Update app.py

app.py

@@ -1,18 +1,18 @@
 import os
-from huggingface_hub import InferenceClient
+import requests
 
-client = InferenceClient(
-    provider="auto",
-    api_key=os.environ["HF_API_TOKEN"],
-)
+API_URL = "https://router.huggingface.co/hf-inference/models/HuggingFaceTB/SmolLM3-3B"
+headers = {
+    "Authorization": f"Bearer {os.environ['HF_API_TOKEN']}",
+}
 
-stream = client.chat.completions.create(
-    model="meta-llama/Llama-3.2-1B-Instruct",
-    messages=[],
-    temperature=0.5,
-    top_p=0.7,
-    stream=True,
-)
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
 
-for chunk in stream:
-    print(chunk.choices[0].delta.content, end="")
+response = query({
+    "messages": "\"Can you please let us know more details about your \"",
+    "model": "meta-llama/Llama-3.1-8B-Instruct:hf-inference"
+})
+
+print(response["choices"][0]["message"])