app.py

import gradio as gr
import pandas as pd
from surprise import SVD, Dataset, Reader
from src.data_preprocessing import load_parquet_data
from src.collaborative_filtering import train_user_based_cf, get_top_n_recommendations

# Load and preprocess sample data
df_raw = pd.read_csv("train_sample.csv")
df_raw['user_id'] = df_raw['user_id'].astype(int)
df_raw['product_id'] = df_raw['product_id'].astype(str)

# Train the collaborative filtering model using SVD
algo, df_clean = train_user_based_cf(df_raw, algorithm=SVD(), show_rmse=True)

# Generate top-N recommendations for all users
top_n = get_top_n_recommendations(algo, df_clean, n=5)

def recommend(user_id_input):
    try:
        user_id = int(float(user_id_input))  # Converts strings or floats safely
    except ValueError:
        return "❌ Invalid user ID format."

    if user_id not in top_n:
        return "🚫 User not found or not enough data."

    recs = top_n[user_id]
    return "\n".join([f"👉 Product {pred.iid} (predicted rating: {pred.est:.3f})" for pred in recs])

# Build Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("🎁 **E-commerce Recommender (SVD)**")
    gr.Markdown("Enter a user ID to get top-5 product recommendations using collaborative filtering (SVD).")

    with gr.Row():
        user_id_input = gr.Textbox(label="Enter User ID", placeholder="e.g., 553033744")
        output = gr.Textbox(label="Top 5 Recommendations")

    with gr.Row():
        clear_btn = gr.Button("Clear")
        submit_btn = gr.Button("Submit")

    submit_btn.click(fn=recommend, inputs=user_id_input, outputs=output)
    clear_btn.click(fn=lambda: ("", ""), inputs=[], outputs=[user_id_input, output])

if __name__ == "__main__":
    demo.launch()