oagi-core 0.10.0__tar.gz → 0.10.1__tar.gz

{oagi_core-0.10.0 → oagi_core-0.10.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: oagi-core
-Version: 0.10.0
+Version: 0.10.1
 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'

{oagi_core-0.10.0 → oagi_core-0.10.1}/examples/tasker_agent_example.py

@@ -42,16 +42,10 @@ async def main():
         "Click on the official Python.org website link",
     ]
 
-    # Define deliverables to achieve
-    deliverables = [
-        "Python.org website is opened",
-    ]
-
     # Set the task
     tasker.set_task(
         task=task_description,
         todos=todos,
-        deliverables=deliverables,
     )
 
     image_provider = AsyncScreenshotMaker()

{oagi_core-0.10.0 → oagi_core-0.10.1}/metapackage/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "oagi"
-version = "0.10.0"
+version = "0.10.1"
 description = "Official API of OpenAGI Foundation (metapackage with all features)"
 readme = "README.md"
 license = "MIT"
@@ -16,7 +16,7 @@ authors = [
 requires-python = ">= 3.10"
 
 dependencies = [
-    "oagi-core[desktop,server]==0.10.0",
+    "oagi-core[desktop,server]==0.10.1",
 ]
 
 [project.urls]

{oagi_core-0.10.0 → oagi_core-0.10.1}/metapackage/uv.lock

@@ -397,27 +397,27 @@ sdist = { url = "https://files.pythonhosted.org/packages/28/fa/b2ba8229b9381e8f6
 
 [[package]]
 name = "oagi"
-version = "0.10.0"
+version = "0.10.1"
 source = { editable = "." }
 dependencies = [
     { name = "oagi-core", extra = ["desktop", "server"] },
 ]
 
 [package.metadata]
-requires-dist = [{ name = "oagi-core", extras = ["desktop", "server"], specifier = "==0.9.2" }]
+requires-dist = [{ name = "oagi-core", extras = ["desktop", "server"], specifier = "==0.10.0" }]
 
 [[package]]
 name = "oagi-core"
-version = "0.9.2"
+version = "0.10.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "httpx" },
     { name = "pydantic" },
     { name = "rich" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/4b/68/382a3096645de041f252f5ee8f01d02439f5a419476ab8e84fec638c3592/oagi_core-0.9.2.tar.gz", hash = "sha256:eb18c49f364561c94a26b09c98b855225ac19c19a456170e7a69c1467166f4e5", size = 253368, upload-time = "2025-11-19T04:29:14.85Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/2d/3f/6fa48d2d207051d5f3a8eedf8f7d9cd97f7463746f6578b6655511a4cf47/oagi_core-0.10.0.tar.gz", hash = "sha256:d50c6dac28eb4d395e6910bd24cf412845ffbad25c1be7a2a6a04511691470e9", size = 259893, upload-time = "2025-11-24T04:15:08.503Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/56/15/f6e55511b78622fea2265eb9137d9c03ebbac102b66ebbf49f26bcd713ab/oagi_core-0.9.2-py3-none-any.whl", hash = "sha256:6f14abeae8338f1703fd30ff7e7ae7316be12c025cb3b30307dfb2d35efbf4fc", size = 75648, upload-time = "2025-11-19T04:29:13.961Z" },
+    { url = "https://files.pythonhosted.org/packages/44/9f/582af4d42c1074fe647a3fdbdaddb0bd7b4402ac85e42ea9dcfdf7b52528/oagi_core-0.10.0-py3-none-any.whl", hash = "sha256:f7e19c744586dfb22cfcef0ddc031a457ad4605e6441198cdc6bc4f27bebbb1a", size = 83329, upload-time = "2025-11-24T04:15:07.201Z" },
 ]
 
 [package.optional-dependencies]

{oagi_core-0.10.0 → oagi_core-0.10.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "oagi-core"
-version = "0.10.0"
+version = "0.10.1"
 description = "Official API of OpenAGI Foundation"
 readme = "README.md"
 license = { file = "LICENSE" }
@@ -31,6 +31,7 @@ desktop = [
     "pillow>=11.3.0",
     "pyautogui>=0.9.54",
     "pyobjc-framework-Quartz>=9.0; sys_platform == 'darwin'",
+    "pyobjc-framework-ApplicationServices>=9.0; sys_platform == 'darwin'",
 ]
 server = [
     "fastapi[standard]>=0.115.0",

{oagi_core-0.10.0 → oagi_core-0.10.1}/src/oagi/agent/tasker/__init__.py

@@ -9,7 +9,6 @@
 from .memory import PlannerMemory
 from .models import (
     Action,
-    Deliverable,
     PlannerOutput,
     ReflectionOutput,
     Todo,
@@ -27,7 +26,6 @@ __all__ = [
     "Planner",
     "Todo",
     "TodoStatus",
-    "Deliverable",
     "Action",
     "TodoHistory",
     "PlannerOutput",

{oagi_core-0.10.0 → oagi_core-0.10.1}/src/oagi/agent/tasker/memory.py

@@ -8,7 +8,7 @@
 
 from typing import Any
 
-from .models import Action, Deliverable, Todo, TodoHistory, TodoStatus
+from .models import Action, Todo, TodoHistory, TodoStatus
 
 
 class PlannerMemory:
@@ -16,7 +16,7 @@ class PlannerMemory:
 
     This class manages the hierarchical task execution state for TaskerAgent.
     It provides methods for:
-    - Task/todo/deliverable management
+    - Task/todo management
     - Execution history tracking
     - Memory state serialization
 
@@ -27,7 +27,6 @@ class PlannerMemory:
         """Initialize empty memory."""
         self.task_description: str = ""
         self.todos: list[Todo] = []
-        self.deliverables: list[Deliverable] = []
         self.history: list[TodoHistory] = []
         self.task_execution_summary: str = ""
         self.todo_execution_summaries: dict[int, str] = {}
@@ -36,14 +35,12 @@ class PlannerMemory:
         self,
         task_description: str,
         todos: list[str] | list[Todo],
-        deliverables: list[str] | list[Deliverable] | None = None,
     ) -> None:
-        """Set the task, todos, and deliverables.
+        """Set the task and todos.
 
         Args:
             task_description: Overall task description
             todos: List of todo items (strings or Todo objects)
-            deliverables: Optional list of deliverables (strings or Deliverable objects)
         """
         self.task_description = task_description
 
@@ -55,15 +52,6 @@ class PlannerMemory:
             else:
                 self.todos.append(todo)
 
-        # Convert deliverables
-        self.deliverables = []
-        if deliverables:
-            for deliverable in deliverables:
-                if isinstance(deliverable, str):
-                    self.deliverables.append(Deliverable(description=deliverable))
-                else:
-                    self.deliverables.append(deliverable)
-
     def get_current_todo(self) -> tuple[Todo | None, int]:
         """Get the next pending or in-progress todo.
 
@@ -133,10 +121,6 @@ class PlannerMemory:
                 {"index": i, "description": t.description, "status": t.status}
                 for i, t in enumerate(self.todos)
             ],
-            "deliverables": [
-                {"description": d.description, "achieved": d.achieved}
-                for d in self.deliverables
-            ],
             "history": [
                 {
                     "todo_index": h.todo_index,
@@ -174,11 +158,3 @@ class PlannerMemory:
             description: Description of the new todo
         """
         self.todos.append(Todo(description=description))
-
-    def append_deliverable(self, description: str) -> None:
-        """Append a new deliverable to the list.
-
-        Args:
-            description: Description of the new deliverable
-        """
-        self.deliverables.append(Deliverable(description=description))

{oagi_core-0.10.0 → oagi_core-0.10.1}/src/oagi/agent/tasker/models.py

@@ -28,13 +28,6 @@ class Todo(BaseModel):
     status: TodoStatus = TodoStatus.PENDING
 
 
-class Deliverable(BaseModel):
-    """A deliverable or goal to be achieved."""
-
-    description: str
-    achieved: bool = False
-
-
 class Action(BaseModel):
     """An action taken during execution."""
 

{oagi_core-0.10.0 → oagi_core-0.10.1}/src/oagi/agent/tasker/planner.py

@@ -62,7 +62,7 @@ class Planner:
         memory: PlannerMemory | None,
         context: dict[str, Any],
         todo_index: int | None = None,
-    ) -> tuple[str, list, list, list, str | None, str]:
+    ) -> tuple[str, list, list, str | None, str]:
         """Extract memory data for API calls.
 
         Args:
@@ -71,7 +71,7 @@ class Planner:
             todo_index: Optional todo index for extracting overall_todo
 
         Returns:
-            Tuple of (task_description, todos, deliverables, history,
+            Tuple of (task_description, todos, history,
                      task_execution_summary, overall_todo)
         """
         if memory and todo_index is not None:
@@ -86,7 +86,6 @@ class Planner:
                 }
                 for i, t in enumerate(memory.todos)
             ]
-            deliverables = [d.model_dump() for d in memory.deliverables]
             history = [
                 {
                     "todo_index": h.todo_index,
@@ -103,7 +102,6 @@ class Planner:
             # Fallback to basic context
             task_description = context.get("task_description", "")
             todos = context.get("todos", [])
-            deliverables = context.get("deliverables", [])
             history = context.get("history", [])
             task_execution_summary = None
             overall_todo = context.get("current_todo", "")
@@ -111,7 +109,6 @@ class Planner:
         return (
             task_description,
             todos,
-            deliverables,
             history,
             task_execution_summary,
             overall_todo,
@@ -150,7 +147,6 @@ class Planner:
         (
             task_description,
             todos,
-            deliverables,
             history,
             task_execution_summary,
             _,  # overall_todo not needed here, we use the `todo` parameter
@@ -162,7 +158,6 @@ class Planner:
             overall_todo=todo,
             task_description=task_description,
             todos=todos,
-            deliverables=deliverables,
             history=history,
             current_todo_index=todo_index,
             task_execution_summary=task_execution_summary,
@@ -209,7 +204,6 @@ class Planner:
         (
             task_description,
             todos,
-            deliverables,
             history,
             task_execution_summary,
             overall_todo,
@@ -245,7 +239,6 @@ class Planner:
             overall_todo=overall_todo,
             task_description=task_description,
             todos=todos,
-            deliverables=deliverables,
             history=history,
             current_todo_index=todo_index,
             task_execution_summary=task_execution_summary,
@@ -284,7 +277,6 @@ class Planner:
         (
             task_description,
             todos,
-            deliverables,
             history,
             task_execution_summary,
             overall_todo,
@@ -302,7 +294,6 @@ class Planner:
             overall_todo=overall_todo,
             task_description=task_description,
             todos=todos,
-            deliverables=deliverables,
             history=history,
             current_todo_index=todo_index,
             task_execution_summary=task_execution_summary,

{oagi_core-0.10.0 → oagi_core-0.10.1}/src/oagi/agent/tasker/tasker_agent.py

@@ -73,20 +73,15 @@ 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,
@@ -327,12 +322,3 @@ 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_core-0.10.0 → oagi_core-0.10.1}/src/oagi/cli/agent.py

@@ -66,10 +66,106 @@ def add_agent_parser(subparsers: argparse._SubParsersAction) -> None:
         help="Output file path for export (default: execution_report.[md|html|json])",
     )
 
+    # 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 +173,9 @@ 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
 

{oagi_core-0.10.0 → oagi_core-0.10.1}/src/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_core-0.10.0 → oagi_core-0.10.1}/src/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_core-0.10.0 → oagi_core-0.10.1}/src/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_core-0.10.0 → oagi_core-0.10.1}/src/oagi/handler/pyautogui_action_handler.py

@@ -171,8 +171,14 @@ class PyautoguiActionHandler:
         """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_core-0.10.0 → oagi_core-0.10.1}/src/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 → oagi_core-0.10.1}/tests/test_planner.py

@@ -30,7 +30,6 @@ class TestPlanner:
         memory.set_task(
             task_description="Test task",
             todos=["Todo 1", "Todo 2"],
-            deliverables=["Deliverable 1"],
         )
         return memory
 
@@ -48,7 +47,7 @@ class TestPlanner:
             response='{"reasoning": "Start with clicking", "subtask": "Click on button"}',
             prompt_tokens=100,
             completion_tokens=50,
-            cost=0.001,
+            cost=0.0,
         )
 
         screenshot = b"fake_image_bytes"
@@ -72,7 +71,7 @@ class TestPlanner:
             response='{"reasoning": "No visual needed", "subtask": "Type text"}',
             prompt_tokens=100,
             completion_tokens=50,
-            cost=0.001,
+            cost=0.0,
         )
 
         result = await planner.initial_plan(
@@ -111,7 +110,7 @@ class TestPlanner:
             response='{"assessment": "Good progress", "success": "yes", "reflection": "Task completed", "subtask_instruction": ""}',
             prompt_tokens=150,
             completion_tokens=75,
-            cost=0.002,
+            cost=0.0,
         )
 
         result = await planner.reflect(
@@ -133,7 +132,7 @@ class TestPlanner:
             response='{"success": "no", "subtask_instruction": "Try different approach", "reflection": "Need to pivot"}',
             prompt_tokens=100,
             completion_tokens=50,
-            cost=0.001,
+            cost=0.0,
         )
 
         result = await planner.reflect(
@@ -152,7 +151,7 @@ class TestPlanner:
             response='{"task_summary": "Successfully completed the task"}',
             prompt_tokens=80,
             completion_tokens=20,
-            cost=0.001,
+            cost=0.0,
         )
 
         result = await planner.summarize(