diff --git a/omnitool/gradio/app.py b/omnitool/gradio/app.py
index 54cca8a0..4b85b4cc 100644
--- a/omnitool/gradio/app.py
+++ b/omnitool/gradio/app.py
@@ -22,9 +22,11 @@
 import requests
 from requests.exceptions import RequestException
 import base64
+import glob
 
 CONFIG_DIR = Path("~/.anthropic").expanduser()
 API_KEY_FILE = CONFIG_DIR / "api_key"
+IMG_DIR = "./tmp/outputs/"
 
 INTRO_TEXT = '''
 OmniParser lets you turn any vision-langauge model into an AI agent. We currently support **OpenAI (4o/o1/o3-mini), DeepSeek (R1), Qwen (2.5VL) or Anthropic Computer Use (Sonnet).**
@@ -192,8 +194,20 @@ def valid_params(user_input, state):
     
     for server_name, url in [('Windows Host', 'localhost:5000'), ('OmniParser Server', args.omniparser_server_url)]:
         try:
-            url = f'http://{url}/probe'
-            response = requests.get(url, timeout=3)
+            if server_name == "OmniParser Server":
+                try:
+                    url_test = f'http://{url}/probe'
+                    response = requests.get(url_test, timeout=3)
+                    if response.status_code != 200:
+                        errors.append(f"{server_name} is not responding")
+                except RequestException:
+                    url_test = f'https://{url}/probe/'
+                    response = requests.get(url_test, timeout=3)
+                    if response.status_code != 200:
+                        errors.append(f"{server_name} is not responding")
+            else:
+                url = f'http://{url}/probe'
+                response = requests.get(url, timeout=3)
             if response.status_code != 200:
                 errors.append(f"{server_name} is not responding")
         except RequestException as e:
@@ -202,6 +216,9 @@ def valid_params(user_input, state):
     if not state["api_key"].strip():
         errors.append("LLM API Key is not set")
 
+    if state["provider"] == "azure" and not state.get("azure_resource_name", "").strip():
+        errors.append("Azure Resource Name is required when using Azure OpenAI")
+
     if not user_input:
         errors.append("no computer use request provided")
     
@@ -405,7 +422,14 @@ def update_api_key(api_key_value, state):
         state["api_key"] = api_key_value
         state[f'{state["provider"]}_api_key'] = api_key_value
 
+    def clear_img_cache():
+        files = glob.glob(IMG_DIR+'screenshot_*.png')
+        for f in files:
+            os.remove(f)
+
     def clear_chat(state):
+        # Clear images in the cache
+        clear_img_cache()
         # Reset message-related state
         state["messages"] = []
         state["responses"] = {}
diff --git a/omnitool/gradio/loop.py b/omnitool/gradio/loop.py
index 9ce63169..4edafea4 100644
--- a/omnitool/gradio/loop.py
+++ b/omnitool/gradio/loop.py
@@ -20,6 +20,8 @@
 from agent.vlm_agent import VLMAgent
 from agent.vlm_agent_with_orchestrator import VLMOrchestratedAgent
 from executor.anthropic_executor import AnthropicExecutor
+import requests
+from requests.exceptions import RequestException
 
 BETA_FLAG = "computer-use-2024-10-22"
 
@@ -55,7 +57,19 @@ def sampling_loop_sync(
     Synchronous agentic sampling loop for the assistant/tool interaction of computer use.
     """
     print('in sampling_loop_sync, model:', model)
-    omniparser_client = OmniParserClient(url=f"http://{omniparser_url}/parse/")
+    
+    # Try https with a trailing slash if http fails
+    url = omniparser_url
+    try:
+        url_test1 = f'http://{url}/parse'
+        requests.get(url_test1, timeout=3)
+        url = url_test1
+    except RequestException:
+        url_test2 = f'https://{url}/parse/'
+        requests.get(url_test2, timeout=3)
+        url = url_test2
+
+    omniparser_client = OmniParserClient(url=url)
     if model == "claude-3-5-sonnet-20241022":
         # Register Actor and Executor
         actor = AnthropicActor(
diff --git a/requirements.txt b/requirements.txt
index 901a27fa..95cfb08f 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -3,7 +3,7 @@ easyocr
 torchvision
 supervision==0.18.0
 openai==1.3.5
-transformers
+transformers==4.45.0
 ultralytics==8.3.70
 azure-identity
 numpy==1.26.4