Update app.py

app.py

@@ -35,7 +35,7 @@ PRE_PROMPT = load_decrypted_preprompt()
 
 # Default parameters for the QA chain
 DEFAULT_TEMPERATURE = 0.7
-DEFAULT_MAX_TOKENS = 512
+DEFAULT_MAX_TOKENS = 1024
 DEFAULT_TOP_K = 3
 DEFAULT_TOP_P = 0.95
 
@@ -116,19 +116,49 @@ def update_chat(message, history):
     history.append({"role": "user", "content": message})
     return history, message, ""
 
+# def get_assistant_response(message, history, max_tokens, temperature, top_p, qa_chain_state_dict):
+#     """
+#     Generate the assistant's response using the QA chain (if available) or fallback to plain chat.
+#     The pre-prompt is always included by concatenating it to the user's new question.
+#     """
+#     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
+#         response = qa_chain.invoke({"question": combined_question, "chat_history": formatted_history})
+#         answer = response.get("answer", "")
+#         history.append({"role": "assistant", "content": answer})
+#         return history, {"qa_chain": qa_chain}
+    
+#     # Fallback: Plain Chat Mode using the InferenceClient (pre-prompt already included here)
+#     messages = [{"role": "system", "content": PRE_PROMPT}] + history
+#     response = ""
+#     result = client.chat_completion(
+#         messages,
+#         max_tokens=max_tokens,
+#         stream=False,
+#         temperature=temperature,
+#         top_p=top_p,
+#     )
+#     for token_message in result:
+#         token = token_message.choices[0].delta.content
+#         response += token
+    
+#     history.append({"role": "assistant", "content": response})
+#     return history, {"qa_chain": qa_chain}
+
 def get_assistant_response(message, history, max_tokens, temperature, top_p, qa_chain_state_dict):
-    """
-    Generate the assistant's response using the QA chain (if available) or fallback to plain chat.
-    The pre-prompt is always included by concatenating it to the user's new question.
-    """
     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", "")
         history.append({"role": "assistant", "content": answer})
         return history, {"qa_chain": qa_chain}
@@ -143,6 +173,7 @@ 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
@@ -190,10 +221,19 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue="sky")) as demo:
       document.documentElement.setAttribute('data-theme', 'light');
     </script>
     
+    <style>
+
     <style>
     :root {
         color-scheme: light !important;
+        background-color: #fff !important;
+        color: #333 !important;
     }
+        body, .gradio-container, .chatbot, .hf-chat-input {
+        background-color: #fff !important;
+        color: #333 !important;
+    }
+
     .example-row {
         flex-grow: 1 !important;
         width: 100% !important;