Update app.py

app.py

@@ -54,29 +54,33 @@ def answer_query_from_csv(query, file):
     # Read the file into a DataFrame
     table_data = pd.read_csv(file)
 
-    # Convert object-type columns (text) to lowercase, leaving numeric columns as is
+    # Convert object-type columns to lowercase, ensuring only valid strings are affected
     for column in table_data.columns:
-        if table_data[column].dtype == 'object':  # Only apply to text columns
+        if table_data[column].dtype == 'object':
             table_data[column] = table_data[column].apply(lambda x: x.lower() if isinstance(x, str) else x)
 
-    # Handle datetime columns to extract components
+    
+    
+    # Extract year, month, day, and time components for datetime columns
     for column in table_data.columns:
         if pd.api.types.is_datetime64_any_dtype(table_data[column]):
             table_data[f'{column}_year'] = table_data[column].dt.year
             table_data[f'{column}_month'] = table_data[column].dt.month
             table_data[f'{column}_day'] = table_data[column].dt.day
             table_data[f'{column}_time'] = table_data[column].dt.strftime('%H:%M:%S')
-
-    # Now process the table and query
+     # Ensure all data in the table is converted to string
+    table_data = table_data.astype(str)
+    # Process the CSV file and query using TAPEX
     result_tapex = process_table_query(query, table_data)
 
-    # Process the query using TAPAS pipelines
+    # Process the query using TAPAS pipelines (ensure all cells are strings)
     result_tapas = pipe_tapas(table=table_data, query=query)['cells'][0]
     result_tapas2 = pipe_tapas2(table=table_data, query=query)['cells'][0]
     
     return result_tapex, result_tapas, result_tapas2
 
 
+
 # Create Gradio interface
 with gr.Blocks() as interface:
     gr.Markdown("# Table Question Answering with TAPEX and TAPAS Models")