cilipep/ai-agent-with-tslab/multi-timeframe-features

import numpy as np, pandas as pd, talib
def compute_indicators(df, prefix):
    df = df.copy()
    c, h, l = df["Close"].values.astype(np.float64), df["High"].values.astype(np.float64), df["Low"].values.astype(np.float64)
    df[f"{prefix}ema10"] = talib.EMA(c, 10)
    df[f"{prefix}ema20"] = talib.EMA(c, 20)
    df[f"{prefix}rsi"] = talib.RSI(c, 14)
    macd, sig, hist = talib.MACD(c)
    df[f"{prefix}macd"] = macd
    bu, bm, bl = talib.BBANDS(c, 20)
    df[f"{prefix}bb_w"] = (bu - bl) / bm
    df[f"{prefix}atr"] = talib.ATR(h, l, c, 14)
    df[f"{prefix}adx"] = talib.ADX(h, l, c, 14)
    df[f"{prefix}ret"] = df["Close"].pct_change()
    keep = [col for col in df.columns if col.startswith(prefix) or col == "Close"]
    return df[keep]
def merge_timeframes(daily, tf_data):
    result = daily.copy()
    for tf_name, tf_df in tf_data.items():
        ind = compute_indicators(tf_df, f"{tf_name}_")
        aligned = ind.reindex(result.index, method="ffill")
        close_cols = [c for c in aligned.columns if "Close" in c or c == "Close"]
        if close_cols: aligned = aligned.drop(columns=close_cols)
        result = pd.concat([result, aligned], axis=1)
    result = result.ffill().bfill()
    result = result.loc[:, ~result.columns.duplicated()]
    return result