Update app.py

app.py

@@ -1,67 +1,47 @@
 import gradio as gr
 import pandas as pd
 
-# Load your multimodal dataset
+# Load the dataset
 df = pd.read_csv("multimodal_df.csv")
 
-# Optional: fill NAs for consistency
-df.fillna('', inplace=True)
-
-# Basic semantic scoring (tunable)
-def semantic_search(query, top_k=5):
+def semantic_search(query):
     query = query.lower()
     results = []
 
     for _, row in df.iterrows():
         score = 0
-        combined_text = f"{row['Address']} {row['Location']} {row['ML_Assessment']} {row['predicted_Rent_category']}".lower()
+        text = f"{row['Address']} {row['Location']} {row['ML_Assessment']}"
+
+        if "luxury" in query and row['predicted_Rent_category'].lower() == "high":
+            score += 2
+        elif "affordable" in query and row['predicted_Rent_category'].lower() == "low":
+            score += 2
 
-        if 'affordable' in query or 'cheap' in query:
-            score += 2 if row['Rent'] < 40000 else 0
-        if 'luxury' in query or 'premium' in query:
-            score += 2 if row['Rent'] > 100000 else 0
         if str(row['Beds']) in query:
             score += 1
-        if any(word in combined_text for word in query.split()):
+
+        if any(word in text.lower() for word in query.split()):
             score += 1
 
         if score > 0:
-            results.append((score, row))
+            results.append((row['Address'], row['Rent'], row['Interior_Image_URL']))
 
-    # Sort by score
-    results.sort(key=lambda x: x[0], reverse=True)
-    top_results = results[:top_k]
+    top_results = sorted(results, key=lambda x: -x[1])[:5]  # Sort by rent descending
 
-    # Prepare output
-    outputs = []
-    for score, row in top_results:
-        info = f"🏠 **{row['Address']}**\n"
-        info += f"📍 Location: {row['Location']}\n"
-        info += f"💰 Rent: AED {int(row['Rent']):,}\n"
-        info += f"🛏️ Beds: {int(row['Beds'])}\n"
-        info += f"🧠 ML Assessment: {row['ML_Assessment']}\n"
-        info += f"📊 Predicted Category: {row['predicted_Rent_category']}\n"
-        image_url = row['Interior_Image_URL']
-        outputs.append((info, image_url))
+    images = [f'<img src="{url}" width="300">' for _, _, url in top_results]
+    texts = [f"🏠 **{addr}**  | 💰 AED {rent}" for addr, rent, _ in top_results]
 
-    return outputs
+    display = "<br><br>".join(f"{t}<br>{i}" for t, i in zip(texts, images))
 
-# Gradio UI
-def search_interface(query):
-    results = semantic_search(query)
-    cards = []
-    for info, image_url in results:
-        cards.append(gr.HTML(info))
-        cards.append(gr.Image(value=image_url, label="Interior View"))
-    return cards
+    return display
 
+# Build Gradio UI
 demo = gr.Interface(
-    fn=search_interface,
-    inputs=gr.Textbox(placeholder="e.g. affordable 2 bedroom in Business Bay", label="Enter your property search query"),
-    outputs=gr.Group(),
-    title="🏡 Semantic Real Estate Search (Multimodal)",
-    description="Search Abu Dhabi & Dubai listings using smart natural language queries. Powered by ML, images & embeddings."
+    fn=semantic_search,
+    inputs=gr.Textbox(label="Describe what you're looking for (e.g., 'affordable 2 bed in Dubai')"),
+    outputs=gr.HTML(),
+    title="🏡 Multimodal Property Search",
+    description="Search for properties based on text query. Returns matching homes and photos!"
 )
 
-if __name__ == "__main__":
-    demo.launch()
+demo.launch()