import streamlit as st
import pandas as pd
import pickle
import requests
import os

# Function to download files from Hugging Face
def download_file(url, output_path):
    if not os.path.exists(output_path):
        with open(output_path, "wb") as f:
            response = requests.get(url)
            if response.status_code == 200:
                f.write(response.content)
            else:
                st.error(f"Failed to download {url}. HTTP status code: {response.status_code}")
                st.stop()

# Updated URLs to the model and label encoder files
model_url = "https://huggingface.co/spaces/amornpan/weather-prediction-app/raw/main/random_forest_weather_model.pkl"
encoder_url = "https://huggingface.co/spaces/amornpan/weather-prediction-app/raw/main/label_encoders.pkl"

# Local paths for the downloaded files
model_path = "random_forest_weather_model.pkl"
encoder_path = "label_encoders.pkl"

# Download files if not already present
download_file(model_url, model_path)
download_file(encoder_url, encoder_path)

# Load the model and label encoders
try:
    with open(model_path, "rb") as f:
        model = pickle.load(f)
except Exception as e:
    st.error(f"Error loading model: {e}")
    st.stop()

try:
    with open(encoder_path, "rb") as f:
        label_encoders = pickle.load(f)
except Exception as e:
    st.error(f"Error loading label encoders: {e}")
    st.stop()

# Debug: Check the structure of label_encoders
if not isinstance(label_encoders, dict):
    st.error("Label encoders file is not in the expected format.")
    st.stop()

st.write("Loaded Label Encoders:", label_encoders.keys())

# Define UI
st.title("Weather Prediction App")
st.write("กรอกข้อมูลเพื่อพยากรณ์ประเภทของสภาพอากาศ")

# User inputs
temperature = st.number_input("Temperature (°C)", min_value=-50, max_value=50, value=25)
humidity = st.number_input("Humidity (%)", min_value=0, max_value=100, value=50)
wind_speed = st.number_input("Wind Speed (km/h)", min_value=0.0, max_value=200.0, value=10.0)
precipitation = st.number_input("Precipitation (%)", min_value=0.0, max_value=100.0, value=10.0)

# Handle 'Cloud Cover' selection
cloud_cover_options = label_encoders['Cloud Cover'].classes_ if 'Cloud Cover' in label_encoders else ["Clear", "Partly Cloudy", "Overcast"]
cloud_cover = st.selectbox("Cloud Cover", cloud_cover_options)

# Other inputs
atmospheric_pressure = st.number_input("Atmospheric Pressure (hPa)", min_value=900.0, max_value=1100.0, value=1010.0)
uv_index = st.number_input("UV Index", min_value=0, max_value=10, value=5)
season_options = label_encoders['Season'].classes_ if 'Season' in label_encoders else ["Winter", "Spring", "Summer", "Autumn"]
season = st.selectbox("Season", season_options)
visibility = st.number_input("Visibility (km)", min_value=0.0, max_value=50.0, value=10.0)
location_options = label_encoders['Location'].classes_ if 'Location' in label_encoders else ["Inland", "Coastal", "Mountain"]
location = st.selectbox("Location", location_options)

# Encode categorical inputs
try:
    cloud_cover_encoded = label_encoders['Cloud Cover'].transform([cloud_cover])[0]
    season_encoded = label_encoders['Season'].transform([season])[0]
    location_encoded = label_encoders['Location'].transform([location])[0]
except Exception as e:
    st.error(f"Error encoding categorical inputs: {e}")
    st.stop()

# Prepare input data
input_data = pd.DataFrame({
    "Temperature": [temperature],
    "Humidity": [humidity],
    "Wind Speed": [wind_speed],
    "Precipitation (%)": [precipitation],
    "Cloud Cover": [cloud_cover_encoded],
    "Atmospheric Pressure": [atmospheric_pressure],
    "UV Index": [uv_index],
    "Season": [season_encoded],
    "Visibility (km)": [visibility],
    "Location": [location_encoded]
})

# Predict weather type
if st.button("Predict Weather Type"):
    try:
        prediction = model.predict(input_data)
        weather_type = label_encoders['Weather Type'].inverse_transform(prediction)[0] if 'Weather Type' in label_encoders else "Unknown"
        st.success(f"Predicted Weather Type: {weather_type}")
    except Exception as e:
        st.error(f"Error in prediction: {e}")