oagi-core 0.9.1__py3-none-any.whl → 0.10.0__py3-none-any.whl

oagi/client/base.py

@@ -41,16 +41,12 @@ class BaseClient(Generic[HttpClientT]):
 
     def __init__(self, base_url: str | None = None, api_key: str | None = None):
         # Get from environment if not provided
-        self.base_url = base_url or os.getenv("OAGI_BASE_URL")
+        self.base_url = (
+            base_url or os.getenv("OAGI_BASE_URL") or "https://api.agiopen.org"
+        )
         self.api_key = api_key or os.getenv("OAGI_API_KEY")
 
         # Validate required configuration
-        if not self.base_url:
-            raise ConfigurationError(
-                "OAGI base URL must be provided either as 'base_url' parameter or "
-                "OAGI_BASE_URL environment variable"
-            )
-
         if not self.api_key:
             raise ConfigurationError(
                 "OAGI API key must be provided either as 'api_key' parameter or "

oagi/handler/_macos.py

@@ -0,0 +1,55 @@
+# -----------------------------------------------------------------------------
+#  Copyright (c) OpenAGI Foundation
+#  All rights reserved.
+#
+#  This file is part of the official API project.
+#  Licensed under the MIT License.
+# -----------------------------------------------------------------------------
+
+import pyautogui
+
+from ..exceptions import check_optional_dependency
+
+check_optional_dependency("Quartz", "macOS multiple clicks", "desktop")
+import Quartz  # noqa: E402
+
+
+def macos_click(x: int, y: int, clicks: int = 1) -> None:
+    """
+    Execute a mouse click sequence on macOS with correct click state.
+
+    This avoids the PyAutoGUI bug where multi-clicks are sent as separate
+    single clicks (clickState=1), which macOS interprets as distinct events
+    rather than double/triple clicks.
+
+    Check https://github.com/asweigart/pyautogui/issues/672
+
+    Args:
+        x: X coordinate
+        y: Y coordinate
+        clicks: Number of clicks (1=single, 2=double, 3=triple)
+    """
+    # Move to position first using pyautogui to ensure consistency
+    pyautogui.moveTo(x, y)
+
+    point = Quartz.CGPoint(x=x, y=y)
+
+    # Create and post events for each click in the sequence
+    for i in range(1, clicks + 1):
+        # Create Down/Up events
+        mouse_down = Quartz.CGEventCreateMouseEvent(
+            None, Quartz.kCGEventLeftMouseDown, point, Quartz.kCGMouseButtonLeft
+        )
+        mouse_up = Quartz.CGEventCreateMouseEvent(
+            None, Quartz.kCGEventLeftMouseUp, point, Quartz.kCGMouseButtonLeft
+        )
+
+        # Set the click state (1 for first click, 2 for second, etc.)
+        Quartz.CGEventSetIntegerValueField(
+            mouse_down, Quartz.kCGMouseEventClickState, i
+        )
+        Quartz.CGEventSetIntegerValueField(mouse_up, Quartz.kCGMouseEventClickState, i)
+
+        # Post events
+        Quartz.CGEventPost(Quartz.kCGHIDEventTap, mouse_down)
+        Quartz.CGEventPost(Quartz.kCGHIDEventTap, mouse_up)

oagi/handler/pyautogui_action_handler.py

@@ -7,6 +7,7 @@
 # -----------------------------------------------------------------------------
 
 import re
+import sys
 import time
 
 from pydantic import BaseModel, Field
@@ -17,6 +18,9 @@ from ..types import Action, ActionType
 check_optional_dependency("pyautogui", "PyautoguiActionHandler", "desktop")
 import pyautogui  # noqa: E402
 
+if sys.platform == "darwin":
+    from . import _macos
+
 
 class CapsLockManager:
     """Manages caps lock state for text transformation."""
@@ -76,6 +80,10 @@ class PyautoguiConfig(BaseModel):
         default="session",
         description="Caps lock handling mode: 'session' (internal state) or 'system' (OS-level)",
     )
+    macos_ctrl_to_cmd: bool = Field(
+        default=True,
+        description="Replace 'ctrl' with 'command' in hotkey combinations on macOS",
+    )
 
 
 class PyautoguiActionHandler:
@@ -165,6 +173,9 @@ class PyautoguiActionHandler:
         # Normalize caps lock variations
         if key in ["caps_lock", "caps", "capslock"]:
             return "capslock"
+        # Remap ctrl to command on macOS if enabled
+        if self.config.macos_ctrl_to_cmd and sys.platform == "darwin" and key == "ctrl":
+            return "command"
         return key
 
     def _parse_hotkey(self, args_str: str) -> list[str]:
@@ -186,11 +197,17 @@ class PyautoguiActionHandler:
 
             case ActionType.LEFT_DOUBLE:
                 x, y = self._parse_coords(arg)
-                pyautogui.doubleClick(x, y)
+                if sys.platform == "darwin":
+                    _macos.macos_click(x, y, clicks=2)
+                else:
+                    pyautogui.doubleClick(x, y)
 
             case ActionType.LEFT_TRIPLE:
                 x, y = self._parse_coords(arg)
-                pyautogui.tripleClick(x, y)
+                if sys.platform == "darwin":
+                    _macos.macos_click(x, y, clicks=3)
+                else:
+                    pyautogui.tripleClick(x, y)
 
             case ActionType.RIGHT_SINGLE:
                 x, y = self._parse_coords(arg)

oagi/server/agent_wrappers.py

@@ -9,7 +9,7 @@
 import logging
 from typing import TYPE_CHECKING
 
-from ..types import URLImage
+from ..types import URL
 from ..types.models.action import Action
 from .models import ScreenshotRequestData, ScreenshotResponseData
 
@@ -56,7 +56,7 @@ class SocketIOImageProvider:
         self.oagi_client = oagi_client
         self._last_url: str | None = None
 
-    async def __call__(self) -> URLImage:
+    async def __call__(self) -> URL:
         logger.debug("Requesting screenshot via Socket.IO")
 
         # Get S3 presigned URL from OAGI
@@ -87,12 +87,12 @@ class SocketIOImageProvider:
         self.session.current_screenshot_url = upload_response.download_url
 
         logger.debug(f"Screenshot captured successfully: {upload_response.uuid}")
-        return URLImage(upload_response.download_url)
+        return URL(upload_response.download_url)
 
-    async def last_image(self) -> URLImage:
+    async def last_image(self) -> URL:
         if self._last_url:
             logger.debug("Returning last captured screenshot")
-            return URLImage(self._last_url)
+            return URL(self._last_url)
 
         logger.debug("No previous screenshot, capturing new one")
         return await self()

oagi/server/config.py

@@ -28,11 +28,11 @@ class ServerConfig(BaseSettings):
     session_timeout_seconds: float = Field(default=10.0)
 
     # Model settings
-    default_model: str = Field(default="lux-v1", alias="OAGI_DEFAULT_MODEL")
-    default_temperature: float = Field(default=0.0, ge=0.0, le=2.0)
+    default_model: str = Field(default="lux-actor-1", alias="OAGI_DEFAULT_MODEL")
+    default_temperature: float = Field(default=0.5, ge=0.0, le=2.0)
 
     # Agent settings
-    max_steps: int = Field(default=30, alias="OAGI_MAX_STEPS", ge=1, le=100)
+    max_steps: int = Field(default=20, alias="OAGI_MAX_STEPS", ge=1, le=100)
 
     # Socket.IO settings
     socketio_path: str = Field(default="/socket.io")

oagi/server/models.py

@@ -15,7 +15,7 @@ from pydantic import BaseModel, Field
 class InitEventData(BaseModel):
     instruction: str = Field(...)
     mode: str | None = Field(default="actor")
-    model: str | None = Field(default="lux-v1")
+    model: str | None = Field(default="lux-actor-1")
     temperature: float | None = Field(default=0.1, ge=0.0, le=2.0)
 
 
@@ -75,7 +75,7 @@ class ScreenshotResponseData(BaseModel):
 
 # Action acknowledgement
 class ActionAckData(BaseModel):
-    action_index: int = Field(...)
+    index: int = Field(...)
     success: bool = Field(...)
     error: str | None = Field(None)
     execution_time_ms: int | None = Field(None)

oagi/server/session_store.py

@@ -18,7 +18,7 @@ class Session:
         session_id: str,
         instruction: str,
         mode: str = "actor",
-        model: str = "lux-v1",
+        model: str = "lux-actor-1",
         temperature: float = 0.0,
     ):
         self.session_id: str = session_id
@@ -54,7 +54,7 @@ class SessionStore:
         self,
         instruction: str,
         mode: str = "actor",
-        model: str = "lux-v1",
+        model: str = "lux-actor-1",
         temperature: float = 0.0,
         session_id: str | None = None,
     ) -> str:

oagi/server/socketio_server.py

@@ -224,7 +224,7 @@ class SessionNamespace(socketio.AsyncNamespace):
                 # Emit finish event
                 await self.call(
                     "finish",
-                    FinishEventData(action_index=0, total_actions=1).model_dump(),
+                    FinishEventData(index=0, total=1).model_dump(),
                     to=session.socket_id,
                     timeout=self.config.socketio_timeout,
                 )

oagi/task/async_.py

@@ -9,21 +9,18 @@
 import warnings
 
 from ..client import AsyncClient
-from ..logging import get_logger
-from ..types import Image, Step
-from .base import BaseTask
+from ..types import URL, Image, Step
+from .base import BaseActor
 
-logger = get_logger("async_task")
 
-
-class AsyncActor(BaseTask):
+class AsyncActor(BaseActor):
     """Async base class for task automation with the OAGI API."""
 
     def __init__(
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "vision-model-v1",
+        model: str = "lux-actor-1",
         temperature: float | None = None,
     ):
         super().__init__(api_key, base_url, model, temperature)
@@ -34,19 +31,19 @@ class AsyncActor(BaseTask):
     async def init_task(
         self,
         task_desc: str,
-        max_steps: int = 5,
+        max_steps: int = 20,
     ):
         """Initialize a new task with the given description.
 
         Args:
             task_desc: Task description
-            max_steps: Maximum number of steps (for logging)
+            max_steps: Maximum number of steps allowed
         """
         self._prepare_init_task(task_desc, max_steps)
 
     async def step(
         self,
-        screenshot: Image | bytes,
+        screenshot: Image | URL | bytes,
         instruction: str | None = None,
         temperature: float | None = None,
     ) -> Step:
@@ -60,33 +57,15 @@ class AsyncActor(BaseTask):
         Returns:
             Step: The actions and reasoning for this step
         """
-        self._validate_step_preconditions()
-        self._log_step_execution(prefix="async ")
+        kwargs = self._prepare_step(
+            screenshot, instruction, temperature, prefix="async "
+        )
 
         try:
-            # Use provided temperature or fall back to task default
-            temp = self._get_temperature(temperature)
-
-            # Prepare screenshot kwargs (handles URLImage vs bytes/Image)
-            screenshot_kwargs = self._prepare_screenshot_kwargs(screenshot)
-
-            # Call API with dynamically determined screenshot argument
-            response = await self.client.create_message(
-                model=self.model,
-                task_description=self.task_description,
-                task_id=self.task_id,
-                instruction=instruction,
-                messages_history=self.message_history,
-                temperature=temp,
-                **screenshot_kwargs,
-            )
-
-            # Convert API response to Step (also updates message_history)
+            response = await self.client.create_message(**kwargs)
             return self._build_step_response(response, prefix="Async ")
-
         except Exception as e:
-            logger.error(f"Error during async step execution: {e}")
-            raise
+            self._handle_step_error(e, prefix="async ")
 
     async def close(self):
         """Close the underlying HTTP client to free resources."""
@@ -110,7 +89,7 @@ class AsyncTask(AsyncActor):
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "vision-model-v1",
+        model: str = "lux-actor-1",
         temperature: float | None = None,
     ):
         warnings.warn(

oagi/task/async_short.py

@@ -27,7 +27,7 @@ class AsyncShortTask(AsyncActor, BaseAutoMode):
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "vision-model-v1",
+        model: str = "lux-actor-1",
         temperature: float | None = None,
     ):
         warnings.warn(
@@ -43,7 +43,7 @@ class AsyncShortTask(AsyncActor, BaseAutoMode):
     async def auto_mode(
         self,
         task_desc: str,
-        max_steps: int = 5,
+        max_steps: int = 20,
         executor: AsyncActionHandler = None,
         image_provider: AsyncImageProvider = None,
         temperature: float | None = None,

oagi/task/base.py

@@ -9,14 +9,14 @@
 from uuid import uuid4
 
 from ..logging import get_logger
-from ..types import Image, Step, URLImage
+from ..types import URL, Image, Step
 from ..types.models import LLMResponse
 
 logger = get_logger("task.base")
 
 
-class BaseTask:
-    """Base class with shared task management logic for sync/async tasks."""
+class BaseActor:
+    """Base class with shared task management logic for sync/async actors."""
 
     def __init__(
         self,
@@ -30,6 +30,8 @@ class BaseTask:
         self.model = model
         self.temperature = temperature
         self.message_history: list = []  # OpenAI-compatible message history
+        self.max_steps: int = 20  # Maximum steps allowed
+        self.current_step: int = 0  # Current step counter
         # Client will be set by subclasses
         self.api_key: str | None = None
         self.base_url: str | None = None
@@ -48,11 +50,43 @@ class BaseTask:
         self.task_id = uuid4().hex
         self.task_description = task_desc
         self.message_history = []
+        self.max_steps = max_steps
+        self.current_step = 0
         logger.info(f"Task initialized: '{task_desc}' (max_steps: {max_steps})")
 
-    def _validate_step_preconditions(self):
+    def _validate_and_increment_step(self):
         if not self.task_description:
             raise ValueError("Task description must be set. Call init_task() first.")
+        if self.current_step >= self.max_steps:
+            raise ValueError(
+                f"Max steps limit ({self.max_steps}) reached. "
+                "Call init_task() to start a new task."
+            )
+        self.current_step += 1
+
+    def _prepare_step(
+        self,
+        screenshot: Image | URL | bytes,
+        instruction: str | None,
+        temperature: float | None,
+        prefix: str = "",
+    ) -> dict:
+        self._validate_and_increment_step()
+        self._log_step_execution(prefix=prefix)
+
+        return {
+            "model": self.model,
+            "task_description": self.task_description,
+            "task_id": self.task_id,
+            "instruction": instruction,
+            "messages_history": self.message_history,
+            "temperature": self._get_temperature(temperature),
+            **self._prepare_screenshot_kwargs(screenshot),
+        }
+
+    def _handle_step_error(self, error: Exception, prefix: str = ""):
+        logger.error(f"Error during {prefix}step execution: {error}")
+        raise
 
     def _prepare_screenshot(self, screenshot: Image | bytes) -> bytes:
         if isinstance(screenshot, Image):
@@ -62,9 +96,9 @@ class BaseTask:
     def _get_temperature(self, temperature: float | None) -> float | None:
         return temperature if temperature is not None else self.temperature
 
-    def _prepare_screenshot_kwargs(self, screenshot: Image | bytes) -> dict:
-        if isinstance(screenshot, URLImage):
-            return {"screenshot_url": screenshot.get_url()}
+    def _prepare_screenshot_kwargs(self, screenshot: Image | URL | bytes) -> dict:
+        if isinstance(screenshot, str):
+            return {"screenshot_url": screenshot}
         return {"screenshot": self._prepare_screenshot(screenshot)}
 
     def _handle_response_message_history(self, response: LLMResponse):

oagi/task/short.py

@@ -27,7 +27,7 @@ class ShortTask(Actor, BaseAutoMode):
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "vision-model-v1",
+        model: str = "lux-actor-1",
         temperature: float | None = None,
     ):
         warnings.warn(
@@ -43,7 +43,7 @@ class ShortTask(Actor, BaseAutoMode):
     def auto_mode(
         self,
         task_desc: str,
-        max_steps: int = 5,
+        max_steps: int = 20,
         executor: ActionHandler = None,
         image_provider: ImageProvider = None,
         temperature: float | None = None,

oagi/task/sync.py

@@ -9,21 +9,18 @@
 import warnings
 
 from ..client import SyncClient
-from ..logging import get_logger
-from ..types import Image, Step
-from .base import BaseTask
+from ..types import URL, Image, Step
+from .base import BaseActor
 
-logger = get_logger("task")
 
-
-class Actor(BaseTask):
+class Actor(BaseActor):
     """Base class for task automation with the OAGI API."""
 
     def __init__(
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "vision-model-v1",
+        model: str = "lux-actor-1",
         temperature: float | None = None,
     ):
         super().__init__(api_key, base_url, model, temperature)
@@ -34,19 +31,19 @@ class Actor(BaseTask):
     def init_task(
         self,
         task_desc: str,
-        max_steps: int = 5,
+        max_steps: int = 20,
     ):
         """Initialize a new task with the given description.
 
         Args:
             task_desc: Task description
-            max_steps: Maximum number of steps (for logging)
+            max_steps: Maximum number of steps allowed
         """
         self._prepare_init_task(task_desc, max_steps)
 
     def step(
         self,
-        screenshot: Image | bytes,
+        screenshot: Image | URL | bytes,
         instruction: str | None = None,
         temperature: float | None = None,
     ) -> Step:
@@ -60,33 +57,13 @@ class Actor(BaseTask):
         Returns:
             Step: The actions and reasoning for this step
         """
-        self._validate_step_preconditions()
-        self._log_step_execution()
+        kwargs = self._prepare_step(screenshot, instruction, temperature)
 
         try:
-            # Use provided temperature or fall back to task default
-            temp = self._get_temperature(temperature)
-
-            # Prepare screenshot kwargs (handles URLImage vs bytes/Image)
-            screenshot_kwargs = self._prepare_screenshot_kwargs(screenshot)
-
-            # Call API with dynamically determined screenshot argument
-            response = self.client.create_message(
-                model=self.model,
-                task_description=self.task_description,
-                task_id=self.task_id,
-                instruction=instruction,
-                messages_history=self.message_history,
-                temperature=temp,
-                **screenshot_kwargs,
-            )
-
-            # Convert API response to Step (also updates message_history)
+            response = self.client.create_message(**kwargs)
             return self._build_step_response(response)
-
         except Exception as e:
-            logger.error(f"Error during step execution: {e}")
-            raise
+            self._handle_step_error(e)
 
     def close(self):
         """Close the underlying HTTP client to free resources."""
@@ -110,7 +87,7 @@ class Task(Actor):
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "vision-model-v1",
+        model: str = "lux-actor-1",
         temperature: float | None = None,
     ):
         warnings.warn(

oagi/types/__init__.py

@@ -12,19 +12,39 @@ from .async_image_provider import AsyncImageProvider
 from .image import Image
 from .image_provider import ImageProvider
 from .models import Action, ActionType, ImageConfig, Step
-from .step_observer import AsyncStepObserver
-from .url_image import URLImage
+from .step_observer import (
+    ActionEvent,
+    AsyncObserver,
+    AsyncStepObserver,
+    BaseEvent,
+    ImageEvent,
+    LogEvent,
+    ObserverEvent,
+    PlanEvent,
+    SplitEvent,
+    StepEvent,
+)
+from .url import URL
 
 __all__ = [
     "Action",
+    "ActionEvent",
     "ActionType",
+    "AsyncObserver",
+    "AsyncStepObserver",
+    "BaseEvent",
     "Image",
     "ImageConfig",
+    "ImageEvent",
+    "LogEvent",
+    "ObserverEvent",
+    "PlanEvent",
+    "SplitEvent",
     "Step",
+    "StepEvent",
     "ActionHandler",
     "AsyncActionHandler",
     "ImageProvider",
     "AsyncImageProvider",
-    "AsyncStepObserver",
-    "URLImage",
+    "URL",
 ]

oagi/types/async_image_provider.py

@@ -9,10 +9,11 @@
 from typing import Protocol
 
 from .image import Image
+from .url import URL
 
 
 class AsyncImageProvider(Protocol):
-    async def __call__(self) -> Image:
+    async def __call__(self) -> Image | URL:
         """
         Asynchronously provides an image.
 
@@ -28,7 +29,7 @@ class AsyncImageProvider(Protocol):
             RuntimeError: If an error occurs during image capture or generation.
         """
 
-    async def last_image(self) -> Image:
+    async def last_image(self) -> Image | URL:
         """
         Asynchronously returns the last captured image.
 

oagi/types/image_provider.py

@@ -9,10 +9,11 @@
 from typing import Protocol
 
 from .image import Image
+from .url import URL
 
 
 class ImageProvider(Protocol):
-    def __call__(self) -> Image:
+    def __call__(self) -> Image | URL:
         """
         Represents the functionality to invoke the callable object and produce an Image
         result. Typically used to process or generate images using the defined logic
@@ -22,7 +23,7 @@ class ImageProvider(Protocol):
             Image: The resulting image output from the callable logic.
         """
 
-    def last_image(self) -> Image:
+    def last_image(self) -> Image | URL:
         """
         Returns the last captured image.