"""
Risk Agent — Trade gatekeeper.
100% rule-based (zero LLM tokens).
Filters out bad trades based on liquidity, volatility, concentration, and drawdown.
"""
import logging

from backend.data.market_data import is_indian_ticker
from config import RISK_PARAMS as RP

logger = logging.getLogger(__name__)


def evaluate_risk(signal: dict, latest_features: dict, existing_positions: list[dict] | None = None) -> dict:
    """
    Evaluate risk for a trade signal. Returns approval/rejection decision.

    Args:
        signal: The trade signal dict
        latest_features: Latest feature row for the ticker
        existing_positions: Currently open positions (for concentration check)

    Returns:
        {
            "decision": "approve"|"reject"|"reduce",
            "original_signal": signal,
            "risk_checks": {...},
            "reason": str,
        }
    """
    checks = {}
    reasons = []
    decision = "approve"
    ticker = signal.get("ticker", "")

    # ── 1. Liquidity Check ──
    avg_volume = float(latest_features.get("Volume") or 0)
    close = float(latest_features.get("Close") or 0)
    if avg_volume and close:
        daily_value = avg_volume * close
        threshold = RP["min_liquidity_inr"] if is_indian_ticker(ticker) else RP["min_liquidity_usd"]
        checks["liquidity"] = {
            "daily_value": round(daily_value),
            "threshold": threshold,
            "pass": daily_value >= threshold,
        }
        if not checks["liquidity"]["pass"]:
            decision = "reject"
            reasons.append(f"Illiquid: daily volume value {daily_value:,.0f} < {threshold:,.0f}")

    # ── 2. Volatility Check ──
    atr_pct = latest_features.get("atr_pct")
    if atr_pct:
        checks["volatility"] = {
            "atr_pct": round(atr_pct, 2),
            "threshold": RP["max_atr_pct"],
            "pass": atr_pct <= RP["max_atr_pct"],
        }
        if not checks["volatility"]["pass"]:
            if decision != "reject":
                decision = "reduce"
            reasons.append(f"High volatility: ATR%={atr_pct:.1f}% > {RP['max_atr_pct']}%")

    # ── 3. Risk:Reward Check ──
    rr = signal.get("risk_reward", 0)
    checks["risk_reward"] = {
        "value": rr,
        "threshold": 1.5,
        "pass": rr >= 1.5,
    }
    if not checks["risk_reward"]["pass"]:
        if decision == "approve":
            decision = "reduce"
        reasons.append(f"Poor R:R = {rr:.1f} (min 1.5)")

    # ── 4. Sector Concentration ──
    if existing_positions:
        sector = latest_features.get("sector", "Unknown")
        sector_count = sum(1 for p in existing_positions if p.get("sector") == sector)
        sector_pct = (sector_count + 1) / max(len(existing_positions), 1) * 100
        checks["sector_concentration"] = {
            "sector": sector,
            "current_pct": round(sector_pct, 1),
            "threshold": RP["max_sector_pct"],
            "pass": sector_pct <= RP["max_sector_pct"],
        }
        if not checks["sector_concentration"]["pass"]:
            if decision == "approve":
                decision = "reduce"
            reasons.append(f"Sector concentration: {sector} at {sector_pct:.0f}% > {RP['max_sector_pct']}%")

    # ── 5. Stop Loss Distance ──
    entry = signal.get("entry_price", 0)
    sl = signal.get("stop_loss", 0)
    if entry and sl:
        sl_pct = abs(entry - sl) / entry * 100
        checks["stop_loss_distance"] = {
            "pct": round(sl_pct, 2),
            "max_acceptable": 8.0,
            "pass": sl_pct <= 8.0,
        }
        if not checks["stop_loss_distance"]["pass"]:
            if decision == "approve":
                decision = "reduce"
            reasons.append(f"Wide stop loss: {sl_pct:.1f}% from entry")

    reason = "; ".join(reasons) if reasons else "All risk checks passed"

    result = {
        "decision": decision,
        "ticker": ticker,
        "signal_type": signal.get("signal_type"),
        "risk_checks": checks,
        "reason": reason,
        "agent_name": "risk_agent",
    }

    emoji = {"approve": "✅", "reject": "❌", "reduce": "⚠️"}
    logger.info(f"{emoji.get(decision, '?')} Risk [{ticker}]: {decision} — {reason}")
    return result


def filter_signals_by_risk(signals: list[dict], feature_results: list[dict],
                           existing_positions: list[dict] | None = None) -> list[dict]:
    """
    Filter a list of signals through the risk agent.
    Returns only approved/reduced signals (rejects are dropped).
    """
    # Build a lookup of features by ticker
    features_map = {}
    for result in feature_results:
        features_map[result["ticker"]] = result.get("latest", {})

    approved = []
    for signal in signals:
        ticker = signal.get("ticker", "")
        latest = features_map.get(ticker, {})
        risk = evaluate_risk(signal, latest, existing_positions)

        if risk["decision"] == "approve":
            signal["risk_decision"] = "approved"
            approved.append(signal)
        elif risk["decision"] == "reduce":
            signal["risk_decision"] = "reduce_size"
            signal["risk_reason"] = risk["reason"]
            approved.append(signal)
        # Rejected signals are dropped

    logger.info(f"Risk filter: {len(approved)}/{len(signals)} signals passed")
    return approved