+    app.run(host="0.0.0.0", port=7860, debug=False)