bifitnex-trading/indicators.py

import logging

import pandas as pd
import ta

import config

logger = logging.getLogger(__name__)


def calculate_indicators(candles_df: pd.DataFrame) -> pd.DataFrame:
    """Add technical indicators to a candles DataFrame.

    Expected input columns: timestamp, open, close, high, low, volume.
    Returns the DataFrame with additional indicator columns.
    """
    df = candles_df.copy()
    if df.empty or len(df) < 30:
        logger.warning("Not enough candle data to compute indicators (got %d rows)", len(df))
        return df

    close = df["close"]
    high = df["high"]
    low = df["low"]
    volume = df["volume"]

    # RSI(14)
    df["rsi"] = ta.momentum.rsi(close, window=14)

    # MACD(12, 26, 9)
    macd = ta.trend.MACD(close, window_slow=26, window_fast=12, window_sign=9)
    df["macd"] = macd.macd()
    df["macd_signal"] = macd.macd_signal()
    df["macd_hist"] = macd.macd_diff()

    # Bollinger Bands(20, 2)
    bb = ta.volatility.BollingerBands(close, window=20, window_dev=2)
    df["bb_upper"] = bb.bollinger_hband()
    df["bb_middle"] = bb.bollinger_mavg()
    df["bb_lower"] = bb.bollinger_lband()

    # EMA(9) and EMA(21)
    df["ema9"] = ta.trend.ema_indicator(close, window=9)
    df["ema21"] = ta.trend.ema_indicator(close, window=21)

    # VWAP (cumulative for the session)
    tp = (high + low + close) / 3
    cumvol = volume.cumsum()
    df["vwap"] = (tp * volume).cumsum() / cumvol.replace(0, float("nan"))

    # ATR(14)
    df["atr"] = ta.volatility.average_true_range(high, low, close, window=14)

    # Volume moving average (20)
    df["vol_ma20"] = volume.rolling(window=20).mean()

    # ADX(14) — trend strength
    df["adx"] = ta.trend.adx(high, low, close, window=14)

    # Stochastic RSI(14, 14, 3, 3)
    df["stoch_rsi_k"] = ta.momentum.stochrsi_k(close, window=14, smooth1=3, smooth2=3)
    df["stoch_rsi_d"] = ta.momentum.stochrsi_d(close, window=14, smooth1=3, smooth2=3)

    # OBV (On-Balance Volume) + slope
    df["obv"] = ta.volume.on_balance_volume(close, volume)
    df["obv_slope"] = df["obv"].diff(5)

    # CMF (Chaikin Money Flow, 20-period)
    df["cmf"] = ta.volume.chaikin_money_flow(high, low, close, volume, window=20)

    # Buy pressure (order book imbalance proxy from OHLCV)
    hl_range = high - low
    df["buy_pressure"] = ((close - low) / hl_range).where(hl_range > 0, 0.5)

    return df


def calculate_htf_indicators(candles_df: pd.DataFrame) -> pd.DataFrame:
    """Calculate trend-context indicators on higher timeframe (1h) candles."""
    df = candles_df.copy()
    if df.empty or len(df) < 30:
        return df

    close = df["close"]
    high = df["high"]
    low = df["low"]

    df["ema9"] = ta.trend.ema_indicator(close, window=9)
    df["ema21"] = ta.trend.ema_indicator(close, window=21)
    df["adx"] = ta.trend.adx(high, low, close, window=14)
    df["rsi"] = ta.momentum.rsi(close, window=14)

    return df


def calculate_pivot_points(htf_df: pd.DataFrame) -> dict | None:
    """Calculate classic pivot points from previous day's price action using 1h candles."""
    if htf_df.empty or len(htf_df) < 24:
        return None

    df = htf_df.copy()
    df["date"] = df["timestamp"].dt.date

    dates = sorted(df["date"].unique())
    if len(dates) < 2:
        return None

    prev_day = df[df["date"] == dates[-2]]
    if prev_day.empty:
        return None

    h = prev_day["high"].max()
    l = prev_day["low"].min()
    c = prev_day["close"].iloc[-1]

    pivot = (h + l + c) / 3
    return {
        "pivot": pivot,
        "r1": 2 * pivot - l,
        "r2": pivot + (h - l),
        "r3": h + 2 * (pivot - l),
        "s1": 2 * pivot - h,
        "s2": pivot - (h - l),
        "s3": l - 2 * (h - pivot),
    }


def summarize_indicators(df: pd.DataFrame, symbol: str) -> str:
    """Produce a concise text summary of the latest indicator values for LLM consumption."""
    if df.empty or len(df) < 2:
        return f"{symbol}: insufficient data\n"

    last = df.iloc[-1]
    prev = df.iloc[-2]
    name = config.SYMBOL_NAMES.get(symbol, symbol)

    lines = [f"### {name} ({symbol})"]
    lines.append(f"Price: {last['close']:.6g} | High: {last['high']:.6g} | Low: {last['low']:.6g}")
    lines.append(f"Volume: {last['volume']:.2f} | Vol MA20: {last.get('vol_ma20', 0):.2f}")

    if pd.notna(last.get("rsi")):
        lines.append(f"RSI(14): {last['rsi']:.1f}")
    if pd.notna(last.get("macd")):
        hist_dir = "↑" if last["macd_hist"] > prev.get("macd_hist", 0) else "↓"
        lines.append(
            f"MACD: {last['macd']:.6g} | Signal: {last['macd_signal']:.6g} | "
            f"Hist: {last['macd_hist']:.6g} ({hist_dir})"
        )
    if pd.notna(last.get("bb_upper")):
        lines.append(
            f"BB: Upper={last['bb_upper']:.6g} Mid={last['bb_middle']:.6g} Lower={last['bb_lower']:.6g}"
        )
    if pd.notna(last.get("ema9")):
        trend = "bullish" if last["ema9"] > last["ema21"] else "bearish"
        lines.append(f"EMA9: {last['ema9']:.6g} | EMA21: {last['ema21']:.6g} → {trend}")
    if pd.notna(last.get("atr")):
        lines.append(f"ATR(14): {last['atr']:.6g}")
    if pd.notna(last.get("vwap")):
        lines.append(f"VWAP: {last['vwap']:.6g}")
    if pd.notna(last.get("adx")):
        adx_label = "趨勢" if last["adx"] > 25 else "盤整" if last["adx"] < 20 else "弱"
        lines.append(f"ADX(14): {last['adx']:.1f} ({adx_label})")
    if pd.notna(last.get("stoch_rsi_k")):
        lines.append(f"StochRSI: K={last['stoch_rsi_k']:.2f} D={last['stoch_rsi_d']:.2f}")
    if pd.notna(last.get("obv_slope")):
        obv_dir = "流入" if last["obv_slope"] > 0 else "流出"
        lines.append(f"OBV趨勢: {obv_dir} (slope={last['obv_slope']:.0f})")
    if pd.notna(last.get("cmf")):
        cmf_label = "買壓" if last["cmf"] > 0.05 else "賣壓" if last["cmf"] < -0.05 else "中性"
        lines.append(f"CMF(20): {last['cmf']:.3f} ({cmf_label})")

    return "\n".join(lines) + "\n"


def summarize_htf(htf_by_symbol: dict[str, pd.DataFrame]) -> str:
    """Produce a compact 1h trend context summary for LLM."""
    lines = ["## 1小時趨勢背景"]
    for sym in sorted(htf_by_symbol):
        df = htf_by_symbol[sym]
        if df.empty or len(df) < 2:
            continue
        last = df.iloc[-1]
        name = config.SYMBOL_NAMES.get(sym, sym)

        trend = "多頭" if last.get("ema9", 0) > last.get("ema21", 0) else "空頭"
        adx_val = last.get("adx", 0)
        strength = "強趨勢" if adx_val > 25 else "盤整" if adx_val < 20 else "弱趨勢"
        rsi_1h = last.get("rsi", 50)

        lines.append(f"- {name}: {trend} ({strength}, ADX={adx_val:.0f}, RSI={rsi_1h:.0f})")
    return "\n".join(lines) + "\n"


def summarize_pivots(pivots_by_symbol: dict[str, dict], current_prices: dict) -> str:
    """Summarize pivot point levels relative to current price."""
    lines = ["## 關鍵支撐/阻力 (日線樞紐)"]
    for sym in sorted(pivots_by_symbol):
        p = pivots_by_symbol[sym]
        if not p:
            continue
        name = config.SYMBOL_NAMES.get(sym, sym)
        price = current_prices.get(sym, 0)
        if price <= 0:
            continue

        supports = [("S3", p["s3"]), ("S2", p["s2"]), ("S1", p["s1"]), ("P", p["pivot"])]
        resistances = [("P", p["pivot"]), ("R1", p["r1"]), ("R2", p["r2"]), ("R3", p["r3"])]

        nearest_sup = max([(n, v) for n, v in supports if v < price], key=lambda x: x[1], default=None)
        nearest_res = min([(n, v) for n, v in resistances if v > price], key=lambda x: x[1], default=None)

        sup_str = f"{nearest_sup[0]}={nearest_sup[1]:.6g}" if nearest_sup else "N/A"
        res_str = f"{nearest_res[0]}={nearest_res[1]:.6g}" if nearest_res else "N/A"
        lines.append(f"- {name}: 支撐 {sup_str} | 阻力 {res_str}")

    return "\n".join(lines) + "\n"


def summarize_all(indicators_by_symbol: dict[str, pd.DataFrame]) -> str:
    """Build a combined indicator summary for all symbols."""
    parts = []
    for sym, df in sorted(indicators_by_symbol.items()):
        parts.append(summarize_indicators(df, sym))
    return "\n".join(parts)