Update app.py

app.py

@@ -56,19 +56,15 @@ def get_messages_api(messages):
     return json.dumps(messages, indent=4)
 
 def run_huggingface_model(model, messages, max_tokens, temperature, top_p):
-    model = "https://api-inference.huggingface.co/models/" + model;
-    headers = {};
-    
-    response = requests.post(model, headers=headers, json=messages, timeout=60)
-    if response.status_code != 200:
-        print(f"Error: Failed to get image. Response status: {response.status_code}")
-        print(f"Response content: {response.text}")
-        if response.status_code == 503:
-            raise gr.Error(f"{response.status_code} : The model is being loaded")
-        raise gr.Error(f"{response.status_code}")
-    else:
-        print(response.content)
-    return response.content
+    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
 
 def run_together_model(model, messages, max_tokens, temperature, top_p):
     client = Together()