aria-code 0.52.0__py3-none-any.whl

aria_code/tools/local_finance_tools.py

@@ -0,0 +1,3417 @@
+"""
+local_finance_tools.py — Fully-offline financial tool implementations for Aria Code.
+
+Purpose
+-------
+When running in local_mode (no Arthera backend), this module provides drop-in
+replacements for every ARIA_TOOLS entry using open-source data libraries:
+
+  A股 data    → akshare
+  US/Global   → yfinance
+  Crypto      → ccxt
+  Backtesting → vectorbt (or pandas fallback)
+  Technical   → pandas_ta (or ta-lib if installed)
+  Risk        → scipy / numpy
+
+Each tool follows the same contract as the remote Aria tools:
+  handler(params: dict) -> dict  (always returns a dict, never raises)
+
+Registration
+------------
+Call ``register_local_finance_tools(LOCAL_TOOLS, LOCAL_TOOL_SCHEMAS)`` at
+startup to extend the CLI's tool registry.  The function is idempotent.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import traceback
+from importlib.util import find_spec
+from datetime import datetime, timedelta
+from typing import Any, Dict, List, Optional, Tuple
+
+import numpy as np
+import pandas as pd
+from aria_code.packages.aria_core.paths import aria_home
+
+logger = logging.getLogger(__name__)
+
+# ---------------------------------------------------------------------------
+# Alibaba Cloud data client (optional — degrades gracefully when offline)
+# ---------------------------------------------------------------------------
+
+try:
+    from aria_code.aliyun_data_client import (
+        AliyunDataClient,
+        cloud_get_quote_sync,
+        cloud_get_history_sync,
+        cloud_get_factors_sync,
+        cloud_get_ai_signal_sync,
+    )
+    _HAS_CLOUD = True
+    logger.debug("Alibaba Cloud data client loaded ✓")
+except ImportError:
+    _HAS_CLOUD = False
+    logger.debug("aliyun_data_client not found — cloud features disabled")
+
+# ---------------------------------------------------------------------------
+# Optional dependency guards
+# ---------------------------------------------------------------------------
+
+try:
+    import yfinance as yf
+    _HAS_YF = True
+except ImportError:
+    _HAS_YF = False
+
+try:
+    import akshare as ak
+    _HAS_AK = True
+except ImportError:
+    _HAS_AK = False
+
+
+def _ak_retry(fn, *args, _tries: int = 3, _delay: float = 0.8, **kwargs):
+    """Call an akshare function with retries, bypassing a broken proxy.
+
+    Two transient failure modes are handled:
+      1. akshare hits numbered eastmoney hosts (NN.push2.eastmoney.com) that go
+         down individually — a retry usually lands on a healthy host.
+      2. A misconfigured HTTP(S)_PROXY raises ProxyError even though the source
+         is directly reachable. On a proxy/connection error we retry with the
+         proxy env vars temporarily cleared (akshare uses requests trust_env),
+         then restore them so the rest of the app is unaffected.
+    """
+    import os as _os
+    import time as _t
+    _PROXY_VARS = ("HTTP_PROXY", "HTTPS_PROXY", "http_proxy", "https_proxy", "ALL_PROXY", "all_proxy")
+    last_exc = None
+    for _i in range(_tries):
+        _saved = {}
+        _bypass = _i > 0  # first attempt honours the proxy; retries bypass it
+        if _bypass:
+            for _v in _PROXY_VARS:
+                if _v in _os.environ:
+                    _saved[_v] = _os.environ.pop(_v)
+        try:
+            return fn(*args, **kwargs)
+        except Exception as exc:  # noqa: BLE001 — akshare raises many types
+            last_exc = exc
+            if _i < _tries - 1:
+                _t.sleep(_delay)
+        finally:
+            for _v, _val in _saved.items():
+                _os.environ[_v] = _val
+    raise last_exc
+
+try:
+    import ccxt
+    _HAS_CCXT = True
+except ImportError:
+    _HAS_CCXT = False
+
+_HAS_TA = find_spec("pandas_ta") is not None
+
+try:
+    import vectorbt as vbt
+    _HAS_VBT = True
+except ImportError:
+    _HAS_VBT = False
+
+try:
+    from scipy import stats as sp_stats
+    _HAS_SCIPY = True
+except ImportError:
+    _HAS_SCIPY = False
+
+
+# ---------------------------------------------------------------------------
+# Helper utilities
+# ---------------------------------------------------------------------------
+
+def _safe(fn, params, *args, **kwargs):
+    """Wrap a tool handler so it never raises — returns error dict instead."""
+    try:
+        return fn(params, *args, **kwargs)
+    except Exception as exc:
+        logger.debug("Local finance tool error: %s", traceback.format_exc())
+        return {"success": False, "error": str(exc), "traceback": traceback.format_exc()[-500:]}
+
+
+def _parse_date(s: Optional[str], default_days_back: int = 365) -> str:
+    if s:
+        return s
+    return (datetime.today() - timedelta(days=default_days_back)).strftime("%Y-%m-%d")
+
+
+def _today() -> str:
+    return datetime.today().strftime("%Y-%m-%d")
+
+
+def _is_ashare(symbol: str) -> bool:
+    s = symbol.strip().lower()
+    return (
+        s.startswith("sh") or s.startswith("sz")
+        or (len(s) == 6 and s.isdigit())
+        or s.endswith(".ss") or s.endswith(".sz")
+    )
+
+
+def _normalise_ashare(symbol: str) -> str:
+    s = symbol.strip().lower().replace(".ss", "").replace(".sz", "")
+    s = s.replace("sh", "").replace("sz", "")
+    if len(s) == 6 and s.isdigit():
+        prefix = "sh" if s.startswith("6") else "sz"
+        return prefix + s
+    return s
+
+
+def _get_pandas_ta():
+    if not _HAS_TA:
+        return None
+    try:
+        import pandas_ta as ta
+    except Exception:
+        return None
+    return ta
+
+
+# ---------------------------------------------------------------------------
+# 1. get_market_data
+# ---------------------------------------------------------------------------
+
+def _get_market_data(params: dict) -> dict:
+    symbol   = params.get("symbol", "AAPL").upper()
+    period   = params.get("period", "1y")        # yfinance period
+    interval = params.get("interval", "1d")
+    start    = params.get("start")
+    end      = params.get("end", _today())
+
+    if _is_ashare(symbol):
+        return _get_ashare_data(symbol, start or _parse_date(None, 365), end)
+
+    # ── Try Alibaba Cloud first (for US/global symbols) ────────────────
+    if _HAS_CLOUD:
+        try:
+            cloud_q = cloud_get_quote_sync(symbol)
+            if cloud_q and cloud_q.get("price"):
+                return {
+                    "success":      True,
+                    "symbol":       symbol,
+                    "latest_close": round(float(cloud_q.get("price", 0)), 4),
+                    "change_pct":   round(float(cloud_q.get("change_percent", 0)), 3),
+                    "volume":       int(cloud_q.get("volume", 0)),
+                    "high":         round(float(cloud_q.get("high", 0)), 4),
+                    "low":          round(float(cloud_q.get("low", 0)), 4),
+                    "open":         round(float(cloud_q.get("open", 0)), 4),
+                    "name":         cloud_q.get("name", symbol),
+                    "market":       cloud_q.get("market", "US"),
+                    "provider":     "aliyun_cloud",
+                }
+        except Exception as exc:
+            logger.debug("Cloud quote failed for %s: %s — falling back to yfinance", symbol, exc)
+
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: pip install yfinance"}
+
+    try:
+        tkr  = yf.Ticker(symbol)
+        if start:
+            hist = tkr.history(start=start, end=end, interval=interval)
+        else:
+            hist = tkr.history(period=period, interval=interval)
+
+        if hist.empty:
+            return {"success": False, "error": f"No data for {symbol}"}
+
+        info        = tkr.fast_info
+        latest      = hist.iloc[-1]
+        prev        = hist.iloc[-2] if len(hist) > 1 else latest
+        chg         = (latest["Close"] - prev["Close"]) / prev["Close"] * 100
+
+        return {
+            "success":       True,
+            "symbol":        symbol,
+            "latest_close":  round(float(latest["Close"]), 4),
+            "change_pct":    round(float(chg), 3),
+            "volume":        int(latest["Volume"]),
+            "high_52w":      round(float(info.year_high),  4) if hasattr(info, "year_high")  else None,
+            "low_52w":       round(float(info.year_low),   4) if hasattr(info, "year_low")   else None,
+            "market_cap":    getattr(info, "market_cap", None),
+            "currency":      getattr(info, "currency", "USD"),
+            "bars":          len(hist),
+            "history_tail":  _df_tail(hist, 5),
+            "provider":      "yfinance",
+        }
+    except Exception as exc:
+        return {"success": False, "error": str(exc)}
+
+
+def _get_ashare_data(symbol: str, start: str, end: str) -> dict:
+    # ── Try Alibaba Cloud akshare_data_server first ────────────────────
+    if _HAS_CLOUD:
+        try:
+            cloud_hist = cloud_get_history_sync(symbol, start=start, end=end)
+            if cloud_hist and cloud_hist.get("data"):
+                rows = cloud_hist["data"]
+                if rows:
+                    latest   = rows[-1]
+                    prev     = rows[-2] if len(rows) > 1 else latest
+                    cl, pc   = float(latest.get("close", 0)), float(prev.get("close", 0) or 0.0001)
+                    chg      = (cl - pc) / pc * 100
+                    # Build history_tail
+                    tail = [
+                        {k: v for k, v in r.items() if k in ("date", "open", "high", "low", "close", "volume")}
+                        for r in rows[-5:]
+                    ]
+                    return {
+                        "success":       True,
+                        "symbol":        symbol,
+                        "latest_close":  round(cl, 3),
+                        "change_pct":    round(chg, 3),
+                        "volume":        int(latest.get("volume", 0)),
+                        "bars":          len(rows),
+                        "history_tail":  tail,
+                        "provider":      "aliyun_data",
+                    }
+        except Exception as exc:
+            logger.debug("Cloud history failed for %s: %s — falling back to akshare", symbol, exc)
+
+    if not _HAS_AK:
+        return {"success": False, "error": "akshare not installed: pip install akshare"}
+
+    norm = _normalise_ashare(symbol)
+    code = norm[2:]  # strip sh/sz prefix
+    try:
+        df = ak.stock_zh_a_hist(symbol=code, period="daily",
+                                 start_date=start.replace("-", ""),
+                                 end_date=end.replace("-", ""),
+                                 adjust="qfq")
+        if df is None or df.empty:
+            return {"success": False, "error": f"No A-share data for {symbol}"}
+
+        df = df.rename(columns={
+            "日期": "date", "开盘": "open", "收盘": "close",
+            "最高": "high", "最低": "low", "成交量": "volume",
+            "成交额": "amount", "换手率": "turnover_rate",
+        })
+        latest = df.iloc[-1]
+        prev   = df.iloc[-2] if len(df) > 1 else latest
+        chg    = (latest["close"] - prev["close"]) / prev["close"] * 100
+
+        return {
+            "success":       True,
+            "symbol":        symbol,
+            "latest_close":  round(float(latest["close"]), 3),
+            "change_pct":    round(float(chg), 3),
+            "volume":        int(latest["volume"]),
+            "turnover_rate": float(latest.get("turnover_rate", 0) or 0),
+            "bars":          len(df),
+            "history_tail":  _df_tail(df.rename(columns={"close": "Close", "volume": "Volume"}), 5),
+            "provider":      "akshare",
+        }
+    except Exception as exc:
+        return {"success": False, "error": str(exc)}
+
+
+# ---------------------------------------------------------------------------
+# 2. get_crypto_data
+# ---------------------------------------------------------------------------
+
+def _get_crypto_data(params: dict) -> dict:
+    symbol   = params.get("symbol", "BTC/USDT").upper().replace("-", "/")
+    exchange = params.get("exchange", "binance")
+    timeframe = params.get("timeframe", "1d")
+    limit    = int(params.get("limit", 100))
+
+    if not _HAS_CCXT:
+        # Fallback to yfinance for common crypto tickers
+        if _HAS_YF:
+            yf_sym = symbol.replace("/", "-") + ("" if symbol.endswith("USD") else "")
+            return _get_market_data({"symbol": yf_sym, "period": "3mo"})
+        return {"success": False, "error": "ccxt not installed: pip install ccxt"}
+
+    def _yf_crypto_fallback(reason: str) -> dict:
+        """Fall back to yfinance when the exchange is unreachable (region block, etc.)."""
+        if not _HAS_YF:
+            return {"success": False, "error": reason}
+        # BTC/USDT → BTC-USD ; strip stablecoin quote to USD for yfinance
+        base = symbol.split("/")[0]
+        yf_sym = f"{base}-USD"
+        res = _get_market_data({"symbol": yf_sym, "period": "3mo"})
+        if res.get("success"):
+            res["provider"] = "yfinance (ccxt fallback)"
+            res["note"] = f"{exchange} 不可用，已回退 yfinance: {reason[:60]}"
+        return res
+
+    try:
+        ex_class = getattr(ccxt, exchange.lower(), None)
+        if ex_class is None:
+            return {"success": False, "error": f"Unknown exchange: {exchange}"}
+        ex = ex_class({"enableRateLimit": True})
+        try:
+            ohlcv = ex.fetch_ohlcv(symbol, timeframe=timeframe, limit=limit)
+        except Exception as net_exc:
+            # Network / region block (e.g. Binance 451) → yfinance fallback
+            return _yf_crypto_fallback(str(net_exc))
+        if not ohlcv:
+            return _yf_crypto_fallback("Empty OHLCV data")
+
+        df = pd.DataFrame(ohlcv, columns=["ts", "open", "high", "low", "close", "volume"])
+        df["date"] = pd.to_datetime(df["ts"], unit="ms")
+        latest   = df.iloc[-1]
+        prev     = df.iloc[-2]
+        chg      = (latest["close"] - prev["close"]) / prev["close"] * 100
+        vol_avg  = df["volume"].tail(20).mean()
+
+        # Ticker for bid/ask
+        try:
+            ticker = ex.fetch_ticker(symbol)
+        except Exception:
+            ticker = {}
+
+        return {
+            "success":       True,
+            "symbol":        symbol,
+            "exchange":      exchange,
+            "latest_close":  round(float(latest["close"]), 6),
+            "change_pct_24h": round(float(chg), 3),
+            "volume_24h":    float(latest["volume"]),
+            "volume_avg_20d": round(float(vol_avg), 2),
+            "bid":           ticker.get("bid"),
+            "ask":           ticker.get("ask"),
+            "bars":          len(df),
+            "provider":      "ccxt",
+        }
+    except Exception as exc:
+        return _yf_crypto_fallback(str(exc))
+
+
+# ---------------------------------------------------------------------------
+# 3. get_forex_data
+# ---------------------------------------------------------------------------
+
+def _get_forex_data(params: dict) -> dict:
+    pair   = params.get("pair", "EURUSD=X")
+    period = params.get("period", "3mo")
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+    # Normalise to yfinance format
+    p = pair.replace("/", "").upper()
+    if not p.endswith("=X"):
+        p += "=X"
+    return _get_market_data({"symbol": p, "period": period})
+
+
+# ---------------------------------------------------------------------------
+# 4. calculate_factors
+# ---------------------------------------------------------------------------
+
+def _calculate_factors(params: dict) -> dict:
+    symbol = params.get("symbol", "AAPL")
+    period = params.get("period", "1y")
+
+    # ── Try Alibaba Cloud enhanced factor engine first ─────────────────
+    if _HAS_CLOUD:
+        try:
+            cloud_factors = cloud_get_factors_sync(symbol)
+            if cloud_factors and cloud_factors.get("success"):
+                cloud_factors["provider"] = "aliyun_cloud"
+                return cloud_factors
+        except Exception as exc:
+            logger.debug("Cloud factors failed for %s: %s — computing locally", symbol, exc)
+
+    if _is_ashare(symbol) and _HAS_AK:
+        data_result = _get_ashare_data(
+            symbol,
+            _parse_date(None, 365),
+            _today(),
+        )
+    elif _HAS_YF:
+        data_result = _get_market_data({"symbol": symbol, "period": period})
+    else:
+        return {"success": False, "error": "No data source available"}
+
+    if not data_result.get("success"):
+        return data_result
+
+    # Re-fetch full history for factor computation
+    if _is_ashare(symbol) and _HAS_AK:
+        norm = _normalise_ashare(symbol)
+        code = norm[2:]
+        df = ak.stock_zh_a_hist(symbol=code, period="daily",
+                                 start_date=_parse_date(None, 365).replace("-", ""),
+                                 end_date=_today().replace("-", ""),
+                                 adjust="qfq")
+        df = df.rename(columns={"收盘": "Close", "成交量": "Volume",
+                                 "开盘": "Open", "最高": "High", "最低": "Low"})
+    else:
+        tkr = yf.Ticker(symbol)
+        df  = tkr.history(period=period)
+
+    if df is None or len(df) < 20:
+        return {"success": False, "error": "Insufficient history for factors"}
+
+    close  = df["Close"].astype(float)
+    volume = df["Volume"].astype(float)
+    ret    = close.pct_change().dropna()
+
+    factors: Dict[str, Any] = {"symbol": symbol}
+
+    # ── Price momentum ─────────────────────────────────────────────────────
+    for n in (5, 10, 20, 60, 120):
+        if len(close) > n:
+            factors[f"return_{n}d"] = round(float(close.pct_change(n).iloc[-1]), 5)
+
+    # ── Moving averages ────────────────────────────────────────────────────
+    for n in (5, 10, 20, 60, 120, 200):
+        if len(close) >= n:
+            ma = close.rolling(n).mean().iloc[-1]
+            factors[f"ma_{n}"] = round(float(ma), 4)
+            factors[f"ma_{n}_gap"] = round(float(close.iloc[-1] / ma - 1), 5)
+
+    # ── Volatility ─────────────────────────────────────────────────────────
+    for n in (10, 20, 60):
+        if len(ret) >= n:
+            factors[f"volatility_{n}d"] = round(float(ret.tail(n).std() * np.sqrt(252)), 5)
+
+    # ── Volume ─────────────────────────────────────────────────────────────
+    if len(volume) >= 20:
+        vol_ma20 = volume.rolling(20).mean().iloc[-1]
+        factors["volume_ratio_20d"] = round(float(volume.iloc[-1] / vol_ma20), 3) if vol_ma20 > 0 else None
+
+    # ── RSI ────────────────────────────────────────────────────────────────
+    if _HAS_TA and len(close) >= 14:
+        ta = _get_pandas_ta()
+        if ta is not None:
+            rsi = ta.rsi(close, length=14)
+            if rsi is not None and not rsi.empty:
+                factors["rsi_14"] = round(float(rsi.iloc[-1]), 2)
+            else:
+                factors["rsi_14"] = None
+        else:
+            factors["rsi_14"] = None
+    else:
+        # Manual RSI
+        delta = close.diff()
+        gain  = delta.clip(lower=0).rolling(14).mean()
+        loss  = (-delta.clip(upper=0)).rolling(14).mean()
+        rs    = gain / loss.replace(0, np.nan)
+        rsi_v = (100 - 100 / (1 + rs)).iloc[-1]
+        factors["rsi_14"] = round(float(rsi_v), 2) if not np.isnan(rsi_v) else None
+
+    # ── MACD ───────────────────────────────────────────────────────────────
+    if len(close) >= 26:
+        ema12 = close.ewm(span=12, adjust=False).mean()
+        ema26 = close.ewm(span=26, adjust=False).mean()
+        macd  = ema12 - ema26
+        sig   = macd.ewm(span=9, adjust=False).mean()
+        factors["macd"]          = round(float(macd.iloc[-1]), 5)
+        factors["macd_signal"]   = round(float(sig.iloc[-1]),  5)
+        factors["macd_hist"]     = round(float((macd - sig).iloc[-1]), 5)
+
+    # ── Bollinger Bands ────────────────────────────────────────────────────
+    if len(close) >= 20:
+        bb_ma  = close.rolling(20).mean()
+        bb_std = close.rolling(20).std()
+        bb_up  = bb_ma + 2 * bb_std
+        bb_lo  = bb_ma - 2 * bb_std
+        prc    = close.iloc[-1]
+        factors["bb_position"] = round(float((prc - bb_lo.iloc[-1]) /
+                                              (bb_up.iloc[-1] - bb_lo.iloc[-1] + 1e-8)), 4)
+        factors["bb_upper"] = round(float(bb_up.iloc[-1]), 4)
+        factors["bb_lower"] = round(float(bb_lo.iloc[-1]), 4)
+
+    # ── Beta vs market ─────────────────────────────────────────────────────
+    if _HAS_YF and not _is_ashare(symbol) and len(ret) >= 60:
+        try:
+            bench = yf.Ticker("SPY").history(period=period)["Close"].pct_change().dropna()
+            aligned = pd.DataFrame({"asset": ret, "bench": bench}).dropna().tail(60)
+            if len(aligned) >= 30:
+                beta_v = float(np.cov(aligned["asset"], aligned["bench"])[0, 1] /
+                                np.var(aligned["bench"]))
+                factors["beta_60d"] = round(beta_v, 3)
+        except Exception:
+            pass
+
+    # ── Trend score ────────────────────────────────────────────────────────
+    trend = 0
+    for ma_key, w in [("ma_5_gap", 0.15), ("ma_20_gap", 0.35), ("ma_60_gap", 0.50)]:
+        v = factors.get(ma_key, 0.0) or 0.0
+        trend += w * (1 if v > 0 else -1)
+    factors["trend_score"] = round(trend, 3)
+
+    factors["provider"] = "local"
+    return {"success": True, **factors}
+
+
+# ---------------------------------------------------------------------------
+# 5. backtest_strategy
+# ---------------------------------------------------------------------------
+
+def _backtest_strategy(params: dict) -> dict:
+    """
+    Simple backtest engine.  Supports:
+      strategy = "sma_cross" | "rsi_mean_revert" | "momentum" | "buy_hold"
+    """
+    symbol   = params.get("symbol", "AAPL")
+    strategy = params.get("strategy", "sma_cross").lower().replace(" ", "_").replace("-", "_")
+    start    = params.get("start", _parse_date(None, 3 * 365))
+    end      = params.get("end", _today())
+    fast     = int(params.get("fast_period", 20))
+    slow     = int(params.get("slow_period", 60))
+    rsi_lo   = float(params.get("rsi_oversold", 30))
+    rsi_hi   = float(params.get("rsi_overbought", 70))
+    mom_n    = int(params.get("momentum_period", 20))
+
+    # Fetch data
+    if _is_ashare(symbol) and _HAS_AK:
+        norm = _normalise_ashare(symbol)
+        code = norm[2:]
+        df = ak.stock_zh_a_hist(symbol=code, period="daily",
+                                 start_date=start.replace("-", ""),
+                                 end_date=end.replace("-", ""),
+                                 adjust="qfq")
+        df = df.rename(columns={"收盘": "Close", "成交量": "Volume",
+                                 "日期": "Date", "开盘": "Open",
+                                 "最高": "High", "最低": "Low"})
+        df["Date"] = pd.to_datetime(df["Date"])
+        df = df.set_index("Date").sort_index()
+    elif _HAS_YF:
+        df = yf.Ticker(symbol).history(start=start, end=end)
+    else:
+        return {"success": False, "error": "No data source available"}
+
+    if df is None or len(df) < max(slow, 60):
+        return {"success": False, "error": f"Insufficient data: {len(df) if df is not None else 0} bars"}
+
+    close = df["Close"].astype(float)
+
+    # ── Signal generation ─────────────────────────────────────────────────
+    if strategy in ("sma_cross", "ma_cross"):
+        ma_f = close.rolling(fast).mean()
+        ma_s = close.rolling(slow).mean()
+        signal = (ma_f > ma_s).astype(int)
+
+    elif strategy in ("rsi_mean_revert", "rsi"):
+        delta = close.diff()
+        gain  = delta.clip(lower=0).rolling(14).mean()
+        loss  = (-delta.clip(upper=0)).rolling(14).mean()
+        rsi   = 100 - 100 / (1 + gain / loss.replace(0, np.nan))
+        signal = pd.Series(0, index=rsi.index)
+        in_pos = False
+        for i in range(len(rsi)):
+            r = rsi.iloc[i]
+            if np.isnan(r):
+                signal.iloc[i] = 0
+            elif r < rsi_lo and not in_pos:
+                in_pos = True
+                signal.iloc[i] = 1
+            elif r > rsi_hi and in_pos:
+                in_pos = False
+                signal.iloc[i] = 0
+            else:
+                signal.iloc[i] = int(in_pos)
+
+    elif strategy in ("momentum", "mom"):
+        ret_n  = close.pct_change(mom_n)
+        signal = (ret_n > 0).astype(int)
+
+    else:  # buy_hold
+        signal = pd.Series(1, index=close.index)
+
+    # ── Vectorbt backtest (if available) ─────────────────────────────────
+    if _HAS_VBT:
+        try:
+            pf = vbt.Portfolio.from_signals(
+                close, signal.shift(1).fillna(0) == 1,
+                signal.shift(1).fillna(0) == 0,
+                freq="D",
+            )
+            stats = pf.stats()
+            return {
+                "success":          True,
+                "symbol":           symbol,
+                "strategy":         strategy,
+                "start":            str(df.index[0].date()),
+                "end":              str(df.index[-1].date()),
+                "bars":             len(df),
+                "total_return":     round(float(stats.get("Total Return [%]", 0) / 100), 4),
+                "annual_return":    round(float(stats.get("Annualized Return [%]", 0) / 100), 4),
+                "sharpe_ratio":     round(float(stats.get("Sharpe Ratio", 0) or 0), 3),
+                "sortino_ratio":    round(float(stats.get("Sortino Ratio", 0) or 0), 3),
+                "max_drawdown":     round(float(stats.get("Max Drawdown [%]", 0) / 100), 4),
+                "win_rate":         round(float(stats.get("Win Rate [%]", 0) / 100), 3),
+                "total_trades":     int(stats.get("Total Trades", 0) or 0),
+                "provider":         "vectorbt",
+            }
+        except Exception as exc:
+            logger.debug("vectorbt failed, falling back to manual: %s", exc)
+
+    # ── Manual pandas backtest fallback ──────────────────────────────────
+    sig      = signal.shift(1).fillna(0)
+    ret      = close.pct_change().fillna(0)
+    port_ret = (ret * sig).fillna(0)
+    equity   = (1 + port_ret).cumprod()
+
+    total_r   = float(equity.iloc[-1] - 1)
+    n_years   = max((df.index[-1] - df.index[0]).days / 365.25, 0.01)
+    annual_r  = float((1 + total_r) ** (1 / n_years) - 1)
+    rf        = 0.04  # risk-free rate
+    excess    = port_ret - rf / 252
+    sharpe    = float(excess.mean() / port_ret.std() * np.sqrt(252)) if port_ret.std() > 0 else 0.0
+    neg_ret   = port_ret[port_ret < 0]
+    sortino   = float(excess.mean() / neg_ret.std() * np.sqrt(252)) if len(neg_ret) > 0 and neg_ret.std() > 0 else 0.0
+
+    # Max drawdown
+    peak    = equity.cummax()
+    dd      = (equity - peak) / peak
+    max_dd  = float(dd.min())
+
+    # Trade stats
+    trades     = (sig.diff() != 0) & (sig == 1)
+    exits      = (sig.diff() != 0) & (sig == 0)
+    n_trades   = int(trades.sum())
+    trade_rets = []
+    entry_date = None
+    entry_price = None
+    for date, row in sig.items():
+        if row == 1 and entry_price is None:
+            entry_price = float(close.loc[date])
+            entry_date  = date
+        elif row == 0 and entry_price is not None:
+            trade_rets.append(float(close.loc[date]) / entry_price - 1)
+            entry_price = None
+
+    win_rate = sum(1 for r in trade_rets if r > 0) / len(trade_rets) if trade_rets else 0.0
+
+    # Benchmark: buy-and-hold
+    bh_ret  = float(close.iloc[-1] / close.iloc[0] - 1)
+
+    return {
+        "success":          True,
+        "symbol":           symbol,
+        "strategy":         strategy,
+        "start":            str(df.index[0].date()),
+        "end":              str(df.index[-1].date()),
+        "bars":             len(df),
+        "total_return":     round(total_r, 4),
+        "annual_return":    round(annual_r, 4),
+        "sharpe_ratio":     round(sharpe, 3),
+        "sortino_ratio":    round(sortino, 3),
+        "max_drawdown":     round(max_dd, 4),
+        "win_rate":         round(win_rate, 3),
+        "total_trades":     n_trades,
+        "benchmark_return": round(bh_ret, 4),
+        "alpha":            round(annual_r - bh_ret / n_years, 4),
+        "provider":         "pandas",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 6. get_risk_metrics
+# ---------------------------------------------------------------------------
+
+def _get_risk_metrics(params: dict) -> dict:
+    symbol     = params.get("symbol", "AAPL")
+    period     = params.get("period", "1y")
+    conf_level = float(params.get("confidence", 0.95))
+
+    if _is_ashare(symbol) and _HAS_AK:
+        data = _get_ashare_data(symbol, _parse_date(None, 365), _today())
+        if not data.get("success"):
+            return data
+        norm = _normalise_ashare(symbol)
+        code = norm[2:]
+        df = ak.stock_zh_a_hist(symbol=code, period="daily",
+                                 start_date=_parse_date(None, 365).replace("-", ""),
+                                 end_date=_today().replace("-", ""),
+                                 adjust="qfq")
+        df = df.rename(columns={"收盘": "Close"})
+        close = df["Close"].astype(float)
+    elif _HAS_YF:
+        close = yf.Ticker(symbol).history(period=period)["Close"].astype(float)
+    else:
+        return {"success": False, "error": "No data source"}
+
+    if close is None or len(close) < 30:
+        return {"success": False, "error": "Insufficient data"}
+
+    ret  = close.pct_change().dropna()
+    mu   = float(ret.mean())
+    sig  = float(ret.std())
+
+    # VaR (parametric)
+    if _HAS_SCIPY:
+        var_daily = float(-sp_stats.norm.ppf(1 - conf_level, mu, sig))
+    else:
+        var_daily = float(-np.percentile(ret, (1 - conf_level) * 100))
+
+    var_monthly = var_daily * np.sqrt(21)
+
+    # CVaR (Expected Shortfall)
+    losses = -ret
+    cvar   = float(losses[losses >= var_daily].mean())
+
+    # Max drawdown
+    equity = (1 + ret).cumprod()
+    peak   = equity.cummax()
+    dd     = (equity - peak) / peak
+    max_dd = float(dd.min())
+
+    # Calmar
+    annual_ret = mu * 252
+    calmar     = annual_ret / abs(max_dd) if max_dd != 0 else 0.0
+
+    # Downside deviation (Sortino denominator)
+    neg_ret     = ret[ret < 0]
+    down_dev    = float(neg_ret.std() * np.sqrt(252)) if len(neg_ret) > 0 else 0.0
+
+    # Skewness / kurtosis
+    skew_v = float(ret.skew()) if _HAS_SCIPY else 0.0
+    kurt_v = float(ret.kurtosis()) if _HAS_SCIPY else 0.0
+
+    return {
+        "success":             True,
+        "symbol":              symbol,
+        "confidence_level":    conf_level,
+        "var_daily":           round(var_daily, 5),
+        "var_monthly":         round(float(var_monthly), 5),
+        "cvar_daily":          round(cvar, 5),
+        "max_drawdown":        round(max_dd, 5),
+        "annual_volatility":   round(float(sig * np.sqrt(252)), 5),
+        "annual_return":       round(float(annual_ret), 5),
+        "sharpe_ratio":        round(float((annual_ret - 0.04) / (sig * np.sqrt(252))), 3) if sig > 0 else 0.0,
+        "calmar_ratio":        round(float(calmar), 3),
+        "downside_deviation":  round(down_dev, 5),
+        "skewness":            round(skew_v, 3),
+        "kurtosis":            round(kurt_v, 3),
+        "provider":            "local",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 7. optimize_positions
+# ---------------------------------------------------------------------------
+
+def _optimize_positions(params: dict) -> dict:
+    symbols  = params.get("symbols", ["AAPL", "MSFT", "GOOGL"])
+    period   = params.get("period", "1y")
+    method   = params.get("method", "max_sharpe")  # max_sharpe | min_var | equal_weight
+    rf       = float(params.get("risk_free_rate", 0.04))
+
+    if isinstance(symbols, str):
+        symbols = [s.strip() for s in symbols.split(",")]
+
+    # Fetch returns
+    prices = {}
+    for sym in symbols:
+        if _is_ashare(sym) and _HAS_AK:
+            norm = _normalise_ashare(sym)
+            code = norm[2:]
+            df = ak.stock_zh_a_hist(symbol=code, period="daily",
+                                     start_date=_parse_date(None, 365).replace("-", ""),
+                                     end_date=_today().replace("-", ""),
+                                     adjust="qfq")
+            df = df.rename(columns={"收盘": "Close"})
+            prices[sym] = df["Close"].astype(float).values
+        elif _HAS_YF:
+            prices[sym] = yf.Ticker(sym).history(period=period)["Close"].astype(float).values
+
+    if not prices:
+        return {"success": False, "error": "Could not fetch prices"}
+
+    # Align length
+    min_len = min(len(v) for v in prices.values())
+    ret_mat = np.column_stack([prices[s][-min_len:] for s in symbols])
+    ret_mat = np.diff(np.log(ret_mat), axis=0)
+
+    mu_vec  = ret_mat.mean(axis=0) * 252
+    cov_mat = np.cov(ret_mat.T) * 252
+    n       = len(symbols)
+
+    if method == "equal_weight":
+        weights = np.ones(n) / n
+    elif method == "min_var":
+        # Analytical min-variance
+        try:
+            inv_cov = np.linalg.inv(cov_mat + 1e-8 * np.eye(n))
+            ones    = np.ones(n)
+            w       = inv_cov @ ones
+            weights = w / w.sum()
+        except np.linalg.LinAlgError:
+            weights = np.ones(n) / n
+    else:  # max_sharpe — gradient-free grid search
+        best_sharpe = -np.inf
+        weights     = np.ones(n) / n
+        rng         = np.random.default_rng(42)
+        for _ in range(10000):
+            w_try  = rng.dirichlet(np.ones(n))
+            p_ret  = float(w_try @ mu_vec)
+            p_vol  = float(np.sqrt(w_try @ cov_mat @ w_try))
+            sharpe = (p_ret - rf) / p_vol if p_vol > 0 else -np.inf
+            if sharpe > best_sharpe:
+                best_sharpe = sharpe
+                weights     = w_try
+
+    # Portfolio metrics
+    p_ret  = float(weights @ mu_vec)
+    p_vol  = float(np.sqrt(weights @ cov_mat @ weights))
+    sharpe = (p_ret - rf) / p_vol if p_vol > 0 else 0.0
+
+    return {
+        "success":          True,
+        "method":           method,
+        "symbols":          symbols,
+        "weights":          {sym: round(float(w), 4) for sym, w in zip(symbols, weights)},
+        "portfolio_return": round(p_ret, 4),
+        "portfolio_vol":    round(p_vol, 4),
+        "sharpe_ratio":     round(sharpe, 3),
+        "provider":         "local",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 8. get_sector_performance (A股 + US)
+# ---------------------------------------------------------------------------
+
+def _get_sector_performance(params: dict) -> dict:
+    market = params.get("market", "cn").lower()  # cn | us
+
+    if market == "cn" and _HAS_AK:
+        try:
+            df = _ak_retry(ak.stock_board_industry_name_em)
+            if df is None or df.empty:
+                raise ValueError("empty sector data")
+            df = df.rename(columns={
+                "板块名称": "sector", "最新价": "price",
+                "涨跌幅": "change_pct", "成交额": "amount",
+                "上涨家数": "rising", "下跌家数": "falling",
+            })
+            top    = df.nlargest(5,  "change_pct")[["sector", "change_pct"]].to_dict("records")
+            bottom = df.nsmallest(5, "change_pct")[["sector", "change_pct"]].to_dict("records")
+            return {
+                "success":   True,
+                "market":    "cn",
+                "date":      _today(),
+                "top_sectors":    top,
+                "bottom_sectors": bottom,
+                "total_sectors":  len(df),
+                "provider":  "akshare",
+            }
+        except Exception as exc:
+            return {"success": False, "error": str(exc)}
+
+    elif _HAS_YF:
+        sector_etfs = {
+            "Technology":    "XLK",  "Healthcare":    "XLV",
+            "Financials":    "XLF",  "Consumer Disc": "XLY",
+            "Industrials":   "XLI",  "Energy":        "XLE",
+            "Utilities":     "XLU",  "Real Estate":   "XLRE",
+            "Materials":     "XLB",  "Comm Services": "XLC",
+            "Consumer Staples": "XLP",
+        }
+        perf = []
+        for name, etf in sector_etfs.items():
+            try:
+                h = yf.Ticker(etf).history(period="5d")
+                if len(h) >= 2:
+                    chg = (h["Close"].iloc[-1] - h["Close"].iloc[-2]) / h["Close"].iloc[-2] * 100
+                    perf.append({"sector": name, "etf": etf, "change_pct": round(float(chg), 2)})
+            except Exception:
+                pass
+        perf.sort(key=lambda x: x["change_pct"], reverse=True)
+        return {
+            "success":       True,
+            "market":        "us",
+            "date":          _today(),
+            "sectors":       perf,
+            "top_sectors":   perf[:3],
+            "bottom_sectors": perf[-3:],
+            "provider":      "yfinance",
+        }
+
+    return {"success": False, "error": "akshare / yfinance not available: 运行 pip install akshare yfinance 或 /install akshare yfinance"}
+
+
+# ---------------------------------------------------------------------------
+# 9. A股 northbound fund flow (北向资金)
+# ---------------------------------------------------------------------------
+
+def _get_northbound_flow(params: dict) -> dict:
+    if not _HAS_AK:
+        return {"success": False, "error": "akshare not installed: 运行 pip install akshare 或 /install akshare"}
+    try:
+        # stock_hsgt_fund_flow_summary_em returns today's 沪深港通 summary.
+        # 成交净买额 is already in 亿元 — no further scaling needed.
+        df = _ak_retry(ak.stock_hsgt_fund_flow_summary_em)
+        north = df[df["资金方向"] == "北向"]
+        if north.empty:
+            return {"success": False, "error": "No northbound data in response"}
+        sh_flow = float(north[north["板块"] == "沪股通"]["成交净买额"].sum())
+        sz_flow = float(north[north["板块"] == "深股通"]["成交净买额"].sum())
+        total   = round(sh_flow + sz_flow, 2)
+        return {
+            "success":           True,
+            "latest_net_buy_yi": total,
+            "sh_net_buy_yi":     round(sh_flow, 2),
+            "sz_net_buy_yi":     round(sz_flow, 2),
+            "total_net_buy_yi":  total,
+            "trend":             "inflow" if total > 0 else "outflow",
+            "provider":          "akshare",
+        }
+    except Exception as exc:
+        return {"success": False, "error": str(exc)}
+
+
+# ---------------------------------------------------------------------------
+# 10. screen_ashare — 选股筛选器
+# ---------------------------------------------------------------------------
+
+def _screen_ashare(params: dict) -> dict:
+    """Screen A-share stocks by fundamental & technical criteria."""
+    min_roe         = float(params.get("min_roe", 10))
+    max_pe          = float(params.get("max_pe", 50))
+    min_revenue_gr  = float(params.get("min_revenue_growth", 10))
+    min_market_cap  = float(params.get("min_market_cap_yi", 0))    # 亿元
+    limit           = int(params.get("limit", 20))
+
+    # Primary: direct eastmoney clist (host-rotating, proxy-resilient, small
+    # paged query). Far more reliable than akshare's full-market spot endpoint.
+    try:
+        from aria_code.market_data_client import screen_ashare as _em_screen
+        _em = _em_screen(max_pe=max_pe, min_market_cap_yi=min_market_cap, limit=limit)
+        if _em.get("success") and _em.get("stocks"):
+            _em["criteria"] = params
+            return _em
+    except Exception:
+        pass  # fall through to akshare
+
+    if not _HAS_AK:
+        return {"success": False, "error": "akshare not installed: 运行 pip install akshare 或 /install akshare"}
+
+    try:
+        # A股实时行情
+        df = _ak_retry(ak.stock_zh_a_spot_em)
+        if df is None or df.empty:
+            return {"success": False, "error": "No spot data"}
+
+        df = df.rename(columns={
+            "代码": "code", "名称": "name", "最新价": "price",
+            "涨跌幅": "change_pct", "总市值": "market_cap",
+            "市盈率-动态": "pe_dynamic", "市净率": "pb",
+            "成交量": "volume", "换手率": "turnover_rate",
+        })
+
+        # Basic filters
+        df = df[~df["name"].str.contains("ST|退", na=False)]
+        df = df[df["price"].notna() & (df["price"] > 0)]
+
+        if "pe_dynamic" in df.columns:
+            df = df[df["pe_dynamic"].between(0.1, max_pe, inclusive="both")]
+
+        if "market_cap" in df.columns and min_market_cap > 0:
+            df = df[df["market_cap"] >= min_market_cap * 1e8]
+
+        # Score on momentum
+        if "change_pct" in df.columns:
+            df["score"] = df["change_pct"].fillna(0)
+            df = df.nlargest(limit, "score")
+
+        cols = ["code", "name", "price", "change_pct", "pe_dynamic", "pb",
+                "market_cap", "turnover_rate"]
+        cols = [c for c in cols if c in df.columns]
+        result_df = df[cols].head(limit)
+
+        # Format market_cap to 亿元
+        if "market_cap" in result_df.columns:
+            result_df = result_df.copy()
+            result_df["market_cap_yi"] = (result_df["market_cap"] / 1e8).round(1)
+
+        stocks = result_df.to_dict("records")
+        return {
+            "success":      True,
+            "count":        len(stocks),
+            "stocks":       stocks,
+            "criteria":     params,
+            "provider":     "akshare",
+        }
+    except Exception as exc:
+        return {"success": False, "error": str(exc)}
+
+
+# ---------------------------------------------------------------------------
+# 11. get_limit_up_pool — 涨停板池
+# ---------------------------------------------------------------------------
+
+def _get_limit_up_pool(params: dict) -> dict:
+    date = params.get("date", _today())
+    if not _HAS_AK:
+        return {"success": False, "error": "akshare not installed: 运行 pip install akshare 或 /install akshare"}
+    try:
+        date_str = date.replace("-", "")
+        df = _ak_retry(ak.stock_zt_pool_em, date=date_str)
+        if df is None or df.empty:
+            return {"success": True, "count": 0, "stocks": [], "date": date}
+
+        df = df.rename(columns={
+            "代码": "code", "名称": "name",
+            "涨停统计": "limit_streak", "连续涨停": "consecutive",
+            "首次封板时间": "first_lock_time", "涨停类型": "limit_type",
+        })
+        cols = [c for c in ["code", "name", "limit_streak", "consecutive",
+                             "first_lock_time", "limit_type"] if c in df.columns]
+        return {
+            "success":  True,
+            "date":     date,
+            "count":    len(df),
+            "stocks":   df[cols].head(50).to_dict("records"),
+            "provider": "akshare",
+        }
+    except Exception as exc:
+        return {"success": False, "error": str(exc)}
+
+
+# ---------------------------------------------------------------------------
+# 12. get_market_indices
+# ---------------------------------------------------------------------------
+
+def _get_market_indices(params: dict) -> dict:
+    indices = {
+        # US
+        "S&P 500": "^GSPC", "NASDAQ": "^IXIC", "Dow Jones": "^DJI",
+        "VIX": "^VIX", "Russell 2000": "^RUT",
+        # CN
+        "上证综指": "000001.SS", "深证成指": "399001.SZ", "创业板": "399006.SZ",
+        # Global
+        "Nikkei 225": "^N225", "FTSE 100": "^FTSE", "DAX": "^GDAXI",
+        "Hang Seng": "^HSI",
+        # Commodities
+        "Gold": "GC=F", "Crude Oil": "CL=F", "Bitcoin": "BTC-USD",
+    }
+
+    if not _HAS_YF:
+        if _HAS_AK:
+            # A股 indices only
+            try:
+                df = ak.stock_zh_index_spot_em()
+                return {"success": True, "indices": df.head(10).to_dict("records"), "provider": "akshare"}
+            except Exception as exc:
+                return {"success": False, "error": str(exc)}
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    results = []
+    for name, ticker in indices.items():
+        try:
+            h = yf.Ticker(ticker).history(period="2d")
+            if len(h) >= 1:
+                latest = float(h["Close"].iloc[-1])
+                chg    = (float(h["Close"].iloc[-1]) - float(h["Close"].iloc[0])) / float(h["Close"].iloc[0]) * 100 if len(h) >= 2 else 0.0
+                results.append({"name": name, "ticker": ticker,
+                                 "price": round(latest, 2), "change_pct": round(chg, 2)})
+        except Exception:
+            pass
+
+    return {"success": True, "indices": results, "date": _today(), "provider": "yfinance"}
+
+
+# ---------------------------------------------------------------------------
+# 13. analyze_news (local sentiment via keyword scoring)
+# ---------------------------------------------------------------------------
+
+def _get_data_key(service: str) -> str:
+    """Read a data service API key from env var or ~/.arthera/providers.json."""
+    _DATA_ENV = {
+        "finnhub":      "FINNHUB_API_KEY",
+        "newsapi":      "NEWS_API_KEY",
+        "brave":        "BRAVE_SEARCH_API_KEY",
+        "alphavantage": "ALPHA_VANTAGE_API_KEY",
+        "coingecko":    "COINGECKO_API_KEY",
+        "twelvedata":   "TWELVEDATA_API_KEY",
+    }
+    env_var = _DATA_ENV.get(service, "")
+    if env_var:
+        val = os.getenv(env_var, "")
+        if val:
+            return val
+    # Fall back to providers.json
+    try:
+        import pathlib as _pl
+        import json as _json
+        pf = aria_home() / "providers.json"
+        if pf.exists():
+            raw = _json.loads(pf.read_text(encoding="utf-8"))
+            entry = raw.get("data", {}).get(service, {})
+            if entry.get("api_key"):
+                return entry["api_key"]
+    except Exception:
+        pass
+    return ""
+
+
+def _fetch_news_finnhub(symbol: str, limit: int) -> list:
+    """Fetch stock news from Finnhub API."""
+    key = _get_data_key("finnhub")
+    if not key:
+        return []
+    try:
+        import requests as _req
+        import datetime as _dt
+        end_dt   = _dt.date.today().isoformat()
+        start_dt = (_dt.date.today() - _dt.timedelta(days=7)).isoformat()
+        url = (f"https://finnhub.io/api/v1/company-news"
+               f"?symbol={symbol}&from={start_dt}&to={end_dt}&token={key}")
+        resp = _req.get(url, timeout=8)
+        if resp.status_code == 200:
+            items = resp.json()[:limit]
+            return [{"title": a.get("headline", ""), "source": a.get("source", ""),
+                     "time": str(a.get("datetime", "")), "url": a.get("url", "")}
+                    for a in items]
+    except Exception:
+        pass
+    return []
+
+
+def _fetch_news_newsapi(query: str, limit: int) -> list:
+    """Fetch news from NewsAPI.org."""
+    key = _get_data_key("newsapi")
+    if not key:
+        return []
+    try:
+        import requests as _req
+        url = (f"https://newsapi.org/v2/everything"
+               f"?q={query}&language=en&sortBy=publishedAt&pageSize={limit}&apiKey={key}")
+        resp = _req.get(url, timeout=8)
+        if resp.status_code == 200:
+            data = resp.json()
+            return [{"title": a.get("title", ""), "source": a.get("source", {}).get("name", ""),
+                     "time": a.get("publishedAt", "")[:10], "url": a.get("url", "")}
+                    for a in data.get("articles", [])[:limit]]
+    except Exception:
+        pass
+    return []
+
+
+def _analyze_news(params: dict) -> dict:
+    symbol  = params.get("symbol", "")
+    query   = params.get("query", params.get("topic", symbol))
+    topic   = query or symbol
+    limit   = int(params.get("limit", 5))
+
+    # ── 1. A股 news via akshare (no key needed) ───────────────────────────────
+    if _HAS_AK and topic and _is_ashare(topic):
+        try:
+            norm = _normalise_ashare(topic) if _is_ashare(topic) else topic
+            code = norm[2:] if len(norm) > 2 else norm
+            df   = ak.stock_news_em(symbol=code)
+            if df is not None and not df.empty:
+                df = df.head(limit)
+                news_list = []
+                for _, row in df.iterrows():
+                    title = str(row.get("新闻标题", row.get("title", "")))
+                    score = _score_sentiment(title)
+                    news_list.append({
+                        "title":     title,
+                        "time":      str(row.get("发布时间", "")),
+                        "sentiment": "positive" if score > 0 else ("negative" if score < 0 else "neutral"),
+                        "score":     score,
+                    })
+                avg_score = sum(n["score"] for n in news_list) / len(news_list) if news_list else 0
+                return {
+                    "success": True, "symbol": topic, "news": news_list,
+                    "overall_sentiment": "positive" if avg_score > 0.1 else ("negative" if avg_score < -0.1 else "neutral"),
+                    "avg_score": round(avg_score, 3), "provider": "akshare",
+                }
+        except Exception as exc:
+            pass  # fall through to other providers
+
+    # ── 2. Finnhub (if key available) ────────────────────────────────────────
+    if symbol and not _is_ashare(symbol):
+        articles = _fetch_news_finnhub(symbol.upper(), limit)
+        if articles:
+            news_list = []
+            for a in articles:
+                score = _score_sentiment(a["title"])
+                news_list.append({**a, "sentiment": "positive" if score > 0 else ("negative" if score < 0 else "neutral"), "score": score})
+            avg_score = sum(n["score"] for n in news_list) / len(news_list) if news_list else 0
+            return {
+                "success": True, "symbol": symbol, "news": news_list,
+                "overall_sentiment": "positive" if avg_score > 0.1 else ("negative" if avg_score < -0.1 else "neutral"),
+                "avg_score": round(avg_score, 3), "provider": "finnhub",
+            }
+
+    # ── 3. NewsAPI (if key available) ─────────────────────────────────────────
+    search_query = topic or symbol or "market"
+    articles = _fetch_news_newsapi(search_query, limit)
+    if articles:
+        news_list = []
+        for a in articles:
+            score = _score_sentiment(a["title"])
+            news_list.append({**a, "sentiment": "positive" if score > 0 else ("negative" if score < 0 else "neutral"), "score": score})
+        avg_score = sum(n["score"] for n in news_list) / len(news_list) if news_list else 0
+        return {
+            "success": True, "symbol": topic, "news": news_list,
+            "overall_sentiment": "positive" if avg_score > 0.1 else ("negative" if avg_score < -0.1 else "neutral"),
+            "avg_score": round(avg_score, 3), "provider": "newsapi",
+        }
+
+    # ── 4. yfinance news (free, no key, US/HK/global stocks) ─────────────────
+    yf_sym = symbol.upper() if symbol else ""
+    if yf_sym and not _is_ashare(yf_sym):
+        try:
+            import yfinance as yf
+            ticker = yf.Ticker(yf_sym)
+            raw_news = ticker.news or []
+            news_list = []
+            for item in raw_news[:limit]:
+                content = item.get("content", {})
+                title = (
+                    content.get("title")
+                    or item.get("title", "")
+                )
+                pub = (
+                    content.get("pubDate")
+                    or item.get("providerPublishTime", "")
+                )
+                url = (
+                    content.get("canonicalUrl", {}).get("url")
+                    or item.get("link", "")
+                )
+                provider = (
+                    content.get("provider", {}).get("displayName")
+                    or item.get("publisher", "")
+                )
+                if not title:
+                    continue
+                score = _score_sentiment(title)
+                news_list.append({
+                    "title":     title,
+                    "time":      str(pub),
+                    "url":       url,
+                    "publisher": provider,
+                    "sentiment": "positive" if score > 0 else ("negative" if score < 0 else "neutral"),
+                    "score":     score,
+                })
+            if news_list:
+                avg_score = sum(n["score"] for n in news_list) / len(news_list)
+                return {
+                    "success": True, "symbol": yf_sym, "news": news_list,
+                    "overall_sentiment": "positive" if avg_score > 0.1 else ("negative" if avg_score < -0.1 else "neutral"),
+                    "avg_score": round(avg_score, 3), "provider": "yfinance",
+                }
+        except Exception:
+            pass
+
+    # ── 5. web_search fallback — search "[symbol] news" ──────────────────────
+    ws_query = f"{topic or symbol} stock news latest" if (topic or symbol) else ""
+    if ws_query:
+        ws_result = _web_search({"query": ws_query, "max_results": limit})
+        if ws_result.get("success") and ws_result.get("results"):
+            news_list = []
+            for item in ws_result["results"]:
+                title = item.get("title", "")
+                score = _score_sentiment(title)
+                news_list.append({
+                    "title":     title,
+                    "url":       item.get("url", ""),
+                    "publisher": item.get("source", ""),
+                    "sentiment": "positive" if score > 0 else ("negative" if score < 0 else "neutral"),
+                    "score":     score,
+                })
+            if news_list:
+                avg_score = sum(n["score"] for n in news_list) / len(news_list)
+                return {
+                    "success": True, "symbol": topic or symbol, "news": news_list,
+                    "overall_sentiment": "positive" if avg_score > 0.1 else ("negative" if avg_score < -0.1 else "neutral"),
+                    "avg_score": round(avg_score, 3), "provider": ws_result.get("provider", "web_search"),
+                }
+
+    # ── 6. No data available ──────────────────────────────────────────────────
+    tip = "配置数据服务 key: /apikey set finnhub <key> 或 /apikey set newsapi <key>；或设置 BRAVE_SEARCH_API_KEY 启用网页搜索"
+    return {"success": False, "error": tip}
+
+
+def _score_sentiment(text: str) -> float:
+    """Keyword-based sentiment scorer (0.0 = neutral, +1 = very positive, -1 = very negative)."""
+    pos = ["上涨", "涨停", "突破", "创新高", "利好", "增长", "盈利", "bull", "beat", "growth", "record", "profit", "rally", "buy", "upgrade"]
+    neg = ["下跌", "跌停", "亏损", "利空", "减少", "违规", "被罚", "风险", "bear", "miss", "loss", "decline", "sell", "downgrade", "fraud"]
+    t   = text.lower()
+    score = sum(1 for w in pos if w in t) - sum(1 for w in neg if w in t)
+    return float(max(-1, min(1, score * 0.25)))
+
+
+# ---------------------------------------------------------------------------
+# 14. get_commodities_data — gold, silver, oil, gas, copper, wheat, etc.
+# ---------------------------------------------------------------------------
+
+# Common commodity keywords → yfinance futures tickers
+_COMMODITY_MAP: Dict[str, str] = {
+    # Precious metals
+    "gold":       "GC=F",   "silver":     "SI=F",   "platinum":   "PL=F",   "palladium":  "PA=F",
+    # Energy
+    "oil":        "CL=F",   "crude":      "CL=F",   "crude oil":  "CL=F",
+    "brent":      "BZ=F",   "natural gas": "NG=F",  "natgas":     "NG=F",   "gas":        "NG=F",
+    "gasoline":   "RB=F",   "heating oil": "HO=F",
+    # Base metals
+    "copper":     "HG=F",   "aluminum":   "ALI=F",  "nickel":     "NI=F",   "zinc":       "ZNC=F",
+    # Agricultural
+    "wheat":      "ZW=F",   "corn":       "ZC=F",   "soybean":    "ZS=F",   "soybeans":   "ZS=F",
+    "coffee":     "KC=F",   "cocoa":      "CC=F",   "sugar":      "SB=F",   "cotton":     "CT=F",
+    "rice":       "ZR=F",   "oats":       "ZO=F",
+    # Livestock
+    "cattle":     "LE=F",   "hogs":       "HE=F",
+    # Softs / other
+    "lumber":     "LBS=F",  "rubber":     "rubber", "iron ore":   "TIO=F",
+    # China-traded (A股 futures) — map to closest international proxy
+    "螺纹钢":     "HG=F",   "铁矿石":     "TIO=F",  "铜":         "HG=F",
+    "黄金":       "GC=F",   "白银":       "SI=F",   "原油":       "CL=F",   "天然气":     "NG=F",
+}
+
+
+def _get_commodities_data(params: dict) -> dict:
+    """Commodity spot/futures price lookup via yfinance."""
+    commodity = str(params.get("commodity", "gold")).strip().lower()
+
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: pip install yfinance"}
+
+    # Resolve ticker
+    ticker = _COMMODITY_MAP.get(commodity)
+    if ticker is None:
+        # Try direct uppercase (user may have passed ticker like GC=F)
+        ticker = commodity.upper()
+
+    try:
+        tkr  = yf.Ticker(ticker)
+        hist = tkr.history(period="5d")
+        if hist.empty:
+            # Fallback: try without =F suffix
+            alt = commodity.upper().replace("=F", "") + "=F"
+            hist = yf.Ticker(alt).history(period="5d")
+            if hist.empty:
+                return {
+                    "success": False,
+                    "error":   f"No data for commodity '{commodity}' (ticker={ticker}). "
+                               "Try using the yfinance ticker directly, e.g. GC=F for gold.",
+                }
+            ticker = alt
+
+        latest   = float(hist["Close"].iloc[-1])
+        prev     = float(hist["Close"].iloc[-2]) if len(hist) > 1 else latest
+        chg      = (latest - prev) / prev * 100 if prev else 0.0
+        high_5d  = float(hist["High"].max())
+        low_5d   = float(hist["Low"].min())
+        vol_5d   = float(hist["Volume"].mean())
+
+        # Longer term context
+        hist_1y  = tkr.history(period="1y")
+        high_52w = float(hist_1y["High"].max())  if not hist_1y.empty else None
+        low_52w  = float(hist_1y["Low"].min())   if not hist_1y.empty else None
+        ret_1y   = float((hist_1y["Close"].iloc[-1] / hist_1y["Close"].iloc[0] - 1)) \
+                   if len(hist_1y) > 1 else None
+
+        info      = tkr.fast_info
+        currency  = getattr(info, "currency", "USD")
+
+        return {
+            "success":       True,
+            "commodity":     commodity,
+            "ticker":        ticker,
+            "latest_price":  round(latest, 4),
+            "change_pct":    round(chg, 3),
+            "currency":      currency,
+            "high_5d":       round(high_5d, 4),
+            "low_5d":        round(low_5d, 4),
+            "volume_5d_avg": int(vol_5d),
+            "high_52w":      round(high_52w, 4) if high_52w else None,
+            "low_52w":       round(low_52w, 4) if low_52w else None,
+            "return_1y":     round(ret_1y, 4) if ret_1y is not None else None,
+            "pct_from_52w_high": round((latest / high_52w - 1), 4) if high_52w else None,
+            "provider":      "yfinance",
+        }
+    except Exception as exc:
+        return {"success": False, "error": str(exc)}
+
+
+# ---------------------------------------------------------------------------
+# 15. get_futures_data — generic futures via yfinance
+# ---------------------------------------------------------------------------
+
+def _get_futures_data(params: dict) -> dict:
+    """Futures contract data (equity index futures, VIX, etc.)"""
+    contract = str(params.get("contract", params.get("symbol", "ES=F"))).strip().upper()
+
+    # Common index futures shortcuts
+    _FUTURES_MAP = {
+        "SP500": "ES=F", "SPX": "ES=F", "S&P": "ES=F",
+        "NQ": "NQ=F", "NASDAQ": "NQ=F",
+        "DOW": "YM=F", "DJIA": "YM=F",
+        "RUSSELL": "RTY=F", "RUT": "RTY=F",
+        "VIX": "^VIX",
+        "NIKKEI": "NK=F", "DAX": "FDAX=F",
+        "HSI": "HSI=F",
+    }
+    ticker = _FUTURES_MAP.get(contract, contract)
+    if not ticker.endswith("=F") and not ticker.startswith("^"):
+        ticker = ticker + "=F"
+
+    return _get_market_data({"symbol": ticker, "period": "5d"})
+
+
+# ---------------------------------------------------------------------------
+# 15b. get_bonds_data (US Treasury yields via yfinance)
+# ---------------------------------------------------------------------------
+
+def _get_bonds_data(params: dict) -> dict:
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+    tickers = {
+        "2Y": "^IRX", "5Y": "^FVX", "10Y": "^TNX", "30Y": "^TYX",
+    }
+    results = {}
+    for tenor, sym in tickers.items():
+        try:
+            h = yf.Ticker(sym).history(period="5d")
+            if not h.empty:
+                results[tenor] = round(float(h["Close"].iloc[-1]), 3)
+        except Exception:
+            pass
+    if not results:
+        return {"success": False, "error": "Could not fetch yield data"}
+    # Yield curve shape
+    if "2Y" in results and "10Y" in results:
+        results["10Y_2Y_spread"] = round(results["10Y"] - results["2Y"], 3)
+        results["curve_shape"]   = "normal" if results["10Y_2Y_spread"] > 0 else "inverted"
+    return {"success": True, "yields": results, "provider": "yfinance"}
+
+
+# ---------------------------------------------------------------------------
+# 16. get_ai_signal — Alibaba Cloud DeepSeek-powered signal
+# ---------------------------------------------------------------------------
+
+def _get_ai_signal(params: dict) -> dict:
+    """
+    AI trading signal from the Alibaba Cloud quant backend.
+    Returns: action (BUY/SELL/HOLD), confidence, reasoning, stop_loss, take_profit.
+    Falls back to calculate_factors when cloud is unavailable.
+    """
+    symbol = params.get("symbol", "600519")
+    market = params.get("market", "CN" if _is_ashare(symbol) else "US")
+
+    if _HAS_CLOUD:
+        try:
+            result = cloud_get_ai_signal_sync(symbol, market=market)
+            if result and result.get("action"):
+                return {"success": True, **result, "provider": "aliyun_cloud"}
+        except Exception as exc:
+            logger.debug("Cloud AI signal failed: %s", exc)
+
+    # Local fallback: derive a simple signal from factors
+    factors = _calculate_factors({"symbol": symbol})
+    if not factors.get("success"):
+        return factors
+
+    rsi     = factors.get("rsi_14", 50)
+    macd_h  = factors.get("macd_hist", 0) or 0
+    trend   = factors.get("trend_score", 0) or 0
+    vol_r   = factors.get("volume_ratio_20d", 1.0) or 1.0
+
+    score = 0
+    if rsi is not None:
+        score += (0.4 if rsi < 40 else -0.4 if rsi > 70 else 0)
+    score += (0.3 if macd_h > 0 else -0.3 if macd_h < 0 else 0)
+    score += trend * 0.3
+
+    action     = "BUY" if score > 0.3 else "SELL" if score < -0.3 else "HOLD"
+    confidence = min(abs(score), 1.0)
+
+    return {
+        "success":    True,
+        "symbol":     symbol,
+        "action":     action,
+        "confidence": round(confidence, 3),
+        "reasoning":  f"RSI={rsi}, MACD_hist={macd_h:.5f}, trend={trend:.3f}, vol_ratio={vol_r:.2f}",
+        "stop_loss":  None,
+        "take_profit": None,
+        "provider":   "local_fallback",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 17. get_market_insights — AI narrative analysis
+# ---------------------------------------------------------------------------
+
+def _get_market_insights(params: dict) -> dict:
+    """
+    Narrative market insights from Alibaba Cloud AI service.
+    symbols: list of stock codes or comma-separated string.
+    """
+    raw = params.get("symbols", params.get("symbol", "sh600519"))
+    if isinstance(raw, str):
+        symbols = [s.strip() for s in raw.replace(",", " ").split() if s.strip()]
+    else:
+        symbols = list(raw)
+    market = params.get("market", "CN")
+
+    if _HAS_CLOUD:
+        try:
+            from aria_code.aliyun_data_client import run_async
+            result = run_async(AliyunDataClient.get().get_market_insights(symbols, market=market))
+            if result:
+                return {"success": True, **result, "provider": "aliyun_cloud"}
+        except Exception as exc:
+            logger.debug("Cloud market insights failed: %s", exc)
+
+    # Local fallback: multi-stock factor summary
+    summaries = []
+    for sym in symbols[:5]:
+        f = _calculate_factors({"symbol": sym})
+        if f.get("success"):
+            summaries.append({
+                "symbol":      sym,
+                "rsi_14":      f.get("rsi_14"),
+                "trend_score": f.get("trend_score"),
+                "macd_hist":   f.get("macd_hist"),
+                "vol_ratio":   f.get("volume_ratio_20d"),
+            })
+    if not summaries:
+        return {"success": False, "error": "Could not compute local factors"}
+
+    return {
+        "success":   True,
+        "symbols":   symbols,
+        "summaries": summaries,
+        "note":      "Cloud AI not available — showing local factor summary",
+        "provider":  "local_fallback",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 18. get_predictions — ML model predictions from cloud
+# ---------------------------------------------------------------------------
+
+def _get_predictions(params: dict) -> dict:
+    """
+    ML-powered stock return predictions from Alibaba Cloud QuantEngine.
+    Falls back to momentum signal locally.
+    """
+    raw = params.get("symbols", params.get("symbol", "sh600519"))
+    if isinstance(raw, str):
+        symbols = [s.strip() for s in raw.replace(",", " ").split() if s.strip()]
+    else:
+        symbols = list(raw)
+    days   = int(params.get("prediction_days", 5))
+    market = params.get("market", "CN")
+
+    if _HAS_CLOUD:
+        try:
+            from aria_code.aliyun_data_client import run_async
+            result = run_async(AliyunDataClient.get().get_predictions(symbols, prediction_days=days, market=market))
+            if result and result.get("predictions"):
+                return {"success": True, **result, "provider": "aliyun_cloud"}
+        except Exception as exc:
+            logger.debug("Cloud predictions failed: %s", exc)
+
+    # Local fallback: simple momentum prediction
+    preds = []
+    for sym in symbols[:5]:
+        f = _calculate_factors({"symbol": sym})
+        if f.get("success"):
+            r5  = f.get("return_5d", 0) or 0
+            r20 = f.get("return_20d", 0) or 0
+            predicted_return = round((r5 * 0.4 + r20 * 0.6) * (days / 20), 4)
+            preds.append({
+                "symbol":           sym,
+                "predicted_return": predicted_return,
+                "confidence":       0.5,
+                "method":           "momentum",
+            })
+
+    if not preds:
+        return {"success": False, "error": "No data for predictions"}
+
+    return {
+        "success":         True,
+        "predictions":     preds,
+        "prediction_days": days,
+        "provider":        "local_fallback",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 19. cloud_backtest — advanced ML-powered backtest via Alibaba Cloud
+# ---------------------------------------------------------------------------
+
+def _cloud_backtest(params: dict) -> dict:
+    """
+    Full ML-powered backtest via Alibaba Cloud QuantEngine.
+    Falls back to local pandas backtest when cloud is unavailable.
+    """
+    raw = params.get("symbols", params.get("symbol", "sh600519"))
+    if isinstance(raw, str):
+        symbols = [s.strip() for s in raw.replace(",", " ").split() if s.strip()]
+    else:
+        symbols = list(raw)
+
+    strategy_cfg = params.get("strategy_config", {
+        "model_type":              params.get("model_type", "lightgbm"),
+        "backtest_period_months":  params.get("months", 12),
+        "rebalance_freq":          params.get("rebalance_freq", "weekly"),
+        "top_k":                   params.get("top_k", 3),
+        "use_enhanced_factors":    True,
+        "use_dynamic_position":    True,
+    })
+    start  = params.get("start", "")
+    end    = params.get("end", "")
+    market = params.get("market", "CN")
+
+    if _HAS_CLOUD:
+        try:
+            from aria_code.aliyun_data_client import run_async
+            result = run_async(
+                AliyunDataClient.get().run_backtest(
+                    symbols, strategy_cfg,
+                    start_date=start, end_date=end, market=market,
+                )
+            )
+            if result and result.get("status") in ("completed", "running"):
+                out = {"success": True, "provider": "aliyun_cloud"}
+                r = result.get("result") or {}
+                perf = r.get("performance") or {}
+                out.update({
+                    "backtest_id":     result.get("backtest_id"),
+                    "status":          result.get("status"),
+                    "total_return":    perf.get("total_return"),
+                    "annual_return":   perf.get("annualized_return"),
+                    "sharpe_ratio":    perf.get("sharpe_ratio"),
+                    "max_drawdown":    perf.get("max_drawdown"),
+                    "win_rate":        perf.get("win_rate"),
+                    "total_trades":    perf.get("total_trades"),
+                    "equity_curve":    r.get("equity_curve"),
+                })
+                return out
+        except Exception as exc:
+            logger.debug("Cloud backtest failed: %s", exc)
+
+    # Local fallback: run simple pandas backtest for first symbol
+    return _backtest_strategy({
+        "symbol":   symbols[0] if symbols else "sh600519",
+        "strategy": params.get("strategy", "sma_cross"),
+        "start":    start or _parse_date(None, 730),
+        "end":      end   or _today(),
+    })
+
+
+# ---------------------------------------------------------------------------
+# Helper: format dataframe tail for display
+# ---------------------------------------------------------------------------
+
+def _df_tail(df: pd.DataFrame, n: int = 5) -> List[Dict]:
+    cols = [c for c in ["date", "Close", "Open", "High", "Low", "Volume"]
+            if c in df.columns]
+    sub  = df[cols].tail(n)
+    records = []
+    for _, row in sub.iterrows():
+        rec = {}
+        for c in cols:
+            v = row[c]
+            try:
+                if isinstance(v, (float, np.floating)):
+                    rec[c] = round(float(v), 4)
+                elif hasattr(v, "item"):
+                    rec[c] = v.item()
+                else:
+                    rec[c] = str(v)
+            except Exception:
+                rec[c] = str(v)
+        records.append(rec)
+    return records
+
+
+
+# OpenAI/Ollama tool schemas for local finance tools
+LOCAL_FINANCE_TOOL_SCHEMAS = [
+    {
+        "type": "function",
+        "function": {
+            "name": "get_market_data",
+            "description": "Get stock/ETF quotes and OHLCV history. Supports US stocks (AAPL), A-share (600519 or sh600519), ETFs, indices.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol":   {"type": "string", "description": "Ticker symbol, e.g. AAPL, sh600519, BTC-USD"},
+                    "period":   {"type": "string", "description": "yfinance period: 1d 5d 1mo 3mo 6mo 1y 2y 5y ytd max"},
+                    "interval": {"type": "string", "description": "Bar interval: 1m 5m 15m 30m 1h 1d 1wk 1mo"},
+                    "start":    {"type": "string", "description": "Start date YYYY-MM-DD (overrides period)"},
+                    "end":      {"type": "string", "description": "End date YYYY-MM-DD"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_crypto_data",
+            "description": "Cryptocurrency OHLCV data. Uses ccxt if available, else yfinance.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol":    {"type": "string", "description": "Pair, e.g. BTC/USDT or ETH/USDT"},
+                    "exchange":  {"type": "string", "description": "Exchange name (binance, okx, bybit, coinbase)"},
+                    "timeframe": {"type": "string", "description": "Candle timeframe: 1m 5m 1h 4h 1d"},
+                    "limit":     {"type": "integer", "description": "Number of bars"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "calculate_factors",
+            "description": "Compute technical and quantitative factors: RSI, MACD, Bollinger, MA gaps, volatility, trend score, beta.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Ticker symbol"},
+                    "period": {"type": "string", "description": "Lookback period (yfinance format)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "backtest_strategy",
+            "description": "Run a strategy backtest. Returns Sharpe, max drawdown, win rate, trade count, alpha vs benchmark.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol":          {"type": "string"},
+                    "strategy":        {"type": "string", "description": "sma_cross | rsi_mean_revert | momentum | buy_hold"},
+                    "start":           {"type": "string", "description": "Start date YYYY-MM-DD"},
+                    "end":             {"type": "string", "description": "End date YYYY-MM-DD"},
+                    "fast_period":     {"type": "integer", "description": "Fast MA period (default 20)"},
+                    "slow_period":     {"type": "integer", "description": "Slow MA period (default 60)"},
+                    "rsi_oversold":    {"type": "number",  "description": "RSI oversold threshold (default 30)"},
+                    "rsi_overbought":  {"type": "number",  "description": "RSI overbought threshold (default 70)"},
+                    "momentum_period": {"type": "integer", "description": "Momentum lookback bars (default 20)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "run_portfolio_backtest",
+            "description": (
+                "Event-driven, multi-asset backtest. Unlike backtest_strategy (single-symbol "
+                "templates), this runs a strategy across a basket with next-open fills (no "
+                "look-ahead), commission+slippage, and an equal-weight buy-and-hold benchmark. "
+                "Returns total/annual return, Sharpe, max drawdown, win rate, trade count, and "
+                "alpha vs buy-hold. Use for portfolio/multi-symbol strategy evaluation."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbols":       {"type": "string", "description": "One or more tickers, space/comma separated (e.g. 'AAPL MSFT 600519')"},
+                    "strategy":      {"type": "string", "description": "sma_cross | buy_hold (default sma_cross)"},
+                    "days":          {"type": "integer", "description": "Lookback window in days (default 365)"},
+                    "fast":          {"type": "integer", "description": "Fast MA period for sma_cross (default 20)"},
+                    "slow":          {"type": "integer", "description": "Slow MA period for sma_cross (default 60)"},
+                    "starting_cash": {"type": "number",  "description": "Initial capital (default 100000)"},
+                    "commission":    {"type": "number",  "description": "Per-trade commission rate (default 0.0005)"},
+                    "slippage":      {"type": "number",  "description": "Per-trade slippage rate (default 0.0005)"},
+                },
+                "required": ["symbols"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_risk_metrics",
+            "description": "Portfolio / stock risk metrics: VaR, CVaR, drawdown, Sharpe, Sortino, Calmar, skewness.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol":     {"type": "string"},
+                    "period":     {"type": "string", "description": "Data lookback (default 1y)"},
+                    "confidence": {"type": "number",  "description": "VaR confidence level (default 0.95)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "optimize_positions",
+            "description": "Portfolio weight optimisation using Markowitz / max-Sharpe / min-variance.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbols":          {"type": "array",  "items": {"type": "string"}, "description": "List of ticker symbols"},
+                    "method":           {"type": "string", "description": "max_sharpe | min_var | equal_weight"},
+                    "period":           {"type": "string", "description": "History period (default 1y)"},
+                    "risk_free_rate":   {"type": "number", "description": "Annual risk-free rate (default 0.04)"},
+                },
+                "required": ["symbols"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_sector_performance",
+            "description": "Sector performance ranking. market='cn' uses akshare industry data; market='us' uses SPDR ETFs.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "market": {"type": "string", "description": "cn | us (default cn)"},
+                },
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_northbound_flow",
+            "description": "北向资金 (Shanghai-HK / Shenzhen-HK Connect) net buy amount in 亿元 via akshare.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "days": {"type": "integer", "description": "Number of trading days to show (default 10)"},
+                },
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "screen_ashare",
+            "description": "A股选股筛选器. Filters by PE, market cap, ST/退市 exclusion, momentum ranking.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "max_pe":              {"type": "number",  "description": "Max dynamic PE ratio (default 50)"},
+                    "min_market_cap_yi":   {"type": "number",  "description": "Min market cap in 亿元 (default 0)"},
+                    "limit":               {"type": "integer", "description": "Max stocks to return (default 20)"},
+                },
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_limit_up_pool",
+            "description": "Today's A股 limit-up stock pool (涨停板) via akshare. Includes consecutive limit-up count.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "date": {"type": "string", "description": "Date YYYY-MM-DD (default today)"},
+                },
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_market_indices",
+            "description": "Major global market indices: S&P 500, NASDAQ, 上证综指, Nikkei, Gold, BTC, etc.",
+            "parameters": {"type": "object", "properties": {}, "required": []},
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "analyze_news",
+            "description": "Fetch and sentiment-score recent news for a stock or topic.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Stock code or company name"},
+                    "limit":  {"type": "integer", "description": "Number of articles (default 5)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_bonds_data",
+            "description": "US Treasury yield curve (2Y, 5Y, 10Y, 30Y) and 10Y-2Y spread.",
+            "parameters": {"type": "object", "properties": {}, "required": []},
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_commodities_data",
+            "description": (
+                "Commodity spot/futures price, 52-week range and 1-year return. "
+                "Supports: gold, silver, oil, crude, brent, natgas, copper, wheat, corn, soybean, coffee, "
+                "cotton, cattle, lumber, 黄金, 原油, 铜, etc. "
+                "Also accepts yfinance tickers directly (e.g. GC=F, CL=F, NG=F)."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "commodity": {
+                        "type": "string",
+                        "description": "Commodity name (gold, oil, copper…) or yfinance ticker (GC=F)"
+                    },
+                },
+                "required": ["commodity"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_futures_data",
+            "description": "Equity index and other futures: S&P (ES=F), NASDAQ (NQ=F), Nikkei, VIX, etc.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "contract": {
+                        "type": "string",
+                        "description": "Futures contract name or ticker: SP500, NQ, DOW, VIX, ES=F, NQ=F…"
+                    },
+                },
+                "required": ["contract"],
+            },
+        },
+    },
+    # ── New cloud-backed tools ────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "get_ai_signal",
+            "description": (
+                "AI-powered trading signal (BUY/SELL/HOLD) with confidence, reasoning, stop-loss and "
+                "take-profit levels. Uses Alibaba Cloud DeepSeek + QuantEngine. "
+                "Falls back to local factor-based signal when cloud unavailable."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Stock code, e.g. sh600519, AAPL"},
+                    "market": {"type": "string", "description": "CN | US (auto-detected from symbol)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_market_insights",
+            "description": (
+                "AI narrative market insights for a basket of stocks. Returns sentiment, key risks, "
+                "opportunities and macro context. Powered by Alibaba Cloud AI."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbols": {
+                        "type": "array",
+                        "items": {"type": "string"},
+                        "description": "List of stock codes, e.g. ['sh600519', 'sz000858']"
+                    },
+                    "market": {"type": "string", "description": "CN | US (default CN)"},
+                },
+                "required": ["symbols"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_predictions",
+            "description": (
+                "ML model return predictions for a list of stocks. "
+                "Uses Alibaba Cloud LightGBM/XGBoost ensemble. "
+                "Returns predicted_return and confidence for each symbol."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbols": {
+                        "type": "array",
+                        "items": {"type": "string"},
+                        "description": "List of stock codes"
+                    },
+                    "prediction_days": {"type": "integer", "description": "Forecast horizon in days (default 5)"},
+                    "market": {"type": "string", "description": "CN | US (default CN)"},
+                },
+                "required": ["symbols"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "cloud_backtest",
+            "description": (
+                "Full ML-powered backtest on Alibaba Cloud QuantEngine. "
+                "Supports rebalance_freq (daily/weekly/monthly), dynamic position sizing, "
+                "model_type (lightgbm/xgboost/ensemble). Falls back to local pandas backtest."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbols": {
+                        "type": "array",
+                        "items": {"type": "string"},
+                        "description": "List of stock codes for backtest universe"
+                    },
+                    "model_type":      {"type": "string",  "description": "lightgbm | xgboost | ensemble (default lightgbm)"},
+                    "months":          {"type": "integer", "description": "Backtest period in months (default 12)"},
+                    "rebalance_freq":  {"type": "string",  "description": "daily | weekly | monthly (default weekly)"},
+                    "top_k":           {"type": "integer", "description": "Top-K stocks to hold per period (default 3)"},
+                    "start":           {"type": "string",  "description": "Start date YYYY-MM-DD"},
+                    "end":             {"type": "string",  "description": "End date YYYY-MM-DD"},
+                    "market":          {"type": "string",  "description": "CN | US (default CN)"},
+                },
+                "required": [],
+            },
+        },
+    },
+    # ── web_search ────────────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "web_search",
+            "description": (
+                "Search the web for current information. "
+                "USE THIS PROACTIVELY when the user asks about: recent news, latest earnings, "
+                "new IPO stocks or unverified tickers, price targets, analyst upgrades/downgrades, "
+                "M&A deals, regulatory decisions, macro events, or anything that may have changed "
+                "after your training cutoff. Do NOT rely on training data for current events — "
+                "always search first. Chain with web_fetch to read full articles."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "query":       {"type": "string",  "description": "Search query, include ticker and topic, e.g. 'AAPL earnings Q1 2026'"},
+                    "max_results": {"type": "integer", "description": "Number of results (default 5, max 10)"},
+                },
+                "required": ["query"],
+            },
+        },
+    },
+    # ── web_fetch ─────────────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "web_fetch",
+            "description": (
+                "Fetch a URL and return the page text. Use after web_search to read full article content, "
+                "SEC filings, earnings reports, or any webpage. Automatically strips HTML tags."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "url":      {"type": "string",  "description": "Full URL to fetch"},
+                    "timeout":  {"type": "integer", "description": "Timeout seconds (default 15)"},
+                },
+                "required": ["url"],
+            },
+        },
+    },
+    # ── get_forex_data ────────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "get_forex_data",
+            "description": "Get exchange rate data for currency pairs (e.g. USD/CNY, EUR/USD, USD/JPY). Returns OHLCV and current rate.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "pair":   {"type": "string", "description": "Currency pair e.g. USDCNY=X, EURUSD=X, USDJPY=X"},
+                    "period": {"type": "string", "description": "1d | 5d | 1mo | 3mo | 1y (default 1mo)"},
+                },
+                "required": ["pair"],
+            },
+        },
+    },
+    # ── get_options_chain ─────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "get_options_chain",
+            "description": "Retrieve options chain for a US stock: calls & puts with strike, expiry, IV, delta, volume, OI. Use for options strategy analysis.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol":  {"type": "string",  "description": "US stock ticker e.g. AAPL, TSLA, NVDA"},
+                    "expiry":  {"type": "string",  "description": "Expiration date YYYY-MM-DD or leave blank for nearest"},
+                    "option_type": {"type": "string", "description": "call | put | both (default both)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    # ── peer_comparison ───────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "peer_comparison",
+            "description": "Compare a stock against sector peers on valuation (PE/PB), profitability (ROE), market cap. Automatically selects peers if not specified.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string",      "description": "Target stock ticker"},
+                    "peers":  {"type": "array", "items": {"type": "string"}, "description": "Peer tickers (optional, auto-selected if omitted)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    # ── piotroski_fscore ──────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "piotroski_fscore",
+            "description": "Calculate Piotroski F-Score (0-9) for financial health assessment. Score ≥7 = strong, ≤2 = weak. Use for fundamental stock screening.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Stock ticker"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    # ── altman_zscore ─────────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "altman_zscore",
+            "description": "Calculate Altman Z-Score for bankruptcy risk. Z>2.99 = safe, 1.81-2.99 = grey zone, <1.81 = distress. Use when user asks about company financial risk.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Stock ticker"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    # ── calculate_ichimoku ────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "calculate_ichimoku",
+            "description": "Calculate Ichimoku Cloud indicators (Tenkan, Kijun, Senkou A/B, Chikou). Provides cloud support/resistance levels and trend signals.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Stock or crypto ticker"},
+                    "period": {"type": "string", "description": "Price history period: 3mo | 6mo | 1y (default 6mo)"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    # ── get_fear_greed_index ──────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "get_fear_greed_index",
+            "description": "Get CNN Fear & Greed Index (0-100) for overall market sentiment. Use when user asks about market sentiment, risk appetite, or broad market mood.",
+            "parameters": {
+                "type": "object",
+                "properties": {},
+                "required": [],
+            },
+        },
+    },
+    # ── get_funding_rates ─────────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "get_funding_rates",
+            "description": "Get perpetual futures funding rates for crypto assets (BTC, ETH, etc.) from major exchanges. Positive rate = longs pay shorts (bullish); negative = shorts pay longs (bearish).",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol": {"type": "string", "description": "Crypto symbol e.g. BTC, ETH, SOL"},
+                },
+                "required": ["symbol"],
+            },
+        },
+    },
+    # ── get_funding_rates_compare ─────────────────────────────────────────────
+    # The handler has been in LOCAL_FINANCE_TOOL_REGISTRY all along with no
+    # schema beside it, so it was registered, callable, and invisible: the
+    # model was never told it exists.
+    {
+        "type": "function",
+        "function": {
+            "name": "get_funding_rates_compare",
+            "description": (
+                "Compare perpetual funding rates for the same assets across Binance, OKX "
+                "and Bybit in one call. Use this to spot cross-exchange arbitrage: a spread "
+                "wider than about 0.02% between venues is worth attention. Requires ccxt."
+            ),
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbols": {
+                        "type": "array",
+                        "items": {"type": "string"},
+                        "description": "Perpetual pairs, e.g. [\"BTC/USDT\", \"ETH/USDT\"]. Defaults to BTC, ETH and SOL.",
+                    },
+                },
+                "required": [],
+            },
+        },
+    },
+    # ── walk_forward_backtest ─────────────────────────────────────────────────
+    {
+        "type": "function",
+        "function": {
+            "name": "walk_forward_backtest",
+            "description": "Run walk-forward validation backtest: train on rolling windows then test out-of-sample to avoid overfitting. More robust than simple backtest.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "symbol":        {"type": "string", "description": "Stock ticker"},
+                    "strategy":      {"type": "string", "description": "Strategy name: sma_cross | rsi_reversal | macd_trend"},
+                    "train_months":  {"type": "integer", "description": "Training window in months (default 12)"},
+                    "test_months":   {"type": "integer", "description": "Test window in months (default 3)"},
+                },
+                "required": ["symbol", "strategy"],
+            },
+        },
+    },
+]
+
+
+# ---------------------------------------------------------------------------
+# 19. Piotroski F-Score  (基本面质量评分, 0–9)
+# ---------------------------------------------------------------------------
+
+def _piotroski_fscore(params: dict) -> dict:
+    """
+    Piotroski F-Score: 9项二元信号综合判断财务质量。
+    ≥7 = 高质量(做多信号), ≤3 = 低质量(做空信号), 4-6 = 中性。
+    """
+    symbol = params.get("symbol", "AAPL")
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    try:
+        tkr  = yf.Ticker(symbol)
+        info = tkr.info or {}
+        bs_annual   = tkr.balance_sheet    if hasattr(tkr, "balance_sheet")    else None
+        is_annual   = tkr.income_stmt      if hasattr(tkr, "income_stmt")      else None
+        cf_annual   = tkr.cashflow         if hasattr(tkr, "cashflow")         else None
+
+        def _get(df, row, col=0):
+            try:
+                if df is None or df.empty: return None
+                matches = [r for r in df.index if row.lower() in str(r).lower()]
+                if not matches: return None
+                val = df.loc[matches[0]].iloc[col]
+                return float(val) if val is not None and str(val) not in ("nan","None") else None
+            except Exception:
+                return None
+
+        scores: Dict[str, Any] = {}
+
+        # ── Profitability (4 signals) ──────────────────────────────────────
+        roa = info.get("returnOnAssets") or _get(is_annual, "Net Income")
+        scores["F1_ROA_positive"]     = int((roa or 0) > 0)
+
+        cfo = _get(cf_annual, "Operating Cash Flow") or _get(cf_annual, "Total Cash From Operating")
+        scores["F2_CFO_positive"]     = int((cfo or 0) > 0)
+
+        # ROA change (current vs prior year)
+        net_inc_cur  = _get(is_annual, "Net Income", 0)
+        net_inc_prev = _get(is_annual, "Net Income", 1)
+        ta_cur       = _get(bs_annual, "Total Assets", 0)
+        ta_prev      = _get(bs_annual, "Total Assets", 1)
+        roa_cur  = (net_inc_cur  / ta_cur)  if (ta_cur  and net_inc_cur  is not None) else None
+        roa_prev = (net_inc_prev / ta_prev) if (ta_prev and net_inc_prev is not None) else None
+        scores["F3_ROA_increasing"]   = int(roa_cur > roa_prev) if (roa_cur is not None and roa_prev is not None) else 0
+
+        # Accruals: CFO > ROA × Total Assets
+        scores["F4_CFO_gt_ROA"]       = int((cfo or 0) > (roa or 0) * (ta_cur or 1))
+
+        # ── Leverage, Liquidity (3 signals) ───────────────────────────────
+        ltd_cur  = _get(bs_annual, "Long Term Debt", 0)
+        ltd_prev = _get(bs_annual, "Long Term Debt", 1)
+        scores["F5_Leverage_lower"]   = int((ltd_cur or 0) < (ltd_prev or 0)) if ltd_prev is not None else 0
+
+        ca_cur   = _get(bs_annual, "Current Assets", 0)
+        ca_prev  = _get(bs_annual, "Current Assets", 1)
+        cl_cur   = _get(bs_annual, "Current Liabilities", 0) or _get(bs_annual, "Total Current Liabilities", 0)
+        cl_prev  = _get(bs_annual, "Current Liabilities", 1) or _get(bs_annual, "Total Current Liabilities", 1)
+        cr_cur   = (ca_cur  / cl_cur)  if (cl_cur  and ca_cur)  else None
+        cr_prev  = (ca_prev / cl_prev) if (cl_prev and ca_prev) else None
+        scores["F6_CurrentRatio_up"]  = int(cr_cur > cr_prev) if (cr_cur and cr_prev) else 0
+
+        # No dilution: shares outstanding not increasing
+        shares_cur  = info.get("sharesOutstanding") or _get(bs_annual, "Ordinary Shares Number", 0)
+        shares_prev = _get(bs_annual, "Ordinary Shares Number", 1)
+        scores["F7_NoDilution"]       = int((shares_cur or 1) <= (shares_prev or 1)) if shares_prev else 1
+
+        # ── Operating Efficiency (2 signals) ──────────────────────────────
+        rev_cur  = _get(is_annual, "Total Revenue", 0)
+        rev_prev = _get(is_annual, "Total Revenue", 1)
+        gp_cur   = _get(is_annual, "Gross Profit", 0)
+        gp_prev  = _get(is_annual, "Gross Profit", 1)
+        gm_cur   = (gp_cur  / rev_cur)  if (rev_cur  and gp_cur)  else None
+        gm_prev  = (gp_prev / rev_prev) if (rev_prev and gp_prev) else None
+        scores["F8_GrossMargin_up"]   = int(gm_cur > gm_prev) if (gm_cur is not None and gm_prev is not None) else 0
+
+        at_cur  = (rev_cur  / ta_cur)  if (ta_cur  and rev_cur)  else None
+        at_prev = (rev_prev / ta_prev) if (ta_prev and rev_prev) else None
+        scores["F9_AssetTurnover_up"] = int(at_cur > at_prev) if (at_cur is not None and at_prev is not None) else 0
+
+        fscore = sum(scores.values())
+
+        if fscore >= 7:
+            verdict, color = "高质量 — 做多信号", "bullish"
+        elif fscore <= 3:
+            verdict, color = "低质量 — 做空信号", "bearish"
+        else:
+            verdict, color = "中性", "neutral"
+
+        return {
+            "success":  True,
+            "symbol":   symbol,
+            "f_score":  fscore,
+            "verdict":  verdict,
+            "signal":   color,
+            "scores":   scores,
+            "provider": "yfinance",
+        }
+    except Exception as e:
+        return {"success": False, "error": str(e)}
+
+
+# ---------------------------------------------------------------------------
+# 20. Altman Z-Score  (破产风险预测)
+# ---------------------------------------------------------------------------
+
+def _altman_zscore(params: dict) -> dict:
+    """
+    Altman Z''-Score（适合非制造业，使用 4 变量版）。
+    Z > 2.6 = 安全区，1.1–2.6 = 灰色区，< 1.1 = 破产风险。
+    """
+    symbol = params.get("symbol", "AAPL")
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    try:
+        tkr  = yf.Ticker(symbol)
+        info = tkr.info or {}
+        bs   = tkr.balance_sheet if hasattr(tkr, "balance_sheet") else None
+        is_  = tkr.income_stmt   if hasattr(tkr, "income_stmt")   else None
+
+        def _g(df, row, col=0):
+            try:
+                if df is None or df.empty: return None
+                m = [r for r in df.index if row.lower() in str(r).lower()]
+                if not m: return None
+                v = df.loc[m[0]].iloc[col]
+                return float(v) if str(v) not in ("nan","None","") else None
+            except Exception:
+                return None
+
+        ta      = _g(bs, "Total Assets") or info.get("totalAssets")
+        tl      = (_g(bs, "Total Liabilities Net Minority Interest") or
+                   _g(bs, "Total Liabilities"))
+        ca      = _g(bs, "Current Assets")
+        cl      = (_g(bs, "Total Current Liabilities") or _g(bs, "Current Liabilities"))
+        re      = _g(bs, "Retained Earnings")
+        ebit    = _g(is_, "EBIT") or _g(is_, "Operating Income")
+        revenue = _g(is_, "Total Revenue")
+        market_cap = info.get("marketCap")
+        bv_equity   = info.get("bookValue", 0) or 0
+        shares_out  = info.get("sharesOutstanding", 0) or 0
+        book_equity = bv_equity * shares_out
+
+        if not ta or ta == 0:
+            return {"success": False, "error": "无法获取总资产数据"}
+
+        # Working capital / Total Assets  (X1)
+        wc = (ca or 0) - (cl or 0)
+        x1 = wc / ta
+
+        # Retained Earnings / Total Assets  (X2)
+        x2 = (re or 0) / ta
+
+        # EBIT / Total Assets  (X3)
+        x3 = (ebit or 0) / ta
+
+        # Book/Market Value of Equity / Total Liabilities  (X4 — Z'' uses book)
+        bv = book_equity or (market_cap or 0)
+        x4 = bv / (tl or 1)
+
+        # Z'' = 6.56·X1 + 3.26·X2 + 6.72·X3 + 1.05·X4
+        z = 6.56 * x1 + 3.26 * x2 + 6.72 * x3 + 1.05 * x4
+        z = round(z, 3)
+
+        if z > 2.6:
+            zone = "安全区"
+            risk = "low"
+        elif z > 1.1:
+            zone = "灰色区（不确定）"
+            risk = "medium"
+        else:
+            zone = "破产风险区"
+            risk = "high"
+
+        return {
+            "success":  True,
+            "symbol":   symbol,
+            "z_score":  z,
+            "zone":     zone,
+            "risk":     risk,
+            "components": {
+                "X1_working_capital_ratio": round(x1, 4),
+                "X2_retained_earnings_ratio": round(x2, 4),
+                "X3_ebit_ratio": round(x3, 4),
+                "X4_equity_to_debt": round(x4, 4),
+            },
+            "formula":  "Z'' = 6.56·X1 + 3.26·X2 + 6.72·X3 + 1.05·X4",
+            "provider": "yfinance",
+        }
+    except Exception as e:
+        return {"success": False, "error": str(e)}
+
+
+# ---------------------------------------------------------------------------
+# 21. Options Chain  (期权链)
+# ---------------------------------------------------------------------------
+
+def _get_options_chain(params: dict) -> dict:
+    """
+    获取股票期权链（via yfinance）。
+    返回最近到期日的 calls + puts 列表。
+    """
+    symbol  = str(params.get("symbol", "AAPL")).strip().upper()
+    expiry  = params.get("expiry", "")      # "YYYY-MM-DD" or "" = nearest
+    opt_type = params.get("type", "both").lower()  # "calls" | "puts" | "both"
+    limit   = int(params.get("limit", 15))
+
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    try:
+        tkr = yf.Ticker(symbol)
+        dates = tkr.options
+        if not dates:
+            return {"success": False, "error": f"{symbol} 无可用期权数据"}
+
+        if expiry and expiry in dates:
+            exp = expiry
+        else:
+            exp = dates[0]  # nearest expiry
+
+        chain = tkr.option_chain(exp)
+        price = (tkr.info or {}).get("regularMarketPrice") or (tkr.info or {}).get("currentPrice") or 0
+
+        def _fmt(df):
+            if df is None or df.empty:
+                return []
+            cols = [c for c in ["strike","lastPrice","bid","ask","volume","openInterest",
+                                 "impliedVolatility","inTheMoney"] if c in df.columns]
+            df = df[cols].head(limit)
+            rows = df.to_dict("records")
+            for r in rows:
+                iv = r.get("impliedVolatility")
+                r["iv_pct"] = round(iv * 100, 1) if iv else None
+            return rows
+
+        result: Dict[str, Any] = {
+            "success":     True,
+            "symbol":      symbol,
+            "price":       price,
+            "expiry":      exp,
+            "all_expiries": list(dates[:6]),
+            "provider":    "yfinance",
+        }
+        if opt_type in ("calls", "both"):
+            result["calls"] = _fmt(chain.calls)
+        if opt_type in ("puts", "both"):
+            result["puts"]  = _fmt(chain.puts)
+
+        return result
+    except Exception as e:
+        return {"success": False, "error": str(e)}
+
+
+# ---------------------------------------------------------------------------
+# 22. Ichimoku Cloud  (一目均衡表)
+# ---------------------------------------------------------------------------
+
+def _calculate_ichimoku(params: dict) -> dict:
+    """
+    一目均衡表指标计算。
+    返回 Tenkan-sen(转换线), Kijun-sen(基准线), Senkou Span A/B(先行带),
+    Chikou(迟行线), 以及云层厚度与当前信号。
+    """
+    symbol = str(params.get("symbol", "AAPL")).strip().upper()
+    period = params.get("period", "6mo")
+
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    try:
+        df = yf.Ticker(symbol).history(period=period)
+        if df is None or len(df) < 52:
+            return {"success": False, "error": "历史数据不足（至少需要 52 天）"}
+
+        high = df["High"].astype(float)
+        low  = df["Low"].astype(float)
+        close= df["Close"].astype(float)
+
+        def _mid(h, l, n):
+            return (h.rolling(n).max() + l.rolling(n).min()) / 2
+
+        tenkan  = _mid(high, low, 9)
+        kijun   = _mid(high, low, 26)
+        senkou_a = ((tenkan + kijun) / 2).shift(26)
+        senkou_b = _mid(high, low, 52).shift(26)
+        chikou  = close.shift(-26)
+
+        t  = round(float(tenkan.iloc[-1]), 3)
+        k  = round(float(kijun.iloc[-1]), 3)
+        sa = round(float(senkou_a.iloc[-1]), 3) if not pd.isna(senkou_a.iloc[-1]) else None
+        sb = round(float(senkou_b.iloc[-1]), 3) if not pd.isna(senkou_b.iloc[-1]) else None
+        c  = round(float(close.iloc[-1]), 3)
+        ck = round(float(chikou.iloc[-53]) if len(chikou) > 53 else float(chikou.iloc[0]), 3)
+
+        # Signal
+        above_cloud = sa is not None and sb is not None and c > max(sa, sb)
+        below_cloud = sa is not None and sb is not None and c < min(sa, sb)
+        bullish_tk  = t > k
+        cloud_color = "绿云(多)" if (sa and sb and sa > sb) else "红云(空)"
+
+        if above_cloud and bullish_tk:
+            signal = "强势多头"
+        elif above_cloud:
+            signal = "偏多（价格在云上方）"
+        elif below_cloud and not bullish_tk:
+            signal = "强势空头"
+        elif below_cloud:
+            signal = "偏空（价格在云下方）"
+        else:
+            signal = "震荡（价格在云内）"
+
+        return {
+            "success":     True,
+            "symbol":      symbol,
+            "price":       c,
+            "tenkan":      t,
+            "kijun":       k,
+            "senkou_a":    sa,
+            "senkou_b":    sb,
+            "chikou":      ck,
+            "cloud_color": cloud_color,
+            "cloud_thickness": round(abs((sa or 0) - (sb or 0)), 3),
+            "signal":      signal,
+            "above_cloud": above_cloud,
+            "below_cloud": below_cloud,
+            "tk_cross":    "金叉(多)" if bullish_tk else "死叉(空)",
+            "provider":    "yfinance",
+        }
+    except Exception as e:
+        return {"success": False, "error": str(e)}
+
+
+# ---------------------------------------------------------------------------
+# 23. Crypto Fear & Greed Index  + Funding Rates
+# ---------------------------------------------------------------------------
+
+def _get_fear_greed_index(params: dict) -> dict:
+    """加密货币恐惧贪婪指数（来源: alternative.me，无需 API Key）。"""
+    try:
+        import urllib.request
+        import json as _json
+        with urllib.request.urlopen(
+            "https://api.alternative.me/fng/?limit=7&format=json", timeout=6
+        ) as resp:
+            data = _json.loads(resp.read().decode())
+        items = data.get("data", [])
+        if not items:
+            return {"success": False, "error": "No data returned"}
+        latest   = items[0]
+        value    = int(latest.get("value", 0))
+        label_en = latest.get("value_classification", "")
+        label_cn_map = {
+            "Extreme Fear": "极度恐惧",
+            "Fear":         "恐惧",
+            "Neutral":      "中性",
+            "Greed":        "贪婪",
+            "Extreme Greed":"极度贪婪",
+        }
+        label_cn = label_cn_map.get(label_en, label_en)
+        history  = [{"date": i.get("timestamp",""), "value": int(i.get("value",0)),
+                     "label": i.get("value_classification","")} for i in items]
+        return {
+            "success":  True,
+            "value":    value,
+            "label":    label_cn,
+            "label_en": label_en,
+            "history":  history,
+            "signal":   "做空" if value >= 75 else "做多" if value <= 25 else "中性",
+            "provider": "alternative.me",
+        }
+    except Exception as e:
+        return {"success": False, "error": str(e)}
+
+
+def _get_funding_rates(params: dict) -> dict:
+    """
+    获取永续合约资金费率 (via ccxt)。
+    支持 binance, okx, bybit 等主流交易所。
+    高正费率 → 多头过多 → 看空信号；负费率 → 空头过多 → 看多信号。
+    """
+    exchange_id = params.get("exchange", "binance").lower()
+    symbols = params.get("symbols", ["BTC/USDT", "ETH/USDT", "SOL/USDT"])
+    if isinstance(symbols, str):
+        symbols = [s.strip() for s in symbols.replace(",", " ").split()]
+
+    try:
+        import ccxt as _ccxt
+    except ImportError:
+        return {"success": False, "error": "ccxt 未安装: pip install ccxt"}
+
+    # Try exchanges in order; fall back if load_markets fails (network/region issues)
+    _all_exchanges = ["binance", "okx", "bybit"]
+    _exchange_fallback = [exchange_id] + [e for e in _all_exchanges if e != exchange_id]
+
+    for _exid in _exchange_fallback:
+        try:
+            exchange_cls = getattr(_ccxt, _exid, None)
+            if not exchange_cls:
+                continue
+            ex = exchange_cls({"options": {"defaultType": "future"}})
+            try:
+                ex.load_markets()
+            except Exception as _lm_err:
+                _lm_msg = str(_lm_err)
+                if len(_lm_msg) > 100 or "http" in _lm_msg or "GET" in _lm_msg or "POST" in _lm_msg:
+                    _lm_msg = f"无法连接 {_exid} 期货市场（网络或区域限制）"
+                if _exid == _exchange_fallback[-1]:
+                    return {"success": False, "error": f"{_exid}: {_lm_msg}"}
+                continue
+
+            results = []
+            for sym in symbols:
+                try:
+                    fi = ex.fetch_funding_rate(sym)
+                    rate = fi.get("fundingRate") or fi.get("funding_rate") or 0
+                    next_time = fi.get("fundingDatetime") or fi.get("nextFundingDatetime") or ""
+                    annualized = round(float(rate) * 3 * 365 * 100, 2)  # 8h intervals
+                    results.append({
+                        "symbol":       sym,
+                        "rate":         round(float(rate) * 100, 4),
+                        "rate_pct":     f"{float(rate)*100:.4f}%",
+                        "annualized":   f"{annualized:.1f}%",
+                        "next_funding": str(next_time)[:16],
+                        "signal":       "空" if float(rate) > 0.0005 else "多" if float(rate) < -0.0001 else "中性",
+                    })
+                except Exception:
+                    pass
+
+            if not results:
+                if _exid == _exchange_fallback[-1]:
+                    return {"success": False, "error": f"已尝试 {', '.join(_exchange_fallback)}，均未能获取资金费率数据"}
+                continue
+
+            avg_rate = sum(r["rate"] for r in results) / len(results)
+            return {
+                "success":     True,
+                "exchange":    _exid,
+                "rates":       results,
+                "avg_rate":    round(avg_rate, 4),
+                "market_bias": "多头过热(偏空)" if avg_rate > 0.05 else "空头过多(偏多)" if avg_rate < -0.01 else "均衡",
+                "provider":    "ccxt",
+            }
+        except Exception as e:
+            _err_msg = str(e)
+            if len(_err_msg) > 100 or "http" in _err_msg or "GET" in _err_msg or "POST" in _err_msg:
+                _err_msg = f"无法连接 {_exid}（网络或区域限制）"
+            if _exid == _exchange_fallback[-1]:
+                return {"success": False, "error": _err_msg}
+
+    return {"success": False, "error": "所有备用交易所均连接失败"}
+
+
+def _get_funding_rates_compare(params: dict) -> dict:
+    """
+    并行查询 binance / okx / bybit，返回三所资金费率横向对比。
+    用于发现跨所套利机会（同一标的费率差 > 0.02% 值得关注）。
+    """
+    symbols = params.get("symbols", ["BTC/USDT", "ETH/USDT", "SOL/USDT"])
+    if isinstance(symbols, str):
+        symbols = [s.strip() for s in symbols.replace(",", " ").split()]
+
+    try:
+        import ccxt as _ccxt  # noqa: F401
+    except ImportError:
+        return {"success": False, "error": "ccxt 未安装: pip install ccxt"}
+
+    import concurrent.futures as _fut
+
+    _exchanges = ["binance", "okx", "bybit"]
+
+    def _fetch(exid):
+        return exid, _get_funding_rates({"exchange": exid, "symbols": symbols})
+
+    ex_results: dict = {}
+    with _fut.ThreadPoolExecutor(max_workers=3) as pool:
+        for exid, r in pool.map(_fetch, _exchanges):
+            ex_results[exid] = r
+
+    comparison = []
+    for sym in symbols:
+        row: dict = {"symbol": sym}
+        for exid in _exchanges:
+            r = ex_results.get(exid, {})
+            if r.get("success"):
+                match = next((x for x in r.get("rates", []) if x["symbol"] == sym), None)
+                if match:
+                    row[exid] = match
+        if len(row) > 1:
+            comparison.append(row)
+
+    if not comparison:
+        return {"success": False, "error": "三所均无数据，请检查网络或 VPN"}
+
+    # Find max cross-exchange spread per symbol
+    spreads = []
+    for row in comparison:
+        rates = [row[e]["rate"] for e in _exchanges if e in row]
+        if len(rates) >= 2:
+            spreads.append(round(max(rates) - min(rates), 4))
+
+    max_spread = max(spreads) if spreads else 0.0
+    arb_note = (
+        "⚠ 套利机会：最大价差 > 0.02%" if max_spread > 0.02
+        else "价差正常，无明显套利空间"
+    )
+
+    return {
+        "success":    True,
+        "comparison": comparison,
+        "exchanges":  _exchanges,
+        "max_spread": max_spread,
+        "arb_note":   arb_note,
+        "provider":   "ccxt_compare",
+    }
+
+
+# ---------------------------------------------------------------------------
+# 24. Walk-Forward Backtest  (滚动验证)
+# ---------------------------------------------------------------------------
+
+def _walk_forward_backtest(params: dict) -> dict:
+    """
+    Walk-Forward 滚动回测：将历史分成 N 个窗口，每窗口 in-sample 优化、
+    out-of-sample 验证，评估策略真实泛化能力。
+    """
+    symbol   = params.get("symbol", "AAPL")
+    strategy = params.get("strategy", "sma_crossover")
+    periods  = int(params.get("periods", 5))      # number of WF windows
+    train_r  = float(params.get("train_ratio", 0.7))
+    period   = params.get("period", "5y")
+
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    try:
+        import numpy as np
+        tkr = yf.Ticker(symbol)
+        df  = tkr.history(period=period)
+        if df is None or len(df) < 200:
+            return {"success": False, "error": "历史数据不足（需要至少 200 天）"}
+
+        close = df["Close"].astype(float).values
+        n     = len(close)
+        window_size = n // periods
+
+        window_results = []
+        for i in range(periods):
+            start = i * window_size
+            end   = start + window_size if i < periods - 1 else n
+            split = start + int((end - start) * train_r)
+
+            train = close[start:split]
+            test  = close[split:end]
+
+            if len(test) < 20:
+                continue
+
+            # Simple parameter: SMA crossover with different windows
+            best_sharpe = -np.inf
+            best_fast, best_slow = 10, 30
+
+            if strategy in ("sma_crossover", "ma_crossover"):
+                for fast in (5, 10, 15, 20):
+                    for slow in (20, 30, 40, 60):
+                        if fast >= slow or len(train) <= slow:
+                            continue
+                        sig = np.where(
+                            np.convolve(train, np.ones(fast)/fast, mode="valid")
+                            [-(len(train)-slow+1):] >
+                            np.convolve(train, np.ones(slow)/slow, mode="valid"),
+                            1, 0
+                        )
+                        if len(sig) < 2: continue
+                        rets = np.diff(train[-len(sig):]) / train[-len(sig):-1]
+                        strat_rets = rets * sig[:-1]
+                        sr = (np.mean(strat_rets) / (np.std(strat_rets) + 1e-8)) * np.sqrt(252)
+                        if sr > best_sharpe:
+                            best_sharpe, best_fast, best_slow = sr, fast, slow
+
+            # Out-of-sample evaluation with best params
+            if len(test) <= best_slow:
+                continue
+            fast_ma = np.convolve(test, np.ones(best_fast)/best_fast, mode="valid")
+            slow_ma = np.convolve(test, np.ones(best_slow)/best_slow, mode="valid")
+            n_sig   = min(len(fast_ma), len(slow_ma))
+            signals = np.where(fast_ma[-n_sig:] > slow_ma[-n_sig:], 1, 0)
+            rets_test = np.diff(test[-n_sig:]) / test[-n_sig:-1]
+            strat_rets_oos = rets_test * signals[:-1]
+            bh_rets        = rets_test
+
+            oos_total  = float(np.prod(1 + strat_rets_oos) - 1)
+            bh_total   = float(np.prod(1 + bh_rets) - 1)
+            oos_sharpe = float(np.mean(strat_rets_oos) / (np.std(strat_rets_oos) + 1e-8)) * np.sqrt(252)
+            dd_vals    = np.maximum.accumulate(np.cumprod(1 + strat_rets_oos)) - np.cumprod(1 + strat_rets_oos)
+            max_dd     = float(np.max(dd_vals) / np.maximum.accumulate(np.cumprod(1 + strat_rets_oos))[-1])
+
+            window_results.append({
+                "window":         i + 1,
+                "train_bars":     split - start,
+                "test_bars":      end - split,
+                "best_fast":      best_fast,
+                "best_slow":      best_slow,
+                "oos_return":     round(oos_total, 4),
+                "bh_return":      round(bh_total, 4),
+                "oos_sharpe":     round(oos_sharpe, 3),
+                "max_drawdown":   round(max_dd, 4),
+                "alpha":          round(oos_total - bh_total, 4),
+            })
+
+        if not window_results:
+            return {"success": False, "error": "回测窗口计算失败"}
+
+        avg_oos_ret   = sum(w["oos_return"] for w in window_results) / len(window_results)
+        avg_sharpe    = sum(w["oos_sharpe"] for w in window_results) / len(window_results)
+        avg_alpha     = sum(w["alpha"] for w in window_results) / len(window_results)
+        pct_win       = sum(1 for w in window_results if w["oos_return"] > 0) / len(window_results)
+
+        verdict = (
+            "策略泛化能力强" if avg_alpha > 0.02 and avg_sharpe > 0.5
+            else "策略泛化能力中等" if avg_alpha > 0
+            else "策略泛化能力弱（过拟合风险）"
+        )
+
+        return {
+            "success":           True,
+            "symbol":            symbol,
+            "strategy":          strategy,
+            "windows":           window_results,
+            "avg_oos_return":    round(avg_oos_ret, 4),
+            "avg_sharpe":        round(avg_sharpe, 3),
+            "avg_alpha":         round(avg_alpha, 4),
+            "win_rate_windows":  round(pct_win, 2),
+            "verdict":           verdict,
+            "provider":          "local",
+        }
+    except Exception as e:
+        return {"success": False, "error": str(e)}
+
+
+# ---------------------------------------------------------------------------
+# 25. Peer Comparison  (同行对比)
+# ---------------------------------------------------------------------------
+
+def _peer_comparison(params: dict) -> dict:
+    """
+    同行估值与表现对比。
+    返回 PE/PB/ROE/YTD收益/股息率/市值 横向对比表。
+    """
+    symbol = str(params.get("symbol", "AAPL")).strip().upper()
+    peers  = params.get("peers", [])  # list of ticker strings
+    if isinstance(peers, str):
+        peers = [p.strip().upper() for p in peers.replace(",", " ").split() if p.strip()]
+
+    # Auto-suggest peers from yfinance sector info if not provided
+    if not peers and _HAS_YF:
+        try:
+            info = yf.Ticker(symbol).info or {}
+            # yfinance doesn't give peers directly; use sector to build manual map
+            _SECTOR_PEERS = {
+                "Technology":   ["AAPL","MSFT","GOOGL","META","NVDA","AMZN"],
+                "Financials":   ["JPM","BAC","GS","MS","WFC","C"],
+                "Healthcare":   ["JNJ","LLY","ABBV","MRK","PFE","UNH"],
+                "Consumer Cyclical": ["AMZN","TSLA","HD","MCD","NKE","SBUX"],
+                "Energy":       ["XOM","CVX","COP","SLB","EOG","OXY"],
+            }
+            sector = info.get("sector", "")
+            default_list = _SECTOR_PEERS.get(sector, ["SPY"])
+            peers = [p for p in default_list if p != symbol][:5]
+        except Exception:
+            pass
+
+    if not peers:
+        return {"success": False, "error": "请提供 peers 参数，如: peers=['MSFT','GOOGL','META']"}
+
+    all_symbols = [symbol] + [p for p in peers if p != symbol]
+
+    if not _HAS_YF:
+        return {"success": False, "error": "yfinance not installed: 运行 pip install yfinance 或 /install yfinance"}
+
+    rows = []
+    for sym in all_symbols[:8]:
+        try:
+            info = yf.Ticker(sym).info or {}
+            price  = info.get("regularMarketPrice") or info.get("currentPrice") or 0
+            prev   = info.get("regularMarketPreviousClose") or price
+            pe     = info.get("trailingPE") or info.get("forwardPE")
+            pb     = info.get("priceToBook")
+            roe    = info.get("returnOnEquity")
+            dy     = info.get("dividendYield")
+            mc     = info.get("marketCap")
+            ytd    = (price / prev - 1) if prev else None
+            rows.append({
+                "symbol":    sym,
+                "name":      (info.get("shortName") or sym)[:12],
+                "price":     round(price, 2),
+                "pe":        round(pe, 1)  if pe  else None,
+                "pb":        round(pb, 2)  if pb  else None,
+                "roe_pct":   round(roe * 100, 1) if roe else None,
+                "div_yield": round(dy * 100, 2)  if dy  else None,
+                "market_cap_b": round(mc / 1e9, 1) if mc else None,
+                "is_target": sym == symbol,
+            })
+        except Exception:
+            pass
+
+    if not rows:
+        return {"success": False, "error": "无法获取对比数据"}
+
+    # Relative rankings
+    pe_vals  = [r["pe"]   for r in rows if r["pe"] is not None]
+    pb_vals  = [r["pb"]   for r in rows if r["pb"] is not None]
+    roe_vals = [r["roe_pct"] for r in rows if r["roe_pct"] is not None]
+
+    target_row = next((r for r in rows if r["is_target"]), rows[0])
+    analysis = []
+    if target_row.get("pe") and pe_vals:
+        med_pe = sorted(pe_vals)[len(pe_vals)//2]
+        vs = "高估" if target_row["pe"] > med_pe * 1.2 else "低估" if target_row["pe"] < med_pe * 0.8 else "合理"
+        analysis.append(f"PE {target_row['pe']:.1f}x vs 同行中位数 {med_pe:.1f}x → {vs}")
+    if target_row.get("roe_pct") and roe_vals:
+        avg_roe = sum(roe_vals) / len(roe_vals)
+        vs = "优于同行" if target_row["roe_pct"] > avg_roe else "低于同行"
+        analysis.append(f"ROE {target_row['roe_pct']:.1f}% vs 同行均值 {avg_roe:.1f}% → {vs}")
+
+    return {
+        "success":   True,
+        "symbol":    symbol,
+        "peers":     peers,
+        "table":     rows,
+        "analysis":  analysis,
+        "provider":  "yfinance",
+    }
+
+
+def _resolve_search_key(env_var: str, provider: str) -> str:
+    """Resolve a search-API key: env var first, then ~/.arthera/providers.json.
+
+    Keys added via `/apikey set <provider> <key>` are stored in providers.json
+    (data section), not exported as env vars — so a web search must check both
+    or it silently falls through to DuckDuckGo despite a configured key.
+    """
+    val = os.getenv(env_var, "")
+    if val:
+        return val
+    try:
+        import json as _json
+        import pathlib as _pl
+        for _loc in ("~/.arthera/providers.json", "~/.aria/providers.json"):
+            p = _pl.Path(_loc).expanduser()
+            if not p.exists():
+                continue
+            raw = _json.loads(p.read_text(encoding="utf-8"))
+            for section in ("data", "llm"):
+                entry = raw.get(section, {}).get(provider.lower(), {})
+                if isinstance(entry, dict) and entry.get("api_key"):
+                    return entry["api_key"]
+                if isinstance(entry, str) and entry:
+                    return entry
+    except Exception:
+        pass
+    return ""
+
+
+def _web_search(params: dict) -> dict:
+    """Web search: Brave → Tavily → DuckDuckGo fallback chain."""
+    if not isinstance(params, dict):
+        params = {"query": str(params)}
+    query = (
+        params.get("query")
+        or params.get("q")
+        or params.get("search")
+        or params.get("keyword")
+        or params.get("keywords")
+        or params.get("prompt")
+        or params.get("text")
+        or params.get("symbol")
+        or params.get("ticker")
+        or ""
+    )
+    if isinstance(query, (list, tuple)):
+        query = " ".join(str(x) for x in query if x)
+    elif isinstance(query, dict):
+        query = " ".join(str(v) for v in query.values() if v)
+    query = str(query).strip().strip("'\"`")
+    if not query:
+        # Fallback to general market news query if empty
+        query = "stock market news"
+
+    num = min(int(params.get("num_results", params.get("max_results", 5)) or 5), 10)
+
+    # ── 1. Brave Search API ───────────────────────────────────────────────────
+    brave_key = _resolve_search_key("BRAVE_SEARCH_API_KEY", "brave")
+    if brave_key:
+        try:
+            import urllib.request as _req
+            import urllib.parse as _parse
+            import gzip as _gzip
+            # search_lang must be a valid Brave code. "zh" is INVALID (→ HTTP 422);
+            # use "zh-hans" for Chinese queries and omit it otherwise (auto-detect).
+            _q_params = {"q": query, "count": num, "safesearch": "moderate"}
+            if any("一" <= _c <= "鿿" for _c in query):
+                _q_params["search_lang"] = "zh-hans"
+            url = "https://api.search.brave.com/res/v1/web/search?" + _parse.urlencode(_q_params)
+            req = _req.Request(url, headers={
+                "Accept":               "application/json",
+                "Accept-Encoding":      "gzip",
+                "X-Subscription-Token": brave_key,
+            })
+            with _req.urlopen(req, timeout=10) as r:
+                raw = r.read()
+                if r.headers.get("Content-Encoding") == "gzip":
+                    raw = _gzip.decompress(raw)
+                data = json.loads(raw)
+            results = []
+            for item in data.get("web", {}).get("results", [])[:num]:
+                results.append({
+                    "title":   item.get("title", ""),
+                    "url":     item.get("url", ""),
+                    "snippet": item.get("description", ""),
+                })
+            if results:
+                return {"success": True, "query": query, "results": results, "provider": "brave"}
+        except Exception as e:
+            # Decompress gzip error bodies so the log is readable, not garbage
+            _detail = str(e)
+            try:
+                import urllib.error as _uerr
+                if isinstance(e, _uerr.HTTPError):
+                    _body = e.read()
+                    try:
+                        import gzip as _gz2
+                        if e.headers.get("Content-Encoding") == "gzip":
+                            _body = _gz2.decompress(_body)
+                    except Exception:
+                        pass
+                    _detail = f"HTTP {e.code}: {_body.decode('utf-8', 'ignore')[:200]}"
+            except Exception:
+                pass
+            logger.debug("Brave search failed: %s; trying next provider", _detail)
+
+    # ── 2. Tavily API (designed for AI agents, generous free tier) ───────────
+    tavily_key = _resolve_search_key("TAVILY_API_KEY", "tavily")
+    if tavily_key:
+        try:
+            import urllib.request as _req2
+            import urllib.parse as _parse2
+            req2 = _req2.Request(
+                "https://api.tavily.com/search",
+                data=json.dumps({"api_key": tavily_key, "query": query, "max_results": num, "search_depth": "basic"}).encode(),
+                headers={"Content-Type": "application/json"},
+            )
+            with _req2.urlopen(req2, timeout=10) as r2:
+                data2 = json.loads(r2.read())
+            results = [
+                {"title": item.get("title", ""), "url": item.get("url", ""), "snippet": item.get("content", "")[:300]}
+                for item in data2.get("results", [])[:num]
+            ]
+            if results:
+                return {"success": True, "query": query, "results": results, "provider": "tavily"}
+        except Exception as e:
+            logger.debug("Tavily search failed: %s", e)
+
+    # ── 3. Google Programmable Search (Custom Search JSON API) ───────────────
+    # GOOGLE_SEARCH_API_KEY / GOOGLE_SEARCH_ENGINE_ID were documented in
+    # .env.example but never read by any code path, so configuring them did
+    # nothing and the chain fell through to rate-limited DuckDuckGo.
+    google_key = _resolve_search_key("GOOGLE_SEARCH_API_KEY", "google")
+    google_cx = _resolve_search_key("GOOGLE_SEARCH_ENGINE_ID", "google_cx")
+    if google_key and google_cx:
+        try:
+            import urllib.request as _req3
+            import urllib.parse as _parse3
+
+            _g_params = {"key": google_key, "cx": google_cx, "q": query, "num": num}
+            if any("\u4e00" <= _c <= "\u9fff" for _c in query):
+                _g_params["lr"] = "lang_zh-CN"
+            url3 = "https://www.googleapis.com/customsearch/v1?" + _parse3.urlencode(_g_params)
+            with _req3.urlopen(url3, timeout=10) as r3:
+                data3 = json.loads(r3.read())
+            results = [
+                {
+                    "title":   item.get("title", ""),
+                    "url":     item.get("link", ""),
+                    "snippet": item.get("snippet", ""),
+                }
+                for item in (data3.get("items") or [])[:num]
+            ]
+            if results:
+                return {"success": True, "query": query, "results": results, "provider": "google"}
+        except Exception as e:
+            logger.debug("Google Programmable Search failed: %s; trying next provider", e)
+
+    # ── 4. DuckDuckGo (free, no key, but rate-limited) ────────────────────────
+    try:
+        import warnings as _w
+        with _w.catch_warnings():
+            _w.simplefilter("ignore")
+            try:
+                from ddgs import DDGS
+            except ImportError:
+                from duckduckgo_search import DDGS  # type: ignore[no-redef]
+        results = []
+        for item in DDGS().text(query, max_results=num):
+            results.append({
+                "title":   item.get("title", ""),
+                "url":     item.get("href", ""),
+                "snippet": item.get("body", ""),
+            })
+        if results:
+            return {"success": True, "query": query, "results": results, "provider": "duckduckgo"}
+        return {
+            "success": False,
+            "query":   query,
+            "results": [],
+            "error":   (
+                "DuckDuckGo returned no results (rate-limited). "
+                "推荐配置: BRAVE_SEARCH_API_KEY (免费2000次/月)、TAVILY_API_KEY (AI专用, 免费1000次/月) "
+                "或 GOOGLE_SEARCH_API_KEY + GOOGLE_SEARCH_ENGINE_ID (免费100次/天)"
+            ),
+        }
+    except ImportError:
+        pass
+    except Exception as e:
+        logger.debug("duckduckgo_search failed: %s", e)
+
+    return {
+        "success": False,
+        "query":   query,
+        "results": [],
+        "error":   (
+            "无可用搜索服务。推荐配置:\n"
+            "  BRAVE_SEARCH_API_KEY — https://brave.com/search/api/ (免费2000次/月)\n"
+            "  TAVILY_API_KEY       — https://tavily.com (AI专用, 免费1000次/月)\n"
+            "  GOOGLE_SEARCH_API_KEY + GOOGLE_SEARCH_ENGINE_ID\n"
+            "                       — https://developers.google.com/custom-search (免费100次/天)\n"
+            "  或安装: pip install duckduckgo-search"
+        ),
+    }
+
+
+def _run_portfolio_backtest(params: dict) -> dict:
+    """Event-driven, multi-asset backtest via backtest_engine.
+
+    Unlike backtest_strategy (single-symbol templates), this runs a Strategy over
+    a portfolio of symbols with next-open fills (no look-ahead), commission +
+    slippage, and an equal-weight buy-hold benchmark. Returns a compact result.
+    """
+    raw = params.get("symbols") or params.get("symbol") or ""
+    if isinstance(raw, str):
+        symbols = [s.strip().upper() for s in raw.replace(",", " ").split() if s.strip()]
+    else:
+        symbols = [str(s).strip().upper() for s in (raw or []) if str(s).strip()]
+    if not symbols:
+        return {"success": False, "error": "symbols is required"}
+
+    try:
+        from aria_code.backtest_engine import BacktestEngine, get_strategy, load_bars
+    except Exception as exc:
+        return {"success": False, "error": f"backtest engine unavailable: {exc}"}
+
+    strategy_name = str(params.get("strategy", "sma_cross") or "sma_cross")
+    skw: dict = {}
+    for k in ("fast", "slow"):
+        if params.get(k) is not None:
+            try:
+                skw[k] = int(params[k])
+            except (TypeError, ValueError):
+                pass
+    if params.get("weight") is not None:
+        try:
+            skw["weight"] = float(params["weight"])
+        except (TypeError, ValueError):
+            pass
+    try:
+        strategy = get_strategy(strategy_name, **skw)
+    except ValueError as e:
+        return {"success": False, "error": str(e)}
+
+    try:
+        days = max(30, min(int(params.get("days", 365) or 365), 2000))
+    except (TypeError, ValueError):
+        days = 365
+
+    data, missing = {}, []
+    for s in symbols:
+        try:
+            bars = load_bars(s, days=days)
+        except Exception:
+            bars = []
+        (data.__setitem__(s, bars) if bars else missing.append(s))
+    if not data:
+        return {"success": False, "error": "无法获取历史数据，请稍后重试",
+                "symbols": symbols, "missing": missing}
+
+    def _f(key, default):
+        try:
+            return float(params.get(key, default) or default)
+        except (TypeError, ValueError):
+            return default
+
+    eng = BacktestEngine(starting_cash=_f("starting_cash", 100000.0),
+                         commission=_f("commission", 0.0005),
+                         slippage=_f("slippage", 0.0005))
+    res = eng.run(data, strategy)
+
+    ec = res.equity_curve
+    step = max(1, len(ec) // 20)
+    sample = ec[::step]
+    if ec and sample[-1] is not ec[-1]:
+        sample.append(ec[-1])
+    return {
+        "success": True,
+        "strategy": res.strategy,
+        "symbols": res.symbols,
+        "missing": missing,
+        "starting_cash": res.starting_cash,
+        "metrics": res.metrics,
+        "benchmark": res.benchmark,
+        "alpha_vs_buyhold": round(res.metrics["total_return"] - res.benchmark.get("total_return", 0.0), 6),
+        "n_points": len(ec),
+        "equity_sample": sample,
+        "recent_trades": res.trades[-10:],
+    }
+
+
+# ---------------------------------------------------------------------------
+# Tool registry  (must be after all function definitions)
+# ---------------------------------------------------------------------------
+
+LOCAL_FINANCE_TOOL_REGISTRY: Dict[str, Tuple] = {
+    # ── Market data (cloud → local fallback) ─────────────────────────────
+    "get_market_data":        (_get_market_data,        "Stock/ETF quotes and OHLCV history (A股/US/global, cloud-backed)"),
+    "get_crypto_data":        (_get_crypto_data,        "Cryptocurrency OHLCV and ticker data via ccxt/yfinance"),
+    "get_forex_data":         (_get_forex_data,         "Foreign exchange rates (yfinance)"),
+    "get_commodities_data":   (_get_commodities_data,   "Commodity futures prices: gold, oil, copper, wheat, etc. (yfinance)"),
+    "get_futures_data":       (_get_futures_data,       "Equity index futures: S&P, NASDAQ, VIX, Nikkei, etc. (yfinance)"),
+    # ── Factor & signal (cloud → local fallback) ─────────────────────────
+    "calculate_factors":      (_calculate_factors,      "Technical factors: RSI, MACD, MA gaps, volatility, momentum, trend score (cloud-enhanced)"),
+    "get_ai_signal":          (_get_ai_signal,          "AI trading signal: BUY/SELL/HOLD with confidence and reasoning (Alibaba Cloud DeepSeek)"),
+    "get_market_insights":    (_get_market_insights,    "AI narrative market insights for a basket of symbols (Alibaba Cloud)"),
+    "get_predictions":        (_get_predictions,        "ML-powered 5/10-day return predictions (Alibaba Cloud LightGBM/XGBoost)"),
+    # ── Backtest (cloud ML → local pandas fallback) ───────────────────────
+    "backtest_strategy":      (_backtest_strategy,      "Run a local pandas backtest: sma_cross | rsi_mean_revert | momentum | buy_hold"),
+    "cloud_backtest":         (_cloud_backtest,         "Full ML-powered backtest via Alibaba Cloud QuantEngine (rebalance freq, dynamic position)"),
+    # ── Risk & portfolio ──────────────────────────────────────────────────
+    "get_risk_metrics":       (_get_risk_metrics,       "VaR, CVaR, max drawdown, Sharpe, Calmar, skew, kurtosis"),
+    "optimize_positions":     (_optimize_positions,     "Portfolio optimisation: max_sharpe | min_var | equal_weight"),
+    # ── A股 data services ─────────────────────────────────────────────────
+    "get_sector_performance": (_get_sector_performance, "Sector performance ranking (A股 industry / US SPDR ETFs)"),
+    "get_northbound_flow":    (_get_northbound_flow,    "北向资金 (沪深港通) net buy flow via akshare"),
+    "screen_ashare":          (_screen_ashare,          "A股选股筛选: PE, ROE, market cap, momentum"),
+    "get_limit_up_pool":      (_get_limit_up_pool,      "A股涨停板池 (today's limit-up stocks via akshare)"),
+    "get_market_indices":     (_get_market_indices,     "Global market indices: US, CN, EU, crypto, commodities"),
+    # ── News & macro ──────────────────────────────────────────────────────
+    "analyze_news":           (_analyze_news,           "News sentiment analysis (A股 via akshare)"),
+    "get_bonds_data":         (_get_bonds_data,         "US Treasury yield curve (yfinance)"),
+    # ── Quality scores ────────────────────────────────────────────────────
+    "piotroski_fscore":       (_piotroski_fscore,       "Piotroski F-Score (0-9): 财务质量评分，≥7 高质量做多，≤3 低质量做空"),
+    "altman_zscore":          (_altman_zscore,          "Altman Z''-Score: 企业破产风险预测，>2.6安全，<1.1高风险"),
+    # ── Options ───────────────────────────────────────────────────────────
+    "get_options_chain":      (_get_options_chain,      "期权链: 获取股票的 calls/puts，含行权价、隐含波动率、未平仓量"),
+    # ── Technical indicators ──────────────────────────────────────────────
+    "calculate_ichimoku":     (_calculate_ichimoku,     "一目均衡表 Ichimoku Cloud: 转换线/基准线/先行带/迟行线，含信号判断"),
+    # ── Crypto ────────────────────────────────────────────────────────────
+    "get_fear_greed_index":   (_get_fear_greed_index,   "加密恐惧贪婪指数 (0-100)，>75极度贪婪/做空信号，<25极度恐惧/做多信号"),
+    "get_funding_rates":      (_get_funding_rates,      "永续合约资金费率 (ccxt): 高正费率=多头过热，负费率=空头过多"),
+    "get_funding_rates_compare": (_get_funding_rates_compare, "三所费率横向对比 (binance/okx/bybit): 并行查询，发现跨所套利机会"),
+    # ── Portfolio / backtesting ───────────────────────────────────────────
+    "walk_forward_backtest":  (_walk_forward_backtest,  "Walk-Forward 滚动回测：N个窗口验证策略泛化能力，避免过拟合"),
+    "run_portfolio_backtest": (_run_portfolio_backtest, "事件驱动多资产回测：next-open 成交(无未来函数)+成本+买入持有基准，同一策略可驱动实盘"),
+    "peer_comparison":        (_peer_comparison,        "同行对比: PE/PB/ROE/市值/股息率横向比较，自动识别同行业股票"),
+    # ── Web search ────────────────────────────────────────────────────────
+    "web_search":             (_web_search,             "Web search via Brave Search API or DuckDuckGo fallback; set BRAVE_SEARCH_API_KEY for higher quota"),
+}
+
+# ---------------------------------------------------------------------------
+# Registration helper
+# ---------------------------------------------------------------------------
+
+def register_local_finance_tools(
+    tool_registry: Dict,
+    schema_registry: List,
+) -> int:
+    """
+    Add local finance tools to the CLI's LOCAL_TOOLS and LOCAL_TOOL_SCHEMAS.
+
+    Only registers tools whose names are NOT already present (never overwrites
+    existing tools, so remote Aria tools take precedence when backend is up).
+
+    Returns number of tools newly registered.
+    """
+    added = 0
+    for name, (handler, description) in LOCAL_FINANCE_TOOL_REGISTRY.items():
+        if name not in tool_registry:
+            tool_registry[name] = (
+                lambda p, h=handler: _safe(h, p),
+                description,
+            )
+            added += 1
+
+    existing_schema_names = {
+        s.get("function", {}).get("name") for s in schema_registry
+    }
+    for schema in LOCAL_FINANCE_TOOL_SCHEMAS:
+        sname = schema.get("function", {}).get("name", "")
+        if sname and sname not in existing_schema_names:
+            schema_registry.append(schema)
+
+    return added
+
+    return added