import gradio as gr
from PIL import Image
import requests
import hopsworks
import joblib
import pandas as pd
import datetime
import os

year = datetime.date.today().year - 1

project = hopsworks.login()
fs = project.get_feature_store()
mr = project.get_model_registry()
model_url = f"valuation_model_v{year-1}_{year}"
model = mr.get_model(model_url)
model_dir = model.download()
model = joblib.load(f"{model_dir}/valuation_model.pkl")
scaler = joblib.load(f"{model_dir}/scaler.pkl")
print("Model downloaded")

def valuation_predictor(goals, assists, y_cards, r_cards, 
                    m_played, height, age, mo_left, 
                    t_made_goals, t_conceded_goals, t_clean_sheets,
                    pos, league):
    
    features_for_prediction = [
        y_cards, r_cards, goals, assists, m_played, height, age, mo_left, t_made_goals, t_conceded_goals, t_clean_sheets, league=="La Liga", league=="League 1", league=="Premier league", league=="Serie A", league=="Bundesliga", pos=="Attack", pos=="Defender", pos=="Goalkeeper", pos=="Midfield"
        ]
    
    column_names = [
        "yellow_cards", "red_cards", "goals", "assists", "minutes_played", "height_in_cm", "age", "months_left", "own_goals", "opponent_goals", "clean_sheets", "league_es1","league_fr1", "league_gb1","league_it1","league_l1", "position_attack","position_defender", "position_goalkeeper","position_midfield"
        ]
    
    
    # Create DataFrame from the input features
    X_predict = pd.DataFrame([features_for_prediction], columns=column_names)
    # Apply the scaler to the relevant columns
    scaled_columns = ['age', 'height_in_cm', 'minutes_played']
    X_predict[scaled_columns] = scaler.transform(X_predict[scaled_columns])
    estimated_valuation = model.predict(X_predict)
    
    # Reformatting to the nearest thousand with commas
    formatted_output_thousand = f"{round(estimated_valuation[0], -3):,} EUR"
    return formatted_output_thousand


with gr.Blocks(theme=gr.themes.Soft()) as demo:
    with gr.Row():
        with gr.Column():
            gr.Label(value="Player specific stats", show_label=False)
            goals = gr.Number(value=7.0, label="Goals")
            assists = gr.Number(value=5, label="Assists")
            y_cards = gr.Number(value=5, label="Yellow cards")
            r_cards = gr.Number(value=0, label="Red cards")
            m_played = gr.Number(value=1500, label="Minutes played")
            height = gr.Number(value=175, label="Height (cm)")
            age = gr.Number(value=25, label="Age")
            mo_left = gr.Number(value=36, label="Months left on contract")
            
        with gr.Column():
            gr.Label(value="Team specific stats", show_label=False)
            t_made_goals = gr.Number(value=60, label="Goals made by player's team during the season")
            t_conceded_goals = gr.Number(value=43, label="Goals conceded by player's team during the season")
            t_clean_sheets = gr.Number(value=15, label="Number of times player's team has let in 0 goals during a game")
        
            gr.Label(value="League and position data", show_label=False)
            pos = gr.Dropdown(choices=["Attack", "Defender", "Goalkeeper", "Midfield"])
            league = gr.Dropdown(choices=["La Liga","League 1", "Premier league", "Serie A", "Bundesliga"])
            
    with gr.Row():
        btn = gr.Button("Predict player value")
        estimated_player_value = gr.Textbox(label="Estimated player value...")
        btn.click(fn=valuation_predictor, 
                  inputs=[
                    goals, assists, y_cards, r_cards, 
                    m_played, height, age, mo_left, 
                    t_made_goals, t_conceded_goals, t_clean_sheets,
                    pos, league], 
                  outputs=estimated_player_value
                  )
        
            
demo.launch()