hyperbrowser 0.67.0__tar.gz → 0.69.0__tar.gz

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hyperbrowser
-Version: 0.67.0
+Version: 0.69.0
 Summary: Python SDK for hyperbrowser
 License: MIT
 License-File: LICENSE

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/client/managers/async_manager/computer_action.py

@@ -16,6 +16,7 @@ from hyperbrowser.models import (
     MouseDownActionParams,
     MouseUpActionParams,
     ComputerActionMouseButton,
+    GetClipboardTextActionParams,
 )
 
 
@@ -154,3 +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-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/client/managers/sync_manager/computer_action.py

@@ -16,6 +16,7 @@ from hyperbrowser.models import (
     MouseDownActionParams,
     MouseUpActionParams,
     ComputerActionMouseButton,
+    GetClipboardTextActionParams,
 )
 
 
@@ -154,3 +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-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/__init__.py

@@ -5,6 +5,26 @@ 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,
@@ -64,6 +84,7 @@ from .consts import (
     SessionEventLogType,
     State,
     SessionRegion,
+    BrowserUseVersion,
 )
 from .crawl import (
     CrawledPage,
@@ -82,6 +103,7 @@ from .extract import (
     ExtractJobStatusResponse,
     StartExtractJobParams,
     StartExtractJobResponse,
+    ExtractJobMetadata,
 )
 from .profile import (
     CreateProfileParams,
@@ -105,6 +127,7 @@ from .scrape import (
     StartBatchScrapeJobResponse,
     StartScrapeJobParams,
     StartScrapeJobResponse,
+    StorageStateOptions,
 )
 from .computer_action import (
     ClickActionParams,
@@ -122,6 +145,9 @@ from .computer_action import (
     ScrollActionParams,
     TypeTextActionParams,
     ComputerActionMouseButton,
+    GetClipboardTextActionParams,
+    ComputerActionResponseDataClipboardText,
+    ComputerActionResponseData,
 )
 from .session import (
     BasicResponse,
@@ -166,6 +192,7 @@ __all__ = [
     "SessionEventLogType",
     "State",
     "SessionRegion",
+    "BrowserUseVersion",
     # agents
     "HyperAgentTaskStatus",
     "HyperAgentActionOutput",
@@ -182,6 +209,26 @@ __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",
@@ -225,6 +272,7 @@ __all__ = [
     "ExtractJobStatusResponse",
     "StartExtractJobParams",
     "StartExtractJobResponse",
+    "ExtractJobMetadata",
     # profile
     "CreateProfileParams",
     "CreateProfileResponse",
@@ -246,6 +294,7 @@ __all__ = [
     "StartBatchScrapeJobResponse",
     "StartScrapeJobParams",
     "StartScrapeJobResponse",
+    "StorageStateOptions",
     # session
     "BasicResponse",
     "CreateSessionParams",
@@ -285,4 +334,7 @@ __all__ = [
     "ScrollActionParams",
     "TypeTextActionParams",
     "ComputerActionMouseButton",
+    "GetClipboardTextActionParams",
+    "ComputerActionResponseDataClipboardText",
+    "ComputerActionResponseData",
 ]

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/agents/browser_use.py

@@ -2,7 +2,7 @@ from typing import Any, Dict, List, Literal, Optional, Type, Union
 
 from pydantic import BaseModel, ConfigDict, Field
 
-from ..consts import BrowserUseLlm
+from ..consts import BrowserUseLlm, BrowserUseVersion
 from ..session import CreateSessionParams
 
 BrowserUseTaskStatus = Literal["pending", "running", "completed", "failed", "stopped"]
@@ -32,6 +32,9 @@ class StartBrowserUseTaskParams(BaseModel):
     )
 
     task: str
+    version: Optional[BrowserUseVersion] = Field(
+        default=None, serialization_alias="version"
+    )
     llm: Optional[BrowserUseLlm] = Field(default=None, serialization_alias="llm")
     session_id: Optional[str] = Field(default=None, serialization_alias="sessionId")
     validate_output: Optional[bool] = Field(
@@ -193,15 +196,79 @@ class BrowserUseAgentHistory(BaseModel):
     metadata: Optional[BrowserUseStepMetadata] = None
 
 
+# ===== 0.7.10-specific response models =====
+
+
+class BrowserUseAgentOutputV0710(BaseModel):
+    thinking: Optional[str] = None
+    evaluation_previous_goal: Optional[str] = None
+    memory: Optional[str] = None
+    next_goal: Optional[str] = None
+    action: list[dict]
+
+
+class BrowserUseActionResultV0710(BaseModel):
+    is_done: Optional[bool] = False
+    success: Optional[bool] = None
+    error: Optional[str] = None
+    metadata: Optional[dict] = None
+    attachments: Optional[list[str]] = None
+    long_term_memory: Optional[str] = None
+    extracted_content: Optional[str] = None
+    include_extracted_content_only_once: Optional[bool] = False
+    include_in_memory: bool = False
+
+
+class BrowserUseBrowserStateHistoryV0710(BaseModel):
+    url: str
+    title: str
+    tabs: list[dict]
+    interacted_element: Union[list[Union[dict, None]], list[None]]
+
+
+class BrowserUseStepMetadataV0710(BaseModel):
+    step_start_time: float
+    step_end_time: float
+    step_number: int
+
+
+class BrowserUseAgentHistoryV0710(BaseModel):
+    model_output: Union[BrowserUseAgentOutputV0710, None]
+    result: list[BrowserUseActionResultV0710]
+    state: BrowserUseBrowserStateHistoryV0710
+    metadata: Optional[BrowserUseStepMetadataV0710] = None
+
+
+# ===== latest response models =====
+
+BrowserUseAgentHistoryLatest = Dict[str, Any]
+
+BrowserUseStep = Union[
+    BrowserUseAgentHistory, BrowserUseAgentHistoryV0710, BrowserUseAgentHistoryLatest
+]
+
+
 class BrowserUseTaskData(BaseModel):
     model_config = ConfigDict(
         populate_by_alias=True,
     )
 
-    steps: list[BrowserUseAgentHistory]
+    steps: list[BrowserUseStep]
     final_result: Optional[str] = Field(default=None, alias="finalResult")
 
 
+class BrowserUseTaskMetadata(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    input_tokens: Optional[int] = Field(default=None, alias="inputTokens")
+    output_tokens: Optional[int] = Field(default=None, alias="outputTokens")
+    num_task_steps_completed: Optional[int] = Field(
+        default=None, alias="numTaskStepsCompleted"
+    )
+
+
 class BrowserUseTaskResponse(BaseModel):
     """
     Response from a browser use task.
@@ -213,6 +280,20 @@ class BrowserUseTaskResponse(BaseModel):
 
     job_id: str = Field(alias="jobId")
     status: BrowserUseTaskStatus
+    metadata: Optional[BrowserUseTaskMetadata] = Field(default=None, alias="metadata")
     data: Optional[BrowserUseTaskData] = Field(default=None, alias="data")
     error: Optional[str] = Field(default=None, alias="error")
     live_url: Optional[str] = Field(default=None, alias="liveUrl")
+
+
+def cast_steps_for_version(
+    steps: List[BrowserUseStep], version: BrowserUseVersion
+) -> List[BrowserUseStep]:
+    if version == "0.1.40":
+        return [BrowserUseAgentHistory.model_validate(s) for s in steps]
+    elif version == "0.7.10":
+        return [BrowserUseAgentHistoryV0710.model_validate(s) for s in steps]
+    elif version == "latest":
+        return steps
+    else:
+        raise ValueError(f"Invalid version: {version}")

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/agents/claude_computer_use.py

@@ -104,6 +104,18 @@ class ClaudeComputerUseTaskData(BaseModel):
     final_result: Optional[str] = Field(default=None, alias="finalResult")
 
 
+class ClaudeComputerUseTaskMetadata(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    input_tokens: Optional[int] = Field(default=None, alias="inputTokens")
+    output_tokens: Optional[int] = Field(default=None, alias="outputTokens")
+    num_task_steps_completed: Optional[int] = Field(
+        default=None, alias="numTaskStepsCompleted"
+    )
+
+
 class ClaudeComputerUseTaskResponse(BaseModel):
     """
     Response from a Claude Computer Use task.
@@ -115,6 +127,9 @@ class ClaudeComputerUseTaskResponse(BaseModel):
 
     job_id: str = Field(alias="jobId")
     status: ClaudeComputerUseTaskStatus
+    metadata: Optional[ClaudeComputerUseTaskMetadata] = Field(
+        default=None, alias="metadata"
+    )
     data: Optional[ClaudeComputerUseTaskData] = Field(default=None, alias="data")
     error: Optional[str] = Field(default=None, alias="error")
     live_url: Optional[str] = Field(default=None, alias="liveUrl")

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/agents/cua.py

@@ -134,6 +134,18 @@ class CuaTaskData(BaseModel):
     final_result: Optional[str] = Field(default=None, alias="finalResult")
 
 
+class CuaTaskMetadata(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    input_tokens: Optional[int] = Field(default=None, alias="inputTokens")
+    output_tokens: Optional[int] = Field(default=None, alias="outputTokens")
+    num_task_steps_completed: Optional[int] = Field(
+        default=None, alias="numTaskStepsCompleted"
+    )
+
+
 class CuaTaskResponse(BaseModel):
     """
     Response from a CUA task.
@@ -145,6 +157,7 @@ class CuaTaskResponse(BaseModel):
 
     job_id: str = Field(alias="jobId")
     status: CuaTaskStatus
+    metadata: Optional[CuaTaskMetadata] = Field(default=None, alias="metadata")
     data: Optional[CuaTaskData] = Field(default=None, alias="data")
     error: Optional[str] = Field(default=None, alias="error")
     live_url: Optional[str] = Field(default=None, alias="liveUrl")

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/agents/gemini_computer_use.py

@@ -104,6 +104,18 @@ class GeminiComputerUseTaskData(BaseModel):
     final_result: Optional[str] = Field(default=None, alias="finalResult")
 
 
+class GeminiComputerUseTaskMetadata(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    input_tokens: Optional[int] = Field(default=None, alias="inputTokens")
+    output_tokens: Optional[int] = Field(default=None, alias="outputTokens")
+    num_task_steps_completed: Optional[int] = Field(
+        default=None, alias="numTaskStepsCompleted"
+    )
+
+
 class GeminiComputerUseTaskResponse(BaseModel):
     """
     Response from a Gemini Computer Use task.
@@ -115,6 +127,9 @@ class GeminiComputerUseTaskResponse(BaseModel):
 
     job_id: str = Field(alias="jobId")
     status: GeminiComputerUseTaskStatus
+    metadata: Optional[GeminiComputerUseTaskMetadata] = Field(
+        default=None, alias="metadata"
+    )
     data: Optional[GeminiComputerUseTaskData] = Field(default=None, alias="data")
     error: Optional[str] = Field(default=None, alias="error")
     live_url: Optional[str] = Field(default=None, alias="liveUrl")

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/agents/hyper_agent.py

@@ -126,6 +126,16 @@ class HyperAgentTaskData(BaseModel):
     final_result: Optional[str] = Field(default=None, alias="finalResult")
 
 
+class HyperAgentTaskMetadata(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    num_task_steps_completed: Optional[int] = Field(
+        default=None, alias="numTaskStepsCompleted"
+    )
+
+
 class HyperAgentTaskResponse(BaseModel):
     """
     Response from a HyperAgent task.
@@ -137,6 +147,7 @@ class HyperAgentTaskResponse(BaseModel):
 
     job_id: str = Field(alias="jobId")
     status: HyperAgentTaskStatus
+    metadata: Optional[HyperAgentTaskMetadata] = Field(default=None, alias="metadata")
     data: Optional[HyperAgentTaskData] = Field(default=None, alias="data")
     error: Optional[str] = Field(default=None, alias="error")
     live_url: Optional[str] = Field(default=None, alias="liveUrl")

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/computer_action.py

@@ -16,6 +16,7 @@ class ComputerAction(str, Enum):
     SCREENSHOT = "screenshot"
     SCROLL = "scroll"
     TYPE_TEXT = "type_text"
+    GET_CLIPBOARD_TEXT = "get_clipboard_text"
 
 
 ComputerActionMouseButton = Literal[
@@ -154,6 +155,19 @@ class TypeTextActionParams(BaseModel):
     )
 
 
+class GetClipboardTextActionParams(BaseModel):
+    """Parameters for get clipboard text action."""
+
+    model_config = ConfigDict(use_enum_values=True)
+
+    action: Literal[ComputerAction.GET_CLIPBOARD_TEXT] = (
+        ComputerAction.GET_CLIPBOARD_TEXT
+    )
+    return_screenshot: bool = Field(
+        serialization_alias="returnScreenshot", default=False
+    )
+
+
 ComputerActionParams = Union[
     ClickActionParams,
     DragActionParams,
@@ -162,9 +176,24 @@ ComputerActionParams = Union[
     ScreenshotActionParams,
     ScrollActionParams,
     TypeTextActionParams,
+    HoldKeyActionParams,
+    MouseDownActionParams,
+    MouseUpActionParams,
+    GetClipboardTextActionParams,
 ]
 
 
+class ComputerActionResponseDataClipboardText(BaseModel):
+    """Data for get clipboard text action."""
+
+    model_config = ConfigDict(populate_by_alias=True)
+
+    clipboard_text: Optional[str] = Field(default=None, alias="clipboardText")
+
+
+ComputerActionResponseData = Union[ComputerActionResponseDataClipboardText]
+
+
 class ComputerActionResponse(BaseModel):
     """Response from computer action API."""
 
@@ -174,5 +203,6 @@ class ComputerActionResponse(BaseModel):
 
     success: bool
     screenshot: Optional[str] = None
+    data: Optional[ComputerActionResponseData] = None
     error: Optional[str] = None
     message: Optional[str] = None

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/consts.py

@@ -16,6 +16,8 @@ SessionEventLogType = Literal[
 
 POLLING_ATTEMPTS = 5
 
+BrowserUseVersion = Literal["0.1.40", "0.7.10", "latest"]
+
 BrowserUseLlm = Literal[
     "gpt-4o",
     "gpt-4o-mini",

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/extract.py

@@ -55,6 +55,16 @@ class ExtractJobStatusResponse(BaseModel):
     status: ExtractJobStatus
 
 
+class ExtractJobMetadata(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    input_tokens: Optional[int] = Field(default=None, alias="inputTokens")
+    output_tokens: Optional[int] = Field(default=None, alias="outputTokens")
+    num_pages_scraped: Optional[int] = Field(default=None, alias="numPagesScraped")
+
+
 class ExtractJobResponse(BaseModel):
     """
     Response from a extract job.
@@ -67,4 +77,5 @@ class ExtractJobResponse(BaseModel):
     job_id: str = Field(alias="jobId")
     status: ExtractJobStatus
     error: Optional[str] = None
+    metadata: Optional[ExtractJobMetadata] = None
     data: Optional[dict] = None

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/hyperbrowser/models/scrape.py

@@ -23,6 +23,15 @@ class ScreenshotOptions(BaseModel):
     )
 
 
+class StorageStateOptions(BaseModel):
+    local_storage: Optional[dict[str, str]] = Field(
+        default=None, serialization_alias="localStorage"
+    )
+    session_storage: Optional[dict[str, str]] = Field(
+        default=None, serialization_alias="sessionStorage"
+    )
+
+
 class ScrapeOptions(BaseModel):
     """
     Options for scraping a page.
@@ -46,6 +55,9 @@ class ScrapeOptions(BaseModel):
     screenshot_options: Optional[ScreenshotOptions] = Field(
         default=None, serialization_alias="screenshotOptions"
     )
+    storage_state: Optional[StorageStateOptions] = Field(
+        default=None, serialization_alias="storageState"
+    )
 
 
 class StartScrapeJobParams(BaseModel):

{hyperbrowser-0.67.0 → hyperbrowser-0.69.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "hyperbrowser"
-version = "0.67.0"
+version = "0.69.0"
 description = "Python SDK for hyperbrowser"
 authors = ["Nikhil Shahi <nshahi1998@gmail.com>"]
 license = "MIT"