from fastapi import FastAPI
from pydantic import BaseModel
import numpy as np
import pandas as pd
import tensorflow as tf
from tensorflow.keras.models import load_model
import json
import os

app = FastAPI(
    title="Forex LSTM Prediction API",
    description="Prediksi harga EUR/USD H+1 dengan LSTM menggunakan scaler harian",
    version="2.0"
)

# ==========================================================
# LOAD MODEL DAN KONFIGURASI
# ==========================================================
MODEL_PATH = "lstm_model.h5"
PARAMS_PATH = "best_params.json"
SCALER_FILE = "scaler_config.json"

print("📥 Loading LSTM model...")
model = load_model(MODEL_PATH, compile=False)

print("📥 Loading best parameters...")
with open(PARAMS_PATH, "r") as f:
    best_params = json.load(f)

LOOKBACK = best_params.get("lookback", 7)
FEATURE_ORDER = best_params.get("features", [
    "mood_score", "t_pos", "t_neg", "c_pos", "c_neg",
    "norm_ema20", "norm_ema50", "norm_close"
])

# ==========================================================
# LOAD SCALER CONFIG
# ==========================================================
def load_scaler_config():
    if not os.path.exists(SCALER_FILE):
        print("⚠️ Scaler config not found, using default values.")
        return {"CLOSE_MIN": 1.05, "CLOSE_MAX": 1.15}
    with open(SCALER_FILE, "r") as f:
        return json.load(f)

scaler_cfg = load_scaler_config()
CLOSE_MIN = scaler_cfg["CLOSE_MIN"]
CLOSE_MAX = scaler_cfg["CLOSE_MAX"]

print(f"✅ Scaler range loaded: {CLOSE_MIN:.5f} - {CLOSE_MAX:.5f}")

# ==========================================================
# INPUT SCHEMA
# ==========================================================
class LSTMInput(BaseModel):
    data: list

# ==========================================================
# HELPER FUNCTIONS
# ==========================================================
def prepare_input(data):
    df = pd.DataFrame(data)
    missing_cols = [f for f in FEATURE_ORDER if f not in df.columns]
    if missing_cols:
        raise ValueError(f"Missing columns: {missing_cols}")

    X = df[FEATURE_ORDER].values[-LOOKBACK:]
    if X.shape[0] < LOOKBACK:
        raise ValueError(f"Need at least {LOOKBACK} timesteps, got {X.shape[0]}")
    X = np.expand_dims(X, axis=0)
    return X, df


def inverse_scale(norm_value):
    """Denormalisasi nilai close dari [0,1] ke skala asli"""
    return (norm_value * (CLOSE_MAX - CLOSE_MIN)) + CLOSE_MIN

# ==========================================================
# ENDPOINT
# ==========================================================
@app.post("/predict")
def predict_price(input_data: LSTMInput):
    try:
        X, df = prepare_input(input_data.data)
        pred_norm = model.predict(X)[0][0]
        pred_close = inverse_scale(pred_norm)

        last_date = pd.to_datetime(df["date"].iloc[-1])
        next_date = (last_date + pd.Timedelta(days=1)).strftime("%Y-%m-%d")

        result = {
            "next_date": next_date,
            "predicted_norm_close": float(pred_norm),
            "predicted_close": float(pred_close),
            "scaler_used": {"min": CLOSE_MIN, "max": CLOSE_MAX},
            "features_used": FEATURE_ORDER
        }
        return {"status": "ok", "result": result}

    except Exception as e:
        return {"status": "error", "message": str(e)}

@app.get("/")
def root():
    return {"message": "Forex LSTM Prediction API is active!"}