Create app.py

app.py

@@ -0,0 +1,26 @@
+import os
+import gradio as gr
+from huggingface_hub import InferenceClient
+
+client = InferenceClient(provider="novita", api_key=os.environ["HF_TOKEN"])
+models = ["Qwen/Qwen3-Next-80B-A3B-Instruct", "meta-llama/Llama-3.1-70B-Instruct", "mistralai/Mixtral-8x7B-Instruct-v0.1"]
+
+def respond(message, history, model):
+    completion = client.chat.completions.create(
+        model=model,
+        messages=[{"role": "user", "content": message}]
+    )
+    reply = completion.choices[0].message.content
+    history.append((message, reply))
+    return "", history
+
+with gr.Blocks() as demo:
+    gr.Markdown("## AI")
+    chatbot = gr.Chatbot(height=400)
+    msg = gr.Textbox(label="Ask me smth")
+    model_dd = gr.Dropdown(models, label="Model", value=models[0])
+    clear = gr.Button("Clear")
+    msg.submit(respond, [msg, chatbot, model_dd], [msg, chatbot])
+    clear.click(lambda: None, None, chatbot, queue=False)
+
+demo.launch()