Update app.py

app.py

@@ -56,15 +56,24 @@ def get_messages_api(messages):
     return json.dumps(messages, indent=4)
 
 def run_huggingface_model(model, messages, max_tokens, temperature, top_p):
-    client = InferenceClient(model)
-    response = client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=False,
-        temperature=temperature,
-        top_p=top_p,
-    )
-    return response.choices[0].message.content
+    API_TOKEN = random.choice([os.getenv("HF_READ_TOKEN")])
+    headers = {"Authorization": f"Bearer {API_TOKEN}"}
+    
+    payload = {
+        "inputs": messages,
+        "seed": random.randint(1, 1000000000),
+        "parameters": {
+            "max_tokens": max_tokens,
+            "temperature": temperature,
+            "top_p": top_p
+        }
+    }
+
+    model = "https://api-inference.huggingface.co/models/" + model;
+    response = requests.post(model, headers=headers, json=payload, timeout=timeout)
+    
+    print(response.content)
+    return response.content
 
 def run_together_model(model, messages, max_tokens, temperature, top_p):
     client = Together()