Update app.py

app.py

@@ -225,21 +225,18 @@ answer = ""  # Initialize ai_response with a default value
 
 
 
-# Update the ask_alans_ai function to handle Claude Sonnet
 def ask_alans_ai(query, vectordb, chat_history, aoc_qa):
     try:
         # Use the ConversationalRetrievalChain directly
         result = aoc_qa.invoke({"question": query})
         answer = result["answer"]
-        source_documents = result.get("source_documents", [])
-
-        # You can use source_documents if needed, e.g., to display sources
+        source_documents = result.get("source_documents", [])  # Use .get() with a default empty list
 
         chat_history.append((query, answer))
-        return answer
+        return answer, source_documents  # Return both answer and source_documents
     except Exception as e:
         st.error(f"An error occurred: {str(e)}")
-        return "I'm sorry, but I encountered an error while processing your request. Please try again later."
+        return "I'm sorry, but I encountered an error while processing your request. Please try again later.", []  # Return empty list as source_documents
 
 
 
@@ -647,11 +644,13 @@ def main():
         # Submit button
         submitted = st.form_submit_button(label="Ask", use_container_width=True)
 
+        answer = ""  # Initialize answer
+        source_documents = []  # Initialize source_documents as an empty list
+
         if submitted and message:
-            # Process the query and get the response
             with st.spinner('Thinking...'):
-                response = ask_alans_ai(message, vectordb, st.session_state.chat_history, aoc_qa)
-               
+                response, source_documents = ask_alans_ai(message, vectordb, st.session_state.chat_history, aoc_qa)
+
                
                     
                     
@@ -669,14 +668,32 @@ def main():
             </div>
             """, unsafe_allow_html=True)
 
-            # Custom HTML for the answer with assistant icon
+            # Display the response
             st.markdown(f"""
-            <div id="{answer_id}" style="display: flex; align-items: flex-start; margin-bottom: 20px;">
+                <div id="{answer_id}" style="display: flex; align-items: flex-start; margin-bottom: 20px;">
                 <img src='data:image/png;base64,{ci_icon}' style='width: 50px; height: 50px; margin-right: 10px;'>
-                <span>{answer}</span>
+                <span>{response}</span>
             </div>
             """, unsafe_allow_html=True)
 
+            # Display metadata for unique source documents
+            if source_documents:
+                unique_sources = []
+                seen_urls = set()
+            
+                for doc in source_documents:
+                    url = doc.metadata.get('url', 'N/A')
+                    if url not in seen_urls:
+                        unique_sources.append(doc)
+                        seen_urls.add(url)
+            
+                if unique_sources:
+                    st.markdown("### Sources:")
+                    for doc in unique_sources:
+                        st.markdown(f"""
+                        - [{doc.metadata.get('title', 'No title available')}]({doc.metadata.get('url', '#')})
+                        """)
+
         # JavaScript to scroll to the latest answer
         if st.session_state.chat_history:
             latest_answer_id = f"answer-{len(st.session_state.chat_history) - 1}"
@@ -693,8 +710,10 @@ def main():
 
 
     # Your combined string with the current date included
-    combined_string = f"Question: {message}\n\nAnswer: {answer}\n\nDate: {date_string}\n\nhttps://www.citizensinformation.ie/"
-    # Create a list with the three strings
+    combined_string = f"Question: {message}\n\nAnswer: {answer}\n\nDate: {date_string}\n\n"
+    for doc in source_documents:
+        combined_string += f"Source: {doc.metadata.get('title', 'N/A')} - {doc.metadata.get('url', 'N/A')}\n"
+    combined_string += "https://www.citizensinformation.ie/"
     
     message_clean = clean_string(message)
     answer_clean = clean_string(answer)
@@ -706,9 +725,19 @@ def main():
     answer_clean = answer_clean[:max_length]
     date_string_clean = date_string_clean[:max_length]
 
-    # Append the cleaned data to the worksheet
-    data_to_append = [message_clean, answer_clean, date_string, str(ai_temp), st.session_state['session_id'], st.session_state['selected_model'], str(k_value), selected_directory, selected_search_type]
-
+    data_to_append = [
+    message_clean, 
+    answer_clean, 
+    date_string, 
+    str(ai_temp), 
+    st.session_state['session_id'], 
+    st.session_state['selected_model'], 
+    str(k_value), 
+    selected_directory, 
+    selected_search_type,
+    ", ".join([doc.metadata.get('title', 'N/A') for doc in source_documents]),
+    ", ".join([doc.metadata.get('url', 'N/A') for doc in source_documents])
+]
    
     # Create and display the copy button only if answer has content
     if answer: