fix pyautogui resolution scaling issue, tool input get coordinate issue

main.py

@@ -6,7 +6,6 @@ import pyautogui
 import subprocess
 from io import BytesIO
 from PIL import Image
-import quartz_doubleclick
 
 ANTHROPIC_API_URL = "https://api.anthropic.com/v1/messages"
 MODEL = "claude-3-7-sonnet-20250219"
@@ -104,8 +103,7 @@ def execute_computer_tool(tool_input):
             return {"type": "text", "text": "Error: Missing coordinates for double click action"}
         
         # Perform double click
-        # pyautogui.doubleClick(x, y, interval=0.2) this doesn't work
-        quartz_doubleclick.double_click(x, y)
+        pyautogui.doubleClick(x, y, interval=0.2)
         return {"type": "text", "text": f"Double-clicked at coordinates ({x}, {y})"}
     
     elif action == "type":
@@ -116,17 +114,12 @@ def execute_computer_tool(tool_input):
     
     elif action == "key":
         # Press a key or key combination
-        text = tool_input.get("text", "")
+        key = tool_input.get("key", "")
         try:
-            if '+' in text:
-                # Handle key combinations like "command+c"
-                keys = text.replace('super', 'command').split('+')
-                pyautogui.hotkey(*keys, interval=0.05) # interval is required
-            else:
-                pyautogui.press(text)
-            return {"type": "text", "text": f"Pressed key: {text}"}
+            pyautogui.press(key)
+            return {"type": "text", "text": f"Pressed key: {key}"}
         except Exception as e:
-            return {"type": "text", "text": f"Error pressing key {text}: {str(e)}"}
+            return {"type": "text", "text": f"Error pressing key {key}: {str(e)}"}
     
     elif action == "scroll":
         # Scroll action

quartz_doubleclick.py

@@ -1,39 +0,0 @@
-'''
-This script simulates a double-click at the given mouse cursor position.
-'''
-
-import Quartz
-from time import sleep
-
-def post_mouse_event(type, pos, click_state):
-    event = Quartz.CGEventCreateMouseEvent(
-        None, type, pos, Quartz.kCGMouseButtonLeft
-    )
-    Quartz.CGEventSetIntegerValueField(event, Quartz.kCGMouseEventClickState, click_state)
-    Quartz.CGEventPost(Quartz.kCGHIDEventTap, event)
-
-def double_click (x, y):
-    pos = None
-    if x is None or y is None:
-        # Get current mouse position if no coordinates are provided
-        loc = Quartz.CGEventGetLocation(Quartz.CGEventCreate(None))
-        pos = (loc.x, loc.y)
-    else:
-        # Use provided coordinates
-        pos = (x, y)
-
-    # First click
-    post_mouse_event(Quartz.kCGEventLeftMouseDown, pos, 1)
-    post_mouse_event(Quartz.kCGEventLeftMouseUp, pos, 1)
-
-    sleep(0.05)  # Short delay within double-click threshold
-
-    # Second click
-    post_mouse_event(Quartz.kCGEventLeftMouseDown, pos, 2)
-    post_mouse_event(Quartz.kCGEventLeftMouseUp, pos, 2)
-
-
-if __name__ == "__main__":
-    print('test: double clicking at cursor position in 3 seconds...')
-    sleep(3)
-    double_click()