Create app.py

app.py

@@ -0,0 +1,67 @@
+# app.py
+# Simple Gradio interface: upload CSV -> run pipeline -> show clusters + sample customers
+
+
+import gradio as gr
+import pandas as pd
+import numpy as np
+import os
+import subprocess
+
+
+from data_prep import load_data, basic_clean, feature_engineer, prepare_features
+from embed import build_text_for_embedding, embed_texts
+from clustering import reduce_and_cluster
+
+
+
+
+def run_pipeline(uploaded_csv, k=6, use_hdbscan=False):
+# save upload
+csv_path = 'data/uploaded.csv'
+os.makedirs('data', exist_ok=True)
+uploaded_csv.save(csv_path)
+
+
+df = load_data(csv_path)
+df = basic_clean(df)
+df = feature_engineer(df)
+features = prepare_features(df)
+
+
+texts = build_text_for_embedding(df)
+embs = embed_texts(texts)
+
+
+labels, arts = reduce_and_cluster(embs, k=int(k), use_hdbscan=use_hdbscan)
+df['cluster'] = labels
+
+
+# return simple summary
+summary = df.groupby('cluster').agg({'customer_id':'count'}).to_dict()
+sample = df.groupby('cluster').head(3).to_dict(orient='records')
+return f"Clusters created: {len(set(labels))}", pd.DataFrame(sample)
+
+
+
+
+def main():
+with gr.Blocks() as demo:
+gr.Markdown('# Customer Segmentation — Hugging Face Space')
+with gr.Row():
+csv_in = gr.File(label='Upload customers CSV')
+k = gr.Slider(minimum=2, maximum=20, step=1, label='K (for KMeans)')
+use_hdbscan = gr.Checkbox(label='Use HDBSCAN (instead of KMeans)')
+out_text = gr.Textbox()
+out_table = gr.Dataframe()
+
+
+run_btn = gr.Button('Run pipeline')
+run_btn.click(fn=run_pipeline, inputs=[csv_in, k, use_hdbscan], outputs=[out_text, out_table])
+
+
+demo.launch()
+
+
+if __name__ == '__main__':
+main()