import os
os.environ['CUDA_VISIBLE_DEVICES'] = '-1'  # Disable all GPUs
os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'  # Optional: disable oneDNN
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"

import tensorflow as tf
tf.config.set_visible_devices([], 'GPU')

import gradio as gr
import numpy as np
from tensorflow.keras.preprocessing import image
from fastapi import FastAPI, File, UploadFile
from fastapi.responses import JSONResponse
from fastapi.middleware.cors import CORSMiddleware
from io import BytesIO
from PIL import Image
import logging
from tensorflow.keras.models import load_model, model_from_json
from tensorflow.keras import mixed_precision
from tensorflow.keras.saving import get_custom_objects, register_keras_serializable
from tensorflow.keras.mixed_precision import Policy

# Import deskripsi dan lokasi
from description import description
from location import location

# @register_keras_serializable(package="keras")
# class DTypePolicy(Policy):
#     pass

# from tensorflow.keras.saving import get_custom_objects
# get_custom_objects()["DTypePolicy"] = DTypePolicy

# Nonaktifkan GPU (jika tidak digunakan)
tf.config.set_visible_devices([], 'GPU')

# Inisialisasi logger
# logging.basicConfig(level=logging.INFO)
# logger = logging.getLogger(__name__)

# ========== Fungsi Load Model dari File JSON + H5 ==========
def load_model_from_file(json_path, h5_path):
    with open(json_path, "r") as f:
        json_config = f.read()
        model = model_from_json(json_config)
    model.load_weights(h5_path)
    return model

# ========== Load Model ==========
model = load_model_from_file("model.json", "my_model.h5")


# Daftar label
labels = [
    "Benteng Vredeburg", "Candi Borobudur", "Candi Prambanan", "Gedung Agung Istana Kepresidenan",
    "Masjid Gedhe Kauman", "Monumen Serangan 1 Maret", "Museum Gunungapi Merapi",
    "Situs Ratu Boko", "Taman Sari", "Tugu Yogyakarta"
]

# Fungsi klasifikasi
def classify_image(img):
    try:
        img = img.resize((224, 224))
        img_array = image.img_to_array(img)
        img_array = np.expand_dims(img_array, axis=0)
        img_array = img_array / 255.0

        pred = model.predict(img_array)[0]
        confidence = np.max(pred)
        predicted_label = labels[np.argmax(pred)]

        akurasi = float(confidence)
        if confidence < 0.8:
            label_output = "Tidak dapat dikenali (Confidence: {:.2f}%)".format(confidence * 100)
            deskripsi = (
                "Tolong arahkan ke objek yang jelas agar bisa diidentifikasikan. "
                "Pastikan anda berada di salah satu tempat seperti:\n"
                "- Benteng Vredeburg\n- Candi Borobudur\n- Candi Prambanan\n"
                "- Gedung Agung Istana Kepresidenan Yogyakarta\n- Masjid Gedhe Kauman\n"
                "- Monumen Serangan 1 Maret\n- Museum Gunungapi Merapi\n- Situs Ratu Boko\n"
                "- Taman Sari\n- Tugu Yogyakarta"
            )
            lokasi = "-"
        else:
            label_output = f"{predicted_label} (Confidence: {confidence * 100:.2f}%)"
            deskripsi = description.get(predicted_label, "Deskripsi belum tersedia.")
            lokasi = location.get(predicted_label, None)
            if lokasi:
                lokasi = f'<a href="{lokasi}" target="_blank">Lihat Lokasi di Google Maps</a>'
            else:
                lokasi = "Lokasi tidak ditemukan"

        return label_output, deskripsi, lokasi, akurasi

    except Exception as e:
        return "Error", str(e), "-"

# Fungsi untuk membuat FastAPI app
def create_app():
    app = FastAPI()
    app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # atau daftar domain yang sah
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)



    # @app.post("/api/predict")
    # async def predict(file: UploadFile = File(...)):
    #     contents = await file.read()
    #     img = Image.open(BytesIO(contents)).convert("RGB")
    #     label_output, deskripsi, lokasi, akurasi = classify_image(img)
    #     return JSONResponse(content={
    #         "label_output": label_output,
    #         "deskripsi": deskripsi,
    #         "lokasi": lokasi,
    #         "confidence": akurasi
    #     })

    gradio_app = gr.Interface(
        fn=classify_image,
        inputs=gr.Image(type="pil", label="Upload Gambar"),
        # outputs=["text", "text", "html"],
        outputs=[
            gr.Textbox(label="Output Klasifikasi"),
            gr.Textbox(label="Deskripsi Lengkap", lines=20, max_lines=50),
            gr.HTML(label="Link Lokasi"),
        ],
        # flagging_mode="never",
        title="Klasifikasi Gambar",
        description="Upload gambar, sistem akan mengklasifikasikan dan memberikan deskripsi mengenai gambar tersebut."
    )

    app = gr.mount_gradio_app(app, gradio_app, path="/gradio")
    return app

app = create_app()

# Run server jika dijalankan langsung
# if __name__ == "__main__":
#     import uvicorn
#     # app = create_app()
#     uvicorn.run(app, host="127.0.0.1", port=8000)