Update app.py

app.py

@@ -12,14 +12,18 @@ def respond(message, history):
         messages.extend(history)
     messages.append({"role": "user", "content": message})
 
-    response = client.chat_completion(
+    response = ""
+    for message in client.chat_completion(
         messages,
         max_tokens=100, # change the length of message
         temperature = 0.1,
-        top_p = 0.7
-    )
+        top_p = 0.7,
+        stream = True,
+    ):
+        token = messages.choices[0].delta.content
+        response += token
 
-    return response['choices'][0]['message']['content'].strip()
+    yield response['choices'][0]['message']['content'].strip()
 
 chatbot = gr.ChatInterface(respond, type = "messages", title = "SadhanaGPT for KWK", theme = gr.themes.Glass(), examples = ["How's the weather today?", "Who won the match?", "Is the sky green?"])