Merge branch 'main' into zdy

main.py

@@ -1,59 +1,50 @@
-#from pprint import pprint
-from desktop_env.envs.desktop_env import DesktopEnv, Action, MouseClick
-
-def get_human_action():
-    """
-    Prompts the human player for an action and returns a structured action.
-    """
-    print("\nAvailable actions:", [action.name for action in Action])
-    action_type = None
-    while action_type not in [action.value for action in Action]:
-        action_type = Action[input("Enter the type of action: ".strip())].value
-
-    action = {"action_type": action_type}
-
-    if action_type == Action.CLICK.value or action_type == Action.MOUSE_DOWN.value or action_type == Action.MOUSE_UP.value:
-        print("\n Available clicks:", [action.name for action in MouseClick])
-        click_type = input("Enter click type: ")
-        action["click_type"] = MouseClick[click_type].value
-
-    if action_type == Action.MOUSE_MOVE.value:
-        x = int(input("Enter x-coordinate for mouse move: "))
-        y = int(input("Enter y-coordinate for mouse move: "))
-        action["x"] = x
-        action["y"] = y
-
-    if action_type == Action.KEY.value:
-        key = input("Enter the key to press: ")
-        action["key"] = [ord(c) for c in key]
-
-    if action_type == Action.TYPE.value:
-        text = input("Enter the text to type: ")
-        action["text"] = [ord(c) for c in text]
-
-    return action
+import json
+from desktop_env.envs.desktop_env import DesktopEnv
 
 
 def human_agent():
     """
     Runs the Gym environment with human input.
     """
+
+    with open("evaluation_examples/examples/37608790-6147-45d0-9f20-1137bb35703d.json", "r") as f:
+        example = json.load(f)
+
     #env = DesktopEnv( path_to_vm="/home/yuri/vmware/Windows 10 x64/Windows 10 x64.vmx"
                     #  path_to_vm="/home/yuri/vmware/Ubuntu 64-bit/Ubuntu 64-bit.vmx",
     env = DesktopEnv( path_to_vm="/home/david/vmware/KUbuntu 64-bit/KUbuntu 64-bit.vmx"
-                    , username="david"
-                    , password="123456"
-                    , host="192.168.174.129"
-                     #host="http://192.168.7.129:5000",
-                     #vm_os="windows")
-                    , vm_os="ubuntu"
+                    , action_space="computer_13"
+                    , snapshot_path="base_setup"
+                    , instruction=example["instruction"]
+                     #, config=example["config"]
+                     #, evaluator=example["evaluator"]
                     )
+
+    # reset the environment to certain snapshot
     observation = env.reset()
     done = False
 
-    while not done:
-        action = get_human_action()
-        observation, reward, done, info = env.step(action)
+    trajectory = [
+        {
+            "action_type": "MOVE_TO",
+            "parameters": {
+                "x": 754,
+                "y": 1057
+            }
+        },
+        {"action_type": "CLICK", "parameters": {"button": "right", "num_clicks": 1}}
+    ]
+
+    for i in range(len(trajectory)):
+        # action = get_human_action()
+
+        # action = {
+        #     "action_type": 0,
+        #     "click_type": 3,
+        # }
+        print(trajectory[i])
+
+        observation, reward, done, info = env.step(trajectory[i], pause=5)
         print("Observation:", observation)
         print("Reward:", reward)
         print("Info:", info)
@@ -64,8 +55,12 @@ def human_agent():
             print("The episode is done.")
             break
 
+    result = env.evaluate()
+    print("Result:", result)
+
     env.close()
     print("Environment closed.")
 
+
 if __name__ == "__main__":
     human_agent()