Update app.py

app.py

@@ -1,9 +1,11 @@
-import gradio as gr
+import { InferenceClient } from "@huggingface/inference";
 
-gr.load(
-   "models/Qwen/Qwen3-235B-A22B",
-   provider="hf-inference",
-   messages: [
+let out = "";
+
+const stream = client.chatCompletionStream({
+    provider: "hf-inference",
+    model: "Qwen/Qwen3-235B-A22B",
+    messages: [
         {
             role: "user",
             content: "你是谁？",
@@ -17,6 +19,12 @@ gr.load(
     max_tokens: 50000,
     top_p: 0.6,
     response_format: undefined
-).launch()
-
+});
 
+for await (const chunk of stream) {
+	if (chunk.choices && chunk.choices.length > 0) {
+		const newContent = chunk.choices[0].delta.content;
+		out += newContent;
+		console.log(newContent);
+	}
+}