oagi-core 0.10.2__tar.gz → 0.11.0__tar.gz

{oagi_core-0.10.2 → oagi_core-0.11.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: oagi-core
-Version: 0.10.2
+Version: 0.11.0
 Summary: Official API of OpenAGI Foundation
 Project-URL: Homepage, https://github.com/agiopen-org/oagi
 Author-email: OpenAGI Foundation <contact@agiopen.org>
@@ -116,6 +116,39 @@ config = PyautoguiConfig(
 action_handler = AsyncPyautoguiActionHandler(config=config)
 ```
 
+### Command Line Interface
+
+Run agents directly from the terminal:
+
+```bash
+# Run with actor model
+oagi agent run "Go to nasdaq.com, search for AAPL. Under More, go to Insider Activity" --model lux-actor-1
+
+# Run with thinker mode (uses lux-thinker-1 model with more steps)
+oagi agent run "Look up the store hours for the nearest Apple Store to zip code 23456 using the Apple Store Locator" --model lux-thinker-1
+
+# Run pre-configured tasker workflows (no instruction needed)
+oagi agent run --mode tasker:software_qa
+
+# List all available modes
+oagi agent modes
+
+# Check macOS permissions (screen recording & accessibility)
+oagi agent permission
+
+# Export execution history
+oagi agent run "Complete the form" --export html --export-file report.html
+```
+
+CLI options:
+- `--mode`: Agent mode (default: actor). Use `oagi agent modes` to list available modes
+- `--model`: Override the model (default: determined by mode)
+- `--max-steps`: Maximum steps (default: determined by mode)
+- `--temperature`: Sampling temperature (default: determined by mode)
+- `--step-delay`: Delay after each action before next screenshot (default: 0.3s)
+- `--export`: Export format (markdown, html, json)
+- `--export-file`: Output file path for export
+
 ### Image Processing
 
 Process and optimize images before sending to API:

{oagi_core-0.10.2 → oagi_core-0.11.0}/README.md

@@ -73,6 +73,39 @@ config = PyautoguiConfig(
 action_handler = AsyncPyautoguiActionHandler(config=config)
 ```
 
+### Command Line Interface
+
+Run agents directly from the terminal:
+
+```bash
+# Run with actor model
+oagi agent run "Go to nasdaq.com, search for AAPL. Under More, go to Insider Activity" --model lux-actor-1
+
+# Run with thinker mode (uses lux-thinker-1 model with more steps)
+oagi agent run "Look up the store hours for the nearest Apple Store to zip code 23456 using the Apple Store Locator" --model lux-thinker-1
+
+# Run pre-configured tasker workflows (no instruction needed)
+oagi agent run --mode tasker:software_qa
+
+# List all available modes
+oagi agent modes
+
+# Check macOS permissions (screen recording & accessibility)
+oagi agent permission
+
+# Export execution history
+oagi agent run "Complete the form" --export html --export-file report.html
+```
+
+CLI options:
+- `--mode`: Agent mode (default: actor). Use `oagi agent modes` to list available modes
+- `--model`: Override the model (default: determined by mode)
+- `--max-steps`: Maximum steps (default: determined by mode)
+- `--temperature`: Sampling temperature (default: determined by mode)
+- `--step-delay`: Delay after each action before next screenshot (default: 0.3s)
+- `--export`: Export format (markdown, html, json)
+- `--export-file`: Output file path for export
+
 ### Image Processing
 
 Process and optimize images before sending to API:

{oagi_core-0.10.2 → oagi_core-0.11.0}/examples/tasker_agent_example.py

@@ -27,7 +27,7 @@ async def main():
     tasker = TaskerAgent(
         api_key=os.getenv("OAGI_API_KEY"),
         base_url=os.getenv("OAGI_BASE_URL", "https://api.agiopen.org"),
-        model="sft-bigs1-1027-s2-1113-mixoc-1107-32b",
+        model="lux-actor-1",
         max_steps=30,
         temperature=0.5,
         step_observer=observer,

{oagi_core-0.10.2 → oagi_core-0.11.0}/metapackage/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "oagi"
-version = "0.10.2"
+version = "0.11.0"
 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]==0.10.2",
+    "oagi-core[desktop,server]==0.11.0",
 ]
 
 [project.urls]

{oagi_core-0.10.2 → oagi_core-0.11.0}/metapackage/uv.lock

@@ -397,27 +397,27 @@ sdist = { url = "https://files.pythonhosted.org/packages/28/fa/b2ba8229b9381e8f6
 
 [[package]]
 name = "oagi"
-version = "0.10.2"
+version = "0.11.0"
 source = { editable = "." }
 dependencies = [
     { name = "oagi-core", extra = ["desktop", "server"] },
 ]
 
 [package.metadata]
-requires-dist = [{ name = "oagi-core", extras = ["desktop", "server"], specifier = "==0.10.1" }]
+requires-dist = [{ name = "oagi-core", extras = ["desktop", "server"], specifier = "==0.10.3" }]
 
 [[package]]
 name = "oagi-core"
-version = "0.10.1"
+version = "0.10.3"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "httpx" },
     { name = "pydantic" },
     { name = "rich" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/9c/2d/83e5e2efe2591858e2aceaed4329d995c78cc9602b6a9fd9331096559351/oagi_core-0.10.1.tar.gz", hash = "sha256:264b88fb5b1e24f7284af3e82243d846817f6b1d0be108280491383e671e5b8d", size = 261938, upload-time = "2025-11-25T13:19:31.714Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/60/97/c54ecd43969132f902086d4f7fa7eb8d1f5e5087774c81df463a27b3017f/oagi_core-0.10.3.tar.gz", hash = "sha256:46417fde3b20427338d7e2798246960fd4bc6515e93f95bd32b8236337d6cfcd", size = 268408, upload-time = "2025-11-30T11:34:46.494Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/54/bb/3242e763f93360eeb6a5a5b1e61845cd3e27a485c442aeb9a918f6e4f3f6/oagi_core-0.10.1-py3-none-any.whl", hash = "sha256:5438e4310fab1aee3ffbe3d43963f04fc10d3747d24897b89f34df850bb76f89", size = 83830, upload-time = "2025-11-25T13:19:30.237Z" },
+    { url = "https://files.pythonhosted.org/packages/81/22/47271b7d2ac5b7bcaa9d58819b00dfa8e3c0aa57ea3db6e2f6fd681fd0ce/oagi_core-0.10.3-py3-none-any.whl", hash = "sha256:fc91a7bb29ffdcf490bb7edc2574826229ba86726ec6acd505379bdf09645721", size = 88927, upload-time = "2025-11-30T11:34:45.531Z" },
 ]
 
 [package.optional-dependencies]

{oagi_core-0.10.2 → oagi_core-0.11.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "oagi-core"
-version = "0.10.2"
+version = "0.11.0"
 description = "Official API of OpenAGI Foundation"
 readme = "README.md"
 license = { file = "LICENSE" }

{oagi_core-0.10.2 → oagi_core-0.11.0}/src/oagi/agent/default.py

@@ -10,6 +10,13 @@ import asyncio
 import logging
 
 from .. import AsyncActor
+from ..constants import (
+    DEFAULT_MAX_STEPS,
+    DEFAULT_STEP_DELAY,
+    DEFAULT_TEMPERATURE,
+    MODEL_ACTOR,
+)
+from ..handler import reset_handler
 from ..types import (
     ActionEvent,
     AsyncActionHandler,
@@ -36,11 +43,11 @@ class AsyncDefaultAgent:
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "lux-actor-1",
-        max_steps: int = 20,
-        temperature: float | None = 0.5,
+        model: str = MODEL_ACTOR,
+        max_steps: int = DEFAULT_MAX_STEPS,
+        temperature: float | None = DEFAULT_TEMPERATURE,
         step_observer: AsyncObserver | None = None,
-        step_delay: float = 0.3,
+        step_delay: float = DEFAULT_STEP_DELAY,
     ):
         self.api_key = api_key
         self.base_url = base_url
@@ -62,6 +69,9 @@ class AsyncDefaultAgent:
             logger.info(f"Starting async task execution: {instruction}")
             await self.actor.init_task(instruction, max_steps=self.max_steps)
 
+            # Reset handler state at automation start
+            reset_handler(action_handler)
+
             for i in range(self.max_steps):
                 step_num = i + 1
                 logger.debug(f"Executing step {step_num}/{self.max_steps}")

oagi_core-0.11.0/src/oagi/agent/factories.py

@@ -0,0 +1,162 @@
+# -----------------------------------------------------------------------------
+#  Copyright (c) OpenAGI Foundation
+#  All rights reserved.
+#
+#  This file is part of the official API project.
+#  Licensed under the MIT License.
+# -----------------------------------------------------------------------------
+from oagi.agent.tasker import TaskerAgent
+from oagi.constants import (
+    DEFAULT_MAX_STEPS,
+    DEFAULT_MAX_STEPS_TASKER,
+    DEFAULT_MAX_STEPS_THINKER,
+    DEFAULT_REFLECTION_INTERVAL_TASKER,
+    DEFAULT_STEP_DELAY,
+    DEFAULT_TEMPERATURE_LOW,
+    MODEL_ACTOR,
+    MODEL_THINKER,
+)
+from oagi.types import AsyncStepObserver
+
+from .default import AsyncDefaultAgent
+from .protocol import AsyncAgent
+from .registry import async_agent_register
+
+
+@async_agent_register(mode="actor")
+def create_default_agent(
+    api_key: str | None = None,
+    base_url: str | None = None,
+    model: str = MODEL_ACTOR,
+    max_steps: int = DEFAULT_MAX_STEPS,
+    temperature: float = DEFAULT_TEMPERATURE_LOW,
+    step_observer: AsyncStepObserver | None = None,
+    step_delay: float = DEFAULT_STEP_DELAY,
+) -> AsyncAgent:
+    return AsyncDefaultAgent(
+        api_key=api_key,
+        base_url=base_url,
+        model=model,
+        max_steps=max_steps,
+        temperature=temperature,
+        step_observer=step_observer,
+        step_delay=step_delay,
+    )
+
+
+@async_agent_register(mode="thinker")
+def create_thinker_agent(
+    api_key: str | None = None,
+    base_url: str | None = None,
+    model: str = MODEL_THINKER,
+    max_steps: int = DEFAULT_MAX_STEPS_THINKER,
+    temperature: float = DEFAULT_TEMPERATURE_LOW,
+    step_observer: AsyncStepObserver | None = None,
+    step_delay: float = DEFAULT_STEP_DELAY,
+) -> AsyncAgent:
+    return AsyncDefaultAgent(
+        api_key=api_key,
+        base_url=base_url,
+        model=model,
+        max_steps=max_steps,
+        temperature=temperature,
+        step_observer=step_observer,
+        step_delay=step_delay,
+    )
+
+
+@async_agent_register(mode="tasker:cvs_appointment")
+def create_cvs_appointment_agent(
+    api_key: str | None = None,
+    base_url: str | None = None,
+    model: str = MODEL_ACTOR,
+    max_steps: int = DEFAULT_MAX_STEPS_TASKER,
+    temperature: float = DEFAULT_TEMPERATURE_LOW,
+    reflection_interval: int = DEFAULT_REFLECTION_INTERVAL_TASKER,
+    step_observer: AsyncStepObserver | None = None,
+    step_delay: float = DEFAULT_STEP_DELAY,
+    # CVS-specific parameters
+    first_name: str = "First",
+    last_name: str = "Last",
+    email: str = "user@example.com",
+    birthday: str = "01-01-1990",  # MM-DD-YYYY
+    zip_code: str = "00000",
+) -> AsyncAgent:
+    tasker = TaskerAgent(
+        api_key=api_key,
+        base_url=base_url,
+        model=model,
+        max_steps=max_steps,
+        temperature=temperature,
+        reflection_interval=reflection_interval,
+        step_observer=step_observer,
+        step_delay=step_delay,
+    )
+
+    month, day, year = birthday.split("-")
+    instruction = (
+        f"Schedule an appointment at CVS for {first_name} {last_name} "
+        f"with email {email} and birthday {birthday}"
+    )
+    todos = [
+        "Open a new tab, go to www.cvs.com, type 'flu shot' in the search bar and press enter, "
+        "wait for the page to load, then click on the button of Schedule vaccinations on the "
+        "top of the page",
+        f"Enter the first name '{first_name}', last name '{last_name}', and email '{email}' "
+        "in the form. Do not use any suggested autofills. Make sure the mobile phone number "
+        "is empty.",
+        f"Slightly scroll down to see the date of birth, enter Month '{month}', Day '{day}', "
+        f"and Year '{year}' in the form",
+        "Click on 'Continue as guest' button, wait for the page to load with wait, "
+        "click on 'Add vaccines' button, select 'Flu' and click on 'Add vaccines'",
+        f"Click on 'next' to enter the page with recommendation vaccines, then click on "
+        f"'next' again, until on the page of entering zip code, enter '{zip_code}', select "
+        "the first option from the dropdown menu, and click on 'Search'",
+    ]
+
+    tasker.set_task(instruction, todos)
+    return tasker
+
+
+@async_agent_register(mode="tasker:software_qa")
+def create_software_qa_agent(
+    api_key: str | None = None,
+    base_url: str | None = None,
+    model: str = MODEL_ACTOR,
+    max_steps: int = DEFAULT_MAX_STEPS_TASKER,
+    temperature: float = DEFAULT_TEMPERATURE_LOW,
+    reflection_interval: int = DEFAULT_REFLECTION_INTERVAL_TASKER,
+    step_observer: AsyncStepObserver | None = None,
+    step_delay: float = DEFAULT_STEP_DELAY,
+) -> AsyncAgent:
+    tasker = TaskerAgent(
+        api_key=api_key,
+        base_url=base_url,
+        model=model,
+        max_steps=max_steps,
+        temperature=temperature,
+        reflection_interval=reflection_interval,
+        step_observer=step_observer,
+        step_delay=step_delay,
+    )
+
+    instruction = "QA: click through every sidebar button in the Nuclear Player UI"
+    todos = [
+        "Click on 'Dashboard' in the left sidebar",
+        "Click on 'Downloads' in the left sidebar",
+        "Click on 'Lyrics' in the left sidebar",
+        "Click on 'Plugins' in the left sidebar",
+        "Click on 'Search Results' in the left sidebar",
+        "Click on 'Settings' in the left sidebar",
+        "Click on 'Equalizer' in the left sidebar",
+        "Click on 'Visualizer' in the left sidebar",
+        "Click on 'Listening History' in the left sidebar",
+        "Click on 'Favorite Albums' in the left sidebar",
+        "Click on 'Favorite Tracks' in the left sidebar",
+        "Click on 'Favorite Artists' in the left sidebar",
+        "Click on 'Local Library' in the left sidebar",
+        "Click on 'Playlists' in the left sidebar",
+    ]
+
+    tasker.set_task(instruction, todos)
+    return tasker

{oagi_core-0.10.2 → oagi_core-0.11.0}/src/oagi/agent/tasker/planner.py

@@ -10,7 +10,8 @@ import json
 from typing import Any
 
 from ...client import AsyncClient
-from ...types import URL, Image
+from ...constants import DEFAULT_REFLECTION_INTERVAL
+from ...types import URL, Image, extract_uuid_from_url
 from .memory import PlannerMemory
 from .models import Action, PlannerOutput, ReflectionOutput
 
@@ -137,11 +138,16 @@ class Planner:
         # Ensure we have a client
         client = self._ensure_client()
 
-        # Upload screenshot if provided
+        # Get screenshot UUID - either extract from URL or upload
         screenshot_uuid = None
         if screenshot:
-            upload_response = await client.put_s3_presigned_url(screenshot)
-            screenshot_uuid = upload_response.uuid
+            # Check if screenshot is already a URL (already uploaded to S3)
+            if isinstance(screenshot, str):
+                screenshot_uuid = extract_uuid_from_url(screenshot)
+            # If not a URL or UUID extraction failed, upload the image
+            if not screenshot_uuid:
+                upload_response = await client.put_s3_presigned_url(screenshot)
+                screenshot_uuid = upload_response.uuid
 
         # Extract memory data if provided
         (
@@ -175,7 +181,7 @@ class Planner:
         memory: PlannerMemory | None = None,
         todo_index: int | None = None,
         current_instruction: str | None = None,
-        reflection_interval: int = 4,
+        reflection_interval: int = DEFAULT_REFLECTION_INTERVAL,
     ) -> ReflectionOutput:
         """Reflect on recent actions and progress.
 
@@ -194,11 +200,16 @@ class Planner:
         # Ensure we have a client
         client = self._ensure_client()
 
-        # Upload screenshot if provided
+        # Get screenshot UUID - either extract from URL or upload
         result_screenshot_uuid = None
         if screenshot:
-            upload_response = await client.put_s3_presigned_url(screenshot)
-            result_screenshot_uuid = upload_response.uuid
+            # Check if screenshot is already a URL (already uploaded to S3)
+            if isinstance(screenshot, str):
+                result_screenshot_uuid = extract_uuid_from_url(screenshot)
+            # If not a URL or UUID extraction failed, upload the image
+            if not result_screenshot_uuid:
+                upload_response = await client.put_s3_presigned_url(screenshot)
+                result_screenshot_uuid = upload_response.uuid
 
         # Extract memory data if provided
         (

{oagi_core-0.10.2 → oagi_core-0.11.0}/src/oagi/agent/tasker/taskee_agent.py

@@ -12,6 +12,14 @@ from datetime import datetime
 from typing import Any
 
 from oagi import AsyncActor
+from oagi.constants import (
+    DEFAULT_MAX_STEPS,
+    DEFAULT_REFLECTION_INTERVAL,
+    DEFAULT_STEP_DELAY,
+    DEFAULT_TEMPERATURE,
+    MODEL_ACTOR,
+)
+from oagi.handler import reset_handler
 from oagi.types import (
     URL,
     ActionEvent,
@@ -21,6 +29,7 @@ from oagi.types import (
     Image,
     PlanEvent,
     StepEvent,
+    extract_uuid_from_url,
 )
 
 from ..protocol import AsyncAgent
@@ -52,15 +61,15 @@ class TaskeeAgent(AsyncAgent):
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "lux-actor-1",
-        max_steps: int = 20,
-        reflection_interval: int = 4,
-        temperature: float = 0.5,
+        model: str = MODEL_ACTOR,
+        max_steps: int = DEFAULT_MAX_STEPS,
+        reflection_interval: int = DEFAULT_REFLECTION_INTERVAL,
+        temperature: float = DEFAULT_TEMPERATURE,
         planner: Planner | None = None,
         external_memory: PlannerMemory | None = None,
         todo_index: int | None = None,
         step_observer: AsyncObserver | None = None,
-        step_delay: float = 0.3,
+        step_delay: float = DEFAULT_STEP_DELAY,
     ):
         """Initialize the taskee agent.
 
@@ -114,6 +123,9 @@ class TaskeeAgent(AsyncAgent):
         Returns:
             True if successful, False otherwise
         """
+        # Reset handler state at todo execution start
+        reset_handler(action_handler)
+
         self.current_todo = instruction
         self.actions = []
         self.total_actions = 0
@@ -249,11 +261,21 @@ class TaskeeAgent(AsyncAgent):
             # Capture screenshot
             screenshot = await image_provider()
 
-            # Upload screenshot first to get UUID (avoids re-upload in actor.step)
+            # Get screenshot UUID - either extract from URL or upload
             try:
-                upload_response = await client.put_s3_presigned_url(screenshot)
-                screenshot_uuid = upload_response.uuid
-                screenshot_url = upload_response.download_url
+                screenshot_uuid = None
+                screenshot_url = None
+
+                # Check if screenshot is already a URL (from SocketIOImageProvider)
+                if isinstance(screenshot, str):
+                    screenshot_uuid = extract_uuid_from_url(screenshot)
+                    screenshot_url = screenshot
+
+                # If not a URL or UUID extraction failed, upload the image
+                if not screenshot_uuid:
+                    upload_response = await client.put_s3_presigned_url(screenshot)
+                    screenshot_uuid = upload_response.uuid
+                    screenshot_url = upload_response.download_url
             except Exception as e:
                 logger.error(f"Error uploading screenshot: {e}")
                 self._record_action(

{oagi_core-0.10.2 → oagi_core-0.11.0}/src/oagi/agent/tasker/tasker_agent.py

@@ -9,6 +9,14 @@
 import logging
 from typing import Any
 
+from oagi.constants import (
+    DEFAULT_MAX_STEPS_TASKER,
+    DEFAULT_REFLECTION_INTERVAL,
+    DEFAULT_STEP_DELAY,
+    DEFAULT_TEMPERATURE,
+    MODEL_ACTOR,
+)
+from oagi.handler import reset_handler
 from oagi.types import AsyncActionHandler, AsyncImageProvider, AsyncObserver, SplitEvent
 
 from ..protocol import AsyncAgent
@@ -34,13 +42,13 @@ class TaskerAgent(AsyncAgent):
         self,
         api_key: str | None = None,
         base_url: str | None = None,
-        model: str = "lux-actor-1",
-        max_steps: int = 60,
-        temperature: float = 0.5,
-        reflection_interval: int = 4,
+        model: str = MODEL_ACTOR,
+        max_steps: int = DEFAULT_MAX_STEPS_TASKER,
+        temperature: float = DEFAULT_TEMPERATURE,
+        reflection_interval: int = DEFAULT_REFLECTION_INTERVAL,
         planner: Planner | None = None,
         step_observer: AsyncObserver | None = None,
-        step_delay: float = 0.3,
+        step_delay: float = DEFAULT_STEP_DELAY,
     ):
         """Initialize the tasker agent.
 
@@ -105,6 +113,9 @@ class TaskerAgent(AsyncAgent):
         Returns:
             True if all todos completed successfully, False otherwise
         """
+        # Reset handler state at automation start
+        reset_handler(action_handler)
+
         overall_success = True
 
         # Execute todos until none remain