import pandas as pd
import numpy as np
import pickle
from typing import Dict, Any
from routes.predict import build_long_format_input, modelPro
from routes.predict_bag_of_heroes import build_bag_of_heroes_features, modelBagOfHeroes
from routes.predict_with_players import build_player_features, modelWithPlayers

# Загрузка мета-модели (Logistic Regression)
with open("artifacts/model_stacking.pkl", 'rb') as f:
    modelStacking = pickle.load(f)

def predict_stacking(payload: Dict[str, Any]) -> Dict[str, float]:
    """
    Делает предсказание с использованием стекинг-модели.

    Сначала получает предсказания от всех базовых моделей,
    затем использует их как признаки для мета-модели.

    Возвращает:
    {
        "radiant_win": вероятность победы Radiant (0-100),
        "dire_win": вероятность победы Dire (0-100)
    }
    """
    # === Предсказание модели 1: Heroes + Positions ===
    X_with_pos = build_long_format_input(payload)
    pred1 = float(modelPro.predict_proba(X_with_pos)[0, 1])

    # === Предсказание модели 2: Bag of Heroes ===
    X_bag = build_bag_of_heroes_features(payload)
    pred2 = float(modelBagOfHeroes.predict_proba(X_bag)[0, 1])

    # === Предсказание модели 3: With Players ===
    X_players = build_player_features(payload)
    pred3 = float(modelWithPlayers.predict_proba(X_players)[0, 1])

    # === Мета-модель ===
    X_meta = pd.DataFrame([{
        "pred_with_positions": pred1,
        "pred_bag_of_heroes": pred2,
        "pred_with_players": pred3
    }])

    proba = modelStacking.predict_proba(X_meta)[0, 1]

    radiant_win = round(float(np.clip(proba * 100.0, 0.0, 100.0)))
    dire_win = 100.0 - radiant_win

    return {
        "radiant_win": radiant_win,
        "dire_win": dire_win
    }