Update app.py

app.py

@@ -463,33 +463,60 @@ def bot(history, choice, tts_choice, retrieval_mode, model_choice):
 
 import re
 
-def clean_response(response_text):
-    # Remove system and user tags
+# def clean_response(response_text):
+#     # Remove system and user tags
+#     response_text = re.sub(r'<\|system\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
+#     response_text = re.sub(r'<\|user\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
+#     response_text = re.sub(r'<\|assistant\|>', '', response_text, flags=re.DOTALL)
+#     #-------
+#     response_text = re.sub(r'\\u[0-9A-Fa-f]{4}', '', response_text)
+#     # Add spaces between words and dates (e.g., "born04/04/1963" becomes "born 04/04/1963")
+#     response_text = re.sub(r'([a-zA-Z])(\d)', r'\1 \2', response_text)
+#     response_text = re.sub(r'(\d)([a-zA-Z])', r'\1 \2', response_text)
+#     #--------
+#     # Clean up the text by removing extra whitespace
+#     cleaned_response = response_text.strip()
+#     cleaned_response = re.sub(r'\s+', ' ', cleaned_response)
+
+#     # Ensure the response is conversational and organized
+#     cleaned_response = cleaned_response.replace('1.', '\n1.').replace('2.', '\n2.').replace('3.', '\n3.').replace('4.', '\n4.').replace('5.', '\n5.')
+
+#     return cleaned_response
+
+def clean_response(response_text, document_name, page_number):
+    #Remove System and user Tag
     response_text = re.sub(r'<\|system\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
     response_text = re.sub(r'<\|user\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
     response_text = re.sub(r'<\|assistant\|>', '', response_text, flags=re.DOTALL)
-    #-------
+    
+    # Remove metadata and unwanted parts (like 'Document(metadata...' etc.)
+    response_text = re.sub(r'Document\(metadata=\{.*?\},page_content="', '', response_text)
+    
+    # Remove any unwanted escape characters like \u and \u00
     response_text = re.sub(r'\\u[0-9A-Fa-f]{4}', '', response_text)
-    # Add spaces between words and dates (e.g., "born04/04/1963" becomes "born 04/04/1963")
+    
+    # Add spaces between words and numbers (e.g., "born04/04/1963" becomes "born 04/04/1963")
     response_text = re.sub(r'([a-zA-Z])(\d)', r'\1 \2', response_text)
     response_text = re.sub(r'(\d)([a-zA-Z])', r'\1 \2', response_text)
-    #--------
-    # Clean up the text by removing extra whitespace
-    cleaned_response = response_text.strip()
-    cleaned_response = re.sub(r'\s+', ' ', cleaned_response)
-
-    # Ensure the response is conversational and organized
-    cleaned_response = cleaned_response.replace('1.', '\n1.').replace('2.', '\n2.').replace('3.', '\n3.').replace('4.', '\n4.').replace('5.', '\n5.')
-
-    return cleaned_response
-
-
-
-
-
+    
+    # Remove any extra quotation marks at the end
+    response_text = response_text.rstrip('"')
 
+    # Ensure spaces between words
+    response_text = re.sub(r'([a-z])([A-Z])', r'\1 \2', response_text)
 
+    # Clean up any extra spaces that may result from the replacements
+    response_text = re.sub(r'\s+', ' ', response_text).strip()
 
+    # Now, format the output as required, directly within the clean_response function
+    formatted_response = f"""
+Sure! The Responses are as follows:
+Document name - {document_name}
+Page No - {page_number}
+Response - {response_text}
+"""
+    
+    return formatted_response
 
 
 # Define a new template specifically for GPT-4o-mini in VDB Details mode