Update space

app.py

@@ -15,28 +15,26 @@ def respond(
     temperature,
     top_p,
 ):
-    messages = [{"role": "system", "content": system_message}]
-
+    # Format the conversation history into a prompt
+    prompt = f"{system_message}\n\n"
     for val in history:
         if val[0]:
-            messages.append({"role": "user", "content": val[0]})
+            prompt += f"User: {val[0]}\n"
         if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
+            prompt += f"Assistant: {val[1]}\n"
+    prompt += f"User: {message}\nAssistant:"
 
     response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
+    
+    # Use text generation instead of chat completion
+    for message in client.text_generation(
+        prompt,
+        max_new_tokens=max_tokens,
         stream=True,
         temperature=temperature,
         top_p=top_p,
     ):
-        token = message.choices[0].delta.content
-
-        response += token
+        response += message
         yield response