hyperbrowser 1.9.3__tar.gz → 1.10.0__tar.gz

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hyperbrowser
-Version: 1.9.3
+Version: 1.10.0
 Summary: Python SDK for hyperbrowser
 License: MIT
 License-File: LICENSE

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/client/managers/async_manager/computer_action.py

@@ -12,6 +12,8 @@ from hyperbrowser.models import (
     PressKeysActionParams,
     MoveMouseActionParams,
     ScreenshotActionParams,
+    CursorPositionActionParams,
+    ScrollAtCursorActionParams,
     ScrollActionParams,
     TypeTextActionParams,
     Coordinate,
@@ -38,6 +40,7 @@ _ACTION_PARAM_MODELS = {
     ComputerAction.MOVE_MOUSE.value: MoveMouseActionParams,
     ComputerAction.PRESS_KEYS.value: PressKeysActionParams,
     ComputerAction.SCREENSHOT.value: ScreenshotActionParams,
+    ComputerAction.CURSOR_POSITION.value: CursorPositionActionParams,
     ComputerAction.SCROLL.value: ScrollActionParams,
     ComputerAction.TYPE_TEXT.value: TypeTextActionParams,
     ComputerAction.GET_CLIPBOARD_TEXT.value: GetClipboardTextActionParams,
@@ -47,6 +50,9 @@ _ACTION_PARAM_MODELS = {
 
 
 def _action_param_model(params):
+    if isinstance(params, ScrollAtCursorActionParams):
+        return ScrollAtCursorActionParams
+
     for model in _ACTION_PARAM_MODELS.values():
         if isinstance(params, model):
             return model
@@ -56,6 +62,8 @@ def _action_param_model(params):
         if isinstance(action, ComputerAction):
             action = action.value
         model = _ACTION_PARAM_MODELS.get(action)
+        if model is ScrollActionParams and "x" not in params and "y" not in params:
+            return ScrollAtCursorActionParams
         if model is not None:
             return model
 
@@ -71,18 +79,13 @@ class ComputerActionManager:
         session: Union[SessionDetail, str],
         params: Union[ComputerActionParamsDict, ComputerActionParams],
     ) -> ComputerActionResponse:
+        payload = dump_request(params, _action_param_model(params), name="params")
         if isinstance(session, str):
             session = await self._client.sessions.get(session)
 
         if not session.computer_action_endpoint:
             raise ValueError("Computer action endpoint not available for this session")
 
-        payload = dump_request(
-            params,
-            _action_param_model(params),
-            name="params",
-        )
-
         response = await self._client.transport.post(
             session.computer_action_endpoint,
             data=payload,
@@ -97,12 +100,14 @@ class ComputerActionManager:
         button: ComputerActionMouseButton = "left",
         num_clicks: int = 1,
         return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
     ) -> ComputerActionResponse:
         params = ClickActionParams(
             x=x,
             y=y,
             button=button,
             num_clicks=num_clicks,
+            keys=keys,
             return_screenshot=return_screenshot,
         )
         return await self._execute_request(session, params)
@@ -123,6 +128,14 @@ class ComputerActionManager:
         params = ScreenshotActionParams()
         return await self._execute_request(session, params)
 
+    async def cursor_position(
+        self,
+        session: Union[SessionDetail, str],
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = CursorPositionActionParams(return_screenshot=return_screenshot)
+        return await self._execute_request(session, params)
+
     async def press_keys(
         self,
         session: Union[SessionDetail, str],
@@ -169,6 +182,7 @@ class ComputerActionManager:
         session: Union[SessionDetail, str],
         path: List[Union[CoordinateDict, Coordinate]],
         return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
     ) -> ComputerActionResponse:
         params = DragActionParams(
             path=[
@@ -176,6 +190,7 @@ class ComputerActionManager:
                 for coordinate in path
             ],
             return_screenshot=return_screenshot,
+            keys=keys,
         )
         return await self._execute_request(session, params)
 
@@ -197,12 +212,30 @@ class ComputerActionManager:
         scroll_x: int,
         scroll_y: int,
         return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
     ) -> ComputerActionResponse:
         params = ScrollActionParams(
             x=x,
             y=y,
             scroll_x=scroll_x,
             scroll_y=scroll_y,
+            keys=keys,
+            return_screenshot=return_screenshot,
+        )
+        return await self._execute_request(session, params)
+
+    async def scroll_at_cursor(
+        self,
+        session: Union[SessionDetail, str],
+        scroll_x: int = 0,
+        scroll_y: int = 0,
+        return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
+    ) -> ComputerActionResponse:
+        params = ScrollAtCursorActionParams(
+            scroll_x=scroll_x,
+            scroll_y=scroll_y,
+            keys=keys,
             return_screenshot=return_screenshot,
         )
         return await self._execute_request(session, params)

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/client/managers/sync_manager/computer_action.py

@@ -12,6 +12,8 @@ from hyperbrowser.models import (
     PressKeysActionParams,
     MoveMouseActionParams,
     ScreenshotActionParams,
+    CursorPositionActionParams,
+    ScrollAtCursorActionParams,
     ScrollActionParams,
     TypeTextActionParams,
     Coordinate,
@@ -38,6 +40,7 @@ _ACTION_PARAM_MODELS = {
     ComputerAction.MOVE_MOUSE.value: MoveMouseActionParams,
     ComputerAction.PRESS_KEYS.value: PressKeysActionParams,
     ComputerAction.SCREENSHOT.value: ScreenshotActionParams,
+    ComputerAction.CURSOR_POSITION.value: CursorPositionActionParams,
     ComputerAction.SCROLL.value: ScrollActionParams,
     ComputerAction.TYPE_TEXT.value: TypeTextActionParams,
     ComputerAction.GET_CLIPBOARD_TEXT.value: GetClipboardTextActionParams,
@@ -47,6 +50,9 @@ _ACTION_PARAM_MODELS = {
 
 
 def _action_param_model(params):
+    if isinstance(params, ScrollAtCursorActionParams):
+        return ScrollAtCursorActionParams
+
     for model in _ACTION_PARAM_MODELS.values():
         if isinstance(params, model):
             return model
@@ -56,6 +62,8 @@ def _action_param_model(params):
         if isinstance(action, ComputerAction):
             action = action.value
         model = _ACTION_PARAM_MODELS.get(action)
+        if model is ScrollActionParams and "x" not in params and "y" not in params:
+            return ScrollAtCursorActionParams
         if model is not None:
             return model
 
@@ -71,18 +79,13 @@ class ComputerActionManager:
         session: Union[SessionDetail, str],
         params: Union[ComputerActionParamsDict, ComputerActionParams],
     ) -> ComputerActionResponse:
+        payload = dump_request(params, _action_param_model(params), name="params")
         if isinstance(session, str):
             session = self._client.sessions.get(session)
 
         if not session.computer_action_endpoint:
             raise ValueError("Computer action endpoint not available for this session")
 
-        payload = dump_request(
-            params,
-            _action_param_model(params),
-            name="params",
-        )
-
         response = self._client.transport.post(
             session.computer_action_endpoint,
             data=payload,
@@ -97,12 +100,14 @@ class ComputerActionManager:
         button: ComputerActionMouseButton = "left",
         num_clicks: int = 1,
         return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
     ) -> ComputerActionResponse:
         params = ClickActionParams(
             x=x,
             y=y,
             button=button,
             num_clicks=num_clicks,
+            keys=keys,
             return_screenshot=return_screenshot,
         )
         return self._execute_request(session, params)
@@ -123,6 +128,14 @@ class ComputerActionManager:
         params = ScreenshotActionParams()
         return self._execute_request(session, params)
 
+    def cursor_position(
+        self,
+        session: Union[SessionDetail, str],
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = CursorPositionActionParams(return_screenshot=return_screenshot)
+        return self._execute_request(session, params)
+
     def press_keys(
         self,
         session: Union[SessionDetail, str],
@@ -169,6 +182,7 @@ class ComputerActionManager:
         session: Union[SessionDetail, str],
         path: List[Union[CoordinateDict, Coordinate]],
         return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
     ) -> ComputerActionResponse:
         params = DragActionParams(
             path=[
@@ -176,6 +190,7 @@ class ComputerActionManager:
                 for coordinate in path
             ],
             return_screenshot=return_screenshot,
+            keys=keys,
         )
         return self._execute_request(session, params)
 
@@ -197,12 +212,30 @@ class ComputerActionManager:
         scroll_x: int,
         scroll_y: int,
         return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
     ) -> ComputerActionResponse:
         params = ScrollActionParams(
             x=x,
             y=y,
             scroll_x=scroll_x,
             scroll_y=scroll_y,
+            keys=keys,
+            return_screenshot=return_screenshot,
+        )
+        return self._execute_request(session, params)
+
+    def scroll_at_cursor(
+        self,
+        session: Union[SessionDetail, str],
+        scroll_x: int = 0,
+        scroll_y: int = 0,
+        return_screenshot: bool = False,
+        keys: Optional[List[str]] = None,
+    ) -> ComputerActionResponse:
+        params = ScrollAtCursorActionParams(
+            scroll_x=scroll_x,
+            scroll_y=scroll_y,
+            keys=keys,
             return_screenshot=return_screenshot,
         )
         return self._execute_request(session, params)

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/models/__init__.py

@@ -274,6 +274,8 @@ from .computer_action import (
     MoveMouseActionParams,
     PressKeysActionParams,
     ScreenshotActionParams,
+    CursorPositionActionParams,
+    ScrollAtCursorActionParams,
     ScrollActionParams,
     TypeTextActionParams,
     ComputerActionMouseButton,
@@ -283,6 +285,7 @@ from .computer_action import (
     ComputerActionWindow,
     ComputerActionResponseDataClipboardText,
     ComputerActionResponseDataListWindows,
+    ComputerActionResponseDataCursorPosition,
     ComputerActionResponseData,
 )
 from .session import (
@@ -779,6 +782,8 @@ __all__ = [
     "MoveMouseActionParams",
     "PressKeysActionParams",
     "ScreenshotActionParams",
+    "CursorPositionActionParams",
+    "ScrollAtCursorActionParams",
     "ScrollActionParams",
     "TypeTextActionParams",
     "ComputerActionMouseButton",
@@ -788,6 +793,7 @@ __all__ = [
     "ComputerActionWindow",
     "ComputerActionResponseDataClipboardText",
     "ComputerActionResponseDataListWindows",
+    "ComputerActionResponseDataCursorPosition",
     "ComputerActionResponseData",
     # web
     "StartBatchFetchJobParams",

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/models/computer_action.py

@@ -14,6 +14,7 @@ class ComputerAction(str, Enum):
     MOVE_MOUSE = "move_mouse"
     PRESS_KEYS = "press_keys"
     SCREENSHOT = "screenshot"
+    CURSOR_POSITION = "cursor_position"
     SCROLL = "scroll"
     TYPE_TEXT = "type_text"
     GET_CLIPBOARD_TEXT = "get_clipboard_text"
@@ -43,6 +44,7 @@ class ClickActionParams(BaseModel):
     y: Optional[int] = Field(default=None)
     button: ComputerActionMouseButton = Field(default="left")
     num_clicks: int = Field(serialization_alias="numClicks", default=1)
+    keys: Optional[List[str]] = None
     return_screenshot: bool = Field(
         serialization_alias="returnScreenshot", default=False
     )
@@ -55,6 +57,7 @@ class DragActionParams(BaseModel):
 
     action: Literal[ComputerAction.DRAG] = ComputerAction.DRAG
     path: List[Coordinate]
+    keys: Optional[List[str]] = None
     return_screenshot: bool = Field(
         serialization_alias="returnScreenshot", default=False
     )
@@ -130,6 +133,17 @@ class ScreenshotActionParams(BaseModel):
     action: Literal[ComputerAction.SCREENSHOT] = ComputerAction.SCREENSHOT
 
 
+class CursorPositionActionParams(BaseModel):
+    """Parameters for reading the desktop cursor position."""
+
+    model_config = ConfigDict(use_enum_values=True)
+
+    action: Literal[ComputerAction.CURSOR_POSITION] = ComputerAction.CURSOR_POSITION
+    return_screenshot: bool = Field(
+        serialization_alias="returnScreenshot", default=False
+    )
+
+
 class ScrollActionParams(BaseModel):
     """Parameters for scroll action."""
 
@@ -138,8 +152,23 @@ class ScrollActionParams(BaseModel):
     action: Literal[ComputerAction.SCROLL] = ComputerAction.SCROLL
     x: int
     y: int
+    keys: Optional[List[str]] = None
+    scroll_x: int = Field(serialization_alias="scrollX")
+    scroll_y: int = Field(serialization_alias="scrollY")
+    return_screenshot: bool = Field(
+        serialization_alias="returnScreenshot", default=False
+    )
+
+
+class ScrollAtCursorActionParams(BaseModel):
+    """Parameters for scrolling without moving the desktop cursor."""
+
+    model_config = ConfigDict(use_enum_values=True)
+
+    action: Literal[ComputerAction.SCROLL] = ComputerAction.SCROLL
     scroll_x: int = Field(serialization_alias="scrollX")
     scroll_y: int = Field(serialization_alias="scrollY")
+    keys: Optional[List[str]] = None
     return_screenshot: bool = Field(
         serialization_alias="returnScreenshot", default=False
     )
@@ -201,7 +230,9 @@ ComputerActionParams = Union[
     PressKeysActionParams,
     MoveMouseActionParams,
     ScreenshotActionParams,
+    CursorPositionActionParams,
     ScrollActionParams,
+    ScrollAtCursorActionParams,
     TypeTextActionParams,
     HoldKeyActionParams,
     MouseDownActionParams,
@@ -237,7 +268,15 @@ class ComputerActionResponseDataListWindows(BaseModel):
     windows: List[ComputerActionWindow] = Field(default_factory=list)
 
 
+class ComputerActionResponseDataCursorPosition(BaseModel):
+    """Desktop cursor coordinates in physical screen pixels."""
+
+    x: int
+    y: int
+
+
 ComputerActionResponseData = Union[
+    ComputerActionResponseDataCursorPosition,
     ComputerActionResponseDataListWindows,
     ComputerActionResponseDataClipboardText,
 ]

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/models/consts.py

@@ -39,6 +39,7 @@ BrowserUseLlm = Literal[
     "claude-sonnet-5",
     "claude-sonnet-4-6",
     "claude-sonnet-4-5",
+    "claude-haiku-5-5",
     "claude-sonnet-4-20250514",
     "claude-3-7-sonnet-20250219",
     "claude-3-5-sonnet-20241022",
@@ -62,6 +63,7 @@ HyperAgentLlm = Literal[
     "claude-sonnet-5",
     "claude-sonnet-4-6",
     "claude-sonnet-4-5",
+    "claude-haiku-5-5",
     "gemini-2.5-flash",
     "gemini-3.8-flash",
     "gemini-3.7-flash",
@@ -75,6 +77,7 @@ ClaudeComputerUseLlm = Literal[
     "claude-opus-4-6",
     "claude-opus-4-7",
     "claude-opus-4-8",
+    "claude-haiku-5-5",
     "claude-haiku-4-5-20251001",
     "claude-sonnet-5-5",
     "claude-sonnet-5",

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/types/__init__.py

@@ -32,6 +32,8 @@ from .agents import (
     StartMetaComputerUseTaskParams,
 )
 from .computer_action import (
+    CursorPositionActionParams,
+    ScrollAtCursorActionParams,
     ClickActionParams,
     ComputerActionMouseButton,
     ComputerActionParams,
@@ -212,6 +214,8 @@ __all__ = [
     "PressKeysActionParams",
     "PutSelectionTextActionParams",
     "ScreenshotActionParams",
+    "CursorPositionActionParams",
+    "ScrollAtCursorActionParams",
     "ScrollActionParams",
     "TypeTextActionParams",
     "GetCrawlJobParams",

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/hyperbrowser/types/computer_action.py

@@ -23,6 +23,7 @@ class ClickActionParams(TypedDict, total=False):
     y: Optional[int]
     button: ComputerActionMouseButton
     num_clicks: int
+    keys: List[str]
     return_screenshot: bool
 
 
@@ -31,6 +32,7 @@ class DragActionParams(TypedDict, total=False):
 
     action: Required[Literal["drag"]]
     path: Required[List[Coordinate]]
+    keys: List[str]
     return_screenshot: bool
 
 
@@ -82,14 +84,32 @@ class ScreenshotActionParams(TypedDict):
     action: Literal["screenshot"]
 
 
+class CursorPositionActionParams(TypedDict, total=False):
+    """Parameters for reading the desktop cursor position."""
+
+    action: Required[Literal["cursor_position"]]
+    return_screenshot: bool
+
+
 class ScrollActionParams(TypedDict, total=False):
     """Parameters for scrolling at a screen coordinate."""
 
     action: Required[Literal["scroll"]]
     x: Required[int]
     y: Required[int]
+    keys: List[str]
+    scroll_x: Required[int]
+    scroll_y: Required[int]
+    return_screenshot: bool
+
+
+class ScrollAtCursorActionParams(TypedDict, total=False):
+    """Parameters for scrolling without moving the desktop cursor."""
+
+    action: Required[Literal["scroll"]]
     scroll_x: Required[int]
     scroll_y: Required[int]
+    keys: List[str]
     return_screenshot: bool
 
 
@@ -129,7 +149,9 @@ ComputerActionParams: TypeAlias = Union[
     PressKeysActionParams,
     MoveMouseActionParams,
     ScreenshotActionParams,
+    CursorPositionActionParams,
     ScrollActionParams,
+    ScrollAtCursorActionParams,
     TypeTextActionParams,
     HoldKeyActionParams,
     MouseDownActionParams,
@@ -145,6 +167,7 @@ __all__ = [
     "ComputerActionMouseButton",
     "ComputerActionParams",
     "Coordinate",
+    "CursorPositionActionParams",
     "DragActionParams",
     "GetClipboardTextActionParams",
     "HoldKeyActionParams",
@@ -156,5 +179,6 @@ __all__ = [
     "PutSelectionTextActionParams",
     "ScreenshotActionParams",
     "ScrollActionParams",
+    "ScrollAtCursorActionParams",
     "TypeTextActionParams",
 ]

{hyperbrowser-1.9.3 → hyperbrowser-1.10.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "hyperbrowser"
-version = "1.9.3"
+version = "1.10.0"
 description = "Python SDK for hyperbrowser"
 authors = ["Nikhil Shahi <nshahi1998@gmail.com>"]
 license = "MIT"