Update app.py

app.py

@@ -6,7 +6,6 @@ from cerebras.cloud.sdk import Cerebras
 
 # Load environment variables
 load_dotenv()
-
 client = Cerebras(
     api_key=os.environ.get("CEREBRAS_API_KEY")
 )
@@ -29,25 +28,21 @@ def chat(user_message, history):
                 {"role": "system", "content": "You are a helpful assistant."},
                 {"role": "user", "content": user_message}
             ],
-            model="llama3.1-8b",
+            model="gpt-oss-120b",          # ← Updated to free-tier model
             stream=True,
             max_completion_tokens=65536,
             temperature=1,
             top_p=1
         )
-
         response_text = ""
         for chunk in stream:
             if chunk.choices[0].delta.content:
                 response_text += chunk.choices[0].delta.content
-
         response_text = clean_response(response_text)
-
         history = history or []
         history.append({"role": "user", "content": user_message})
         history.append({"role": "assistant", "content": response_text})
         return history
-
     except Exception as e:
         history = history or []
         history.append({"role": "user", "content": user_message})
@@ -56,12 +51,10 @@ def chat(user_message, history):
 
 with gr.Blocks() as demo:
     gr.Markdown("## Chat")
-
     chatbot = gr.Chatbot(
         label="Conversation",
         height=600
     )
-
     with gr.Row():
         msg = gr.Textbox(
             label="Ask something",
@@ -69,7 +62,6 @@ with gr.Blocks() as demo:
             scale=4
         )
         clear = gr.Button("Clear")
-
     msg.submit(chat, [msg, chatbot], chatbot)
     clear.click(lambda: [], None, chatbot)