Create app.py

app.py

@@ -0,0 +1,24 @@
+import gradio as gr
+from huggingface_hub.utils import HfHubHTTPError
+
+def predict(message, history):
+    try:
+        response = gr.Interface.load(
+            "models/meta-llama/Meta-Llama-3.1-8B"
+        ).predict(message, history)
+        history.append((message, response))
+        return "", history
+    except HfHubHTTPError as e:
+        if e.response.status_code == 504:
+            return "Server überlastet. Bitte später erneut versuchen.", history
+        else:
+            raise e
+
+with gr.Blocks() as demo:
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox()
+    clear = gr.ClearButton([msg, chatbot])
+
+    msg.submit(predict, [msg, chatbot], [msg, chatbot])
+
+demo.launch()