Fix the implementation of action 13 of computer

desktop_env/envs/desktop_env.py

@@ -25,6 +25,7 @@ class DesktopEnv(gym.Env):
             path_to_vm: str,
             host: str = "192.168.7.128:5000",
             snapshot_path: str = "base",
+            action_space: str = "pyautogui",
     ):
         # Initialize environment variables
         self.path_to_vm = path_to_vm
@@ -35,7 +36,11 @@ class DesktopEnv(gym.Env):
         print("Initializing...")
         self._start_emulator()
         self.controller = PythonController(http_server=self.host)
-        # todo: define the action space and the observation space as gym did
+
+        # mode: human or machine
+        assert action_space in ["computer_13", "pyautogui"]
+        self.action_space = action_space
+        # todo: define the action space and the observation space as gym did, or extend theirs
 
     def _start_emulator(self):
         while True:
@@ -86,9 +91,12 @@ class DesktopEnv(gym.Env):
         return observation
 
     def step(self, action, pause=0.5):
-        # todo: support both the action space of our-designed space and the executable code space in pyautogui
-        # Our action space is the set of all possible python commands insides `pyautogui`
-        self.controller.execute_python_command(action)
+        if self.action_space == "computer_13":
+            # 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)
 
         # todo: maybe for the better here we need to add a logic to wait until the rendering is done
         time.sleep(pause)