Upload app.py with huggingface_hub

app.py

@@ -1,11 +1,14 @@
 import os
 import gradio as gr
-from huggingface_hub import InferenceClient
+from openai import OpenAI
 
-client = InferenceClient(token=os.getenv("HF_TOKEN"))
+client = OpenAI(
+    base_url="https://router.huggingface.co/v1",
+    api_key=os.getenv("HF_TOKEN")
+)
 
 def chat(message, history):
-    response = client.chat_completion(
+    response = client.chat.completions.create(
         model="HuggingFaceH4/zephyr-7b-beta",
         messages=[{"role": "user", "content": message}],
         max_tokens=256