from flask import Flask, render_template, request, jsonify
import numpy as np

app = Flask(__name__)

# -------------------------------
# Load Trained Model Weights
# -------------------------------
best = np.load("best_model.npy")   # make sure this file is present


# -------------------------------
# HARDCODE YOUR MSE VALUES HERE
# -------------------------------
train_mse = 0.005707
test_mse = 0.008744


# -------------------------------
# FCNN Model Functions
# -------------------------------
input_dim = 1
h1 = 32
h2 = 16
output_dim = 1

D = input_dim*h1 + h1 + h1*h2 + h2 + h2*output_dim + output_dim


def decode_theta(theta):
    idx = 0
    W1 = theta[idx:idx + input_dim*h1].reshape(input_dim, h1)
    idx += input_dim*h1
    b1 = theta[idx:idx + h1].reshape(1, h1)
    idx += h1

    W2 = theta[idx:idx + h1*h2].reshape(h1, h2)
    idx += h1*h2
    b2 = theta[idx:idx + h2].reshape(1, h2)
    idx += h2

    W3 = theta[idx:idx + h2*output_dim].reshape(h2, output_dim)
    idx += h2*output_dim
    b3 = theta[idx:idx + output_dim].reshape(1, output_dim)

    return W1, b1, W2, b2, W3, b3


def fcnn_forward(X_batch, theta):
    W1, b1, W2, b2, W3, b3 = decode_theta(theta)

    z1 = X_batch @ W1 + b1
    a1 = np.maximum(z1, 0)

    z2 = a1 @ W2 + b2
    a2 = np.maximum(z2, 0)

    out = a2 @ W3 + b3
    return out


def predict_next(theta, x):
    y = fcnn_forward(np.array([[x]]), theta)
    return float(y[0][0])


def forecast_future(theta, x_start, steps=20):
    preds = []
    x = x_start
    for _ in range(steps):
        y = float(fcnn_forward(np.array([[x]]), theta)[0][0])
        preds.append(y)
        x = y
    return preds


# -------------------------------
# Routes
# -------------------------------

@app.route("/")
def index():
    return render_template("index.html")


@app.route("/predict", methods=["POST"])
def predict():
    data = request.json.get("data", None)
    if data is None or len(data) == 0:
        return jsonify({"error": "No data provided"}), 400

    try:
        series = np.array(data, dtype=float)
    except:
        return jsonify({"error": "Invalid numeric data"}), 400

    last_val = float(series[-1])
    next_val = predict_next(best, last_val)
    future_vals = forecast_future(best, last_val, steps=20)

    return jsonify({
        "last": last_val,
        "next": next_val,
        "future": future_vals,
        "train_mse": train_mse,
        "test_mse": test_mse
    })


if __name__ == "__main__":
    app.run(debug=True)