import gradio as gr
import joblib
import pandas as pd
import numpy as np
from huggingface_hub import hf_hub_download

# Download the model from Hugging Face hub
model_filename = hf_hub_download(repo_id="poudel/Job_Predictor", filename="random_forest_pipeline.pkl")

# Load the model directly
loaded_model = joblib.load(model_filename)

# Download the dataset (CSV) from Hugging Face hub
data_filename = hf_hub_download(repo_id="poudel/Job_Predictor", filename="cleaned_erecruit_open_data.csv")

# Load the CSV dataset
data = pd.read_csv(data_filename)

# Get unique values for dropdowns
position_titles = data['PositionTitle'].unique().tolist()
designations = data['Designation'].unique().tolist()
agencies = data['Agency'].unique().tolist()
vacancy_types = data['VacancyType'].unique().tolist()
employment_categories = data['EmploymentCategory'].unique().tolist()
locations = data['Locations'].unique().tolist()
vacancy_6_months_or_less = data['Vacancy6MonthsOrLess'].unique().tolist()

# Define a function to make predictions based on user input
def predict_applicants(position_title, designation, agency, vacancy_type, employment_category, location, vacancy_6_months_or_less, number_of_vacancies, number_of_successful_applicants):
    # Create a DataFrame from the inputs
    input_data = pd.DataFrame({
        'PositionTitle': [position_title],
        'Designation': [designation],
        'Agency': [agency],
        'VacancyType': [vacancy_type],
        'EmploymentCategory': [employment_category],
        'Locations': [location],
        'Vacancy6MonthsOrLess': [vacancy_6_months_or_less],
        'NumberOfSuccessfulApplicants': [number_of_successful_applicants],
        'NumberOfVacancies': [number_of_vacancies]
    })

    # Calculate additional features
    input_data['Success_Ratio'] = input_data['NumberOfSuccessfulApplicants'] / input_data['NumberOfVacancies'].replace(0, np.nan)
    input_data['Applicants_per_Vacancy'] = input_data['NumberOfVacancies'] / np.where(input_data['NumberOfSuccessfulApplicants'] == 0, np.nan, input_data['NumberOfSuccessfulApplicants'])
    
    # Avoid inplace modification, return to the column
    input_data['Success_Ratio'] = input_data['Success_Ratio'].fillna(0)
    input_data['Applicants_per_Vacancy'] = input_data['Applicants_per_Vacancy'].fillna(0)

    # Make predictions using the loaded model pipeline
    try:
        prediction = loaded_model.predict(input_data)
        return f"Predicted Number of Applicants: {int(prediction[0])}"
    except Exception as e:
        return f"Error during prediction: {str(e)}"

# Create the Gradio Blocks Interface
with gr.Blocks() as interface:
    # Add a title and description
    gr.Markdown("# NT's Job Predictor")
    gr.Markdown("Select the job details below to predict the number of applicants for a given position.")

    with gr.Row():
        position_title_input = gr.Dropdown(choices=position_titles, label="Position Title", value=None)
        designation_input = gr.Dropdown(choices=designations, label="Designation", value=None)
        agency_input = gr.Dropdown(choices=agencies, label="Agency", value=None)

    with gr.Row():
        vacancy_type_input = gr.Dropdown(choices=vacancy_types, label="Vacancy Type", value=None)
        employment_category_input = gr.Dropdown(choices=employment_categories, label="Employment Category", value=None)
        location_input = gr.Dropdown(choices=locations, label="Locations", value=None)
        vacancy_6_months_or_less_input = gr.Dropdown(choices=vacancy_6_months_or_less, label="Vacancy 6 Months or Less", value=None)

    with gr.Row():
        number_of_vacancies_input = gr.Number(label="Past Number of Vacancies", value=None)
        number_of_successful_applicants_input = gr.Number(label="Past Number of Successful Applicants", value=None)

    predict_button = gr.Button("Predict")

    predicted_applicants_output = gr.Textbox(label="Predicted Number of Applicants")

    predict_button.click(
        fn=predict_applicants,
        inputs=[
            position_title_input,
            designation_input,
            agency_input,
            vacancy_type_input,
            employment_category_input,
            location_input,
            vacancy_6_months_or_less_input,
            number_of_vacancies_input,
            number_of_successful_applicants_input
        ],
        outputs=predicted_applicants_output
    )

interface.launch(share=True)