Improve on eval script on web browsing tasks; Add one setup example

desktop_env/envs/desktop_env.py

@@ -316,6 +316,9 @@ class DesktopEnv(gym.Env):
                 return 1
             else:
                 return 0
+        else:
+            if len(self.action_history) > 0 and self.action_history[-1] == "FAIL":
+                return 0
 
         if type(self.metric) == list:
             results = []

desktop_env/evaluators/metrics/chrome.py

@@ -17,7 +17,7 @@ def is_expected_active_tab(active_tab_info: Dict[str, str], rule: Dict[str, Any]
 
     if match_type == "url":
         expected_url = rule['url']
-        actual_url = active_tab_info['url']
+        actual_url = active_tab_info.get('url', None)
         print("expected_url: {}".format(expected_url))
         print("actual_url: {}".format(actual_url))
         return 1 if compare_urls(expected_url, actual_url) else 0

desktop_env/evaluators/metrics/utils.py

@@ -625,6 +625,9 @@ def are_lists_equal(list1, list2, comparison_func):
 
 
 def compare_urls(url1, url2):
+    if url1 is None or url2 is None:
+        return url1 == url2
+
     def normalize_url(url):
         # Parse the URL
         parsed_url = urlparse(url)