Update app.py

app.py

@@ -157,12 +157,16 @@ def respond(
     ]
 
     message_text = message["text"]
+    
 
     client_mixtral = InferenceClient("NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO")
     client_llama = InferenceClient("meta-llama/Meta-Llama-3-8B-Instruct")
     generate_kwargs = dict( max_new_tokens=2000, do_sample=True, stream=True, details=True, return_full_text=False )
+    question_history = ""
+    for msg in history:
+        question_history += f"{str(msg[0])}/n"
 
-    messages.append({"role": "user", "content": f'[SYSTEM]You are a helpful assistant with access to the following functions: \n {str(functions_metadata)}\n\nTo use these functions respond with:\n<functioncall> {{ "name": "function_name", "arguments": {{ "arg_1": "value_1", "arg_1": "value_1", ... }} }} </functioncall> [USER] {message_text} {vqa}'})
+    messages.append({"role": "user", "content": f'[SYSTEM]You are a helpful assistant with access to the following functions: \n {str(functions_metadata)}\n\nTo use these functions respond with:\n<functioncall> {{ "name": "function_name", "arguments": {{ "arg_1": "value_1", "arg_1": "value_1", ... }} }} </functioncall> Choose functions wisely and Also reply wisely, reply with just functioncall only as tell you before, Make function while learning from Prev Questions . [PREV_QUESTIONS] {question_history} [USER] {message_text} {vqa}'})
 
     response = client.chat_completion( messages, max_tokens=150)
     response = str(response)