src/patterns.py

"""
Market Structure, Chart Pattern, and Candlestick Pattern Detector
Provides objective, rule-based detection for the TVTA Vision Dataset.
"""

import numpy as np
import pandas as pd
from typing import List, Dict, Any, Optional, Tuple


# ─── SWING POINT DETECTION ───────────────────────────────────────────────────

def find_swing_highs(lows: pd.Series, highs: pd.Series, order: int = 5) -> List[int]:
    """Find swing high indices: highs[i] is the highest in [i-order, i+order]."""
    swing_indices = []
    for i in range(order, len(highs) - order):
        if highs.iloc[i] == highs.iloc[i - order:i + order + 1].max():
            swing_indices.append(i)
    return swing_indices


def find_swing_lows(lows: pd.Series, highs: pd.Series, order: int = 5) -> List[int]:
    """Find swing low indices: lows[i] is the lowest in [i-order, i+order]."""
    swing_indices = []
    for i in range(order, len(lows) - order):
        if lows.iloc[i] == lows.iloc[i - order:i + order + 1].min():
            swing_indices.append(i)
    return swing_indices


# ─── MARKET STRUCTURE ────────────────────────────────────────────────────────

def detect_market_structure(
    df: pd.DataFrame, swing_order: int = 5
) -> Dict[str, Any]:
    """
    Detect market structure: HH, HL, LH, LL, BOS, CHoCH.
    Uses only data up to the final candle (no future leakage).
    """
    highs = df["high"]
    lows = df["low"]
    closes = df["close"]

    swing_highs = find_swing_highs(lows, highs, swing_order)
    swing_lows = find_swing_lows(lows, highs, swing_order)

    if len(swing_highs) < 2 and len(swing_lows) < 2:
        return {
            "classification": "neutral",
            "sequence": "",
            "break_of_structure": False,
            "change_of_character": False,
            "swing_highs": len(swing_highs),
            "swing_lows": len(swing_lows),
        }

    # Build structure sequence
    structure_seq = []
    all_swings = []
    for idx in swing_highs:
        all_swings.append(("H", idx, highs.iloc[idx]))
    for idx in swing_lows:
        all_swings.append(("L", idx, lows.iloc[idx]))
    all_swings.sort(key=lambda x: x[1])

    # Determine HH/HL/LH/LL
    prev_high = None
    prev_low = None
    seq_labels = []

    for swing_type, idx, price in all_swings:
        if swing_type == "H":
            if prev_high is not None:
                if price > prev_high:
                    seq_labels.append("HH")
                else:
                    seq_labels.append("LH")
            prev_high = price
        else:
            if prev_low is not None:
                if price > prev_low:
                    seq_labels.append("HL")
                else:
                    seq_labels.append("LL")
            prev_low = price

    sequence_str = "_".join(seq_labels) if seq_labels else ""

    # Classification
    hh_count = seq_labels.count("HH")
    ll_count = seq_labels.count("LL")
    hl_count = seq_labels.count("HL")
    lh_count = seq_labels.count("LH")

    if hh_count > lh_count and hl_count >= ll_count:
        classification = "bullish"
    elif ll_count > hl_count and lh_count >= hh_count:
        classification = "bearish"
    elif hh_count == ll_count and len(seq_labels) > 0:
        classification = "mixed"
    else:
        classification = "neutral"

    # Break of Structure (BOS): price breaks the last significant swing level
    bos = False
    if len(all_swings) >= 2:
        last_swing = all_swings[-1]
        prev_swing = all_swings[-2]
        if last_swing[0] == "H" and closes.iloc[-1] > last_swing[2]:
            bos = True
        elif last_swing[0] == "L" and closes.iloc[-1] < last_swing[2]:
            bos = True

    # Change of Character (CHoCH): structure reverses after a clear trend
    choch = False
    if len(seq_labels) >= 3:
        recent = seq_labels[-3:]
        if classification == "bullish" and "LL" in recent[:2]:
            choch = True
        elif classification == "bearish" and "HH" in recent[:2]:
            choch = True

    return {
        "classification": classification,
        "sequence": sequence_str,
        "break_of_structure": bos,
        "change_of_character": choch,
        "swing_highs": len(swing_highs),
        "swing_lows": len(swing_lows),
    }


# ─── TREND CLASSIFICATION ───────────────────────────────────────────────────

def classify_trend(df: pd.DataFrame, idx: int = -1) -> Dict[str, Any]:
    """
    Classify the dominant trend using objective rules.
    """
    if idx < 0:
        idx = len(df) + idx
    if idx < 0 or idx >= len(df):
        idx = len(df) - 1

    row = df.iloc[idx]
    close = row.get("close", 0)

    score = 0
    reasons = []

    # Moving average alignment
    ema20 = row.get("ema_20", 0)
    ema50 = row.get("ema_50", 0)
    ema200 = row.get("ema_200", 0)

    if ema20 > ema50 > ema200:
        score += 2
        reasons.append("ema_bullish_alignment")
    elif ema20 < ema50 < ema200:
        score -= 2
        reasons.append("ema_bearish_alignment")

    # Price vs MAs
    if close > ema20:
        score += 1
    else:
        score -= 1
    if close > ema200:
        score += 1
    else:
        score -= 1

    # RSI
    rsi = row.get("rsi_14", 50)
    if rsi > 60:
        score += 1
    elif rsi < 40:
        score -= 1

    # MACD
    macd = row.get("macd", 0)
    macd_sig = row.get("macd_signal", 0)
    if macd > macd_sig:
        score += 1
    else:
        score -= 1

    # ADX
    adx = row.get("adx", 20)
    trending = adx > 25

    # Price structure (last 20 candles)
    if idx >= 20:
        recent_highs = df["high"].iloc[idx - 20:idx + 1]
        recent_lows = df["low"].iloc[idx - 20:idx + 1]
        if recent_highs.is_monotonic_increasing:
            score += 1
        elif recent_highs.is_monotonic_decreasing:
            score -= 1
        if recent_lows.is_monotonic_increasing:
            score += 1
        elif recent_lows.is_monotonic_decreasing:
            score -= 1

    # Classify
    if score >= 5:
        trend = "strong_bullish"
    elif score >= 3:
        trend = "bullish"
    elif score >= 1:
        trend = "weak_bullish"
    elif score == 0:
        trend = "sideways"
    elif score >= -2:
        trend = "weak_bearish"
    elif score >= -4:
        trend = "bearish"
    else:
        trend = "strong_bearish"

    return {
        "classification": trend,
        "score": score,
        "trending": trending,
        "adx": round(float(adx), 2),
        "reasons": reasons,
    }


# ─── SUPPORT AND RESISTANCE ─────────────────────────────────────────────────

def find_support_resistance(
    df: pd.DataFrame,
    swing_order: int = 5,
    tolerance_pct: float = 0.01,
    min_touches: int = 2,
) -> Dict[str, Any]:
    """
    Find significant support and resistance levels from swing points.
    """
    highs = df["high"]
    lows = df["low"]

    swing_highs = find_swing_highs(lows, highs, swing_order)
    swing_lows = find_swing_lows(lows, highs, swing_order)

    # Cluster swing highs into resistance levels
    resistance_clusters = _cluster_levels(
        [highs.iloc[i] for i in swing_highs], tolerance_pct
    )
    support_clusters = _cluster_levels(
        [lows.iloc[i] for i in swing_lows], tolerance_pct
    )

    # Filter by minimum touches
    resistance_levels = []
    for price, touches in resistance_clusters:
        if touches >= min_touches:
            resistance_levels.append({
                "price": round(float(price), 2),
                "strength": _level_strength(touches),
                "touches": touches,
            })

    support_levels = []
    for price, touches in support_clusters:
        if touches >= min_touches:
            support_levels.append({
                "price": round(float(price), 2),
                "strength": _level_strength(touches),
                "touches": touches,
            })

    # Sort by proximity to current price
    current_price = df["close"].iloc[-1]
    resistance_levels.sort(key=lambda x: x["price"])
    support_levels.sort(key=lambda x: x["price"], reverse=True)

    # Keep only levels near current price
    resistance_levels = [
        r for r in resistance_levels
        if r["price"] > current_price * 0.95 and r["price"] < current_price * 1.15
    ]
    support_levels = [
        s for s in support_levels
        if s["price"] < current_price * 1.05 and s["price"] > current_price * 0.85
    ]

    return {
        "support_levels": support_levels[:5],
        "resistance_levels": resistance_levels[:5],
    }


def _cluster_levels(prices: List[float], tolerance_pct: float) -> List[Tuple[float, int]]:
    """Cluster nearby price levels."""
    if not prices:
        return []

    prices_sorted = sorted(prices)
    clusters = []
    current_cluster = [prices_sorted[0]]

    for p in prices_sorted[1:]:
        cluster_mean = np.mean(current_cluster)
        if abs(p - cluster_mean) / cluster_mean <= tolerance_pct:
            current_cluster.append(p)
        else:
            clusters.append((np.mean(current_cluster), len(current_cluster)))
            current_cluster = [p]

    if current_cluster:
        clusters.append((np.mean(current_cluster), len(current_cluster)))

    return clusters


def _level_strength(touches: int) -> str:
    if touches >= 5:
        return "very_strong"
    elif touches >= 3:
        return "strong"
    elif touches >= 2:
        return "medium"
    else:
        return "weak"


# ─── CANDLESTICK PATTERNS ───────────────────────────────────────────────────

def detect_candlestick_patterns(df: pd.DataFrame) -> List[Dict[str, Any]]:
    """
    Detect common candlestick patterns at the most recent candles.
    """
    patterns = []
    n = len(df)
    if n < 5:
        return patterns

    o = df["open"].values
    h = df["high"].values
    l = df["low"].values
    c = df["close"].values

    # Body and shadow sizes
    body = np.abs(c - o)
    upper_shadow = h - np.maximum(o, c)
    lower_shadow = np.minimum(o, c) - l
    total_range = h - l
    total_range = np.where(total_range == 0, 1e-10, total_range)

    avg_body = np.mean(body[-20:]) if n >= 20 else np.mean(body)

    # Doji: very small body
    for i in range(max(n - 5, 0), n):
        if body[i] < 0.1 * total_range[i]:
            patterns.append({
                "pattern": "doji",
                "index": i,
                "strength": "medium",
            })

    # Hammer: small body at upper end, long lower shadow
    for i in range(max(n - 5, 0), n):
        if (body[i] < 0.3 * total_range[i] and
            lower_shadow[i] > 2 * body[i] and
            upper_shadow[i] < 0.1 * total_range[i]):
            patterns.append({
                "pattern": "hammer",
                "index": i,
                "strength": "medium",
            })

    # Inverted Hammer: small body at lower end, long upper shadow
    for i in range(max(n - 5, 0), n):
        if (body[i] < 0.3 * total_range[i] and
            upper_shadow[i] > 2 * body[i] and
            lower_shadow[i] < 0.1 * total_range[i]):
            patterns.append({
                "pattern": "inverted_hammer",
                "index": i,
                "strength": "medium",
            })

    # Shooting Star: small body at lower end, long upper shadow, uptrend
    for i in range(max(n - 5, 0), n):
        if (body[i] < 0.3 * total_range[i] and
            upper_shadow[i] > 2 * body[i] and
            lower_shadow[i] < 0.1 * total_range[i] and
            c[i - 1] > o[i - 1] if i > 0 else False):
            patterns.append({
                "pattern": "shooting_star",
                "index": i,
                "strength": "medium",
            })

    # Bullish Engulfing
    for i in range(max(n - 5, 1), n):
        if (c[i - 1] < o[i - 1] and  # Previous bearish
            c[i] > o[i] and            # Current bullish
            o[i] <= c[i - 1] and
            c[i] >= o[i - 1]):
            patterns.append({
                "pattern": "bullish_engulfing",
                "index": i,
                "strength": "strong",
            })

    # Bearish Engulfing
    for i in range(max(n - 5, 1), n):
        if (c[i - 1] > o[i - 1] and  # Previous bullish
            c[i] < o[i] and            # Current bearish
            o[i] >= c[i - 1] and
            c[i] <= o[i - 1]):
            patterns.append({
                "pattern": "bearish_engulfing",
                "index": i,
                "strength": "strong",
            })

    # Morning Star (3-candle bullish reversal)
    for i in range(max(n - 5, 2), n):
        if (c[i - 2] < o[i - 2] and                    # Bearish
            body[i - 1] < 0.3 * body[i - 2] and          # Small body
            c[i] > o[i] and                               # Bullish
            c[i] > (o[i - 2] + c[i - 2]) / 2):           # Close above midpoint
            patterns.append({
                "pattern": "morning_star",
                "index": i,
                "strength": "strong",
            })

    # Evening Star (3-candle bearish reversal)
    for i in range(max(n - 5, 2), n):
        if (c[i - 2] > o[i - 2] and                    # Bullish
            body[i - 1] < 0.3 * body[i - 2] and          # Small body
            c[i] < o[i] and                               # Bearish
            c[i] < (o[i - 2] + c[i - 2]) / 2):           # Close below midpoint
            patterns.append({
                "pattern": "evening_star",
                "index": i,
                "strength": "strong",
            })

    # Three White Soldiers (3 consecutive bullish candles)
    for i in range(max(n - 5, 2), n):
        if (c[i - 2] > o[i - 2] and c[i - 1] > o[i - 1] and c[i] > o[i] and
            c[i - 2] < c[i - 1] < c[i] and o[i - 2] < o[i - 1] < o[i]):
            patterns.append({
                "pattern": "three_white_soldiers",
                "index": i,
                "strength": "strong",
            })

    # Three Black Crows (3 consecutive bearish candles)
    for i in range(max(n - 5, 2), n):
        if (c[i - 2] < o[i - 2] and c[i - 1] < o[i - 1] and c[i] < o[i] and
            c[i - 2] > c[i - 1] > c[i] and o[i - 2] > o[i - 1] > o[i]):
            patterns.append({
                "pattern": "three_black_crows",
                "index": i,
                "strength": "strong",
            })

    # Deduplicate: keep only the most recent occurrence of each pattern type
    seen = set()
    unique_patterns = []
    for p in reversed(patterns):
        if p["pattern"] not in seen:
            seen.add(p["pattern"])
            unique_patterns.append(p)
    unique_patterns.reverse()

    return unique_patterns


# ─── CHART PATTERN DETECTION ────────────────────────────────────────────────

def detect_chart_patterns(df: pd.DataFrame) -> List[Dict[str, Any]]:
    """
    Detect chart patterns from price structure.
    Returns patterns only when structural evidence meets criteria.
    """
    patterns = []
    n = len(df)
    if n < 30:
        return patterns

    highs = df["high"].values
    lows = df["low"].values
    closes = df["close"].values
    volumes = df["volume"].values

    # Find swing points
    swing_highs = find_swing_highs(df["low"], df["high"], 5)
    swing_lows = find_swing_lows(df["low"], df["high"], 5)

    # Double Top: two similar highs with a trough in between
    if len(swing_highs) >= 2:
        for i in range(len(swing_highs) - 1):
            h1 = highs[swing_highs[i]]
            h2 = highs[swing_highs[i + 1]]
            if abs(h1 - h2) / h1 < 0.02:  # Within 2%
                trough_idx = swing_highs[i] + np.argmin(lows[swing_highs[i]:swing_highs[i + 1]])
                trough = lows[trough_idx]
                if (h1 - trough) / h1 > 0.03:  # At least 3% drop
                    patterns.append({
                        "pattern": "double_top",
                        "strength": "strong",
                        "neckline": round(float(trough), 2),
                    })
                    break

    # Double Bottom: two similar lows with a peak in between
    if len(swing_lows) >= 2:
        for i in range(len(swing_lows) - 1):
            l1 = lows[swing_lows[i]]
            l2 = lows[swing_lows[i + 1]]
            if abs(l1 - l2) / l1 < 0.02:
                peak_idx = swing_lows[i] + np.argmax(highs[swing_lows[i]:swing_lows[i + 1]])
                peak = highs[peak_idx]
                if (peak - l1) / l1 > 0.03:
                    patterns.append({
                        "pattern": "double_bottom",
                        "strength": "strong",
                        "neckline": round(float(peak), 2),
                    })
                    break

    # Head and Shoulders: three highs with the middle highest
    if len(swing_highs) >= 3:
        for i in range(len(swing_highs) - 2):
            left = highs[swing_highs[i]]
            head = highs[swing_highs[i + 1]]
            right = highs[swing_highs[i + 2]]
            if head > left * 1.02 and head > right * 1.02:
                # Find neckline (lows between shoulders)
                neck_lows = lows[swing_highs[i]:swing_highs[i + 2] + 1]
                neckline = np.min(neck_lows)
                patterns.append({
                    "pattern": "head_and_shoulders",
                    "strength": "strong",
                    "neckline": round(float(neckline), 2),
                })
                break

    # Ascending Triangle: rising flat resistance, higher lows
    if len(swing_highs) >= 2 and len(swing_lows) >= 2:
        recent_highs = [highs[i] for i in swing_highs[-3:]]
        recent_lows = [lows[i] for i in swing_lows[-3:]]
        if (abs(recent_highs[-1] - recent_highs[0]) / recent_highs[0] < 0.01 and
            recent_lows[-1] > recent_lows[0]):
            patterns.append({
                "pattern": "ascending_triangle",
                "strength": "medium",
            })

    # Descending Triangle: falling flat support, lower highs
    if len(swing_highs) >= 2 and len(swing_lows) >= 2:
        recent_highs = [highs[i] for i in swing_highs[-3:]]
        recent_lows = [lows[i] for i in swing_lows[-3:]]
        if (abs(recent_lows[-1] - recent_lows[0]) / recent_lows[0] < 0.01 and
            recent_highs[-1] < recent_highs[0]):
            patterns.append({
                "pattern": "descending_triangle",
                "strength": "medium",
            })

    # Bull Flag: strong uptrend followed by consolidation
    if n >= 20:
        first_half = closes[:n // 2]
        second_half = closes[n // 2:]
        if (first_half[-1] > first_half[0] * 1.05 and  # Strong uptrend
            np.std(second_half) < np.std(first_half) * 0.5):  # Consolidation
            patterns.append({
                "pattern": "bull_flag",
                "strength": "medium",
            })

    # Bear Flag: strong downtrend followed by consolidation
    if n >= 20:
        first_half = closes[:n // 2]
        second_half = closes[n // 2:]
        if (first_half[-1] < first_half[0] * 0.95 and
            np.std(second_half) < np.std(first_half) * 0.5):
            patterns.append({
                "pattern": "bear_flag",
                "strength": "medium",
            })

    # Rectangle/Consolidation: price in a range
    if n >= 20:
        recent_high = np.max(highs[-20:])
        recent_low = np.min(lows[-20:])
        range_pct = (recent_high - recent_low) / recent_low
        if range_pct < 0.05:  # Less than 5% range
            patterns.append({
                "pattern": "rectangle",
                "strength": "medium",
                "range_high": round(float(recent_high), 2),
                "range_low": round(float(recent_low), 2),
            })

    # Deduplicate
    seen = set()
    unique = []
    for p in patterns:
        if p["pattern"] not in seen:
            seen.add(p["pattern"])
            unique.append(p)

    return unique


# ─── DIVERGENCE DETECTION ───────────────────────────────────────────────────

def detect_divergence(df: pd.DataFrame) -> List[Dict[str, Any]]:
    """
    Detect RSI and MACD divergence from price.
    """
    divergences = []
    n = len(df)
    if n < 30:
        return divergences

    if "rsi_14" not in df.columns:
        return divergences

    closes = df["close"].values
    rsi = df["rsi_14"].values

    # Find local swing points in price and RSI
    swing_highs = find_swing_highs(df["low"], df["high"], 5)
    swing_lows = find_swing_lows(df["low"], df["high"], 5)

    # Bullish RSI divergence: price makes lower low, RSI makes higher low
    if len(swing_lows) >= 2:
        for i in range(len(swing_lows) - 1):
            if (closes[swing_lows[i + 1]] < closes[swing_lows[i]] and
                rsi[swing_lows[i + 1]] > rsi[swing_lows[i]]):
                divergences.append({
                    "type": "bullish_rsi_divergence",
                    "indicator": "rsi",
                    "strength": "medium",
                })
                break

    # Bearish RSI divergence: price makes higher high, RSI makes lower high
    if len(swing_highs) >= 2:
        for i in range(len(swing_highs) - 1):
            if (closes[swing_highs[i + 1]] > closes[swing_highs[i]] and
                rsi[swing_highs[i + 1]] < rsi[swing_highs[i]]):
                divergences.append({
                    "type": "bearish_rsi_divergence",
                    "indicator": "rsi",
                    "strength": "medium",
                })
                break

    # MACD divergence
    if "macd" in df.columns:
        macd = df["macd"].values
        if len(swing_lows) >= 2:
            for i in range(len(swing_lows) - 1):
                if (closes[swing_lows[i + 1]] < closes[swing_lows[i]] and
                    macd[swing_lows[i + 1]] > macd[swing_lows[i]]):
                    divergences.append({
                        "type": "bullish_macd_divergence",
                        "indicator": "macd",
                        "strength": "medium",
                    })
                    break

        if len(swing_highs) >= 2:
            for i in range(len(swing_highs) - 1):
                if (closes[swing_highs[i + 1]] > closes[swing_highs[i]] and
                    macd[swing_highs[i + 1]] < macd[swing_highs[i]]):
                    divergences.append({
                        "type": "bearish_macd_divergence",
                        "indicator": "macd",
                        "strength": "medium",
                    })
                    break

    return divergences


# ─── BREAKOUT DETECTION ─────────────────────────────────────────────────────

def detect_breakout(df: pd.DataFrame) -> Dict[str, Any]:
    """
    Detect breakout/breakdown conditions.
    """
    n = len(df)
    if n < 20:
        return {"status": "no_breakout"}

    closes = df["close"].values
    volumes = df["volume"].values
    highs = df["high"].values
    lows = df["low"].values

    # Recent range
    recent_high = np.max(highs[-20:-1])  # Exclude last candle
    recent_low = np.min(lows[-20:-1])
    current_close = closes[-1]
    avg_vol = np.mean(volumes[-20:-1])
    current_vol = volumes[-1]

    result = {"status": "no_breakout"}

    # Upside breakout
    if current_close > recent_high:
        vol_confirm = current_vol > avg_vol * 1.2
        if vol_confirm:
            result = {
                "status": "confirmed_breakout",
                "direction": "up",
                "level": round(float(recent_high), 2),
                "volume_confirmation": True,
            }
        else:
            result = {
                "status": "potential_breakout",
                "direction": "up",
                "level": round(float(recent_high), 2),
                "volume_confirmation": False,
            }

    # Downside breakdown
    elif current_close < recent_low:
        vol_confirm = current_vol > avg_vol * 1.2
        if vol_confirm:
            result = {
                "status": "confirmed_breakdown",
                "direction": "down",
                "level": round(float(recent_low), 2),
                "volume_confirmation": True,
            }
        else:
            result = {
                "status": "potential_breakdown",
                "direction": "down",
                "level": round(float(recent_low), 2),
                "volume_confirmation": False,
            }

    return result


# ─── MARKET REGIME ──────────────────────────────────────────────────────────

def classify_market_regime(df: pd.DataFrame) -> Dict[str, Any]:
    """
    Classify the primary market regime.
    """
    n = len(df)
    if n < 20:
        return {"primary": "unknown", "secondary": []}

    closes = df["close"].values
    volumes = df["volume"].values

    # Volatility
    returns = np.diff(np.log(closes))
    current_vol = np.std(returns[-20:]) * np.sqrt(252)
    avg_vol = np.std(returns) * np.sqrt(252) if len(returns) > 0 else current_vol

    # Trend
    trend_info = classify_trend(df)
    trend = trend_info["classification"]

    # Range detection
    recent_high = np.max(df["high"].values[-20:])
    recent_low = np.min(df["low"].values[-20:])
    range_pct = (recent_high - recent_low) / recent_low

    # Volume pattern
    vol_mean = np.mean(volumes[-30:]) if n >= 30 else np.mean(volumes)
    vol_trend = np.mean(volumes[-10:]) / vol_mean if vol_mean > 0 else 1.0

    regime = "transition"
    secondary = []

    if current_vol > avg_vol * 1.5:
        secondary.append("high_volatility")
    elif current_vol < avg_vol * 0.5:
        secondary.append("low_volatility")

    if range_pct < 0.03 and "sideways" in trend:
        regime = "sideways_range"
    elif "bullish" in trend and vol_trend > 1.2:
        regime = "bull_trend"
    elif "bearish" in trend and vol_trend > 1.2:
        regime = "bear_trend"
    elif "bullish" in trend:
        regime = "accumulation"
    elif "bearish" in trend:
        regime = "distribution"
    elif current_vol > avg_vol * 1.5:
        regime = "high_volatility"

    # Breakout regime
    breakout = detect_breakout(df)
    if "confirmed_breakout" in breakout.get("status", ""):
        secondary.append("breakout_regime")
    elif "confirmed_breakdown" in breakout.get("status", ""):
        secondary.append("breakdown_regime")

    return {
        "primary": regime,
        "secondary": secondary,
        "volatility_annualized": round(float(current_vol), 4),
    }