Update app.py

app.py

@@ -532,17 +532,20 @@ import re
 #     return final_response
 
 
+import re
+
 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)
 
-    # Extract the document name and page number
-    document_match = re.search(r"(\d{6,})_(\d{12,})_V\d+\.pdf,page:(\d+)", response_text)
+    # Adjusted regex pattern to extract document name and page number
+    document_match = re.search(r"(\d{6}_\d{12}_V\d+\.pdf),page:(\d+)", response_text)
+    
     if document_match:
-        document_name = document_match.group(0).split(',')[0]  # Get the document name (before ',page')
-        page_number = document_match.group(3)  # Get the page number
+        document_name = document_match.group(1)  # Get the document name
+        page_number = document_match.group(2)  # Get the page number
     else:
         document_name = "Unknown"
         page_number = "Unknown"
@@ -576,6 +579,7 @@ Sure! Here is the response for your Query:
 
     return final_response
 
+
 # Define a new template specifically for GPT-4o-mini in VDB Details mode
 gpt4o_mini_template_details = f"""
 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.
@@ -1499,42 +1503,7 @@ def fetch_google_flights(departure_id="JFK", arrival_id="BHM", outbound_date=cur
 #     return prompt[0] if prompt else current_text
 
 
-import pandas as pd
 
-# Create a function to extract top 3 responses with document name, page number, and confidence level
-def extract_top_responses(response_text):
-    # Assuming that the response contains document name, page number, and confidence levels in some format
-    # You can adjust the regex or string manipulation based on the actual response format
-
-    # Placeholder for document details
-    document_name = "Unknown"
-    page_number = "Unknown"
-    
-    # Assuming we have the top 3 responses in the response text
-    responses = re.findall(r'Response: (.+?)(?=Confidence:)', response_text)
-    confidences = re.findall(r'Confidence: (\d+\.\d+)', response_text)
-    
-    # Create a DataFrame with top 3 responses based on confidence level
-    data = {
-        "Document Name": [document_name] * len(confidences[:3]),
-        "Page Number": [page_number] * len(confidences[:3]),
-        "Top Responses": responses[:3]
-    }
-    df = pd.DataFrame(data)
-    
-    return df
-
-# Add function to generate and update the table dynamically
-def update_table_based_on_phi_response(history):
-    if not history:
-        return pd.DataFrame()  # Return an empty DataFrame if no history
-    
-    response = history[-1][1]
-    if response:
-        top_responses_df = extract_top_responses(response)
-        return top_responses_df
-    else:
-        return pd.DataFrame()
 
 
 
@@ -1640,10 +1609,7 @@ with gr.Blocks(theme='gradio/soft') as demo:
         #     refresh_button = gr.Button("Refresh Images")
         #     refresh_button.click(fn=update_images, inputs=None, outputs=[image_output_1, image_output_2, image_output_3])
             
-        with gr.Column():
-        # Add the table to display the top 3 responses based on confidence level
-            response_table = gr.Dataframe(headers=["Document Name", "Page Number", "Top Responses"], interactive=False, show_label=False)
-            # response_table.render()
+