hyperbrowser 0.51.0__py3-none-any.whl → 0.73.0__py3-none-any.whl

hyperbrowser/client/managers/async_manager/agents/__init__.py

@@ -2,6 +2,7 @@ from .browser_use import BrowserUseManager
 from .cua import CuaManager
 from .claude_computer_use import ClaudeComputerUseManager
 from .hyper_agent import HyperAgentManager
+from .gemini_computer_use import GeminiComputerUseManager
 
 
 class Agents:
@@ -10,3 +11,4 @@ class Agents:
         self.cua = CuaManager(client)
         self.claude_computer_use = ClaudeComputerUseManager(client)
         self.hyper_agent = HyperAgentManager(client)
+        self.gemini_computer_use = GeminiComputerUseManager(client)

hyperbrowser/client/managers/async_manager/agents/gemini_computer_use.py

@@ -0,0 +1,71 @@
+import asyncio
+
+from hyperbrowser.exceptions import HyperbrowserError
+
+from .....models import (
+    POLLING_ATTEMPTS,
+    BasicResponse,
+    GeminiComputerUseTaskResponse,
+    GeminiComputerUseTaskStatusResponse,
+    StartGeminiComputerUseTaskParams,
+    StartGeminiComputerUseTaskResponse,
+)
+
+
+class GeminiComputerUseManager:
+    def __init__(self, client):
+        self._client = client
+
+    async def start(
+        self, params: StartGeminiComputerUseTaskParams
+    ) -> StartGeminiComputerUseTaskResponse:
+        response = await self._client.transport.post(
+            self._client._build_url("/task/gemini-computer-use"),
+            data=params.model_dump(exclude_none=True, by_alias=True),
+        )
+        return StartGeminiComputerUseTaskResponse(**response.data)
+
+    async def get(self, job_id: str) -> GeminiComputerUseTaskResponse:
+        response = await self._client.transport.get(
+            self._client._build_url(f"/task/gemini-computer-use/{job_id}")
+        )
+        return GeminiComputerUseTaskResponse(**response.data)
+
+    async def get_status(self, job_id: str) -> GeminiComputerUseTaskStatusResponse:
+        response = await self._client.transport.get(
+            self._client._build_url(f"/task/gemini-computer-use/{job_id}/status")
+        )
+        return GeminiComputerUseTaskStatusResponse(**response.data)
+
+    async def stop(self, job_id: str) -> BasicResponse:
+        response = await self._client.transport.put(
+            self._client._build_url(f"/task/gemini-computer-use/{job_id}/stop")
+        )
+        return BasicResponse(**response.data)
+
+    async def start_and_wait(
+        self, params: StartGeminiComputerUseTaskParams
+    ) -> GeminiComputerUseTaskResponse:
+        job_start_resp = await self.start(params)
+        job_id = job_start_resp.job_id
+        if not job_id:
+            raise HyperbrowserError("Failed to start Gemini Computer Use task job")
+
+        failures = 0
+        while True:
+            try:
+                job_response = await self.get_status(job_id)
+                if (
+                    job_response.status == "completed"
+                    or job_response.status == "failed"
+                    or job_response.status == "stopped"
+                ):
+                    return await self.get(job_id)
+                failures = 0
+            except Exception as e:
+                failures += 1
+                if failures >= POLLING_ATTEMPTS:
+                    raise HyperbrowserError(
+                        f"Failed to poll Gemini Computer Use task job {job_id} after {POLLING_ATTEMPTS} attempts: {e}"
+                    )
+            await asyncio.sleep(2)

hyperbrowser/client/managers/async_manager/computer_action.py

@@ -1,5 +1,5 @@
 from pydantic import BaseModel
-from typing import Union, List
+from typing import Union, List, Optional
 from hyperbrowser.models import (
     SessionDetail,
     ComputerActionParams,
@@ -12,6 +12,11 @@ from hyperbrowser.models import (
     ScrollActionParams,
     TypeTextActionParams,
     Coordinate,
+    HoldKeyActionParams,
+    MouseDownActionParams,
+    MouseUpActionParams,
+    ComputerActionMouseButton,
+    GetClipboardTextActionParams,
 )
 
 
@@ -42,9 +47,9 @@ class ComputerActionManager:
     async def click(
         self,
         session: Union[SessionDetail, str],
-        x: int,
-        y: int,
-        button: str = "left",
+        x: Optional[int] = None,
+        y: Optional[int] = None,
+        button: ComputerActionMouseButton = "left",
         num_clicks: int = 1,
         return_screenshot: bool = False,
     ) -> ComputerActionResponse:
@@ -82,6 +87,38 @@ class ComputerActionManager:
         params = PressKeysActionParams(keys=keys, return_screenshot=return_screenshot)
         return await self._execute_request(session, params)
 
+    async def hold_key(
+        self,
+        session: Union[SessionDetail, str],
+        key: str,
+        duration: int,
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = HoldKeyActionParams(
+            key=key, duration=duration, return_screenshot=return_screenshot
+        )
+        return await self._execute_request(session, params)
+
+    async def mouse_down(
+        self,
+        session: Union[SessionDetail, str],
+        button: ComputerActionMouseButton = "left",
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = MouseDownActionParams(
+            button=button, return_screenshot=return_screenshot
+        )
+        return await self._execute_request(session, params)
+
+    async def mouse_up(
+        self,
+        session: Union[SessionDetail, str],
+        button: ComputerActionMouseButton = "left",
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = MouseUpActionParams(button=button, return_screenshot=return_screenshot)
+        return await self._execute_request(session, params)
+
     async def drag(
         self,
         session: Union[SessionDetail, str],
@@ -118,3 +155,11 @@ class ComputerActionManager:
             return_screenshot=return_screenshot,
         )
         return await self._execute_request(session, params)
+
+    async def get_clipboard_text(
+        self,
+        session: Union[SessionDetail, str],
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = GetClipboardTextActionParams(return_screenshot=return_screenshot)
+        return await self._execute_request(session, params)

hyperbrowser/client/managers/async_manager/session.py

@@ -10,12 +10,32 @@ from ....models.session import (
     SessionListResponse,
     SessionRecording,
     UploadFileResponse,
+    SessionEventLogListParams,
+    SessionEventLogListResponse,
+    SessionEventLog,
 )
 
 
+class SessionEventLogsManager:
+    def __init__(self, client):
+        self._client = client
+
+    async def list(
+        self,
+        session_id: str,
+        params: SessionEventLogListParams = SessionEventLogListParams(),
+    ) -> List[SessionEventLog]:
+        response = await self._client.transport.get(
+            self._client._build_url(f"/session/{session_id}/event-logs"),
+            params=params.model_dump(exclude_none=True, by_alias=True),
+        )
+        return SessionEventLogListResponse(**response.data)
+
+
 class SessionManager:
     def __init__(self, client):
         self._client = client
+        self.event_logs = SessionEventLogsManager(client)
 
     async def create(self, params: CreateSessionParams = None) -> SessionDetail:
         response = await self._client.transport.post(

hyperbrowser/client/managers/sync_manager/agents/__init__.py

@@ -2,6 +2,7 @@ from .browser_use import BrowserUseManager
 from .cua import CuaManager
 from .claude_computer_use import ClaudeComputerUseManager
 from .hyper_agent import HyperAgentManager
+from .gemini_computer_use import GeminiComputerUseManager
 
 
 class Agents:
@@ -10,3 +11,4 @@ class Agents:
         self.cua = CuaManager(client)
         self.claude_computer_use = ClaudeComputerUseManager(client)
         self.hyper_agent = HyperAgentManager(client)
+        self.gemini_computer_use = GeminiComputerUseManager(client)

hyperbrowser/client/managers/sync_manager/agents/claude_computer_use.py

@@ -49,7 +49,7 @@ class ClaudeComputerUseManager:
         job_start_resp = self.start(params)
         job_id = job_start_resp.job_id
         if not job_id:
-            raise HyperbrowserError("Failed to start CUA task job")
+            raise HyperbrowserError("Failed to start Claude Computer Use task job")
 
         failures = 0
         while True:

hyperbrowser/client/managers/sync_manager/agents/gemini_computer_use.py

@@ -0,0 +1,71 @@
+import time
+
+from hyperbrowser.exceptions import HyperbrowserError
+
+from .....models import (
+    POLLING_ATTEMPTS,
+    BasicResponse,
+    GeminiComputerUseTaskResponse,
+    GeminiComputerUseTaskStatusResponse,
+    StartGeminiComputerUseTaskParams,
+    StartGeminiComputerUseTaskResponse,
+)
+
+
+class GeminiComputerUseManager:
+    def __init__(self, client):
+        self._client = client
+
+    def start(
+        self, params: StartGeminiComputerUseTaskParams
+    ) -> StartGeminiComputerUseTaskResponse:
+        response = self._client.transport.post(
+            self._client._build_url("/task/gemini-computer-use"),
+            data=params.model_dump(exclude_none=True, by_alias=True),
+        )
+        return StartGeminiComputerUseTaskResponse(**response.data)
+
+    def get(self, job_id: str) -> GeminiComputerUseTaskResponse:
+        response = self._client.transport.get(
+            self._client._build_url(f"/task/gemini-computer-use/{job_id}")
+        )
+        return GeminiComputerUseTaskResponse(**response.data)
+
+    def get_status(self, job_id: str) -> GeminiComputerUseTaskStatusResponse:
+        response = self._client.transport.get(
+            self._client._build_url(f"/task/gemini-computer-use/{job_id}/status")
+        )
+        return GeminiComputerUseTaskStatusResponse(**response.data)
+
+    def stop(self, job_id: str) -> BasicResponse:
+        response = self._client.transport.put(
+            self._client._build_url(f"/task/gemini-computer-use/{job_id}/stop")
+        )
+        return BasicResponse(**response.data)
+
+    def start_and_wait(
+        self, params: StartGeminiComputerUseTaskParams
+    ) -> GeminiComputerUseTaskResponse:
+        job_start_resp = self.start(params)
+        job_id = job_start_resp.job_id
+        if not job_id:
+            raise HyperbrowserError("Failed to start Gemini Computer Use task job")
+
+        failures = 0
+        while True:
+            try:
+                job_response = self.get_status(job_id)
+                if (
+                    job_response.status == "completed"
+                    or job_response.status == "failed"
+                    or job_response.status == "stopped"
+                ):
+                    return self.get(job_id)
+                failures = 0
+            except Exception as e:
+                failures += 1
+                if failures >= POLLING_ATTEMPTS:
+                    raise HyperbrowserError(
+                        f"Failed to poll Gemini Computer Use task job {job_id} after {POLLING_ATTEMPTS} attempts: {e}"
+                    )
+            time.sleep(2)

hyperbrowser/client/managers/sync_manager/computer_action.py

@@ -1,5 +1,5 @@
 from pydantic import BaseModel
-from typing import Union, List
+from typing import Union, List, Optional
 from hyperbrowser.models import (
     SessionDetail,
     ComputerActionParams,
@@ -12,6 +12,11 @@ from hyperbrowser.models import (
     ScrollActionParams,
     TypeTextActionParams,
     Coordinate,
+    HoldKeyActionParams,
+    MouseDownActionParams,
+    MouseUpActionParams,
+    ComputerActionMouseButton,
+    GetClipboardTextActionParams,
 )
 
 
@@ -42,9 +47,9 @@ class ComputerActionManager:
     def click(
         self,
         session: Union[SessionDetail, str],
-        x: int,
-        y: int,
-        button: str = "left",
+        x: Optional[int] = None,
+        y: Optional[int] = None,
+        button: ComputerActionMouseButton = "left",
         num_clicks: int = 1,
         return_screenshot: bool = False,
     ) -> ComputerActionResponse:
@@ -82,6 +87,38 @@ class ComputerActionManager:
         params = PressKeysActionParams(keys=keys, return_screenshot=return_screenshot)
         return self._execute_request(session, params)
 
+    def hold_key(
+        self,
+        session: Union[SessionDetail, str],
+        key: str,
+        duration: int,
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = HoldKeyActionParams(
+            key=key, duration=duration, return_screenshot=return_screenshot
+        )
+        return self._execute_request(session, params)
+
+    def mouse_down(
+        self,
+        session: Union[SessionDetail, str],
+        button: ComputerActionMouseButton = "left",
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = MouseDownActionParams(
+            button=button, return_screenshot=return_screenshot
+        )
+        return self._execute_request(session, params)
+
+    def mouse_up(
+        self,
+        session: Union[SessionDetail, str],
+        button: ComputerActionMouseButton = "left",
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = MouseUpActionParams(button=button, return_screenshot=return_screenshot)
+        return self._execute_request(session, params)
+
     def drag(
         self,
         session: Union[SessionDetail, str],
@@ -118,3 +155,11 @@ class ComputerActionManager:
             return_screenshot=return_screenshot,
         )
         return self._execute_request(session, params)
+
+    def get_clipboard_text(
+        self,
+        session: Union[SessionDetail, str],
+        return_screenshot: bool = False,
+    ) -> ComputerActionResponse:
+        params = GetClipboardTextActionParams(return_screenshot=return_screenshot)
+        return self._execute_request(session, params)

hyperbrowser/client/managers/sync_manager/session.py

@@ -10,12 +10,32 @@ from ....models.session import (
     SessionListResponse,
     SessionRecording,
     UploadFileResponse,
+    SessionEventLogListParams,
+    SessionEventLogListResponse,
+    SessionEventLog,
 )
 
 
+class SessionEventLogsManager:
+    def __init__(self, client):
+        self._client = client
+
+    def list(
+        self,
+        session_id: str,
+        params: SessionEventLogListParams = SessionEventLogListParams(),
+    ) -> SessionEventLogListResponse:
+        response = self._client.transport.get(
+            self._client._build_url(f"/session/{session_id}/event-logs"),
+            params=params.model_dump(exclude_none=True, by_alias=True),
+        )
+        return SessionEventLogListResponse(**response.data)
+
+
 class SessionManager:
     def __init__(self, client):
         self._client = client
+        self.event_logs = SessionEventLogsManager(client)
 
     def create(self, params: CreateSessionParams = None) -> SessionDetail:
         response = self._client.transport.post(

hyperbrowser/models/__init__.py

@@ -5,8 +5,30 @@ from .agents.browser_use import (
     StartBrowserUseTaskParams,
     StartBrowserUseTaskResponse,
     BrowserUseApiKeys,
+    BrowserUseAgentBrain,
+    BrowserUseAgentOutput,
+    BrowserUseActionResult,
+    BrowserUseStepMetadata,
+    BrowserUseTabInfo,
+    BrowserUseCoordinates,
+    BrowserUseCoordinateSet,
+    BrowserUseViewportInfo,
+    BrowserUseDOMHistoryElement,
+    BrowserUseBrowserStateHistory,
+    BrowserUseAgentHistory,
+    BrowserUseAgentOutputV0710,
+    BrowserUseActionResultV0710,
+    BrowserUseBrowserStateHistoryV0710,
+    BrowserUseStepMetadataV0710,
+    BrowserUseAgentHistoryV0710,
+    BrowserUseAgentHistoryLatest,
+    BrowserUseTaskMetadata,
+    BrowserUseStep,
+    cast_steps_for_version,
 )
 from .agents.claude_computer_use import (
+    ClaudeComputerUseTaskStatus,
+    ClaudeComputerUseStepResponse,
     ClaudeComputerUseTaskData,
     ClaudeComputerUseTaskResponse,
     ClaudeComputerUseTaskStatusResponse,
@@ -14,6 +36,16 @@ from .agents.claude_computer_use import (
     StartClaudeComputerUseTaskResponse,
     ClaudeComputerUseApiKeys,
 )
+from .agents.gemini_computer_use import (
+    GeminiComputerUseTaskStatus,
+    GeminiComputerUseStepResponse,
+    GeminiComputerUseTaskData,
+    GeminiComputerUseTaskResponse,
+    GeminiComputerUseTaskStatusResponse,
+    StartGeminiComputerUseTaskParams,
+    StartGeminiComputerUseTaskResponse,
+    GeminiComputerUseApiKeys,
+)
 from .agents.cua import (
     CuaTaskData,
     CuaTaskResponse,
@@ -49,7 +81,10 @@ from .consts import (
     ScrapePageStatus,
     ScrapeScreenshotFormat,
     ScrapeWaitUntil,
+    SessionEventLogType,
     State,
+    SessionRegion,
+    BrowserUseVersion,
 )
 from .crawl import (
     CrawledPage,
@@ -68,6 +103,7 @@ from .extract import (
     ExtractJobStatusResponse,
     StartExtractJobParams,
     StartExtractJobResponse,
+    ExtractJobMetadata,
 )
 from .profile import (
     CreateProfileParams,
@@ -91,6 +127,7 @@ from .scrape import (
     StartBatchScrapeJobResponse,
     StartScrapeJobParams,
     StartScrapeJobResponse,
+    StorageStateOptions,
 )
 from .computer_action import (
     ClickActionParams,
@@ -99,11 +136,18 @@ from .computer_action import (
     ComputerActionResponse,
     Coordinate,
     DragActionParams,
+    HoldKeyActionParams,
+    MouseDownActionParams,
+    MouseUpActionParams,
     MoveMouseActionParams,
     PressKeysActionParams,
     ScreenshotActionParams,
     ScrollActionParams,
     TypeTextActionParams,
+    ComputerActionMouseButton,
+    GetClipboardTextActionParams,
+    ComputerActionResponseDataClipboardText,
+    ComputerActionResponseData,
 )
 from .session import (
     BasicResponse,
@@ -119,7 +163,13 @@ from .session import (
     SessionListResponse,
     SessionRecording,
     SessionStatus,
+    SessionEventLog,
+    SessionEventLogListParams,
+    SessionEventLogListResponse,
+    SessionProfile,
+    SessionLaunchState,
     UploadFileResponse,
+    ImageCaptchaParam,
 )
 from .team import TeamCreditInfo
 
@@ -139,7 +189,10 @@ __all__ = [
     "ScrapePageStatus",
     "ScrapeScreenshotFormat",
     "ScrapeWaitUntil",
+    "SessionEventLogType",
     "State",
+    "SessionRegion",
+    "BrowserUseVersion",
     # agents
     "HyperAgentTaskStatus",
     "HyperAgentActionOutput",
@@ -156,12 +209,40 @@ __all__ = [
     "BrowserUseTaskStatusResponse",
     "StartBrowserUseTaskParams",
     "StartBrowserUseTaskResponse",
+    "BrowserUseAgentBrain",
+    "BrowserUseAgentOutput",
+    "BrowserUseActionResult",
+    "BrowserUseStepMetadata",
+    "BrowserUseTabInfo",
+    "BrowserUseCoordinates",
+    "BrowserUseCoordinateSet",
+    "BrowserUseViewportInfo",
+    "BrowserUseDOMHistoryElement",
+    "BrowserUseBrowserStateHistory",
+    "BrowserUseAgentHistory",
+    "BrowserUseAgentOutputV0710",
+    "BrowserUseActionResultV0710",
+    "BrowserUseBrowserStateHistoryV0710",
+    "BrowserUseStepMetadataV0710",
+    "BrowserUseAgentHistoryV0710",
+    "BrowserUseAgentHistoryLatest",
+    "BrowserUseStep",
+    "BrowserUseTaskMetadata",
+    "cast_steps_for_version",
     "ClaudeComputerUseTaskStatus",
+    "ClaudeComputerUseStepResponse",
     "ClaudeComputerUseTaskData",
     "ClaudeComputerUseTaskResponse",
     "ClaudeComputerUseTaskStatusResponse",
     "StartClaudeComputerUseTaskParams",
     "StartClaudeComputerUseTaskResponse",
+    "GeminiComputerUseTaskStatus",
+    "GeminiComputerUseStepResponse",
+    "GeminiComputerUseTaskData",
+    "GeminiComputerUseTaskResponse",
+    "GeminiComputerUseTaskStatusResponse",
+    "StartGeminiComputerUseTaskParams",
+    "StartGeminiComputerUseTaskResponse",
     "CuaTaskStatus",
     "CuaTaskData",
     "CuaTaskResponse",
@@ -170,6 +251,7 @@ __all__ = [
     "StartCuaTaskResponse",
     "BrowserUseApiKeys",
     "ClaudeComputerUseApiKeys",
+    "GeminiComputerUseApiKeys",
     "CuaApiKeys",
     "HyperAgentApiKeys",
     # crawl
@@ -190,6 +272,7 @@ __all__ = [
     "ExtractJobStatusResponse",
     "StartExtractJobParams",
     "StartExtractJobResponse",
+    "ExtractJobMetadata",
     # profile
     "CreateProfileParams",
     "CreateProfileResponse",
@@ -211,6 +294,7 @@ __all__ = [
     "StartBatchScrapeJobResponse",
     "StartScrapeJobParams",
     "StartScrapeJobResponse",
+    "StorageStateOptions",
     # session
     "BasicResponse",
     "CreateSessionParams",
@@ -225,7 +309,13 @@ __all__ = [
     "SessionListResponse",
     "SessionRecording",
     "SessionStatus",
+    "SessionEventLog",
+    "SessionEventLogListParams",
+    "SessionEventLogListResponse",
+    "SessionProfile",
+    "SessionLaunchState",
     "UploadFileResponse",
+    "ImageCaptchaParam",
     # team
     "TeamCreditInfo",
     # computer action
@@ -235,9 +325,16 @@ __all__ = [
     "ComputerActionResponse",
     "Coordinate",
     "DragActionParams",
+    "HoldKeyActionParams",
+    "MouseDownActionParams",
+    "MouseUpActionParams",
     "MoveMouseActionParams",
     "PressKeysActionParams",
     "ScreenshotActionParams",
     "ScrollActionParams",
     "TypeTextActionParams",
+    "ComputerActionMouseButton",
+    "GetClipboardTextActionParams",
+    "ComputerActionResponseDataClipboardText",
+    "ComputerActionResponseData",
 ]