oagi-core 0.10.0__py3-none-any.whl → 0.10.2__py3-none-any.whl

oagi/agent/tasker/tasker_agent.py

@@ -40,6 +40,7 @@ class TaskerAgent(AsyncAgent):
         reflection_interval: int = 4,
         planner: Planner | None = None,
         step_observer: AsyncObserver | None = None,
+        step_delay: float = 0.3,
     ):
         """Initialize the tasker agent.
 
@@ -52,6 +53,7 @@ class TaskerAgent(AsyncAgent):
             reflection_interval: Actions before reflection
             planner: Planner for planning and reflection
             step_observer: Optional observer for step tracking
+            step_delay: Delay in seconds after actions before next screenshot
         """
         self.api_key = api_key
         self.base_url = base_url
@@ -61,6 +63,7 @@ class TaskerAgent(AsyncAgent):
         self.reflection_interval = reflection_interval
         self.planner = planner or Planner(api_key=api_key, base_url=base_url)
         self.step_observer = step_observer
+        self.step_delay = step_delay
 
         # Memory for tracking workflow
         self.memory = PlannerMemory()
@@ -73,20 +76,15 @@ class TaskerAgent(AsyncAgent):
         self,
         task: str,
         todos: list[str],
-        deliverables: list[str] | None = None,
     ) -> None:
-        """Set the task, todos, and deliverables for the workflow.
+        """Set the task and todos for the workflow.
 
         Args:
             task: Overall task description
             todos: List of todo descriptions
-            deliverables: Optional list of deliverable descriptions
         """
-        self.memory.set_task(task, todos, deliverables)
-        logger.info(
-            f"Task set with {len(todos)} todos and "
-            f"{len(deliverables) if deliverables else 0} deliverables"
-        )
+        self.memory.set_task(task, todos)
+        logger.info(f"Task set with {len(todos)} todos")
 
     async def execute(
         self,
@@ -189,6 +187,7 @@ class TaskerAgent(AsyncAgent):
             external_memory=self.memory,  # Share memory with child
             todo_index=todo_index,  # Pass the todo index
             step_observer=self.step_observer,  # Pass step observer
+            step_delay=self.step_delay,
         )
 
         self.current_todo_index = todo_index
@@ -327,12 +326,3 @@ class TaskerAgent(AsyncAgent):
         """
         self.memory.append_todo(description)
         logger.info(f"Appended new todo: {description}")
-
-    def append_deliverable(self, description: str) -> None:
-        """Dynamically append a new deliverable to the workflow.
-
-        Args:
-            description: Description of the new deliverable
-        """
-        self.memory.append_deliverable(description)
-        logger.info(f"Appended new deliverable: {description}")

oagi/cli/agent.py

@@ -65,11 +65,112 @@ def add_agent_parser(subparsers: argparse._SubParsersAction) -> None:
         type=str,
         help="Output file path for export (default: execution_report.[md|html|json])",
     )
+    run_parser.add_argument(
+        "--step-delay",
+        type=float,
+        help="Delay in seconds after each step before next screenshot (default: 0.3)",
+    )
+
+    # agent permission command
+    agent_subparsers.add_parser(
+        "permission",
+        help="Check macOS permissions for screen recording and accessibility",
+    )
 
 
 def handle_agent_command(args: argparse.Namespace) -> None:
     if args.agent_command == "run":
         run_agent(args)
+    elif args.agent_command == "permission":
+        check_permissions()
+
+
+def check_permissions() -> None:
+    """Check and request macOS permissions for screen recording and accessibility.
+
+    Guides the user through granting permissions one at a time.
+    """
+    if sys.platform != "darwin":
+        print("Warning: Permission check is only applicable on macOS.")
+        print("On other platforms, no special permissions are required.")
+        return
+
+    check_optional_dependency("Quartz", "Permission check", "desktop")
+    check_optional_dependency("ApplicationServices", "Permission check", "desktop")
+
+    import subprocess  # noqa: PLC0415
+
+    from ApplicationServices import AXIsProcessTrusted  # noqa: PLC0415
+    from Quartz import (  # noqa: PLC0415
+        CGPreflightScreenCaptureAccess,
+        CGRequestScreenCaptureAccess,
+    )
+
+    # Check all permissions first to show status
+    screen_recording_granted = CGPreflightScreenCaptureAccess()
+    accessibility_granted = AXIsProcessTrusted()
+
+    print("Checking permissions...")
+    print(f"  {'[OK]' if screen_recording_granted else '[MISSING]'} Screen Recording")
+    print(f"  {'[OK]' if accessibility_granted else '[MISSING]'} Accessibility")
+
+    # Guide user through missing permissions one at a time
+    if not screen_recording_granted:
+        CGRequestScreenCaptureAccess()
+        subprocess.run(
+            [
+                "open",
+                "x-apple.systempreferences:com.apple.preference.security?Privacy_ScreenCapture",
+            ],
+            check=False,
+        )
+        print("\nPlease grant Screen Recording permission in System Preferences.")
+        print("After granting, run this command again to continue.")
+        print("Note: You may need to restart your terminal after granting permissions.")
+        sys.exit(1)
+
+    if not accessibility_granted:
+        subprocess.run(
+            [
+                "open",
+                "x-apple.systempreferences:com.apple.preference.security?Privacy_Accessibility",
+            ],
+            check=False,
+        )
+        print("\nPlease grant Accessibility permission in System Preferences.")
+        print("After granting, run this command again to continue.")
+        print("Note: You may need to restart your terminal after granting permissions.")
+        sys.exit(1)
+
+    print()
+    print("All permissions granted. You can run the agent.")
+
+
+def _warn_missing_permissions() -> None:
+    if sys.platform != "darwin":
+        return
+
+    if not check_optional_dependency(
+        "Quartz", "Permission check", "desktop", raise_error=False
+    ):
+        return
+    if not check_optional_dependency(
+        "ApplicationServices", "Permission check", "desktop", raise_error=False
+    ):
+        return
+
+    from ApplicationServices import AXIsProcessTrusted  # noqa: PLC0415
+    from Quartz import CGPreflightScreenCaptureAccess  # noqa: PLC0415
+
+    missing = []
+    if not CGPreflightScreenCaptureAccess():
+        missing.append("Screen Recording")
+    if not AXIsProcessTrusted():
+        missing.append("Accessibility")
+
+    if missing:
+        print(f"Warning: Missing macOS permissions: {', '.join(missing)}")
+        print("Run 'oagi agent permission' to configure permissions.\n")
 
 
 def run_agent(args: argparse.Namespace) -> None:
@@ -77,6 +178,9 @@ def run_agent(args: argparse.Namespace) -> None:
     check_optional_dependency("pyautogui", "Agent execution", "desktop")
     check_optional_dependency("PIL", "Agent execution", "desktop")
 
+    # Warn about missing macOS permissions (non-blocking)
+    _warn_missing_permissions()
+
     from oagi import AsyncPyautoguiActionHandler, AsyncScreenshotMaker  # noqa: PLC0415
     from oagi.agent import create_agent  # noqa: PLC0415
 
@@ -97,6 +201,7 @@ def run_agent(args: argparse.Namespace) -> None:
     max_steps = args.max_steps or 20
     temperature = args.temperature if args.temperature is not None else 0.5
     mode = args.mode or "actor"
+    step_delay = args.step_delay if args.step_delay is not None else 0.3
     export_format = args.export
     export_file = args.export_file
 
@@ -122,6 +227,7 @@ def run_agent(args: argparse.Namespace) -> None:
         max_steps=max_steps,
         temperature=temperature,
         step_observer=observer,
+        step_delay=step_delay,
     )
 
     # Create handlers
@@ -130,7 +236,8 @@ def run_agent(args: argparse.Namespace) -> None:
 
     print(f"Starting agent with instruction: {args.instruction}")
     print(
-        f"Mode: {mode}, Model: {model}, Max steps: {max_steps}, Temperature: {temperature}"
+        f"Mode: {mode}, Model: {model}, Max steps: {max_steps}, "
+        f"Temperature: {temperature}, Step delay: {step_delay}s"
     )
     print("-" * 60)
 

oagi/client/async_.py

@@ -223,7 +223,6 @@ class AsyncClient(BaseClient[httpx.AsyncClient]):
         overall_todo: str,
         task_description: str,
         todos: list[dict],
-        deliverables: list[dict],
         history: list[dict] | None = None,
         current_todo_index: int | None = None,
         task_execution_summary: str | None = None,
@@ -243,7 +242,6 @@ class AsyncClient(BaseClient[httpx.AsyncClient]):
             overall_todo: Current todo description
             task_description: Overall task description
             todos: List of todo dicts with index, description, status, execution_summary
-            deliverables: List of deliverable dicts with description, achieved
             history: List of history dicts with todo_index, todo_description, action_count, summary, completed
             current_todo_index: Index of current todo being executed
             task_execution_summary: Summary of overall task execution
@@ -269,7 +267,6 @@ class AsyncClient(BaseClient[httpx.AsyncClient]):
             overall_todo=overall_todo,
             task_description=task_description,
             todos=todos,
-            deliverables=deliverables,
             history=history,
             current_todo_index=current_todo_index,
             task_execution_summary=task_execution_summary,

oagi/client/base.py

@@ -348,7 +348,6 @@ class BaseClient(Generic[HttpClientT]):
         overall_todo: str,
         task_description: str,
         todos: list[dict],
-        deliverables: list[dict],
         history: list[dict] | None = None,
         current_todo_index: int | None = None,
         task_execution_summary: str | None = None,
@@ -368,7 +367,6 @@ class BaseClient(Generic[HttpClientT]):
             overall_todo: Current todo description
             task_description: Overall task description
             todos: List of todo dicts with index, description, status, execution_summary
-            deliverables: List of deliverable dicts with description, achieved
             history: List of history dicts with todo_index, todo_description, action_count, summary, completed
             current_todo_index: Index of current todo being executed
             task_execution_summary: Summary of overall task execution
@@ -402,7 +400,6 @@ class BaseClient(Generic[HttpClientT]):
             "overall_todo": overall_todo,
             "task_description": task_description,
             "todos": todos,
-            "deliverables": deliverables,
             "history": history or [],
         }
 
@@ -456,6 +453,5 @@ class BaseClient(Generic[HttpClientT]):
 
         logger.info(
             f"Generate request successful - tokens: {result.prompt_tokens}+{result.completion_tokens}, "
-            f"cost: ${result.cost:.6f}"
         )
         return result

oagi/client/sync.py

@@ -226,7 +226,6 @@ class SyncClient(BaseClient[httpx.Client]):
         overall_todo: str,
         task_description: str,
         todos: list[dict],
-        deliverables: list[dict],
         history: list[dict] | None = None,
         current_todo_index: int | None = None,
         task_execution_summary: str | None = None,
@@ -246,7 +245,6 @@ class SyncClient(BaseClient[httpx.Client]):
             overall_todo: Current todo description
             task_description: Overall task description
             todos: List of todo dicts with index, description, status, execution_summary
-            deliverables: List of deliverable dicts with description, achieved
             history: List of history dicts with todo_index, todo_description, action_count, summary, completed
             current_todo_index: Index of current todo being executed
             task_execution_summary: Summary of overall task execution
@@ -272,7 +270,6 @@ class SyncClient(BaseClient[httpx.Client]):
             overall_todo=overall_todo,
             task_description=task_description,
             todos=todos,
-            deliverables=deliverables,
             history=history,
             current_todo_index=current_todo_index,
             task_execution_summary=task_execution_summary,

oagi/handler/pyautogui_action_handler.py

@@ -6,14 +6,13 @@
 #  Licensed under the MIT License.
 # -----------------------------------------------------------------------------
 
-import re
 import sys
 import time
 
 from pydantic import BaseModel, Field
 
 from ..exceptions import check_optional_dependency
-from ..types import Action, ActionType
+from ..types import Action, ActionType, parse_coords, parse_drag_coords, parse_scroll
 
 check_optional_dependency("pyautogui", "PyautoguiActionHandler", "desktop")
 import pyautogui  # noqa: E402
@@ -136,43 +135,40 @@ class PyautoguiActionHandler:
 
     def _parse_coords(self, args_str: str) -> tuple[int, int]:
         """Extract x, y coordinates from argument string."""
-        match = re.match(r"(\d+),\s*(\d+)", args_str)
-        if not match:
+        coords = parse_coords(args_str)
+        if not coords:
             raise ValueError(f"Invalid coordinates format: {args_str}")
-        x, y = int(match.group(1)), int(match.group(2))
-        return self._denormalize_coords(x, y)
+        return self._denormalize_coords(coords[0], coords[1])
 
     def _parse_drag_coords(self, args_str: str) -> tuple[int, int, int, int]:
         """Extract x1, y1, x2, y2 coordinates from drag argument string."""
-        match = re.match(r"(\d+),\s*(\d+),\s*(\d+),\s*(\d+)", args_str)
-        if not match:
+        coords = parse_drag_coords(args_str)
+        if not coords:
             raise ValueError(f"Invalid drag coordinates format: {args_str}")
-        x1, y1, x2, y2 = (
-            int(match.group(1)),
-            int(match.group(2)),
-            int(match.group(3)),
-            int(match.group(4)),
-        )
-        x1, y1 = self._denormalize_coords(x1, y1)
-        x2, y2 = self._denormalize_coords(x2, y2)
+        x1, y1 = self._denormalize_coords(coords[0], coords[1])
+        x2, y2 = self._denormalize_coords(coords[2], coords[3])
         return x1, y1, x2, y2
 
     def _parse_scroll(self, args_str: str) -> tuple[int, int, str]:
         """Extract x, y, direction from scroll argument string."""
-        match = re.match(r"(\d+),\s*(\d+),\s*(\w+)", args_str)
-        if not match:
+        result = parse_scroll(args_str)
+        if not result:
             raise ValueError(f"Invalid scroll format: {args_str}")
-        x, y = int(match.group(1)), int(match.group(2))
-        x, y = self._denormalize_coords(x, y)
-        direction = match.group(3).lower()
-        return x, y, direction
+        x, y = self._denormalize_coords(result[0], result[1])
+        return x, y, result[2]
 
     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"
+        hotkey_variations_mapping = {
+            "capslock": ["caps_lock", "caps", "capslock"],
+            "pgup": ["page_up", "pageup"],
+            "pgdn": ["page_down", "pagedown"],
+        }
+        for normalized, variations in hotkey_variations_mapping.items():
+            if key in variations:
+                return normalized
         # Remap ctrl to command on macOS if enabled
         if self.config.macos_ctrl_to_cmd and sys.platform == "darwin" and key == "ctrl":
             return "command"

oagi/server/socketio_server.py

@@ -16,7 +16,13 @@ from pydantic import ValidationError
 from ..agent import AsyncDefaultAgent, create_agent
 from ..client import AsyncClient
 from ..exceptions import check_optional_dependency
-from ..types.models.action import Action, ActionType
+from ..types.models.action import (
+    Action,
+    ActionType,
+    parse_coords,
+    parse_drag_coords,
+    parse_scroll,
+)
 from .agent_wrappers import SocketIOActionHandler, SocketIOImageProvider
 from .config import ServerConfig
 from .models import (
@@ -275,31 +281,29 @@ class SessionNamespace(socketio.AsyncNamespace):
                 | ActionType.LEFT_TRIPLE
                 | ActionType.RIGHT_SINGLE
             ):
-                coords = arg.split(",")
-                if len(coords) >= 2:
-                    x, y = int(coords[0]), int(coords[1])
-                else:
+                coords = parse_coords(arg)
+                if not coords:
                     logger.warning(f"Invalid action coordinates: {arg}")
                     return None
 
                 return await self.call(
                     action.type.value,
-                    ClickEventData(**common, x=x, y=y).model_dump(),
+                    ClickEventData(**common, x=coords[0], y=coords[1]).model_dump(),
                     to=session.socket_id,
                     timeout=self.config.socketio_timeout,
                 )
 
             case ActionType.DRAG:
-                coords = arg.split(",")
-                if len(coords) >= 4:
-                    x1, y1, x2, y2 = (int(coords[i]) for i in range(4))
-                else:
+                coords = parse_drag_coords(arg)
+                if not coords:
                     logger.warning(f"Invalid drag coordinates: {arg}")
                     return None
 
                 return await self.call(
                     "drag",
-                    DragEventData(**common, x1=x1, y1=y1, x2=x2, y2=y2).model_dump(),
+                    DragEventData(
+                        **common, x1=coords[0], y1=coords[1], x2=coords[2], y2=coords[3]
+                    ).model_dump(),
                     to=session.socket_id,
                     timeout=self.config.socketio_timeout,
                 )
@@ -326,11 +330,8 @@ class SessionNamespace(socketio.AsyncNamespace):
                 )
 
             case ActionType.SCROLL:
-                parts = arg.split(",")
-                if len(parts) >= 3:
-                    x, y = int(parts[0]), int(parts[1])
-                    direction = parts[2].strip().lower()
-                else:
+                result = parse_scroll(arg)
+                if not result:
                     logger.warning(f"Invalid scroll coordinates: {arg}")
                     return None
 
@@ -340,9 +341,9 @@ class SessionNamespace(socketio.AsyncNamespace):
                     "scroll",
                     ScrollEventData(
                         **common,
-                        x=x,
-                        y=y,
-                        direction=direction,
+                        x=result[0],
+                        y=result[1],
+                        direction=result[2],
                         count=count,  # type: ignore
                     ).model_dump(),
                     to=session.socket_id,

oagi/types/__init__.py

@@ -11,7 +11,15 @@ from .async_action_handler import AsyncActionHandler
 from .async_image_provider import AsyncImageProvider
 from .image import Image
 from .image_provider import ImageProvider
-from .models import Action, ActionType, ImageConfig, Step
+from .models import (
+    Action,
+    ActionType,
+    ImageConfig,
+    Step,
+    parse_coords,
+    parse_drag_coords,
+    parse_scroll,
+)
 from .step_observer import (
     ActionEvent,
     AsyncObserver,
@@ -47,4 +55,7 @@ __all__ = [
     "ImageProvider",
     "AsyncImageProvider",
     "URL",
+    "parse_coords",
+    "parse_drag_coords",
+    "parse_scroll",
 ]

oagi/types/models/__init__.py

@@ -6,7 +6,13 @@
 #  Licensed under the MIT License.
 # -----------------------------------------------------------------------------
 
-from .action import Action, ActionType
+from .action import (
+    Action,
+    ActionType,
+    parse_coords,
+    parse_drag_coords,
+    parse_scroll,
+)
 from .client import (
     ErrorDetail,
     ErrorResponse,
@@ -29,4 +35,7 @@ __all__ = [
     "Step",
     "UploadFileResponse",
     "Usage",
+    "parse_coords",
+    "parse_drag_coords",
+    "parse_scroll",
 ]

oagi/types/models/action.py

@@ -6,6 +6,7 @@
 #  Licensed under the MIT License.
 # -----------------------------------------------------------------------------
 
+import re
 from enum import Enum
 
 from pydantic import BaseModel, Field
@@ -31,3 +32,53 @@ class Action(BaseModel):
     count: int | None = Field(
         default=1, ge=1, description="Number of times to repeat the action"
     )
+
+
+def parse_coords(args_str: str) -> tuple[int, int] | None:
+    """Extract x, y coordinates from argument string.
+
+    Args:
+        args_str: Argument string in format "x, y" (normalized 0-1000 range)
+
+    Returns:
+        Tuple of (x, y) coordinates, or None if parsing fails
+    """
+    match = re.match(r"(\d+),\s*(\d+)", args_str)
+    if not match:
+        return None
+    return int(match.group(1)), int(match.group(2))
+
+
+def parse_drag_coords(args_str: str) -> tuple[int, int, int, int] | None:
+    """Extract x1, y1, x2, y2 coordinates from drag argument string.
+
+    Args:
+        args_str: Argument string in format "x1, y1, x2, y2" (normalized 0-1000 range)
+
+    Returns:
+        Tuple of (x1, y1, x2, y2) coordinates, or None if parsing fails
+    """
+    match = re.match(r"(\d+),\s*(\d+),\s*(\d+),\s*(\d+)", args_str)
+    if not match:
+        return None
+    return (
+        int(match.group(1)),
+        int(match.group(2)),
+        int(match.group(3)),
+        int(match.group(4)),
+    )
+
+
+def parse_scroll(args_str: str) -> tuple[int, int, str] | None:
+    """Extract x, y, direction from scroll argument string.
+
+    Args:
+        args_str: Argument string in format "x, y, direction" (normalized 0-1000 range)
+
+    Returns:
+        Tuple of (x, y, direction) where direction is "up" or "down", or None if parsing fails
+    """
+    match = re.match(r"(\d+),\s*(\d+),\s*(\w+)", args_str)
+    if not match:
+        return None
+    return int(match.group(1)), int(match.group(2)), match.group(3).lower()

oagi/types/models/client.py

@@ -6,7 +6,7 @@
 #  Licensed under the MIT License.
 # -----------------------------------------------------------------------------
 
-from pydantic import BaseModel
+from pydantic import BaseModel, Field
 
 from .action import Action
 
@@ -56,9 +56,13 @@ class UploadFileResponse(BaseModel):
 
 
 class GenerateResponse(BaseModel):
-    """Response from /v2/generate endpoint."""
+    """Response from /v1/generate endpoint."""
 
     response: str
     prompt_tokens: int
     completion_tokens: int
-    cost: float  # in USD
+    cost: float | None = Field(
+        default=None,
+        deprecated=True,
+        description="This field is deprecated",
+    )

{oagi_core-0.10.0.dist-info → oagi_core-0.10.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: oagi-core
-Version: 0.10.0
+Version: 0.10.2
 Summary: Official API of OpenAGI Foundation
 Project-URL: Homepage, https://github.com/agiopen-org/oagi
 Author-email: OpenAGI Foundation <contact@agiopen.org>
@@ -32,6 +32,7 @@ Requires-Dist: rich>=13.0.0
 Provides-Extra: desktop
 Requires-Dist: pillow>=11.3.0; extra == 'desktop'
 Requires-Dist: pyautogui>=0.9.54; extra == 'desktop'
+Requires-Dist: pyobjc-framework-applicationservices>=9.0; (sys_platform == 'darwin') and extra == 'desktop'
 Requires-Dist: pyobjc-framework-quartz>=9.0; (sys_platform == 'darwin') and extra == 'desktop'
 Provides-Extra: server
 Requires-Dist: fastapi[standard]>=0.115.0; extra == 'server'