Update app.py

app.py

@@ -12,11 +12,11 @@ with open(r'vect_values.pickle', 'rb') as file:
 
 df = pd.read_excel('perfume_database_cleaned.xlsx')
 
-# Example columns in df:
-#  - brand
-#  - perfume
-#  - notes
-#  - (optionally) price, gender, season, rating, image_url, etc.
+# Clean the DataFrame:
+# Drop rows where 'perfume' is missing and convert both 'perfume' and 'brand' to strings
+df.dropna(subset=['perfume'], inplace=True)
+df['perfume'] = df['perfume'].astype(str)
+df['brand'] = df['brand'].astype(str)
 
 # Create brand and perfume options for the dropdowns
 brand_options = sorted(df['brand'].unique())
@@ -28,7 +28,7 @@ perfume_options = sorted(df['perfume'].unique())
 
 def perfume_change(brand):
     """
-    When user selects a brand, update the perfume dropdown 
+    When the user selects a brand, update the perfume dropdown 
     with only the perfumes from that brand.
     """
     filtered_perfumes = df.loc[df['brand'] == brand, 'perfume'].unique()
@@ -36,48 +36,33 @@ def perfume_change(brand):
 
 def filter_by_perfume(selected_brand, selected_perfume, top_n, note_search):
     """
-    1) Find the index of the chosen perfume in the DF
-    2) Retrieve the top_n most similar perfumes (based on vect_index and vect_values)
-    3) If note_search is not empty, filter recommendations by that note
-    4) Return a subset of columns + similarity
+    1) Find the index of the chosen perfume in the DataFrame.
+    2) Retrieve the top_n most similar perfumes (based on vect_index and vect_values).
+    3) If note_search is provided, filter recommendations by that note.
+    4) Return a subset of columns plus similarity.
     """
-    # ----------------------------------------------------------------
-    # 1) Locate chosen perfume in the DF
-    # ----------------------------------------------------------------
-    # We assume there's exactly one row that matches brand+perfume.
-    # If there's a possibility of duplicates, handle that carefully.
+    # Locate the chosen perfume in the DataFrame
     row_idx = df.query("brand == @selected_brand and perfume == @selected_perfume").index[0]
     
-    # ----------------------------------------------------------------
-    # 2) Retrieve top_n from vect_index, vect_values
-    # ----------------------------------------------------------------
+    # Retrieve top_n indices and similarity values
     recommended_indices = vect_index[row_idx][:top_n]   # e.g. top_n indices
-    recommended_sims = vect_values[row_idx][:top_n]     # e.g. top_n similarity scores
+    recommended_sims = vect_values[row_idx][:top_n]       # e.g. top_n similarity scores
     
-    # Build a smaller DF of these recommended items
+    # Build a DataFrame of the recommended items
     df_recs = df.iloc[recommended_indices].copy().reset_index(drop=True)
     df_recs['similarity'] = recommended_sims
     df_recs['similarity'] = df_recs['similarity'].map("{:.2%}".format)
     
-    # Clean up notes by sorting & re-joining
+    # Clean up the notes column: split, sort, and join
     df_recs['notes'] = df_recs['notes'].str.split(',').apply(lambda x: sorted(x))
     df_recs['notes'] = df_recs['notes'].apply(lambda x: ', '.join([n.strip() for n in x]))
     
-    # ----------------------------------------------------------------
-    # 3) Optional note search (if user typed something in)
-    # ----------------------------------------------------------------
+    # Optional filtering by a specific note if provided
     if note_search:
-        # For example, filter for rows that contain that note in the 'notes' column
-        # Convert both sides to lowercase for a case-insensitive match
         note_search_lower = note_search.lower()
         df_recs = df_recs[df_recs['notes'].str.lower().str.contains(note_search_lower)]
     
-    # ----------------------------------------------------------------
-    # 4) Return final DataFrame
-    # ----------------------------------------------------------------
-    # You can include any columns you want. 
-    # Example columns: brand, perfume, similarity, notes
-    # If you have images, you could include them in a separate Gallery component, etc.
+    # Return the desired columns
     return df_recs[['brand', 'perfume', 'similarity', 'notes']]
 
 # ----------------------------------------------------------------
@@ -121,7 +106,7 @@ button:hover {
     font-size: 0.9rem !important;
 }
 
-/* Centered text for the top bar */
+/* Centered text for the top header */
 #top-header {
     text-align: center;
     margin-bottom: 20px;
@@ -132,7 +117,7 @@ button:hover {
 # BUILD THE GRADIO INTERFACE
 # ----------------------------------------------------------------
 with gr.Blocks(css=custom_css) as demo:
-    # Top header
+    # Top header with HTML content
     gr.HTML("""
     <div id="top-header">
         <h1>🌸 Perfume Recommendation App 🌸</h1>
@@ -141,17 +126,17 @@ with gr.Blocks(css=custom_css) as demo:
     """)
 
     with gr.Row():
-        # Left column: brand, perfume, note search
+        # Left column: user inputs
         with gr.Column():
             brand_dropdown = gr.Dropdown(
                 choices=brand_options, 
                 label="Select Brand", 
-                value=brand_options[0]  # or None
+                value=brand_options[0]
             )
             perfume_dropdown = gr.Dropdown(
                 choices=perfume_options, 
                 label="Select Perfume", 
-                value=perfume_options[0]  # or None
+                value=perfume_options[0]
             )
             brand_dropdown.change(
                 fn=perfume_change, 
@@ -159,12 +144,12 @@ with gr.Blocks(css=custom_css) as demo:
                 outputs=[perfume_dropdown]
             )
 
-            # Textbox to filter by note (optional)
+            # Optional textbox for note search
             note_search = gr.Textbox(
                 label="Optional: Search for a specific note (e.g., 'rose', 'vanilla')"
             )
 
-            # Slider for how many recommendations to show
+            # Slider to select the number of recommendations to show
             top_n_slider = gr.Slider(
                 minimum=1, 
                 maximum=10, 
@@ -173,19 +158,18 @@ with gr.Blocks(css=custom_css) as demo:
                 label="Number of recommendations to show"
             )
 
-            # Button to trigger the search
+            # Button to trigger the recommendation
             btn_search = gr.Button("Recommend")
 
-        # Right column: results
+        # Right column: display results
         with gr.Column():
-            # The results DataFrame
             output_df = gr.Dataframe(
                 headers=["Brand", "Perfume", "Similarity", "Notes"],
                 label="Recommended Perfumes",
                 wrap=True
             )
 
-    # Connect button to filter_by_perfume function
+    # Link the Recommend button to our function
     btn_search.click(
         fn=filter_by_perfume,
         inputs=[brand_dropdown, perfume_dropdown, top_n_slider, note_search],