Update app.py

app.py

@@ -58,6 +58,8 @@ def load_sentence_transformer():
         print(f"Error loading Sentence Transformer: {e}")
         return None
 
+# Inside app.py, locate this function
+
 def load_google_sheet_data(sheet_id, service_account_key_base64):
     """Authenticates and loads data from Google Sheet."""
     print(f"Attempting to load Google Sheet data from ID: {sheet_id}")
@@ -67,8 +69,22 @@ def load_google_sheet_data(sheet_id, service_account_key_base64):
 
     try:
         print("Decoding base64 key...")
+        # --- Add Debugging Prints Here ---
+        print(f"Raw Base64 string (first 50 chars): {service_account_key_base64[:50]}...")
+        print(f"Raw Base64 string length: {len(service_account_key_base64)}")
+        # --- End Debugging Prints ---
+
         key_bytes = base64.b64decode(service_account_key_base64)
-        key_dict = json.loads(key_bytes)
+
+        # --- Add More Debugging Prints Here ---
+        print(f"Decoded bytes (first 50): {key_bytes[:50]}")
+        print(f"Decoded bytes length: {len(key_bytes)}")
+        # --- End Debugging Prints ---
+
+        # This is the line that is likely failing internally after base64.b64decode
+        # because key_bytes cannot be decoded as UTF-8
+        key_dict = json.loads(key_bytes.decode('utf-8')) # Explicitly decode to see if it throws error here
+
         print("Base64 key decoded and parsed.")
 
         print("Authenticating with service account...")
@@ -78,9 +94,6 @@ def load_google_sheet_data(sheet_id, service_account_key_base64):
         print("Authentication successful.")
 
         print(f"Opening sheet with key '{sheet_id}'...")
-        # *** IMPORTANT: If your sheet is NOT the first sheet, change 'sheet1'
-        # *** For example, if your sheet is named 'Data', use:
-        # sheet = client.open_by_key(sheet_id).worksheet("Data")
         sheet = client.open_by_key(sheet_id).sheet1
         print(f"Successfully opened Google Sheet with ID: {sheet_id}")
 
@@ -98,22 +111,19 @@ def load_google_sheet_data(sheet_id, service_account_key_base64):
 
         if not filtered_data:
             print("Warning: Filtered data is empty after checking for 'Service' and 'Description'.")
-            # Check if headers exist at all if filtered_data is empty but sheet_data isn't
             if sheet_data and ('Service' not in sheet_data[0] or 'Description' not in sheet_data[0]):
                  print("Error: 'Service' or 'Description' headers are missing or misspelled in the sheet.")
             return [], [], torch.tensor([])
 
-        # Re-checking column existence on filtered_data (redundant after filter but safe)
         if 'Service' not in filtered_data[0] or 'Description' not in filtered_data[0]:
              print("Error: Filtered Google Sheet data must contain 'Service' and 'Description' columns. This should not happen if filtering worked.")
              return [], [], torch.tensor([])
 
-
         services = [row["Service"] for row in filtered_data]
         descriptions = [row["Description"] for row in filtered_data]
         print(f"Loaded {len(descriptions)} entries from Google Sheet for embedding.")
 
-        return filtered_data, descriptions, None # Return descriptions, embeddings encoded later
+        return filtered_data, descriptions, None
 
     except gspread.exceptions.SpreadsheetNotFound:
         print(f"Error: Google Sheet with ID '{sheet_id}' not found.")
@@ -121,8 +131,10 @@ def load_google_sheet_data(sheet_id, service_account_key_base64):
         return [], [], torch.tensor([])
     except Exception as e:
         print(f"An error occurred while accessing the Google Sheet: {e}")
+        # Consider adding print(f"Type of error: {type(e)}") to see if it's specifically UnicodeDecodeError
         return [], [], torch.tensor([])
 
+        
 def load_llm_model(model_id, hf_token):
     """Loads the LLM in full precision (for CPU)."""
     print(f"Loading model {model_id} in full precision...")