Update app.py

app.py

@@ -136,7 +136,7 @@ def respond(
 ):
     web_results = search(message)
     web2 = ' '.join([f"Link: {res['link']}\nText: {res['text']}\n\n" for res in web_results])
-    messages1 = "Your name is Chatchat.Answer as Real OpenGPT 4o, Made by 'peterpeter8585', Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses. The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant.If the user asks with Korean, you must answer with Korean.Or if the user asks with English, you have to answer with English.Do not transelate it.This is the web search result, use itto answer the user's questions.:"+web2+"And the instructions:"+system_message
+    messages1 = "Your name is Chatchat.Answer as Real OpenGPT 4o, Made by 'peterpeter8585', Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses. The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant.If the user asks with Korean, you must answer with Korean.Or if the user asks with English, you have to answer with English.Do not transelate it.This is the web search result, use it to answer the user's questions:"+web2+"And the instructions:"+system_message
     messages = [{"role": "system", "content":messages1}]
 
     for val in history:
@@ -150,10 +150,17 @@ def respond(
 
     response = ""
 
-    stream = client.text_generation(formatted_prompt, max_new_tokens=512, stream=True, details=True, return_full_text=False)
-    token="".join([response.token.text for response in stream if response.token.text != "</s>"])
-    response += token
-    yield response
+    for message in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
+        temperature=temperature,
+        top_p=top_p,
+    ):
+        token = message.choices[0].delta.content
+
+        response += token
+        yield response
     
 examples = [
     "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",