Update app.py

app.py

@@ -15,30 +15,31 @@ def respond(
     temperature,
     top_p,
 ):
-    messages = [{"role": "system", "content": system_message}]
-
+    # Prepare the prompt based on the message and history
+    prompt = system_message + "\n"
     for val in history:
         if val[0]:
-            messages.append({"role": "user", "content": val[0]})
+            prompt += "User: " + val[0] + "\n"
         if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+            prompt += "Assistant: " + val[1] + "\n"
 
-    messages.append({"role": "user", "content": message})
+    prompt += "User: " + message + "\nAssistant:"
 
     response = ""
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
+    # Use text-generation instead of chat-completion
+    for message in client.text_generation(
+        prompt=prompt,
+        max_new_tokens=max_tokens,
         temperature=temperature,
         top_p=top_p,
+        stream=True,
     ):
-        token = message.choices[0].delta.content
-
+        token = message['generated_text'].replace(prompt, '')
         response += token
         yield response
 
+
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
 """