from fastapi import FastAPI, UploadFile, File, HTTPException
from tensorflow.keras.models import load_model, Sequential
from tensorflow.keras.layers import Dense, LSTM
from tensorflow.keras.optimizers import Adam
from sklearn.preprocessing import MinMaxScaler
import numpy as np
import tempfile
import os

app = FastAPI()

@app.post("/predict")
async def predict(model: UploadFile = File(...), data: str = None):
    try:
        # Save the uploaded model to a temporary file
        with tempfile.NamedTemporaryFile(delete=False, suffix=".h5") as temp_model_file:
            temp_model_file.write(await model.read())
            temp_model_path = temp_model_file.name

        ds = eval(data)
        ds = np.array(ds).reshape(-1, 1)
        # Normalize the data
        scaler = MinMaxScaler()
        ds_normalized = scaler.fit_transform(ds)
        
        # Load the model
        model = load_model(temp_model_path, compile=False)
        model.compile(optimizer=Adam(learning_rate=0.001), loss='mse', run_eagerly=True)
        print(data)
        # Process the data
        predictions = model.predict(ds_normalized.reshape(1, 12, 1)).tolist()
        predictions_rescaled = scaler.inverse_transform(np.array(predictions).reshape(-1, 1)).flatten()  

        return {"predictions": predictions_rescaled}

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

@app.post("/retrain")
async def retrain(model: UploadFile = File(...), data: str = None):
    try:
        # Save the uploaded model and data to temporary files
        with tempfile.NamedTemporaryFile(delete=False, suffix=".h5") as temp_model_file:
            temp_model_file.write(await model.read())
            temp_model_path = temp_model_file.name


        # Load the model and data
        model = load_model(temp_model_path, compile=False)
        model.compile(optimizer=Adam(learning_rate=0.001), loss='mse', run_eagerly=True)

        ds = eval(data)
        ds = np.array(ds).reshape(-1, 1)
        # Normalize the data
        scaler = MinMaxScaler()
        ds_normalized = scaler.fit_transform(ds)

        x_train = np.array([ds_normalized[i - 12:i] for i in range(12, len(ds_normalized))])
        y_train = ds_normalized[12:]
        
        model.compile(optimizer=Adam(learning_rate=0.001), loss="mse", run_eagerly=True)
        model.fit(x_train, y_train, epochs=1, batch_size=32)

        # Save the updated model to a temporary file
        updated_model_path = temp_model_path.replace(".h5", "_updated.h5")
        model.save(updated_model_path)

        return {"message": "Model retrained successfully.", "updated_model_path": updated_model_path}

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))
    finally:
        # Clean up temporary files
        if os.path.exists(temp_model_path):
            os.remove(temp_model_path)