Update app.py

app.py

@@ -36,7 +36,7 @@ PRE_PROMPT = load_decrypted_preprompt()
 # Default parameters for the QA chain
 DEFAULT_TEMPERATURE = 0.7
 DEFAULT_MAX_TOKENS = 1024
-DEFAULT_TOP_K = 3
+DEFAULT_TOP_K = 10
 DEFAULT_TOP_P = 0.95
 
 def load_vector_db(index_path="faiss_index", model_name="sentence-transformers/all-MiniLM-L6-v2"):
@@ -154,16 +154,21 @@ def get_assistant_response(message, history, max_tokens, temperature, top_p, qa_
     qa_chain = qa_chain_state_dict.get("qa_chain")
     
     if qa_chain is not None:
+        # Format history to the plain-text format expected by the QA chain
         formatted_history = format_chat_history(history)
+        # Prepend the pre-prompt to the current question
         combined_question = PRE_PROMPT + "\n" + message
-        print("Combined Question:", combined_question)  # Debug print
         response = qa_chain.invoke({"question": combined_question, "chat_history": formatted_history})
-        print("QA Chain Response:", response)  # Debug print
-        answer = response.get("answer", "")
+        answer = response.get("answer", "").strip()
+        
+        # Check if the answer is empty and apply a fallback response if needed.
+        if not answer:
+            answer = "I'm sorry, I couldn't retrieve a clear answer. Could you please rephrase your question?"
+        
         history.append({"role": "assistant", "content": answer})
         return history, {"qa_chain": qa_chain}
     
-    # Fallback: Plain Chat Mode using the InferenceClient (pre-prompt already included here)
+    # Fallback: Plain Chat Mode using the InferenceClient
     messages = [{"role": "system", "content": PRE_PROMPT}] + history
     response = ""
     result = client.chat_completion(
@@ -173,10 +178,13 @@ def get_assistant_response(message, history, max_tokens, temperature, top_p, qa_
         temperature=temperature,
         top_p=top_p,
     )
-    print("Chat Completion Result:", result)  # Debug print
     for token_message in result:
         token = token_message.choices[0].delta.content
         response += token
+
+    response = response.strip()
+    if not response:
+        response = "I'm sorry, I couldn't generate a response. Please try asking in a different way. Alterantively, consider contacting Christopher directly: https://gcmarais.com/contact/"
     
     history.append({"role": "assistant", "content": response})
     return history, {"qa_chain": qa_chain}