app.py

import gradio as gr
import pandas as pd
from catboost import CatBoostRegressor
import os

# 1. Load the model
# Make sure 'catboost_skincare_model.cbm' is uploaded to the same Space
model = CatBoostRegressor()
if os.path.exists("catboost_skincare_model.cbm"):
    model.load_model("catboost_skincare_model.cbm")
else:
    print("Model file not found!")

def predict_rating(brand, product, cat1, cat2, origin, age_range, skin_type, concern, ing1, ing2, ing3):
    # 2. Feature Engineering: Age Mapping
    age_map = {
        '18 and Under': 'Teens',
        '19 - 24': 'Young Adult',
        '25 - 29': 'Young Adult',
        '30 - 34': 'Adult',
        '35 - 39': 'Adult',
        '40 - 44': 'Mature',
        '45 and Above': 'Mature'
    }
    age_segment = age_map.get(age_range, 'Young Adult')

    # 3. Feature Engineering: Ingredient Flags
    all_ings = f"{ing1} {ing2} {ing3}".lower()
    
    def check_ing(term):
        return "1" if term in all_ings else "0"

    # 4. Create DataFrame for prediction (Must match training column order)
    input_df = pd.DataFrame([{
        'brand_name': brand,
        'product_name': product,
        '1st_category': cat1,
        '2nd_category': cat2,
        'brand_country_origin': origin,
        'age': age_range,
        'skin_type': skin_type,
        'skin_concern': concern,
        'ingredient_1': ing1,
        'ingredient_2': ing2,
        'ingredient_3': ing3,
        # Engineered features the model expects
        'age_segment': age_segment,
        'has_niacinamide': check_ing('niacinamide'),
        'has_retinol': check_ing('retinol'),
        'has_salicylic_acid': check_ing('salicylic'),
        'has_hyaluronic_acid': check_ing('hyaluronic'),
        'has_vitamin_c': check_ing('vitamin c'),
        'skin_profile': f"{skin_type}_{concern}",
        'product_skin_fit': f"{cat2}_{skin_type}"
    }])

    # 5. Predict
    prediction = model.predict(input_df)
    
    # Return rounded rating
    return round(float(prediction[0]), 2)

# 6. Define the Gradio Interface
inputs = [
    gr.Textbox(label="Brand Name"),
    gr.Textbox(label="Product Name"),
    gr.Textbox(label="1st Category"),
    gr.Textbox(label="2nd Category"),
    gr.Textbox(label="Country of Origin"),
    gr.Dropdown(choices=['18 and Under', '19 - 24', '25 - 29', '30 - 34', '35 - 39', '40 - 44', '45 and Above'], label="Age Range"),
    gr.Textbox(label="Skin Type"),
    gr.Textbox(label="Skin Concern"),
    gr.Textbox(label="Ingredient 1"),
    gr.Textbox(label="Ingredient 2"),
    gr.Textbox(label="Ingredient 3")
]

demo = gr.Interface(
    fn=predict_rating,
    inputs=inputs,
    outputs=gr.Number(label="Predicted Rating"),
    title="Skincare Rating Predictor API",
    api_name="predict_rating"
)

if __name__ == "__main__":
    demo.launch(show_error=True, share=False)