computer-use-ootb-internal 0.0.96.post1__py3-none-any.whl → 0.0.97.post1__py3-none-any.whl

computer_use_ootb_internal/app_teachmode.py

@@ -4,6 +4,8 @@ import json
 from datetime import datetime
 import threading
 import requests
+import platform  # Add platform import
+import subprocess  # Add subprocess import
 from fastapi import FastAPI, Request
 from fastapi.responses import JSONResponse
 from fastapi.middleware.cors import CORSMiddleware
@@ -87,6 +89,30 @@ class SharedState:
 shared_state = None
 rate_limiter = RateLimiter(interval_seconds=2)
 
+# Add the new prepare_environment function here
+def prepare_environment(state):
+    """Prepares the environment before starting the main processing loop, e.g., opening specific apps."""
+    if platform.system() == "Windows":
+        # Assuming Star Rail mode is indicated by user_id containing "star_rail"
+        # You might need to adjust this condition based on the actual logic in run_teachmode_args
+        is_star_rail = "star_rail" in state.user_id.lower() 
+        
+        if is_star_rail:
+            print("Star Rail mode detected on Windows. Opening Edge browser...")
+            url = "https://sr.mihoyo.com/cloud/#/"
+            try:
+                # Use cmd /c start to open the URL in Edge without blocking
+                subprocess.run(["cmd", "/c", "start", "msedge", url], check=True, shell=False)
+                print(f"Successfully requested Edge to open {url}")
+            except FileNotFoundError:
+                print("Error: Microsoft Edge not found. Please ensure it's installed and in PATH.")
+            except Exception as e:
+                print(f"Error opening Edge: {e}")
+    else:
+        # Placeholder for potential preparations on other OS or non-Star Rail modes
+        print("Environment preparation: No specific actions required for this OS/mode.")
+
+
 @app.post("/update_params")
 async def update_parameters(request: Request):
     data = await request.json()
@@ -285,6 +311,9 @@ def process_input():
     shared_state.is_paused = False
     shared_state.stop_event.clear()  # Ensure stop event is cleared at the start
     
+    # Call the preparation function before starting the loop
+    prepare_environment(shared_state)
+    
     print(f"start sampling loop: {shared_state.chatbot_messages}")
     print(f"shared_state.args before sampling loop: {shared_state.args}")
 

computer_use_ootb_internal/computer_use_demo/animation/click_animation.py

@@ -75,7 +75,7 @@ def _ensure_app():
 # 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
+def show_click(x: int, y: int, duration_ms: int = 2000, existing_ms: int = 2000):  # 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:
@@ -116,7 +116,7 @@ def _clean_animation(animation):
 
 
 # ---------- 新增函数 ----------
-def show_move_to(x1: int, y1: int, x2: int, y2: int, duration_ms: int = 1000, existing_ms: int = 1000):
+def show_move_to(x1: int, y1: int, x2: int, y2: int, duration_ms: int = 1000, existing_ms: int = 3000):
     """
     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)，
@@ -179,7 +179,7 @@ def show_move_to(x1: int, y1: int, x2: int, y2: int, duration_ms: int = 1000, ex
 # ---------- 命令行测试 ----------
 if __name__ == "__main__":
     # 确保应用程序实例存在
-    # _ensure_app()
+    _ensure_app()
     
     # 测试点击
     print("Testing non-blocking click animation...")
@@ -191,7 +191,7 @@ if __name__ == "__main__":
     x1, y1 = 200, 200
     x2, y2 = 600, 600
     # show_click(x1, y1)
-    show_move_to(x1, y1, x2, y2)
+    show_move_to(x1, y1, x2, y2, duration_ms=2000)
     
     # # 测试先移动，然后点击
     print("\nTesting sequence with pyautogui simulation...")
@@ -199,7 +199,7 @@ if __name__ == "__main__":
     x4, y4 = 400, 500
     
     # 启动移动动画
-    show_move_to(x3, y3, x4, y4)
+    show_move_to(x3, y3, x4, y4, duration_ms=1500)
     
     # 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))

computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/test_capture.py

@@ -1,8 +1,8 @@
-from PIL import ImageGrab
-
-bbox=(2560, 366, 2560+1920, 366+1080)
-        
-screenshot = ImageGrab.grab(bbox=bbox, all_screens=True)
-
-screenshot = screenshot.convert('RGB')
+from PIL import ImageGrab
+
+bbox=(2560, 366, 2560+1920, 366+1080)
+        
+screenshot = ImageGrab.grab(bbox=bbox, all_screens=True)
+
+screenshot = screenshot.convert('RGB')
 screenshot.save("screenshot.png")

computer_use_ootb_internal/example_websocket_js.html

@@ -0,0 +1,41 @@
+<!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/requirements-lite.txt

@@ -15,5 +15,4 @@ uiautomation
 pywinauto
 textdistance
 matplotlib
-litellm
-PySide6
+litellm

computer_use_ootb_internal/service_teachmode.py

@@ -0,0 +1,194 @@
+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

@@ -0,0 +1,41 @@
+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()

{computer_use_ootb_internal-0.0.96.post1.dist-info → computer_use_ootb_internal-0.0.97.post1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: computer-use-ootb-internal
-Version: 0.0.96.post1
+Version: 0.0.97.post1
 Summary: Computer Use OOTB
 Author-email: Siyuan Hu <siyuan.hu.sg@gmail.com>
 Requires-Python: >=3.11
@@ -14,15 +14,14 @@ Requires-Dist: matplotlib
 Requires-Dist: opencv-python
 Requires-Dist: pre-commit==3.8.0
 Requires-Dist: pyautogui==0.9.54
-Requires-Dist: pyside6
 Requires-Dist: pytest-asyncio==0.23.6
 Requires-Dist: pytest==8.3.3
-Requires-Dist: pywinauto
+Requires-Dist: pywinauto; sys_platform == 'win32'
 Requires-Dist: ruff==0.6.7
 Requires-Dist: screeninfo
 Requires-Dist: streamlit>=1.38.0
 Requires-Dist: textdistance
-Requires-Dist: uiautomation
+Requires-Dist: uiautomation; sys_platform == 'win32'
 Provides-Extra: dev
 Requires-Dist: pytest-asyncio>=0.23.6; extra == 'dev'
 Requires-Dist: pytest>=8.3.3; extra == 'dev'

{computer_use_ootb_internal-0.0.96.post1.dist-info → computer_use_ootb_internal-0.0.97.post1.dist-info}/RECORD

@@ -1,16 +1,19 @@
 computer_use_ootb_internal/README.md,sha256=FxpW95lyub2iX73ZDfK6ML7SdEKg060H5I6Grub7li4,31
-computer_use_ootb_internal/app_teachmode.py,sha256=zmUPvFjqdhysnN1bD2QQhaAKONnAtMFd03Rb9gF1l6c,15646
+computer_use_ootb_internal/app_teachmode.py,sha256=5Ek9cY9-wZvgTw67noYHGvE-oWu28biXU--4kfamjAs,17127
 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/requirements-lite.txt,sha256=5DAHomz4A_P2BmTIXNkNqkHbnIF0AyZ4_1XAlb1LaYs,290
+computer_use_ootb_internal/example_websocket_js.html,sha256=BLYwDExVlgiAX4vXVXW3RuP5KD8FXE4EFXIl54bwF7w,1322
+computer_use_ootb_internal/requirements-lite.txt,sha256=2C4OH_GRzuDsR-c9VpJee_te-VqLOf-KDk3LXKi3qdk,282
 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/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=j3v-CrFp61mZiJ4-eT06txkOoguZipbakrQFj2d_mbk,7759
 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=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
-computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/test_capture.py,sha256=26uxu70di9Me-jseym5ejF7RNIbP35uPEiipN5qamIc,211
+computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/test_capture.py,sha256=YbLpuD-mSFiOU7j6HKrtX05oRP9ciEBxxGIsyse-nPI,204
 computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/uia_parser.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/utils.py,sha256=GEA1ES7vOpHBg_Suxpl99reh34kRG4RQpp072JQBK5c,9787
 computer_use_ootb_internal/computer_use_demo/gui_agent/gui_parser/simple_parser/icon_detection/icon_detection.py,sha256=ysTgvtjE1XM7QSrLLy1HD0i6_7iOb9GME5FWJljrJg0,9752
@@ -35,7 +38,7 @@ computer_use_ootb_internal/computer_use_demo/tools/computer_marbot.py,sha256=zZu
 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.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-0.0.97.post1.dist-info/METADATA,sha256=RodVMKF5EUHUHdzQg84aLdfxiM-PYC0d0gH8eRGWdjg,942
+computer_use_ootb_internal-0.0.97.post1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+computer_use_ootb_internal-0.0.97.post1.dist-info/entry_points.txt,sha256=-AbmawU7IRQuDZihgVMVDrFoY4E6rnXYOUB-5vSeBKs,93
+computer_use_ootb_internal-0.0.97.post1.dist-info/RECORD,,