oagi 0.4.0__py3-none-any.whl → 0.4.2__py3-none-any.whl

oagi/async_short_task.py

@@ -41,14 +41,14 @@ class AsyncShortTask(AsyncTask):
             logger.debug(f"Async auto mode step {i + 1}/{max_steps}")
             image = await image_provider()
             step = await self.step(image)
+            if executor:
+                logger.debug(f"Executing {len(step.actions)} actions asynchronously")
+                await executor(step.actions)
             if step.stop:
                 logger.info(
                     f"Async auto mode completed successfully after {i + 1} steps"
                 )
                 return True
-            if executor:
-                logger.debug(f"Executing {len(step.actions)} actions asynchronously")
-                await executor(step.actions)
 
         logger.warning(
             f"Async auto mode reached max steps ({max_steps}) without completion"

oagi/pyautogui_action_handler.py

@@ -15,6 +15,42 @@ from pydantic import BaseModel, Field
 from .types import Action, ActionType
 
 
+class CapsLockManager:
+    """Manages caps lock state for text transformation."""
+
+    def __init__(self, mode: str = "session"):
+        """Initialize caps lock manager.
+
+        Args:
+            mode: Either "session" (internal state) or "system" (OS-level)
+        """
+        self.mode = mode
+        self.caps_enabled = False
+
+    def toggle(self):
+        """Toggle caps lock state in session mode."""
+        if self.mode == "session":
+            self.caps_enabled = not self.caps_enabled
+
+    def transform_text(self, text: str) -> str:
+        """Transform text based on caps lock state.
+
+        Args:
+            text: Input text to transform
+
+        Returns:
+            Transformed text (uppercase if caps enabled in session mode)
+        """
+        if self.mode == "session" and self.caps_enabled:
+            # Transform letters to uppercase, preserve special characters
+            return "".join(c.upper() if c.isalpha() else c for c in text)
+        return text
+
+    def should_use_system_capslock(self) -> bool:
+        """Check if system-level caps lock should be used."""
+        return self.mode == "system"
+
+
 class PyautoguiConfig(BaseModel):
     """Configuration for PyautoguiActionHandler."""
 
@@ -30,6 +66,13 @@ class PyautoguiConfig(BaseModel):
     action_pause: float = Field(
         default=0.1, description="Pause between PyAutoGUI actions in seconds"
     )
+    hotkey_interval: float = Field(
+        default=0.1, description="Interval between key presses in hotkey combinations"
+    )
+    capslock_mode: str = Field(
+        default="session",
+        description="Caps lock handling mode: 'session' (internal state) or 'system' (OS-level)",
+    )
 
 
 class PyautoguiActionHandler:
@@ -54,6 +97,8 @@ class PyautoguiActionHandler:
         self.screen_width, self.screen_height = pyautogui.size()
         # Set default delay between actions
         pyautogui.PAUSE = self.config.action_pause
+        # Initialize caps lock manager
+        self.caps_manager = CapsLockManager(mode=self.config.capslock_mode)
 
     def _denormalize_coords(self, x: float, y: float) -> tuple[int, int]:
         """Convert coordinates from 0-1000 range to actual screen coordinates."""
@@ -94,12 +139,20 @@ class PyautoguiActionHandler:
         direction = match.group(3).lower()
         return x, y, direction
 
+    def _normalize_key(self, key: str) -> str:
+        """Normalize key names for consistency."""
+        key = key.strip().lower()
+        # Normalize caps lock variations
+        if key in ["caps_lock", "caps", "capslock"]:
+            return "capslock"
+        return key
+
     def _parse_hotkey(self, args_str: str) -> list[str]:
         """Parse hotkey string into list of keys."""
         # Remove parentheses if present
         args_str = args_str.strip("()")
         # Split by '+' to get individual keys
-        keys = [key.strip() for key in args_str.split("+")]
+        keys = [self._normalize_key(key) for key in args_str.split("+")]
         return keys
 
     def _execute_single_action(self, action: Action) -> None:
@@ -115,6 +168,10 @@ class PyautoguiActionHandler:
                 x, y = self._parse_coords(arg)
                 pyautogui.doubleClick(x, y)
 
+            case ActionType.LEFT_TRIPLE:
+                x, y = self._parse_coords(arg)
+                pyautogui.tripleClick(x, y)
+
             case ActionType.RIGHT_SINGLE:
                 x, y = self._parse_coords(arg)
                 pyautogui.rightClick(x, y)
@@ -128,11 +185,25 @@ class PyautoguiActionHandler:
 
             case ActionType.HOTKEY:
                 keys = self._parse_hotkey(arg)
-                pyautogui.hotkey(*keys)
+                # Check if this is a caps lock key press
+                if len(keys) == 1 and keys[0] == "capslock":
+                    if self.caps_manager.should_use_system_capslock():
+                        # System mode: use OS-level caps lock
+                        pyautogui.hotkey(
+                            "capslock", interval=self.config.hotkey_interval
+                        )
+                    else:
+                        # Session mode: toggle internal state
+                        self.caps_manager.toggle()
+                else:
+                    # Regular hotkey combination
+                    pyautogui.hotkey(*keys, interval=self.config.hotkey_interval)
 
             case ActionType.TYPE:
                 # Remove quotes if present
                 text = arg.strip("\"'")
+                # Apply caps lock transformation if needed
+                text = self.caps_manager.transform_text(text)
                 pyautogui.typewrite(text)
 
             case ActionType.SCROLL:

oagi/short_task.py

@@ -41,12 +41,12 @@ class ShortTask(Task):
             logger.debug(f"Auto mode step {i + 1}/{max_steps}")
             image = image_provider()
             step = self.step(image)
-            if step.stop:
-                logger.info(f"Auto mode completed successfully after {i + 1} steps")
-                return True
             if executor:
                 logger.debug(f"Executing {len(step.actions)} actions")
                 executor(step.actions)
+            if step.stop:
+                logger.info(f"Auto mode completed successfully after {i + 1} steps")
+                return True
 
         logger.warning(f"Auto mode reached max steps ({max_steps}) without completion")
         return False

oagi/types/models/action.py

@@ -14,6 +14,7 @@ from pydantic import BaseModel, Field
 class ActionType(str, Enum):
     CLICK = "click"
     LEFT_DOUBLE = "left_double"
+    LEFT_TRIPLE = "left_triple"
     RIGHT_SINGLE = "right_single"
     DRAG = "drag"
     HOTKEY = "hotkey"

{oagi-0.4.0.dist-info → oagi-0.4.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: oagi
-Version: 0.4.0
+Version: 0.4.2
 Summary: Official API of OpenAGI Foundation
 Project-URL: Homepage, https://github.com/agiopen-org/oagi
 Author-email: OpenAGI Foundation <contact@agiopen.org>
@@ -93,6 +93,8 @@ config = PyautoguiConfig(
     scroll_amount=50,       # Larger scroll steps (default: 30)
     wait_duration=2.0,      # Longer waits (default: 1.0)
     action_pause=0.2,       # More pause between actions (default: 0.1)
+    hotkey_interval=0.1,    # Interval between keys in hotkey combinations (default: 0.1)
+    capslock_mode="session" # Caps lock mode: 'session' or 'system' (default: 'session')
 )
 
 executor = PyautoguiActionHandler(config=config)

{oagi-0.4.0.dist-info → oagi-0.4.2.dist-info}/RECORD

@@ -2,15 +2,15 @@ oagi/__init__.py,sha256=m-Z121YCIwQOPXpTC8kd_UIJizcX8QuHyrSSguQ0KE0,2187
 oagi/async_client.py,sha256=oDj4kIdtaV37uopoAeClCFQTxrYRwHV2HwMAcMdVYwE,8455
 oagi/async_pyautogui_action_handler.py,sha256=F-lKyePCONWI03WnSxpX_QwxONbvnfdQu51wTod6mdw,1614
 oagi/async_screenshot_maker.py,sha256=pI-dbLcYOzcO1ffgTmozAdbYJQNBPKA7hmqj1RxEmIY,1688
-oagi/async_short_task.py,sha256=Pn3ggSWfPaxBPuMLox-WnIeD0JdmupxMojXybA8SbtE,1961
+oagi/async_short_task.py,sha256=jvFTbmXTxFlkpAwmWeZlxbTSv_RB7V561hxw6gUcigw,1961
 oagi/async_single_step.py,sha256=QawXO4GyfMz6O9jV8QBC1vKxFuS9vjKQxxJ1nwgHBzI,2838
 oagi/async_task.py,sha256=bclqtgg7mI2WAp-62jOz044tVk4wruycpn9NYDncnA8,4145
 oagi/exceptions.py,sha256=VMwVS8ouE9nHhBpN3AZMYt5_U2kGcihWaTnBhoQLquo,1662
 oagi/logging.py,sha256=CWe89mA5MKTipIvfrqSYkv2CAFNBSwHMDQMDkG_g64g,1350
 oagi/pil_image.py,sha256=Zp7YNwyE_AT25ZEFsWKbzMxbO8JOQsJ1Espph5ye8k8,3804
-oagi/pyautogui_action_handler.py,sha256=T6acM0ZVIOEC8TTI-Qeoxp-B7HHF9xZK7s0HWuGW564,6477
+oagi/pyautogui_action_handler.py,sha256=11-xObprbuJYjoLD1AKbR_4OTdjvXO5ixKaTi_I4gwA,9211
 oagi/screenshot_maker.py,sha256=sVuW7jn-K4FmLhmYI-akdNI-UVcTeBzh9P1_qJhoq1s,1282
-oagi/short_task.py,sha256=fJcirqD7X3_GyINTGdOoe6wi-VFHfP-C8m-zxCvgY5M,1779
+oagi/short_task.py,sha256=9l1PDX70vDUEX2CIJ66yaAtb96P3mK_m95JffspnYFI,1779
 oagi/single_step.py,sha256=djhGOHzA5Y3-9_ity9QiJr_ObZZ04blSmNZsLXXXfkg,2939
 oagi/sync_client.py,sha256=E6EgFIe-H91rdsPhF1puwrBTpOnKaL6JA1WHR4R-CLY,9395
 oagi/task.py,sha256=JfsugIhBrwDmi1xOEVQdqmXsGFK-H4p17-B4rM8kbWs,4001
@@ -21,10 +21,10 @@ oagi/types/async_image_provider.py,sha256=wnhRyPtTmuALt45Qore74-RCkP5yxU9sZGjvOz
 oagi/types/image.py,sha256=KgPCCTJ6D5vHIaGZdbTE7eQEa1WlT6G9tf59ZuUCV2U,537
 oagi/types/image_provider.py,sha256=oYFdOYznrK_VOR9egzOjw5wFM5w8EY2sY01pH0ANAgU,1112
 oagi/types/models/__init__.py,sha256=bVzzGxb6lVxAQyJpy0Z1QknSe-xC3g4OIDr7t-p_3Ys,467
-oagi/types/models/action.py,sha256=8Xd3IcH32ENq7uXczo-mbQ736yUOGxO_TaZTfHVRY7w,935
+oagi/types/models/action.py,sha256=hh6mRRSSWgrW4jpZo71zGMCOcZpV5_COu4148uG6G48,967
 oagi/types/models/image_config.py,sha256=tl6abVg_-IAPLwpaWprgknXu7wRWriMg-AEVyUX73v0,1567
 oagi/types/models/step.py,sha256=RSI4H_2rrUBq_xyCoWKaq7JHdJWNobtQppaKC1l0aWU,471
-oagi-0.4.0.dist-info/METADATA,sha256=haW2dK-avlwzfyw6WjWEm0GwwAEfChy-dhdE2ksk2qk,4620
-oagi-0.4.0.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
-oagi-0.4.0.dist-info/licenses/LICENSE,sha256=sy5DLA2M29jFT4UfWsuBF9BAr3FnRkYtnAu6oDZiIf8,1075
-oagi-0.4.0.dist-info/RECORD,,
+oagi-0.4.2.dist-info/METADATA,sha256=6s93KoSnU41E_vp-CVyPykBA2zxHrTatFs-Ru5OztBc,4799
+oagi-0.4.2.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
+oagi-0.4.2.dist-info/licenses/LICENSE,sha256=sy5DLA2M29jFT4UfWsuBF9BAr3FnRkYtnAu6oDZiIf8,1075
+oagi-0.4.2.dist-info/RECORD,,