Pijush2023 commited on
Commit
f694fcb
·
verified ·
1 Parent(s): e459e9f

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +52 -6
app.py CHANGED
@@ -511,15 +511,63 @@ import re
511
 
512
  # return final_response
513
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
514
  def clean_response(response_text):
515
  # Remove system and user tags
516
  response_text = re.sub(r'<\|system\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
517
  response_text = re.sub(r'<\|user\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
518
  response_text = re.sub(r'<\|assistant\|>', '', response_text, flags=re.DOTALL)
519
 
520
- # Extract the document name and page number
521
- document_match = re.search(r"Document\(metadata=\{'source':'(.+?)','page':(\d+)\}", response_text)
522
- document_match = re.search(r"Document\(metadata=\{'source':'(.+?)','page':(\d+)\}", response_text)
523
  if document_match:
524
  document_name = document_match.group(1).split('/')[-1] # Get the document name
525
  page_number = document_match.group(2) # Get the page number
@@ -551,15 +599,13 @@ def clean_response(response_text):
551
  Sure! Here is the response for your Query:
552
  • Document name - {document_name}
553
  • Page No - {page_number}
554
- Response - {cleaned_response}
555
  """
556
 
557
  return final_response
558
 
559
 
560
 
561
-
562
-
563
  # Define a new template specifically for GPT-4o-mini in VDB Details mode
564
  gpt4o_mini_template_details = f"""
565
  As a highly specialized assistant, I provide precise, detailed, and informative responses. On this bright day of {current_date}, I'm equipped to assist with all your queries about Birmingham, Alabama, offering detailed insights tailored to your needs.
 
511
 
512
  # return final_response
513
 
514
+ # def clean_response(response_text):
515
+ # # Remove system and user tags
516
+ # response_text = re.sub(r'<\|system\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
517
+ # response_text = re.sub(r'<\|user\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
518
+ # response_text = re.sub(r'<\|assistant\|>', '', response_text, flags=re.DOTALL)
519
+
520
+ # # Extract the document name and page number
521
+ # document_match = re.search(r"Document\(metadata=\{'source':'(.+?)','page':(\d+)\}", response_text)
522
+ # # document_match = re.search(r"Document\(metadata=\{'source':'(.+?)','page':(\d+)\}", response_text)
523
+ # if document_match:
524
+ # document_name = document_match.group(1).split('/')[-1] # Get the document name
525
+ # page_number = document_match.group(2) # Get the page number
526
+ # else:
527
+ # document_name = "Unknown"
528
+ # page_number = "Unknown"
529
+
530
+ # # Remove the entire 'Document(metadata=...' and any mention of it from the response
531
+ # response_text = re.sub(r'Document\(metadata=\{.*?\},page_content=', '', response_text, flags=re.DOTALL)
532
+
533
+ # # Remove any mention of "Document:" in the response
534
+ # response_text = re.sub(r'- Document:.*', '', response_text)
535
+
536
+ # # Remove any unwanted escape characters like \u and \u00
537
+ # response_text = re.sub(r'\\u[0-9A-Fa-f]{4}', '', response_text)
538
+
539
+ # # Ensure proper spacing between words and dates
540
+ # response_text = re.sub(r'([a-zA-Z])(\d)', r'\1 \2', response_text)
541
+ # response_text = re.sub(r'(\d)([a-zA-Z])', r'\1 \2', response_text)
542
+
543
+ # # Remove the phrase "Sure! The Responses are as follows:" from the actual content
544
+ # response_text = re.sub(r'Sure! The Responses are as follows:', '', response_text).strip()
545
+
546
+ # # Clean up the text by removing extra whitespace
547
+ # cleaned_response = re.sub(r'\s+', ' ', response_text).strip()
548
+
549
+ # # Format the final response with bullet points
550
+ # final_response = f"""
551
+ # Sure! Here is the response for your Query:
552
+ # • Document name - {document_name}
553
+ # • Page No - {page_number}
554
+ # • Responses - {cleaned_response}
555
+ # """
556
+
557
+ # return final_response
558
+
559
+
560
+ import re
561
+
562
  def clean_response(response_text):
563
  # Remove system and user tags
564
  response_text = re.sub(r'<\|system\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
565
  response_text = re.sub(r'<\|user\|>.*?<\|end\|>', '', response_text, flags=re.DOTALL)
566
  response_text = re.sub(r'<\|assistant\|>', '', response_text, flags=re.DOTALL)
567
 
568
+ # Extract the document name and page number from updated pattern
569
+ document_match = re.search(r"Document\(metadata=\{'source': '(.+?)', 'page': (\d+)\}", response_text)
570
+
571
  if document_match:
572
  document_name = document_match.group(1).split('/')[-1] # Get the document name
573
  page_number = document_match.group(2) # Get the page number
 
599
  Sure! Here is the response for your Query:
600
  • Document name - {document_name}
601
  • Page No - {page_number}
602
+ Responses - {cleaned_response}
603
  """
604
 
605
  return final_response
606
 
607
 
608
 
 
 
609
  # Define a new template specifically for GPT-4o-mini in VDB Details mode
610
  gpt4o_mini_template_details = f"""
611
  As a highly specialized assistant, I provide precise, detailed, and informative responses. On this bright day of {current_date}, I'm equipped to assist with all your queries about Birmingham, Alabama, offering detailed insights tailored to your needs.