computer-use-ootb-internal 0.0.96__py3-none-any.whl → 0.0.96.post1__py3-none-any.whl

computer_use_ootb_internal/app_teachmode.py

@@ -104,8 +104,8 @@ async def update_parameters(request: Request):
     shared_state.model = getattr(shared_state.args, 'model', "teach-mode-gpt-4o")
     shared_state.task = getattr(shared_state.args, 'task', "Create a claim on the SAP system, using Receipt.pdf as attachment.")
     shared_state.selected_screen = getattr(shared_state.args, 'selected_screen', 0)
-    shared_state.user_id = getattr(shared_state.args, 'user_id', "a_test")
-    shared_state.trace_id = getattr(shared_state.args, 'trace_id', "jess_4")
+    shared_state.user_id = getattr(shared_state.args, 'user_id', "hero_cases")
+    shared_state.trace_id = getattr(shared_state.args, 'trace_id', "build_scroll_combat")
     shared_state.api_keys = getattr(shared_state.args, 'api_keys', "sk-proj-1234567890")
     shared_state.server_url = getattr(shared_state.args, 'server_url', "http://ec2-44-234-43-86.us-west-2.compute.amazonaws.com")
     

computer_use_ootb_internal/computer_use_demo/animation/click_animation.py

@@ -6,7 +6,7 @@ show_click(x, y, duration_ms=800)
 """
 import sys, time
 from pathlib import Path
-from PySide6.QtCore import Qt, QPoint, QTimer, QEventLoop, QSize
+from PySide6.QtCore import Qt, QPoint, QTimer, QEventLoop, QSize, QEasingCurve, QPropertyAnimation
 from PySide6.QtGui  import QPainter, QPixmap, QMovie
 from PySide6.QtWidgets import QApplication, QWidget, QLabel
 
@@ -72,8 +72,17 @@ def _ensure_app():
             print("Using existing QApplication instance")
             _app = QApplication.instance()
 
-def show_click(x: int, y: int, duration_ms: int = 2000):  # 增加默认播放时间
-    """阻塞式点击动画：调用后必定肉眼可见"""
+# Keep references to animations to prevent garbage collection
+_active_animations = []
+
+def show_click(x: int, y: int, duration_ms: int = 2000, existing_ms: int = 1000):  # 增加默认播放时间和静止时间
+    """非阻塞式点击动画：立即返回，动画在后台运行
+    
+    Args:
+        x, y          : 屏幕坐标
+        duration_ms   : 动画播放时长
+        existing_ms   : 动画结束后静止显示的时间
+    """
     print(f"Attempting to show click at ({x}, {y})")
     
     if not CLICK_GIF.exists():
@@ -82,45 +91,56 @@ def show_click(x: int, y: int, duration_ms: int = 2000):  # 增加默认播放
     _ensure_app()
     
     try:
-        animation = ClickAnimation(QPoint(x, y), duration_ms)
-
-        # 局部事件循环，动画结束后返回
-        loop = QEventLoop()
-        QTimer.singleShot(duration_ms + 150, loop.quit)  # 增加等待时间
-        loop.exec()
-        print("Click animation completed")
+        # 总生存时间 = 动画时间 + 静止显示时间
+        total_life_ms = duration_ms + existing_ms
+        animation = ClickAnimation(QPoint(x, y), total_life_ms)
+        
+        # Store reference to prevent garbage collection
+        global _active_animations
+        _active_animations.append(animation)
+        
+        # Set up cleanup after animation completes + existing time
+        QTimer.singleShot(total_life_ms + 150, lambda: _clean_animation(animation))
+        
+        print(f"Click animation started (non-blocking, will exist for {total_life_ms}ms)")
     except Exception as e:
         print(f"Error during show_click: {str(e)}")
 
 
-# --- 在原 import 区域追加 ---
-from PySide6.QtCore import QEasingCurve, QPropertyAnimation
-# --------------------------------------------------------
+def _clean_animation(animation):
+    """Remove animation from reference list after it completes"""
+    global _active_animations
+    if animation in _active_animations:
+        _active_animations.remove(animation)
+    print("Animation cleaned up")
 
 
 # ---------- 新增函数 ----------
-def show_move_to(x1: int, y1: int, x2: int, y2: int, duration_ms: int = 1200):
+def show_move_to(x1: int, y1: int, x2: int, y2: int, duration_ms: int = 1000, existing_ms: int = 1000):
     """
-    阻塞式移动动画：在 (x1, y1) 处出现光标 GIF，
-    并在 duration_ms 毫秒内平滑移动到 (x2, y2)。
+    非阻塞式移动动画：在 (x1, y1) 处出现光标 GIF，
+    并在 duration_ms 毫秒内平滑移动到 (x2, y2)，
+    然后在终点静止显示 existing_ms 毫秒。
+    立即返回，动画在后台运行。
 
     Args:
         x1, y1        : 起点屏幕坐标
         x2, y2        : 终点屏幕坐标
         duration_ms   : 移动总时长
+        existing_ms   : 移动结束后在终点静止显示的时间
     """
     print(f"Attempting to move click from ({x1}, {y1}) → ({x2}, {y2}) "
-          f"in {duration_ms} ms")
+          f"in {duration_ms} ms, then stay for {existing_ms} ms")
 
     if not CLICK_GIF.exists():
         raise FileNotFoundError(f"click.gif not found at {CLICK_GIF}")
 
     _ensure_app()
 
-    # 让 widget 的生命周期略长于动画，避免提前销毁
-    life_ms = duration_ms + 200
-    widget  = ClickAnimation(QPoint(x1, y1), life_ms)
-
+    # 总生存时间 = 动画时间 + 静止显示时间
+    total_life_ms = duration_ms + existing_ms
+    widget = ClickAnimation(QPoint(x1, y1), total_life_ms)
+    
     # 用 QPropertyAnimation 平滑移动窗口
     anim = QPropertyAnimation(widget, b"pos")
     anim.setDuration(duration_ms)
@@ -128,27 +148,67 @@ def show_move_to(x1: int, y1: int, x2: int, y2: int, duration_ms: int = 1200):
     anim.setStartValue(QPoint(x1 - 15, y1 - 15))
     anim.setEndValue(QPoint(x2 - 15, y2 - 15))
     anim.setEasingCurve(QEasingCurve.OutQuad)     # 可自行更换缓动曲线
+    
+    # Store references to both widget and animation to prevent garbage collection
+    global _active_animations
+    # Store them as a tuple to keep both references
+    animation_pair = (widget, anim)
+    _active_animations.append(animation_pair)
+    
+    # Clean up both widget and animation after completion of total life time
+    def cleanup():
+        if animation_pair in _active_animations:
+            _active_animations.remove(animation_pair)
+        print("Move animation cleaned up")
+    
+    # Connect finished signal only to print a message
+    anim.finished.connect(lambda: print("Movement finished, now staying still"))
+    
+    # Start the animation
     anim.start()
-
-    # 局部事件循环，直到动画结束
-    loop = QEventLoop()
-    anim.finished.connect(loop.quit)
-    QTimer.singleShot(life_ms, loop.quit)          # 双保险
-    loop.exec()
-
-    print("Move‑to animation completed")
-# ---------------------------------
+    
+    # Process events immediately to kickstart the animation
+    QApplication.processEvents()
+    
+    # Set up final cleanup after animation + existing time
+    QTimer.singleShot(total_life_ms, cleanup)
+    
+    print("Move-to animation started (non-blocking)")
 
 
 # ---------- 命令行测试 ----------
 if __name__ == "__main__":
+    # 确保应用程序实例存在
+    # _ensure_app()
+    
     # 测试点击
+    print("Testing non-blocking click animation...")
     x, y = 500, 500
-    print(f"Testing click at ({x}, {y})")
     show_click(x, y)
-
-    # 测试移动
-    x1, y1 = 400, 400
-    x2, y2 = 800, 600
-    print(f"Testing move from ({x1}, {y1}) → ({x2}, {y2})")
-    show_move_to(x1, y1, x2, y2, duration_ms=2000)
+    
+    # 测试同时运行两个动画
+    print("\nTesting simultaneous animations...")
+    x1, y1 = 200, 200
+    x2, y2 = 600, 600
+    # show_click(x1, y1)
+    show_move_to(x1, y1, x2, y2)
+    
+    # # 测试先移动，然后点击
+    print("\nTesting sequence with pyautogui simulation...")
+    x3, y3 = 800, 300
+    x4, y4 = 400, 500
+    
+    # 启动移动动画
+    show_move_to(x3, y3, x4, y4)
+    
+    # 模拟移动完成后的点击动画（延迟1.5秒）
+    QTimer.singleShot(1500, lambda: show_click(x4, y4))
+    
+    # 保持主程序运行，等待所有动画完成
+    print("\nWaiting for all animations to complete...")
+    loop = QEventLoop()
+    # 等待足够长的时间，确保所有动画都完成（最长的动画是2000ms + 清理时间）
+    QTimer.singleShot(4000, loop.quit)
+    loop.exec()
+    
+    print("All animations completed, exiting test.")

computer_use_ootb_internal/computer_use_demo/executor/teachmode_executor.py

@@ -2,18 +2,12 @@ import ast
 import json
 import asyncio
 from typing import Any, Dict, cast, List, Union
-from collections.abc import Callable
 import uuid
 from anthropic.types.beta import (
-    BetaContentBlock,
-    BetaContentBlockParam,
     BetaImageBlockParam,
-    BetaMessage,
-    BetaMessageParam,
     BetaTextBlockParam,
     BetaToolResultBlockParam,
 )
-from anthropic.types import TextBlock
 from anthropic.types.beta import BetaMessage, BetaTextBlock, BetaToolUseBlock
 from computer_use_ootb_internal.computer_use_demo.tools import BashTool, ComputerTool, EditTool, ToolCollection, ToolResult
 
@@ -221,9 +215,13 @@ class TeachmodeExecutor:
         
     def _add_starrail_alt_actions(self, action_list: list[Dict[str, Any]]) -> list[Dict[str, Any]]:
         parsed_action_list = []
-        parsed_action_list.extend([{"action": "key_down", "text": "alt", "coordinate": None}])
-        parsed_action_list.extend(action_list)
-        parsed_action_list.extend([{"action": "key_up", "text": "alt", "coordinate": None}])
+        for action in action_list:
+            if action["action"] in ["left_click", "mouse_move"]:
+                parsed_action_list.append({"action": "key_down", "text": "alt", "coordinate": None})
+                parsed_action_list.append(action)
+                parsed_action_list.append({"action": "key_up", "text": "alt", "coordinate": None})
+            else:
+                parsed_action_list.append(action)
         return parsed_action_list
 
     def _reformat_starrail_browser_actions(self, action_list: list[Dict[str, Any]]) -> list[Dict[str, Any]]:
@@ -242,9 +240,18 @@ class TeachmodeExecutor:
 
     def _add_starrail_browser_alt_actions(self, action_list: list[Dict[str, Any]]) -> list[Dict[str, Any]]:
         parsed_action_list = []
-        parsed_action_list.extend([{"action": "key_down_windll", "text": "alt", "coordinate": None}])
-        parsed_action_list.extend(action_list)
-        parsed_action_list.extend([{"action": "key_up_windll", "text": "alt", "coordinate": None}])
+
+        for action in action_list:
+            if action["action"] in ["left_click", "mouse_move", "left_click_windll", "mouse_move_windll"]:
+                parsed_action_list.append({"action": "key_down_windll", "text": "alt", "coordinate": None})
+                parsed_action_list.append(action)
+                parsed_action_list.append({"action": "key_up_windll", "text": "alt", "coordinate": None})
+            else:
+                parsed_action_list.append(action)
+
+        # parsed_action_list.extend([{"action": "key_down_windll", "text": "alt", "coordinate": None}])
+        # parsed_action_list.extend(action_list)
+        # parsed_action_list.extend([{"action": "key_up_windll", "text": "alt", "coordinate": None}])
         return parsed_action_list
 
 

computer_use_ootb_internal/computer_use_demo/tools/computer.py

@@ -296,23 +296,36 @@ class ComputerTool(BaseAnthropicTool):
                 raise ToolError(f"coordinate is not accepted for {action}")
 
             if action == "left_click":
+                x, y = pyautogui.position()
+                show_click(x, y)
                 pyautogui.click()
             elif action == "right_click":
+                x, y = pyautogui.position()
+                show_click(x, y)
                 pyautogui.rightClick()
             elif action == "middle_click":
+                x, y = pyautogui.position()
+                show_click(x, y)
                 pyautogui.middleClick()
             elif action == "double_click":
+                x, y = pyautogui.position()
+                show_click(x, y)
                 pyautogui.doubleClick()
             elif action == "left_press":
+                x, y = pyautogui.position()
+                show_click(x, y)
                 pyautogui.mouseDown()
                 time.sleep(1)
                 pyautogui.mouseUp()
+                show_click(x, y)
             elif action == "scroll_down":
                 pyautogui.scroll(-200)  # Adjust scroll amount as needed
                 return ToolResult(output="Scrolled down")
+            
             elif action == "scroll_up":
                 pyautogui.scroll(200)   # Adjust scroll amount as needed
                 return ToolResult(output="Scrolled up")
+            
             elif action == "wait":
                 time.sleep(30)
                 return ToolResult(output="Waited")
@@ -355,14 +368,13 @@ class ComputerTool(BaseAnthropicTool):
             if action == "left_click_windll":
                 if coordinate is None:
                     x, y = pyautogui.position()
-                    # x, y = self.scale_coordinates(ScalingSource.COMPUTER, x, y)
                     show_click(x, y)
                     self.marbot_auto_gui.click()
                 else:
                     x = coordinate[0]+self.offset_x
                     y = coordinate[1]+self.offset_y
-                    self.marbot_auto_gui.click(x=x, y=y)
                     show_click(x, y)
+                    self.marbot_auto_gui.click(x=x, y=y)
             elif action == "mouse_move_windll":
                 if coordinate is None:
                     raise ToolError(f"coordinate is required for {action}")
@@ -371,8 +383,8 @@ class ComputerTool(BaseAnthropicTool):
                 x1 = coordinate[0]+self.offset_x
                 y1 = coordinate[1]+self.offset_y
 
+                show_move_to(x0, y0, x1, y1, duration_ms=1000)
                 self.marbot_auto_gui.moveTo(x=x1, y=y1)
-                show_move_to(x0, y0, x1, y1, duration_ms=2000)
 
             # elif action == "right_click_windll":
             #     self.marbot_auto_gui.rightClick(x=coordinate[0], y=coordinate[1])

computer_use_ootb_internal/run_teachmode_ootb_args.py

@@ -41,7 +41,7 @@ def simple_teachmode_sampling_loop(
     if "star_rail" in user_id or "star_rail" in user_id:
         full_screen_game_mode = 1
     
-    if "star_rail_dev" in trace_id or "star_rail_dev" in user_id:
+    if "star_rail_dev" in trace_id or "star_rail_dev" in user_id or "hero_case" in trace_id:
         full_screen_game_mode = 2
 
     print(f"Full Screen Game Mode: {full_screen_game_mode}")
@@ -102,7 +102,7 @@ def simple_teachmode_sampling_loop(
 
         try:
             step_plan = infer_server_response["generated_plan"]
-            # step_action = json.loads(infer_server_response["generated_action"]["content"])
+            step_info = infer_server_response["generated_action"]["step_info"]
             step_action = infer_server_response["generated_action"]["content"]
             step_traj_idx = infer_server_response["current_traj_step"]
 
@@ -115,7 +115,8 @@ def simple_teachmode_sampling_loop(
             break
 
         # action_history.append(f"Executing Step: {step_count} - Trajectory Step: {step_traj_idx} - Plan: {step_plan} - Action: {step_action};\n")
-        action_history.append(f"Step in Guidance Trajectory: [{step_traj_idx}], Plan: {step_plan}, Action: {step_action};\n")
+
+        action_history.append(f"Executing guidance trajectory step [{step_traj_idx}], Plan: {step_info}, Action: {step_action};\n")
 
         for message in executor({"role": "assistant", "content": step_action}):
             yield message

{computer_use_ootb_internal-0.0.96.dist-info → computer_use_ootb_internal-0.0.96.post1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: computer-use-ootb-internal
-Version: 0.0.96
+Version: 0.0.96.post1
 Summary: Computer Use OOTB
 Author-email: Siyuan Hu <siyuan.hu.sg@gmail.com>
 Requires-Python: >=3.11

{computer_use_ootb_internal-0.0.96.dist-info → computer_use_ootb_internal-0.0.96.post1.dist-info}/RECORD

@@ -1,15 +1,12 @@
 computer_use_ootb_internal/README.md,sha256=FxpW95lyub2iX73ZDfK6ML7SdEKg060H5I6Grub7li4,31
-computer_use_ootb_internal/app_teachmode.py,sha256=9p4a7vRNwJOJI_gkiPVeF4ALqyLuBdcFsEEFQ-rmEbg,15629
+computer_use_ootb_internal/app_teachmode.py,sha256=zmUPvFjqdhysnN1bD2QQhaAKONnAtMFd03Rb9gF1l6c,15646
 computer_use_ootb_internal/app_teachmode_gradio.py,sha256=zAw-n3s20j1Jr0S4TzXHwllKV6APJ8HEHB1KqBuzriY,7907
 computer_use_ootb_internal/dependency_check.py,sha256=y8RMEP6RXQzTgU1MS_1piBLtz4J-Hfn9RjUZg59dyvo,1333
-computer_use_ootb_internal/example_websocket_js.html,sha256=BLYwDExVlgiAX4vXVXW3RuP5KD8FXE4EFXIl54bwF7w,1322
 computer_use_ootb_internal/requirements-lite.txt,sha256=5DAHomz4A_P2BmTIXNkNqkHbnIF0AyZ4_1XAlb1LaYs,290
-computer_use_ootb_internal/run_teachmode_ootb_args.py,sha256=IN20EC5gCJZpaBQyVu5PMBk2aPBPEYIL0meTnYQrpS8,6825
-computer_use_ootb_internal/service_teachmode.py,sha256=e81zp3B7CA2nMjA-3qoKF4P1hlNGRI7P1tBo1j21FBk,8082
-computer_use_ootb_internal/service_teachmode_test.py,sha256=zpfBFFKD9WGLX4m77ajOBfmczpYsCa3_qTBweeSNRV8,1112
-computer_use_ootb_internal/computer_use_demo/animation/click_animation.py,sha256=ZKmlZnMVW4zdxkUVWCTQF2rqq2h70qq5bAuLYVa4qxc,5312
+computer_use_ootb_internal/run_teachmode_ootb_args.py,sha256=djr4E7_G_qG9H6qmRz1mrM9Yrcxf1tnlSPH5ZqykF5Y,6845
+computer_use_ootb_internal/computer_use_demo/animation/click_animation.py,sha256=BsHW1yFLHE5DqLUofnR-D3jnglLcYCaVN34jquWvh94,7725
 computer_use_ootb_internal/computer_use_demo/animation/icons8-select-cursor-transparent-96.gif,sha256=4LfwsfFQnREXrNRs32aJU2jO65JXianJoL_8q7-8elg,30966
-computer_use_ootb_internal/computer_use_demo/executor/teachmode_executor.py,sha256=x6XwhrmSnD0_euotx0X0fFyak2o9khttV4XwaGpzV3E,16121
+computer_use_ootb_internal/computer_use_demo/executor/teachmode_executor.py,sha256=y7lg_PjMif2WwCKWWC7g8Ys2zPRMh08Vtt42fStujY4,16623
 computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/__init__.py,sha256=h2CNeuACklxVpJC65QR8_6AvSybEZLmeO45hY_-lLBs,61
 computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/gui_capture.py,sha256=CxFJbsSb68ERKH7-C4RaaZy7FIhhzrzGx5qQJ4C37cA,13907
 computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/gui_parser.py,sha256=KSTJ0cMwh3ahUMzHRaDgA2sVNUL4MNlF7qEBGN3G0SI,28993
@@ -33,12 +30,12 @@ computer_use_ootb_internal/computer_use_demo/tools/base.py,sha256=QDqpuuKlhUKJT2
 computer_use_ootb_internal/computer_use_demo/tools/bash.py,sha256=rHetQ80_v-TTi-1oxIA7ncFEwJxFTh8FJCErIoZbGeY,4236
 computer_use_ootb_internal/computer_use_demo/tools/collection.py,sha256=8RzHLobL44_Jjt8ltXS6I8XJlEAQOfc75dmnDUaHE-8,922
 computer_use_ootb_internal/computer_use_demo/tools/colorful_text.py,sha256=cvlmnhAImDTwoRRwhT5au7mNFhfAD7ZfeoDEVdVzDKw,892
-computer_use_ootb_internal/computer_use_demo/tools/computer.py,sha256=eXH6-h9YEo-1sSkQNG6Ld43aDm0Ip-ioYNyF1Q60Jd0,25164
+computer_use_ootb_internal/computer_use_demo/tools/computer.py,sha256=bKOkCtE4iYmPpRxldslAAa5yQOs6NkhITJYCfZllWXI,25526
 computer_use_ootb_internal/computer_use_demo/tools/computer_marbot.py,sha256=zZuWz9ArfP3Zss-afnscrPkgCtB5UWbCy7HwAOvO2bo,5970
 computer_use_ootb_internal/computer_use_demo/tools/edit.py,sha256=b0PwUitxckHCQqFP3ZwlthWdqNkn7WETeTHeB6-o98c,11486
 computer_use_ootb_internal/computer_use_demo/tools/run.py,sha256=xhXdnBK1di9muaO44CEirL9hpGy3NmKbjfMpyeVmn8Y,1595
 computer_use_ootb_internal/computer_use_demo/tools/screen_capture.py,sha256=L8qfvtUkPPQGt92N-2Zfw5ZTDBzLsDps39uMnX3_uSA,6857
-computer_use_ootb_internal-0.0.96.dist-info/METADATA,sha256=MY4ASXzlq1JoyiE_JWcQHytdoNWxSv0eXUSx8ap3haY,909
-computer_use_ootb_internal-0.0.96.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-computer_use_ootb_internal-0.0.96.dist-info/entry_points.txt,sha256=-AbmawU7IRQuDZihgVMVDrFoY4E6rnXYOUB-5vSeBKs,93
-computer_use_ootb_internal-0.0.96.dist-info/RECORD,,
+computer_use_ootb_internal-0.0.96.post1.dist-info/METADATA,sha256=45BKygjaYFw3snR8rr1qk9pH9cVJYmjZP9T5O5PSjAo,915
+computer_use_ootb_internal-0.0.96.post1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+computer_use_ootb_internal-0.0.96.post1.dist-info/entry_points.txt,sha256=-AbmawU7IRQuDZihgVMVDrFoY4E6rnXYOUB-5vSeBKs,93
+computer_use_ootb_internal-0.0.96.post1.dist-info/RECORD,,

computer_use_ootb_internal/example_websocket_js.html

@@ -1,41 +0,0 @@
-<!DOCTYPE html>
-<html>
-  <head>
-    <title>Teachmode Client</title>
-    <!-- include socket.io client -->
-    <script src="https://cdn.socket.io/4.5.4/socket.io.min.js"></script>
-  </head>
-  <body>
-    <h1>Teachmode WebSocket Test</h1>
-    <div id="messages"></div>
-
-    <script>
-      const socket = io("http://localhost:5000"); // Your server’s URL/port
-
-      // Listen for partial responses
-      socket.on("partial_response", (data) => {
-        console.log("Got partial:", data);
-        const div = document.createElement("div");
-        div.innerText = "Assistant: " + data.content;
-        document.getElementById("messages").appendChild(div);
-      });
-
-      // Listen for done event
-      socket.on("done", (data) => {
-        console.log("Task completed:", data);
-        const div = document.createElement("div");
-        div.innerText = "TASK DONE!";
-        document.getElementById("messages").appendChild(div);
-      });
-
-      // When connected, emit run_teachmode
-      socket.on("connect", () => {
-        console.log("Connected to server. Emitting run_teachmode event...");
-        socket.emit("run_teachmode", {
-          user_input: "Hello, I'd like to do X, Y, Z."
-          // you can include model, task, user_id, trace_id, etc. if desired
-        });
-      });
-    </script>
-  </body>
-</html>

computer_use_ootb_internal/service_teachmode.py

@@ -1,194 +0,0 @@
-import argparse
-import time
-import json
-import threading
-
-from flask import Flask, request, jsonify
-from flask_socketio import SocketIO, emit
-
-from screeninfo import get_monitors
-from computer_use_ootb_internal.computer_use_demo.tools.computer import get_screen_details
-from computer_use_ootb_internal.run_teachmode_ootb_args import simple_teachmode_sampling_loop
-
-###############################################################################
-#                                Shared State
-###############################################################################
-class SharedState:
-    def __init__(self):
-        self.args = None  # Will hold argparse-like namespace
-        self.messages = []  # If you want to store a global chat or last session
-
-shared_state = SharedState()
-
-###############################################################################
-#                       Flask + SocketIO Application Setup
-###############################################################################
-app = Flask(__name__)
-app.config["SECRET_KEY"] = "some-secret-key"  # In production, change this
-socketio = SocketIO(app, cors_allowed_origins="*")
-
-###############################################################################
-#                            Utility Functions
-###############################################################################
-def setup_default_args():
-    """
-    Creates argparse-like defaults. 
-    You can also parse real CLI args if you wish.
-    """
-    parser = argparse.ArgumentParser(description="Teachmode SocketIO Server.")
-    parser.add_argument("--model", default="teach-mode-gpt-4o")
-    parser.add_argument("--task", default="Help me complete data extraction on YouTube video.")
-    parser.add_argument("--selected_screen", type=int, default=0)
-    parser.add_argument("--user_id", default="liziqi")
-    parser.add_argument("--trace_id", default="default_trace")
-    parser.add_argument("--api_key_file", default="api_key.json")
-    parser.add_argument("--api_keys", default="")
-    parser.add_argument(
-        "--server_url",
-        default="http://ec2-44-234-43-86.us-west-2.compute.amazonaws.com/generate_action",
-        help="Server URL for the session (local='http://localhost:5000/generate_action', \
-              aws='http://ec2-44-234-43-86.us-west-2.compute.amazonaws.com/generate_action').",
-    )
-
-    # If you really want to parse sys.argv, do parser.parse_args().
-    # But you can also return the defaults for this example:
-    return parser.parse_args([])
-
-def apply_args_to_state(args):
-    """
-    Helper that prints or logs relevant arguments and stores them in shared_state.
-    """
-    print("[apply_args_to_state] Applying arguments:", args)
-    shared_state.args = args
-
-def run_teachmode_task(user_input):
-    """
-    Calls simple_teachmode_sampling_loop and emits partial responses over SocketIO.
-    """
-    # 1) Log or store user input
-    print(f"[run_teachmode_task] Received user_input: {user_input}")
-    # Optionally store or reset message history for this session
-    shared_state.messages = [{"role": "user", "content": user_input}]
-
-    # 2) Grab arguments from shared_state
-    args = shared_state.args
-    if not args:
-        print("[run_teachmode_task] No arguments in shared_state, applying defaults.")
-        args = setup_default_args()
-        apply_args_to_state(args)
-
-    # 3) Run the sampling loop
-    print(f"[run_teachmode_task] Starting the sampling loop with task: {args.task}")
-    sampling_loop = simple_teachmode_sampling_loop(
-        model=args.model,
-        task=args.task,
-        selected_screen=args.selected_screen,
-        user_id=args.user_id,
-        trace_id=args.trace_id,
-        api_keys=args.api_keys,
-        server_url=args.server_url
-    )
-
-    # 4) Send partial responses
-    for loop_msg in sampling_loop:
-        print(f"[run_teachmode_task] Emitting partial response: {loop_msg}")
-        # You can store it in shared_state messages
-        shared_state.messages.append({"role": "assistant", "content": loop_msg})
-        # Emit immediately so the client sees partial responses
-        emit("partial_response", {"role": "assistant", "content": loop_msg})
-        time.sleep(1)  # Optional delay to simulate real-time streaming
-
-    # 5) Done event
-    print("[run_teachmode_task] Completed all messages.")
-    emit("done", {"messages": shared_state.messages, "status": "completed"})
-
-###############################################################################
-#                           HTTP Endpoint: update_params
-###############################################################################
-@app.route("/update_params", methods=["POST"])
-def update_parameters():
-    """
-    HTTP endpoint that allows updating the parameters (like Gradio's /update_params).
-    Expects JSON body with fields matching the argparse Namespace (model, task, etc.)
-    """
-    data = request.json
-    if not data:
-        return jsonify({"status": "error", "message": "No JSON provided."}), 400
-
-    # Build an argparse.Namespace from the JSON keys
-    # Fallback to the existing arguments if some keys are missing
-    old_args = shared_state.args or setup_default_args()
-    new_args_dict = {**vars(old_args), **data}  # Merge old with new
-    new_args = argparse.Namespace(**new_args_dict)
-    apply_args_to_state(new_args)
-
-    return jsonify({
-        "status": "success",
-        "message": "Parameters updated",
-        "new_args": vars(new_args)
-    })
-
-###############################################################################
-#                           HTTP Endpoint: get_messages
-###############################################################################
-@app.route("/get_messages", methods=["GET"])
-def get_messages():
-    """
-    Example new function: returns the current chat messages in shared_state.
-    """
-    return jsonify(shared_state.messages)
-
-###############################################################################
-#                           HTTP Endpoint: clear_messages
-###############################################################################
-@app.route("/clear_messages", methods=["POST"])
-def clear_messages():
-    """
-    Example new function: clears the stored chat messages in shared_state.
-    """
-    shared_state.messages = []
-    return jsonify({"status": "success", "message": "Chat history cleared."})
-
-###############################################################################
-#                        SocketIO Event: run_teachmode
-###############################################################################
-@socketio.on("run_teachmode")
-def handle_run_teachmode(data):
-    """
-    Websocket event that starts the teachmode sampling loop.
-    `data` can include e.g. {"user_input": "..."}.
-    """
-    user_input = data.get("user_input", "Hello, let's start!")
-    run_teachmode_task(user_input)
-
-###############################################################################
-#                        SocketIO Event: connect
-###############################################################################
-@socketio.on("connect")
-def on_connect():
-    print("[SocketIO] Client connected.")
-
-@socketio.on("disconnect")
-def on_disconnect():
-    print("[SocketIO] Client disconnected.")
-
-###############################################################################
-#                                   Main
-###############################################################################
-def main():
-    # Pre-populate shared_state with default arguments
-    args = setup_default_args()
-    apply_args_to_state(args)
-
-    # Optional: Preload screen info if needed
-    screens = get_monitors()
-    print("Detected screens:", screens)
-    screen_names, primary_index = get_screen_details()
-    print("Screen names:", screen_names, "Default selected index:", primary_index)
-
-    # Run the Flask-SocketIO app
-    # eventlet is the default async_mode if installed, but we specify it explicitly.
-    socketio.run(app, host="0.0.0.0", port=5001, debug=True)
-
-if __name__ == "__main__":
-    main()

computer_use_ootb_internal/service_teachmode_test.py

@@ -1,41 +0,0 @@
-import socketio
-
-# Create a Socket.IO client instance
-sio = socketio.Client()
-
-@sio.on('connect')
-def on_connect():
-    print("Connected to the server.")
-    # Once connected, send the event to start the teachmode process:
-    data = {
-        "user_input": "Hello, I'd like to open the Chrome browser."
-        # You can add more parameters here if needed, e.g.:
-        # "model": "teach-mode-gpt-4o",
-        # "task": "Some task",
-        # "user_id": "my_user",
-        # etc.
-    }
-    print("Emitting 'run_teachmode' event with data:", data)
-    sio.emit("run_teachmode", data)
-
-@sio.on('partial_response')
-def on_partial_response(data):
-    print("[partial_response] =>", data)
-
-@sio.on('done')
-def on_done(data):
-    print("[done] =>", data)
-    # Since the process is completed, you can disconnect:
-    sio.disconnect()
-
-@sio.on('disconnect')
-def on_disconnect():
-    print("Disconnected from server.")
-
-
-if __name__ == "__main__":
-    # Connect to the Socket.IO server (adapt host/port as needed):
-    sio.connect("http://localhost:5001")
-
-    # Keep the client alive to receive events
-    sio.wait()