Create app.py

app.py

@@ -0,0 +1,61 @@
+import re
+import gradio as gr
+import pandas as pd
+import torch
+import torch.nn.functional as F
+from sentence_transformers import SentenceTransformer
+
+CSV_PATH = "Shopeasy_product_dataset.csv"
+EMB_PATH = "embeddings.pt"
+MODEL_NAME = "all-mpnet-base-v2"
+
+# Load catalog
+data = pd.read_csv(CSV_PATH, low_memory=True)
+
+# Load embeddings (force CPU so it works on Spaces without GPU)
+sentence_embeddings = torch.load(EMB_PATH, map_location="cpu").float()
+
+# Normalize embeddings once (so dot product == cosine similarity)
+sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)
+
+# Load model (CPU by default; will still work fine)
+mpnet_base = SentenceTransformer(MODEL_NAME, device="cpu")
+
+def clean_text(x: str) -> str:
+    x = str(x).lower()
+    x = re.sub(r"[^A-Za-z0-9]+", " ", x).strip()
+    return x
+
+def get_recommendations(query, top_k=10):
+    query = clean_text(query)
+    if not query:
+        return pd.DataFrame(columns=["Rank", "Product Name"])
+
+    query_embedding = mpnet_base.encode(query, convert_to_tensor=True)
+    query_embedding = F.normalize(query_embedding, p=2, dim=0)
+
+    # cosine similarity (since both normalized)
+    similarity_scores = sentence_embeddings @ query_embedding
+
+    top_k = int(top_k)
+    top_scores, top_indices = torch.topk(similarity_scores, k=min(top_k, similarity_scores.shape[0]))
+    top_indices = top_indices.cpu().numpy()
+
+    recs = data.iloc[top_indices]["product_name"].reset_index(drop=True)
+    out = recs.to_frame(name="Product Name")
+    out.insert(0, "Rank", range(1, len(out) + 1))
+    return out
+
+with gr.Blocks() as demo:
+    gr.Markdown("## 🤖 AI-Powered Product Recommendation")
+    gr.Markdown("Enter a query to see the top recommended products (SentenceTransformer embeddings).")
+
+    query = gr.Textbox(label="Query", placeholder="e.g., wireless headphones noise cancelling")
+    top_k = gr.Slider(1, 25, value=10, step=1, label="Top K")
+
+    btn = gr.Button("Recommend")
+    table = gr.Dataframe(label="Recommendations", interactive=False)
+
+    btn.click(get_recommendations, inputs=[query, top_k], outputs=table)
+
+demo.launch()