hyperbrowser 0.61.0__tar.gz → 0.62.0__tar.gz

{hyperbrowser-0.61.0 → hyperbrowser-0.62.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hyperbrowser
-Version: 0.61.0
+Version: 0.62.0
 Summary: Python SDK for hyperbrowser
 License: MIT
 License-File: LICENSE

{hyperbrowser-0.61.0 → hyperbrowser-0.62.0}/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-0.62.0/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-0.61.0 → hyperbrowser-0.62.0}/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-0.61.0 → hyperbrowser-0.62.0}/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-0.62.0/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-0.61.0 → hyperbrowser-0.62.0}/hyperbrowser/models/__init__.py

@@ -7,6 +7,8 @@ from .agents.browser_use import (
     BrowserUseApiKeys,
 )
 from .agents.claude_computer_use import (
+    ClaudeComputerUseTaskStatus,
+    ClaudeComputerUseStepResponse,
     ClaudeComputerUseTaskData,
     ClaudeComputerUseTaskResponse,
     ClaudeComputerUseTaskStatusResponse,
@@ -14,6 +16,16 @@ 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,
@@ -51,6 +63,7 @@ from .consts import (
     ScrapeWaitUntil,
     SessionEventLogType,
     State,
+    SessionRegion,
 )
 from .crawl import (
     CrawledPage,
@@ -126,6 +139,7 @@ from .session import (
     SessionProfile,
     SessionLaunchState,
     UploadFileResponse,
+    ImageCaptchaParam,
 )
 from .team import TeamCreditInfo
 
@@ -147,6 +161,7 @@ __all__ = [
     "ScrapeWaitUntil",
     "SessionEventLogType",
     "State",
+    "SessionRegion",
     # agents
     "HyperAgentTaskStatus",
     "HyperAgentActionOutput",
@@ -164,11 +179,19 @@ __all__ = [
     "StartBrowserUseTaskParams",
     "StartBrowserUseTaskResponse",
     "ClaudeComputerUseTaskStatus",
+    "ClaudeComputerUseStepResponse",
     "ClaudeComputerUseTaskData",
     "ClaudeComputerUseTaskResponse",
     "ClaudeComputerUseTaskStatusResponse",
     "StartClaudeComputerUseTaskParams",
     "StartClaudeComputerUseTaskResponse",
+    "GeminiComputerUseTaskStatus",
+    "GeminiComputerUseStepResponse",
+    "GeminiComputerUseTaskData",
+    "GeminiComputerUseTaskResponse",
+    "GeminiComputerUseTaskStatusResponse",
+    "StartGeminiComputerUseTaskParams",
+    "StartGeminiComputerUseTaskResponse",
     "CuaTaskStatus",
     "CuaTaskData",
     "CuaTaskResponse",
@@ -177,6 +200,7 @@ __all__ = [
     "StartCuaTaskResponse",
     "BrowserUseApiKeys",
     "ClaudeComputerUseApiKeys",
+    "GeminiComputerUseApiKeys",
     "CuaApiKeys",
     "HyperAgentApiKeys",
     # crawl
@@ -238,6 +262,7 @@ __all__ = [
     "SessionProfile",
     "SessionLaunchState",
     "UploadFileResponse",
+    "ImageCaptchaParam",
     # team
     "TeamCreditInfo",
     # computer action

hyperbrowser-0.62.0/hyperbrowser/models/agents/gemini_computer_use.py

@@ -0,0 +1,117 @@
+from typing import Any, Literal, Optional
+
+from pydantic import BaseModel, ConfigDict, Field
+
+from ..session import CreateSessionParams
+from ..consts import GeminiComputerUseLlm
+
+GeminiComputerUseTaskStatus = Literal[
+    "pending", "running", "completed", "failed", "stopped"
+]
+
+
+class GeminiComputerUseApiKeys(BaseModel):
+    """
+    API keys for the Gemini Computer Use task.
+    """
+
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    google: Optional[str] = Field(default=None, serialization_alias="google")
+
+
+class StartGeminiComputerUseTaskParams(BaseModel):
+    """
+    Parameters for creating a new Gemini Computer Use task.
+    """
+
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    task: str
+    llm: Optional[GeminiComputerUseLlm] = Field(default=None, serialization_alias="llm")
+    session_id: Optional[str] = Field(default=None, serialization_alias="sessionId")
+    max_failures: Optional[int] = Field(default=None, serialization_alias="maxFailures")
+    max_steps: Optional[int] = Field(default=None, serialization_alias="maxSteps")
+    keep_browser_open: Optional[bool] = Field(
+        default=None, serialization_alias="keepBrowserOpen"
+    )
+    session_options: Optional[CreateSessionParams] = Field(
+        default=None, serialization_alias="sessionOptions"
+    )
+    use_custom_api_keys: Optional[bool] = Field(
+        default=None, serialization_alias="useCustomApiKeys"
+    )
+    api_keys: Optional[GeminiComputerUseApiKeys] = Field(
+        default=None, serialization_alias="apiKeys"
+    )
+
+
+class StartGeminiComputerUseTaskResponse(BaseModel):
+    """
+    Response from starting a Gemini Computer Use task.
+    """
+
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    job_id: str = Field(alias="jobId")
+    live_url: Optional[str] = Field(default=None, alias="liveUrl")
+
+
+class GeminiComputerUseTaskStatusResponse(BaseModel):
+    """
+    Response from getting a Gemini Computer Use task status.
+    """
+
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    status: GeminiComputerUseTaskStatus
+
+
+class GeminiComputerUseStepResponse(BaseModel):
+    """
+    Response from a single Gemini Computer Use step.
+    """
+
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    candidates: Optional[list[Any]] = Field(
+        default=None, serialization_alias="candidates"
+    )
+    model_version: Optional[str] = Field(
+        default=None, serialization_alias="modelVersion"
+    )
+
+
+class GeminiComputerUseTaskData(BaseModel):
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    steps: list[GeminiComputerUseStepResponse]
+    final_result: Optional[str] = Field(default=None, alias="finalResult")
+
+
+class GeminiComputerUseTaskResponse(BaseModel):
+    """
+    Response from a Gemini Computer Use task.
+    """
+
+    model_config = ConfigDict(
+        populate_by_alias=True,
+    )
+
+    job_id: str = Field(alias="jobId")
+    status: GeminiComputerUseTaskStatus
+    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-0.61.0 → hyperbrowser-0.62.0}/hyperbrowser/models/consts.py

@@ -43,6 +43,7 @@ ClaudeComputerUseLlm = Literal[
     "claude-sonnet-4-20250514",
     "claude-3-7-sonnet-20250219",
 ]
+GeminiComputerUseLlm = Literal["gemini-2.5-computer-use-preview-10-2025",]
 SessionRegion = Literal[
     "us-central",
     "asia-south",

{hyperbrowser-0.61.0 → hyperbrowser-0.62.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "hyperbrowser"
-version = "0.61.0"
+version = "0.62.0"
 description = "Python SDK for hyperbrowser"
 authors = ["Nikhil Shahi <nshahi1998@gmail.com>"]
 license = "MIT"