Update app.py

app.py

@@ -121,15 +121,24 @@ def get_menu():
 def reply_bot(message, history):
     menu = get_menu()
 
+    history = history[-2:]
+
     
     client = InferenceClient(model="https://8cc9-141-3-25-29.ngrok-free.app")
 
-    prompt = "<s>[INST] <<SYS>>\nYou are multilingual chat bot that helps deciding what to eat ina german canteen. In the canteen there are different lines with names. Based on the menu and question, you suggest the user which line they should go to. You respond really briefly and do not generate long responses\n<</SYS>>\n\nMenu:\n" + menu + "\n" + message + " [/INST]"
+    system_prompt = "<s>[INST] <<SYS>>\nYou are multilingual chat bot that helps deciding what to eat ina german canteen. In the canteen there are different lines with names. Based on the menu and question, you suggest the user which line they should go to. You respond really briefly and do not generate long responses\n<</SYS>>\n\nMenu:\n" + menu + "\n"
+
+    for human, ai in history:
+        system_prompt += human + " [/INST]" + ai + "</s><s>[INST]\n"
+
+        curr_prompt = system_prompt + message
+    
+    #prompt = "<s>[INST] <<SYS>>\nYou are multilingual chat bot that helps deciding what to eat ina german canteen. In the canteen there are different lines with names. Based on the menu and question, you suggest the user which line they should go to. You respond really briefly and do not generate long responses\n<</SYS>>\n\nMenu:\n" + menu + "\n" + message + " [/INST]"
 
     try:
         #answer = client.text_generation(prompt=prompt, max_new_tokens=512)
         answer = ""
-        for token in client.text_generation(prompt=prompt, max_new_tokens=512, stream=True):
+        for token in client.text_generation(prompt=curr_prompt, max_new_tokens=512, stream=True):
             answer+=token
             yield answer
     except: