backend/app/services/price_service.py

@@ -1,7 +1,7 @@
 """
 Price data service for loading and processing stock price data
 """
-import pandas as pd
+import polars as pl
 from pathlib import Path
 from typing import List, Dict, Any, Optional
 import logging
@@ -13,7 +13,7 @@ class PriceService:
     """Service for loading and processing price data from data_cache"""
     
     @staticmethod
-    def load_price_data(ticker: str, data_cache_dir: str) -> Optional[pd.DataFrame]:
+    def load_price_data(ticker: str, data_cache_dir: str) -> Optional[pl.DataFrame]:
         """
         Load price data from data_cache CSV files
         
@@ -54,10 +54,10 @@ class PriceService:
             
             logger.info(f"Loading price data from {latest_file}")
             
-            df = pd.read_csv(latest_file)
+            df = pl.read_csv(str(latest_file))
-            df['Date'] = pd.to_datetime(df['Date'])
+            df = df.with_columns(pl.col("Date").str.to_datetime())
             
-            return df.sort_values('Date')
+            return df.sort("Date")
             
         except Exception as e:
             logger.error(f"Error loading price data for {ticker}: {e}")
@@ -82,7 +82,7 @@ class PriceService:
         return cache_age_hours < max_age_hours
     
     @staticmethod
-    def _fetch_and_cache_data(ticker: str, data_cache_dir: str, max_retries: int = 3) -> Optional[pd.DataFrame]:
+    def _fetch_and_cache_data(ticker: str, data_cache_dir: str, max_retries: int = 3) -> Optional[pl.DataFrame]:
         """
         Fetch data from yfinance and cache it
         
@@ -130,10 +130,10 @@ class PriceService:
                 
                 logger.info(f"成功獲取並緩存 {ticker} 數據到 {cache_file}")
                 
-                # Prepare and return DataFrame
+                # Prepare and return DataFrame - convert to polars
-                df = pd.read_csv(cache_file)
+                df = pl.read_csv(str(cache_file))
-                df['Date'] = pd.to_datetime(df['Date'])
+                df = df.with_columns(pl.col("Date").str.to_datetime())
-                return df.sort_values('Date')
+                return df.sort("Date")
                 
             except Exception as e:
                 logger.warning(f"第 {attempt} 次嘗試失敗: {e}")
@@ -149,7 +149,7 @@ class PriceService:
 
     
     @staticmethod
-    def calculate_stats(df: pd.DataFrame) -> Dict[str, Any]:
+    def calculate_stats(df: pl.DataFrame) -> Dict[str, Any]:
         """
         Calculate price statistics
         
@@ -159,22 +159,22 @@ class PriceService:
         Returns:
             Dictionary with statistics
         """
-        start_price = float(df.iloc[0]['Close'])
+        start_price = float(df.row(0, named=True)["Close"])
-        end_price = float(df.iloc[-1]['Close'])
+        end_price = float(df.row(-1, named=True)["Close"])
         growth_rate = ((end_price - start_price) / start_price) * 100
-        duration_days = (df.iloc[-1]['Date'] - df.iloc[0]['Date']).days
+        duration_days = (df.row(-1, named=True)["Date"] - df.row(0, named=True)["Date"]).days
         
         return {
             "growth_rate": round(growth_rate, 2),
             "duration_days": int(duration_days),
-            "start_date": df.iloc[0]['Date'].strftime('%Y-%m-%d'),
+            "start_date": df.row(0, named=True)["Date"].strftime('%Y-%m-%d'),
-            "end_date": df.iloc[-1]['Date'].strftime('%Y-%m-%d'),
+            "end_date": df.row(-1, named=True)["Date"].strftime('%Y-%m-%d'),
             "start_price": round(start_price, 2),
             "end_price": round(end_price, 2),
         }
     
     @staticmethod
-    def prepare_chart_data(df: pd.DataFrame, limit: int = 365) -> List[Dict[str, Any]]:
+    def prepare_chart_data(df: pl.DataFrame, limit: int = 365) -> List[Dict[str, Any]]:
         """
         Prepare price data for charting (limit to recent data)
         
@@ -188,9 +188,9 @@ class PriceService:
         # Get recent data
         recent_df = df.tail(limit)
         
-        # Convert to list of dicts
+        # Convert to list of dicts using polars to_dicts()
         data = []
-        for _, row in recent_df.iterrows():
+        for row in recent_df.iter_rows(named=True):
             data.append({
                 "Date": row['Date'].strftime('%Y-%m-%d'),
                 "Open": round(float(row['Open']), 2),

backend/requirements.txt

@@ -16,7 +16,8 @@ python-dotenv==1.0.0
 typing-extensions
 langchain-openai
 langchain-experimental
-pandas
+polars
+pyarrow
 yfinance
 praw
 feedparser

frontend/package.json

@@ -23,7 +23,7 @@
     "clsx": "^2.1.1",
     "date-fns": "^4.1.0",
     "lucide-react": "^0.554.0",
-    "next": "16.0.3",
+    "next": "16.0.7",
     "next-themes": "^0.4.6",
     "react": "19.2.0",
     "react-day-picker": "^9.11.3",
@@ -42,6 +42,7 @@
     "@types/react": "^19",
     "@types/react-dom": "^19",
     "babel-plugin-react-compiler": "1.0.0",
+    "baseline-browser-mapping": "^2.9.2",
     "eslint": "^9",
     "eslint-config-next": "16.0.3",
     "tailwindcss": "^4",

frontend/pnpm-workspace.yaml

@@ -0,0 +1,3 @@
+ignoredBuiltDependencies:
+  - sharp
+  - unrs-resolver

requirements.txt

@@ -1,7 +1,8 @@
 typing-extensions
 langchain-openai
 langchain-experimental
-pandas
+polars
+pyarrow
 yfinance
 praw
 feedparser

tradingagents/agents/utils/technical_indicators_tools.py

@@ -5,16 +5,31 @@ from tradingagents.dataflows.interface import route_to_vendor
 @tool
 def get_indicators(
     symbol: Annotated[str, "公司的股票代碼"],
-    indicator: Annotated[str, "要獲取分析和報告的技術指標"],
+    indicator: Annotated[str, """技術指標名稱。
+    常用指標：MA (移動平均), RSI, MACD。
+    使用簡寫名稱如：'MA', 'RSI', 'MACD'，
+    或指定期間如：'close_50_sma', 'close_200_sma'。
+    範例：'MA' 或 'RSI' 或 'MACD'"""],
     curr_date: Annotated[str, "您正在交易的當前交易日期，格式為 YYYY-mm-dd"],
     look_back_days: Annotated[int, "回溯天數"] = 30,
 ) -> str:
     """
     檢索給定股票代碼的技術指標。
     使用設定的技術指標供應商。
+    
+    支持的指標：
+    - MA/SMA: 簡單移動平均線（使用 look_back_days 或指定 'close_50_sma', 'close_200_sma'）
+    - EMA: 指數移動平均線
+    - RSI: 相對強弱指數
+    - MACD: 移動平均收斂背離
+    - BOLL: 布林通道
+    - ATR: 平均真實波幅
+    - VWMA: 成交量加權移動平均
+    - MFI: 資金流量指數
+    
     Args:
         symbol (str): 公司的股票代碼，例如 AAPL, TSM
-        indicator (str): 要獲取分析和報告的技術指標
+        indicator (str): 技術指標名稱，使用簡寫（MA, RSI, MACD 等）
         curr_date (str): 您正在交易的當前交易日期，格式為 YYYY-mm-dd
         look_back_days (int): 回溯天數，預設為 30
     Returns:
@@ -23,34 +38,56 @@ def get_indicators(
     # 規範化指標名稱以匹配供應商的預期格式
     indicator_lower = indicator.lower().strip()
     
-    # 處理常見的變體
+    # 處理常見的變體 - 包含 "moving average" 的完整詞
-    if "50" in indicator_lower and ("ma" in indicator_lower or "avg" in indicator_lower):
+    if "50" in indicator_lower and ("ma" in indicator_lower or "avg" in indicator_lower or "moving" in indicator_lower):
         normalized_indicator = "close_50_sma"
-    elif "200" in indicator_lower and ("ma" in indicator_lower or "avg" in indicator_lower):
+    elif "200" in indicator_lower and ("ma" in indicator_lower or "avg" in indicator_lower or "moving" in indicator_lower):
         normalized_indicator = "close_200_sma"
     elif "10" in indicator_lower and "ema" in indicator_lower:
         normalized_indicator = "close_10_ema"
     # 處理通用指標名稱，使用 look_back_days
-    elif indicator_lower in ["sma", "ma"]:
+    elif indicator_lower in ["sma", "ma", "moving average", "simple moving average"]:
         normalized_indicator = f"close_{look_back_days}_sma"
-    elif indicator_lower == "ema":
+    elif indicator_lower in ["ema", "exponential moving average"]:
         normalized_indicator = f"close_{look_back_days}_ema"
     else:
-        # 常見指標名稱映射
+        # 常見指標名稱映射 - 擴充版
         mapping = {
+            # SMA 變體
             "sma50": "close_50_sma",
             "sma200": "close_200_sma",
-            "ema10": "close_10_ema",
-            "bbands": "boll",
-            "bollinger": "boll",
-            "bollinger bands": "boll",
-            "macd_signal": "macds",
-            "macd_hist": "macdh",
             "50-day ma": "close_50_sma",
             "200-day ma": "close_200_sma",
             "50 day ma": "close_50_sma",
             "200 day ma": "close_200_sma",
+            "50-day moving average": "close_50_sma",
+            "200-day moving average": "close_200_sma",
+            "50 day moving average": "close_50_sma",
+            "200 day moving average": "close_200_sma",
+            "50-day simple moving average": "close_50_sma",
+            "200-day simple moving average": "close_200_sma",
+            
+            # EMA 變體
+            "ema10": "close_10_ema",
+            "10-day ema": "close_10_ema",
+            "10 day ema": "close_10_ema",
+            
+            # Bollinger Bands
+            "bbands": "boll",
+            "bollinger": "boll",
+            "bollinger bands": "boll",
             "bb": "boll",
+            
+            # MACD 變體
+            "macd_signal": "macds",
+            "macd signal": "macds",
+            "macd_hist": "macdh",
+            "macd histogram": "macdh",
+            
+            # 其他常見別名
+            "relative strength index": "rsi",
+            "average true range": "atr",
+            "money flow index": "mfi",
         }
         
         # 如果在映射中，使用映射名稱

tradingagents/dataflows/alpha_vantage_common.py

@@ -1,6 +1,6 @@
 import os
 import requests
-import pandas as pd
+import polars as pl
 import json
 from datetime import datetime
 from io import StringIO
@@ -102,20 +102,22 @@ def _filter_csv_by_date_range(csv_data: str, start_date: str, end_date: str) ->
 
     try:
         # 解析 CSV 數據
-        df = pd.read_csv(StringIO(csv_data))
+        df = pl.read_csv(StringIO(csv_data))
 
         # 假設第一欄是日期欄 (時間戳)
         date_col = df.columns[0]
-        df[date_col] = pd.to_datetime(df[date_col])
+        df = df.with_columns(pl.col(date_col).str.to_datetime())
 
         # 按日期範圍過濾
-        start_dt = pd.to_datetime(start_date)
+        start_dt = datetime.strptime(start_date, "%Y-%m-%d")
-        end_dt = pd.to_datetime(end_date)
+        end_dt = datetime.strptime(end_date, "%Y-%m-%d")
 
-        filtered_df = df[(df[date_col] >= start_dt) & (df[date_col] <= end_dt)]
+        filtered_df = df.filter(
+            (pl.col(date_col) >= start_dt) & (pl.col(date_col) <= end_dt)
+        )
 
         # 轉換回 CSV 字串
-        return filtered_df.to_csv(index=False)
+        return filtered_df.write_csv()
 
     except Exception as e:
         # 如果過濾失敗，返回原始數據並附帶警告

tradingagents/dataflows/local.py

@@ -1,5 +1,5 @@
 from typing import Annotated
-import pandas as pd
+import polars as pl
 import os
 from .config import DATA_DIR
 from datetime import datetime
@@ -30,29 +30,28 @@ def get_YFin_data_window(
     start_date = before.strftime("%Y-%m-%d")
 
     # 讀取數據
-    data = pd.read_csv(
+    data = pl.read_csv(
         os.path.join(
             DATA_DIR,
             f"market_data/price_data/{symbol}-YFin-data-2015-01-01-2025-03-25.csv",
         )
     )
 
-    # 僅提取日期部分進行比較
+    # 節取日期部分
-    data["DateOnly"] = data["Date"].str[:10]
+    data = data.with_columns(
+        pl.col("Date").str.slice(0, 10).alias("DateOnly")
+    )
 
     # 過濾指定日期範圍內的數據 (包含起訖日期)
-    filtered_data = data[
+    filtered_data = data.filter(
-        (data["DateOnly"] >= start_date) & (data["DateOnly"] <= curr_date)
+        (pl.col("DateOnly") >= start_date) & (pl.col("DateOnly") <= curr_date)
-    ]
+    )
 
     # 刪除我們創建的臨時欄位
-    filtered_data = filtered_data.drop("DateOnly", axis=1)
+    filtered_data = filtered_data.drop("DateOnly")
 
-    # 設定 pandas 顯示選項以顯示完整的 DataFrame
+    # polars 的字串輸出
-    with pd.option_context(
+    df_string = str(filtered_data)
-        "display.max_rows", None, "display.max_columns", None, "display.width", None
-    ):
-        df_string = filtered_data.to_string()
 
     return (
         f"## {symbol} 從 {start_date} 到 {curr_date} 的原始市場數據：\n\n"
@@ -76,7 +75,7 @@ def get_YFin_data(
         pd.DataFrame: 包含過濾後數據的 DataFrame。
     """
     # 讀取數據
-    data = pd.read_csv(
+    data = pl.read_csv(
         os.path.join(
             DATA_DIR,
             f"market_data/price_data/{symbol}-YFin-data-2015-01-01-2025-03-25.csv",
@@ -88,19 +87,21 @@ def get_YFin_data(
             f"Get_YFin_Data：{end_date} 超出 2015-01-01 到 2025-03-25 的數據範圍"
         )
 
-    # 僅提取日期部分進行比較
+    # 節取日期部分
-    data["DateOnly"] = data["Date"].str[:10]
+    data = data.with_columns(
+        pl.col("Date").str.slice(0, 10).alias("DateOnly")
+    )
 
-    # 過濾指定日期範圍內的數據 (包含起訖日期)
+    # 過濾指定日期範圏內的數據 (包含起訖日期)
-    filtered_data = data[
+    filtered_data = data.filter(
-        (data["DateOnly"] >= start_date) & (data["DateOnly"] <= end_date)
+        (pl.col("DateOnly") >= start_date) & (pl.col("DateOnly") <= end_date)
-    ]
+    )
 
     # 刪除我們創建的臨時欄位
-    filtered_data = filtered_data.drop("DateOnly", axis=1)
+    filtered_data = filtered_data.drop("DateOnly")
 
-    # 從數據框中移除索引
+    # 重置索引 (在 polars 中不需要，但保持一致性)
-    filtered_data = filtered_data.reset_index(drop=True)
+    # filtered_data = filtered_data.with_row_count(name="index", offset=0)
 
     return filtered_data
 
@@ -280,28 +281,34 @@ def get_simfin_balance_sheet(
         "us",
         f"us-balance-{freq}.csv",
     )
-    df = pd.read_csv(data_path, sep=";")
+    df = pl.read_csv(data_path, separator=";")
 
     # 將日期字串轉換為日期時間物件並移除任何時間部分
-    df["Report Date"] = pd.to_datetime(df["Report Date"], utc=True).dt.normalize()
+    df = df.with_columns(
-    df["Publish Date"] = pd.to_datetime(df["Publish Date"], utc=True).dt.normalize()
+        pl.col("Report Date").str.to_datetime().dt.date().alias("Report Date"),
+        pl.col("Publish Date").str.to_datetime().dt.date().alias("Publish Date")
+    )
 
     # 將當前日期轉換為日期時間並標準化
-    curr_date_dt = pd.to_datetime(curr_date, utc=True).normalize()
+    from datetime import datetime as dt
+    curr_date_dt = dt.strptime(curr_date, "%Y-%m-%d").date()
 
     # 過濾 DataFrame，篩選出給定股票代碼且報告發布日期在當前日期或之前的報告
-    filtered_df = df[(df["Ticker"] == ticker) & (df["Publish Date"] <= curr_date_dt)]
+    filtered_df = df.filter(
+        (pl.col("Ticker") == ticker) & (pl.col("Publish Date") <= curr_date_dt)
+    )
 
     # 檢查是否有可用的報告；如果沒有，則返回通知
-    if filtered_df.empty:
+    if filtered_df.is_empty():
         print("在給定的當前日期之前沒有可用的資產負債表。")
         return ""
 
     # 通過選擇具有最新發布日期的行來獲取最新的資產負債表
-    latest_balance_sheet = filtered_df.loc[filtered_df["Publish Date"].idxmax()]
+    max_date_idx = filtered_df.select(pl.col("Publish Date")).arg_max()
+    latest_balance_sheet = filtered_df.row(max_date_idx, named=True)
 
     # 刪除 SimFinID 欄位
-    latest_balance_sheet = latest_balance_sheet.drop("SimFinId")
+    # latest_balance_sheet = latest_balance_sheet.drop("SimFinId")
 
     return (
         f"## {ticker} 於 {str(latest_balance_sheet['Publish Date'])[0:10]} 發布的 {freq} 資產負債表：\n"
@@ -327,25 +334,31 @@ def get_simfin_cashflow(
         "us",
         f"us-cashflow-{freq}.csv",
     )
-    df = pd.read_csv(data_path, sep=";")
+    df = pl.read_csv(data_path, separator=";")
 
     # 將日期字串轉換為日期時間物件並移除任何時間部分
-    df["Report Date"] = pd.to_datetime(df["Report Date"], utc=True).dt.normalize()
+    df = df.with_columns(
-    df["Publish Date"] = pd.to_datetime(df["Publish Date"], utc=True).dt.normalize()
+        pl.col("Report Date").str.to_datetime().dt.date().alias("Report Date"),
+        pl.col("Publish Date").str.to_datetime().dt.date().alias("Publish Date")
+    )
 
     # 將當前日期轉換為日期時間並標準化
-    curr_date_dt = pd.to_datetime(curr_date, utc=True).normalize()
+    from datetime import datetime as dt
+    curr_date_dt = dt.strptime(curr_date, "%Y-%m-%d").date()
 
     # 過濾 DataFrame，篩選出給定股票代碼且報告發布日期在當前日期或之前的報告
-    filtered_df = df[(df["Ticker"] == ticker) & (df["Publish Date"] <= curr_date_dt)]
+    filtered_df = df.filter(
+        (pl.col("Ticker") == ticker) & (pl.col("Publish Date") <= curr_date_dt)
+    )
 
     # 檢查是否有可用的報告；如果沒有，則返回通知
-    if filtered_df.empty:
+    if filtered_df.is_empty():
         print("在給定的當前日期之前沒有可用的現金流量表。")
         return ""
 
     # 通過選擇具有最新發布日期的行來獲取最新的現金流量表
-    latest_cash_flow = filtered_df.loc[filtered_df["Publish Date"].idxmax()]
+    max_date_idx = filtered_df.select(pl.col("Publish Date")).arg_max()
+    latest_cash_flow = filtered_df.row(max_date_idx, named=True)
 
     # 刪除 SimFinID 欄位
     latest_cash_flow = latest_cash_flow.drop("SimFinId")
@@ -374,25 +387,31 @@ def get_simfin_income_statements(
         "us",
         f"us-income-{freq}.csv",
     )
-    df = pd.read_csv(data_path, sep=";")
+    df = pl.read_csv(data_path, separator=";")
 
     # 將日期字串轉換為日期時間物件並移除任何時間部分
-    df["Report Date"] = pd.to_datetime(df["Report Date"], utc=True).dt.normalize()
+    df = df.with_columns(
-    df["Publish Date"] = pd.to_datetime(df["Publish Date"], utc=True).dt.normalize()
+        pl.col("Report Date").str.to_datetime().dt.date().alias("Report Date"),
+        pl.col("Publish Date").str.to_datetime().dt.date().alias("Publish Date")
+    )
 
     # 將當前日期轉換為日期時間並標準化
-    curr_date_dt = pd.to_datetime(curr_date, utc=True).normalize()
+    from datetime import datetime as dt
+    curr_date_dt = dt.strptime(curr_date, "%Y-%m-%d").date()
 
     # 過濾 DataFrame，篩選出給定股票代碼且報告發布日期在當前日期或之前的報告
-    filtered_df = df[(df["Ticker"] == ticker) & (df["Publish Date"] <= curr_date_dt)]
+    filtered_df = df.filter(
+        (pl.col("Ticker") == ticker) & (pl.col("Publish Date") <= curr_date_dt)
+    )
 
     # 檢查是否有可用的報告；如果沒有，則返回通知
-    if filtered_df.empty:
+    if filtered_df.is_empty():
         print("在給定的當前日期之前沒有可用的損益表。")
         return ""
 
     # 通過選擇具有最新發布日期的行來獲取最新的損益表
-    latest_income = filtered_df.loc[filtered_df["Publish Date"].idxmax()]
+    max_date_idx = filtered_df.select(pl.col("Publish Date")).arg_max()
+    latest_income = filtered_df.row(max_date_idx, named=True)
 
     # 刪除 SimFinID 欄位
     latest_income = latest_income.drop("SimFinId")

tradingagents/dataflows/stockstats_utils.py

@@ -1,4 +1,4 @@
-import pandas as pd
+import polars as pl
 import yfinance as yf
 from stockstats import wrap
 from typing import Annotated
@@ -9,6 +9,7 @@ from .config import get_config, DATA_DIR
 class StockstatsUtils:
     """
     一個提供股票統計功能的工具類別。
+    注意: stockstats 函式庫需要 pandas DataFrame，所以需要進行 pandas/polars 轉換。
     """
     @staticmethod
     def get_stock_stats(
@@ -31,6 +32,8 @@ class StockstatsUtils:
         Returns:
             float or str: 指標值或錯誤訊息。
         """
+        from datetime import datetime, timedelta
+        
         # 獲取設定並設定數據目錄路徑
         config = get_config()
         online = config["data_vendors"]["technical_indicators"] != "local"
@@ -40,51 +43,56 @@ class StockstatsUtils:
 
         if not online:
             try:
-                data = pd.read_csv(
+                data = pl.read_csv(
                     os.path.join(
                         DATA_DIR,
                         f"{symbol}-YFin-data-2015-01-01-2025-03-25.csv",
                     )
                 )
-                df = wrap(data)
+                # stockstats 需要 pandas DataFrame
+                data_pd = data.to_pandas()
+                df = wrap(data_pd)
             except FileNotFoundError:
                 raise Exception("Stockstats 失敗：尚未獲取 Yahoo Finance 數據！")
         else:
             # 獲取今天的日期 (YYYY-mm-dd) 以添加到快取
-            today_date = pd.Timestamp.today()
+            today_date = datetime.now()
-            curr_date = pd.to_datetime(curr_date)
+            curr_date_dt = datetime.strptime(curr_date, "%Y-%m-%d")
 
             end_date = today_date
-            start_date = today_date - pd.DateOffset(years=15)
+            start_date = today_date - timedelta(days=365*15)
-            start_date = start_date.strftime("%Y-%m-%d")
+            start_date_str = start_date.strftime("%Y-%m-%d")
-            end_date = end_date.strftime("%Y-%m-%d")
+            end_date_str = end_date.strftime("%Y-%m-%d")
 
             # 獲取設定並確保快取目錄存在
             os.makedirs(config["data_cache_dir"], exist_ok=True)
 
             data_file = os.path.join(
                 config["data_cache_dir"],
-                f"{symbol}-YFin-data-{start_date}-{end_date}.csv",
+                f"{symbol}-YFin-data-{start_date_str}-{end_date_str}.csv",
             )
 
             if os.path.exists(data_file):
-                data = pd.read_csv(data_file)
+                data = pl.read_csv(data_file)
-                data["Date"] = pd.to_datetime(data["Date"])
+                data = data.with_columns(pl.col("Date").str.to_datetime())
             else:
-                data = yf.download(
+                data_yf = yf.download(
                     symbol,
-                    start=start_date,
+                    start=start_date_str,
-                    end=end_date,
+                    end=end_date_str,
                     multi_level_index=False,
                     progress=False,
                     auto_adjust=True,
                 )
-                data = data.reset_index()
+                data_yf = data_yf.reset_index()
-                data.to_csv(data_file, index=False)
+                data_yf.to_csv(data_file, index=False)
+                data = pl.from_pandas(data_yf)
 
-            df = wrap(data)
+            # stockstats 需要 pandas DataFrame
+            data_pd = data.to_pandas()
+            df = wrap(data_pd)
             df["Date"] = df["Date"].dt.strftime("%Y-%m-%d")
-            curr_date = curr_date.strftime("%Y-%m-%d")
+            curr_date = curr_date_dt.strftime("%Y-%m-%d")
 
         df[indicator]  # 觸發 stockstats 計算指標
         matching_rows = df[df["Date"].str.startswith(curr_date)]

tradingagents/dataflows/utils.py

@@ -1,22 +1,22 @@
 import os
 import json
-import pandas as pd
+import polars as pl
 from datetime import date, timedelta, datetime
 from typing import Annotated
 
 SavePathType = Annotated[str, "儲存資料的檔案路徑。如果為 None，則不儲存資料。"]
 
-def save_output(data: pd.DataFrame, tag: str, save_path: SavePathType = None) -> None:
+def save_output(data: pl.DataFrame, tag: str, save_path: SavePathType = None) -> None:
     """
     將 DataFrame 儲存到 CSV 檔案。
 
     Args:
-        data (pd.DataFrame): 要儲存的 DataFrame。
+        data (pl.DataFrame): 要儲存的 DataFrame。
         tag (str): 用於在控制台中打印的標籤。
         save_path (SavePathType, optional): 儲存檔案的路徑。預設為 None。
     """
     if save_path:
-        data.to_csv(save_path)
+        data.write_csv(save_path)
         print(f"{tag} 已儲存至 {save_path}")
 
 

tradingagents/dataflows/y_finance.py

@@ -229,6 +229,7 @@ def _get_stock_stats_bulk(
     返回將日期字串映射到指標值的字典。
     """
     from .config import get_config
+    import polars as pl
     import pandas as pd
     from stockstats import wrap
     import os
@@ -239,22 +240,25 @@ def _get_stock_stats_bulk(
     if not online:
         # 本地數據路徑
         try:
-            data = pd.read_csv(
+            data = pl.read_csv(
                 os.path.join(
                     config.get("data_cache_dir", "data"),
                     f"{symbol}-YFin-data-2015-01-01-2025-03-25.csv",
                 )
             )
-            df = wrap(data)
+            # stockstats 需要 pandas DataFrame
+            data_pd = data.to_pandas()
+            df = wrap(data_pd)
         except FileNotFoundError:
             raise Exception("Stockstats 失敗：尚未獲取 Yahoo Finance 數據！")
     else:
         # 帶有快取的線上數據獲取
-        today_date = pd.Timestamp.today()
+        from datetime import datetime as dt, timedelta
-        curr_date_dt = pd.to_datetime(curr_date)
+        today_date = dt.now()
+        curr_date_dt = dt.strptime(curr_date, "%Y-%m-%d")
         
         end_date = today_date
-        start_date = today_date - pd.DateOffset(years=15)
+        start_date = today_date - timedelta(days=365*15)
         start_date_str = start_date.strftime("%Y-%m-%d")
         end_date_str = end_date.strftime("%Y-%m-%d")
         
@@ -279,8 +283,10 @@ def _get_stock_stats_bulk(
                 logger.info(f"{symbol} 緩存過期（年齡：{cache_age_hours:.1f} 小時），將重新下載")
         
         if cache_valid:
-            data = pd.read_csv(data_file)
+            data_pl = pl.read_csv(data_file)
-            data["Date"] = pd.to_datetime(data["Date"])
+            data_pl = data_pl.with_columns(pl.col("Date").str.to_datetime())
+            # stockstats 需要 pandas DataFrame
+            data = data_pl.to_pandas()
         else:
             # 使用重試機制下載數據
             @retry(max_attempts=3, backoff=2.0)
@@ -305,8 +311,9 @@ def _get_stock_stats_bulk(
                 # 如果下載失敗但有舊緩存，使用舊緩存
                 if os.path.exists(data_file):
                     logger.warning(f"使用過期緩存作為備援")
-                    data = pd.read_csv(data_file)
+                    data_pl = pl.read_csv(data_file)
-                    data["Date"] = pd.to_datetime(data["Date"])
+                    data_pl = data_pl.with_columns(pl.col("Date").str.to_datetime())
+                    data = data_pl.to_pandas()
                 else:
                     raise
         

tradingagents/dataflows/yfin_utils.py

@@ -2,8 +2,7 @@
 
 import yfinance as yf
 from typing import Annotated, Callable, Any, Optional
-from pandas import DataFrame
+import polars as pl
-import pandas as pd
 from functools import wraps
 
 from .utils import save_output, SavePathType, decorate_all_methods
@@ -35,15 +34,18 @@ class YFinanceUtils:
             str, "檢索股價數據的結束日期，格式為 YYYY-mm-dd"
         ],
         save_path: SavePathType = None,
-    ) -> DataFrame:
+    ) -> pl.DataFrame:
         """檢索指定股票代碼的股價數據"""
+        from datetime import datetime, timedelta
         ticker = symbol
         # 將結束日期加一天，使數據範圍包含結束日期
-        end_date = pd.to_datetime(end_date) + pd.DateOffset(days=1)
+        end_date_dt = datetime.strptime(end_date, "%Y-%m-%d") + timedelta(days=1)
-        end_date = end_date.strftime("%Y-%m-%d")
+        end_date = end_date_dt.strftime("%Y-%m-%d")
         stock_data = ticker.history(start=start_date, end=end_date)
-        # save_output(stock_data, f"{ticker.ticker} 的股票數據", save_path)
+        # 轉換為 polars DataFrame
-        return stock_data
+        stock_data_pl = pl.from_pandas(stock_data.reset_index())
+        # save_output(stock_data_pl, f"{ticker.ticker} 的股票數據", save_path)
+        return stock_data_pl
 
     def get_stock_info(
         symbol: Annotated[str, "股票代碼"],
@@ -56,7 +58,7 @@ class YFinanceUtils:
     def get_company_info(
         symbol: Annotated[str, "股票代碼"],
         save_path: Optional[str] = None,
-    ) -> DataFrame:
+    ) -> pl.DataFrame:
         """獲取並以 DataFrame 形式返回公司資訊。"""
         ticker = symbol
         info = ticker.info
@@ -67,41 +69,42 @@ class YFinanceUtils:
             "國家": info.get("country", "N/A"),
             "網站": info.get("website", "N/A"),
         }
-        company_info_df = DataFrame([company_info])
+        company_info_df = pl.DataFrame([company_info])
         if save_path:
-            company_info_df.to_csv(save_path)
+            company_info_df.write_csv(save_path)
             print(f"{ticker.ticker} 的公司資訊已儲存至 {save_path}")
         return company_info_df
 
     def get_stock_dividends(
         symbol: Annotated[str, "股票代碼"],
         save_path: Optional[str] = None,
-    ) -> DataFrame:
+    ) -> pl.DataFrame:
         """獲取並以 DataFrame 形式返回最新的股息數據。"""
         ticker = symbol
         dividends = ticker.dividends
+        dividends_pl = pl.from_pandas(dividends.reset_index())
         if save_path:
-            dividends.to_csv(save_path)
+            dividends_pl.write_csv(save_path)
             print(f"{ticker.ticker} 的股息已儲存至 {save_path}")
-        return dividends
+        return dividends_pl
 
-    def get_income_stmt(symbol: Annotated[str, "股票代碼"]) -> DataFrame:
+    def get_income_stmt(symbol: Annotated[str, "股票代碼"]) -> pl.DataFrame:
         """獲取並以 DataFrame 形式返回公司最新的損益表。"""
         ticker = symbol
         income_stmt = ticker.financials
-        return income_stmt
+        return pl.from_pandas(income_stmt.reset_index())
 
-    def get_balance_sheet(symbol: Annotated[str, "股票代碼"]) -> DataFrame:
+    def get_balance_sheet(symbol: Annotated[str, "股票代碼"]) -> pl.DataFrame:
         """獲取並以 DataFrame 形式返回公司最新的資產負債表。"""
         ticker = symbol
         balance_sheet = ticker.balance_sheet
-        return balance_sheet
+        return pl.from_pandas(balance_sheet.reset_index())
 
-    def get_cash_flow(symbol: Annotated[str, "股票代碼"]) -> DataFrame:
+    def get_cash_flow(symbol: Annotated[str, "股票代碼"]) -> pl.DataFrame:
         """獲取並以 DataFrame 形式返回公司最新的現金流量表。"""
         ticker = symbol
         cash_flow = ticker.cashflow
-        return cash_flow
+        return pl.from_pandas(cash_flow.reset_index())
 
     def get_analyst_recommendations(symbol: Annotated[str, "股票代碼"]) -> tuple:
         """獲取最新的分析師建議，並返回最常見的建議及其計數。"""