Add gemini agent implementation; Add missed requirements; Minor fix some small bugs

desktop_env/envs/desktop_env.py

@@ -231,8 +231,11 @@ class DesktopEnv(gym.Env):
             # the set of all possible actions defined in the action representation
             self.controller.execute_action(action)
         elif self.action_space == "pyautogui":
-            # the set of all possible python commands insides `pyautogui`
-            self.controller.execute_python_command(action)
+            if action in ['WAIT', 'FAIL', 'DONE']:
+                self.controller.execute_action(action)
+            else:
+                # the set of all possible python commands insides `pyautogui`
+                self.controller.execute_python_command(action)
 
         observation = {
             "screenshot": self._get_obs(),