computer-use-ootb-internal 0.0.116__tar.gz → 0.0.118__tar.gz

{computer_use_ootb_internal-0.0.116 → computer_use_ootb_internal-0.0.118}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: computer-use-ootb-internal
-Version: 0.0.116
+Version: 0.0.118
 Summary: Computer Use OOTB
 Author-email: Siyuan Hu <siyuan.hu.sg@gmail.com>
 Requires-Python: >=3.11

{computer_use_ootb_internal-0.0.116 → computer_use_ootb_internal-0.0.118}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "computer-use-ootb-internal"
-version = "0.0.116"
+version = "0.0.118"
 description = "Computer Use OOTB"
 authors = [{ name = "Siyuan Hu", email = "siyuan.hu.sg@gmail.com" }]
 requires-python = ">=3.11"

{computer_use_ootb_internal-0.0.116 → computer_use_ootb_internal-0.0.118}/src/computer_use_ootb_internal/app_teachmode.py

@@ -11,11 +11,48 @@ import os # Import os for path joining
 import logging # Import logging
 import importlib # For dynamic imports
 import pkgutil # To find modules
+import sys # For logging setup
+import traceback # For logging setup
+from logging.handlers import RotatingFileHandler # For logging setup
 from fastapi import FastAPI, Request
 from fastapi.responses import JSONResponse
 from fastapi.middleware.cors import CORSMiddleware
 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
+import uvicorn # Assuming uvicorn is used to run FastAPI
+
+# --- App Logging Setup ---
+try:
+    # Log to user's AppData directory for better accessibility
+    log_dir_base = os.environ.get('APPDATA', os.path.expanduser('~'))
+    log_dir = os.path.join(log_dir_base, 'OOTBAppLogs')
+    os.makedirs(log_dir, exist_ok=True)
+    log_file = os.path.join(log_dir, 'ootb_app.log')
+
+    log_format = '%(asctime)s - %(levelname)s - %(process)d - %(threadName)s - %(message)s'
+    log_level = logging.INFO # Or logging.DEBUG for more detail
+
+    # Use rotating file handler
+    handler = RotatingFileHandler(log_file, maxBytes=5*1024*1024, backupCount=2, encoding='utf-8')
+    handler.setFormatter(logging.Formatter(log_format))
+
+    # Configure root logger
+    logging.basicConfig(level=log_level, handlers=[handler])
+
+    # Add stream handler to see logs if running interactively (optional)
+    # logging.getLogger().addHandler(logging.StreamHandler(sys.stdout))
+
+    logging.info("="*20 + " OOTB App Starting " + "="*20)
+    logging.info(f"Running with args: {sys.argv}")
+    logging.info(f"Python Executable: {sys.executable}")
+    logging.info(f"Working Directory: {os.getcwd()}")
+    logging.info(f"User: {os.getenv('USERNAME')}")
+
+except Exception as log_setup_e:
+    print(f"FATAL: Failed to set up logging: {log_setup_e}")
+    # Fallback logging might be needed here if file logging fails
+
+# --- End App Logging Setup ---
 
 app = FastAPI()
 
@@ -56,6 +93,8 @@ class RateLimiter:
 
 
 def log_ootb_request(server_url, ootb_request_type, data):
+    logging.info(f"OOTB Request: Type={ootb_request_type}, Data={data}")
+    # Keep the requests post for now if it serves a specific purpose
     logging_data = {
         "type": ootb_request_type,
         "data": data,
@@ -65,7 +104,10 @@ def log_ootb_request(server_url, ootb_request_type, data):
         server_logging_url = server_url + "/update_ootb_logging"
     else:
         server_logging_url = server_url
-    requests.post(server_logging_url, json=logging_data)
+    try:
+        requests.post(server_logging_url, json=logging_data, timeout=5)
+    except Exception as req_log_e:
+        logging.warning(f"Could not log ootb request to server {server_logging_url}: {req_log_e}")
 
 
 class SharedState:
@@ -138,35 +180,41 @@ def prepare_environment(state):
 
 @app.post("/update_params")
 async def update_parameters(request: Request):
-    data = await request.json()
-    
-    if 'task' not in data:
-        return JSONResponse(
-            content={"status": "error", "message": "Missing required field: task"},
-            status_code=400
-        )
-    
-    shared_state.args = argparse.Namespace(**data)
-    shared_state.task_updated = True
-    
-    # Update shared state when parameters change
-    shared_state.model = getattr(shared_state.args, 'model', "teach-mode-gpt-4o")
-    shared_state.task = getattr(shared_state.args, 'task', "Following the instructions to complete the task.")
-    shared_state.selected_screen = getattr(shared_state.args, 'selected_screen', 0)
-    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")
-    
-    log_ootb_request(shared_state.server_url, "update_params", data)
+    logging.info("Received request to /update_params")
+    try:
+        data = await request.json()
+        
+        if 'task' not in data:
+            return JSONResponse(
+                content={"status": "error", "message": "Missing required field: task"},
+                status_code=400
+            )
+        
+        shared_state.args = argparse.Namespace(**data)
+        shared_state.task_updated = True
+        
+        # Update shared state when parameters change
+        shared_state.model = getattr(shared_state.args, 'model', "teach-mode-gpt-4o")
+        shared_state.task = getattr(shared_state.args, 'task', "Following the instructions to complete the task.")
+        shared_state.selected_screen = getattr(shared_state.args, 'selected_screen', 0)
+        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")
+        
+        log_ootb_request(shared_state.server_url, "update_params", data)
 
-    # Call the (now dynamic) preparation function here, after parameters are updated
-    prepare_environment(shared_state)
+        # Call the (now dynamic) preparation function here, after parameters are updated
+        prepare_environment(shared_state)
 
-    return JSONResponse(
-        content={"status": "success", "message": "Parameters updated", "new_args": vars(shared_state.args)},
-        status_code=200
-    )
+        logging.info("Parameters updated successfully.")
+        return JSONResponse(
+            content={"status": "success", "message": "Parameters updated", "new_args": vars(shared_state.args)},
+            status_code=200
+        )
+    except Exception as e:
+        logging.error("Error processing /update_params:", exc_info=True)
+        return JSONResponse(content={"status": "error", "message": "Internal server error"}, status_code=500)
 
 @app.post("/update_message")
 async def update_message(request: Request):
@@ -332,6 +380,8 @@ async def get_status(request: Request):
     )
 
 def process_input():
+    global shared_state
+    logging.info("process_input thread started.")
     shared_state.is_processing = True
     shared_state.should_stop = False
     shared_state.is_paused = False
@@ -412,55 +462,75 @@ def process_input():
         shared_state.is_paused = False
         shared_state.stop_event.clear()
         print("Processing completed, ready for new tasks")
+    logging.info("process_input thread finished.")
 
 def main():
-    global app, shared_state, rate_limiter
-    
-    parser = argparse.ArgumentParser(
-        description="Run a synchronous sampling loop for assistant/tool interactions in teach-mode."
-    )
-    parser.add_argument("--model", default="teach-mode-gpt-4o")
-    parser.add_argument("--task", default="Create a claim on the SAP system, using Receipt.pdf as attachment.")
-    parser.add_argument("--selected_screen", type=int, default=0)
-    parser.add_argument("--user_id", default="star_rail_dev")
-    parser.add_argument("--trace_id", default="scroll")
-    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",
-        help="Server URL for the session"
-    )
+    # Logging is set up at the top level now
+    logging.info("App main() function starting setup.")
+    global app, shared_state, rate_limiter # Ensure app is global if needed by uvicorn
+    parser = argparse.ArgumentParser()
+    # Add arguments, but NOT host and port
+    parser.add_argument("--model", type=str, default="teach-mode-gpt-4o", help="Model name")
+    parser.add_argument("--task", type=str, default="Following the instructions to complete the task.", help="Initial task description")
+    parser.add_argument("--selected_screen", type=int, default=0, help="Selected screen index")
+    parser.add_argument("--user_id", type=str, default="hero_cases", help="User ID for the session")
+    parser.add_argument("--trace_id", type=str, default="build_scroll_combat", help="Trace ID for the session")
+    parser.add_argument("--api_keys", type=str, default="sk-proj-1234567890", help="API keys")
+    parser.add_argument("--server_url", type=str, default="http://ec2-44-234-43-86.us-west-2.compute.amazonaws.com", help="Server URL for the session")
 
     args = parser.parse_args()
+
+    # Validate args or set defaults if needed (keep these)
+    if not hasattr(args, 'model'): args.model = "default_model"
+    if not hasattr(args, 'task'): args.task = "default_task"
+    if not hasattr(args, 'selected_screen'): args.selected_screen = 0
+    if not hasattr(args, 'user_id'): args.user_id = "unknown_user"
+    if not hasattr(args, 'trace_id'): args.trace_id = "unknown_trace"
+    if not hasattr(args, 'api_keys'): args.api_keys = "none"
+    if not hasattr(args, 'server_url'): args.server_url = "none"
+
     shared_state = SharedState(args)
-    rate_limiter = RateLimiter(interval_seconds=2)
+    rate_limiter = RateLimiter(interval_seconds=2) # Re-initialize rate limiter
+    logging.info(f"Shared state initialized for user: {args.user_id}")
 
-    import uvicorn
-    import platform
-    import os
-    
-    # Default port
-    port = 7888
+    # --- Restore original port calculation logic --- 
+    port = 7888 # Default port
+    host = "0.0.0.0" # Listen on all interfaces
     
-    # Determine port based on Windows username
     if platform.system() == "Windows":
-        username = os.environ["USERNAME"].lower()
-        if username == "altair":
-            port = 14000
-        elif username.startswith("guest") and username[5:].isdigit():
-            num = int(username[5:])
-            if 1 <= num <= 10:
-                port = 14000 + num
+        try:
+            username = os.environ["USERNAME"].lower()
+            logging.info(f"Determining port based on Windows username: {username}")
+            if username == "altair":
+                port = 14000
+            elif username.startswith("guest") and username[5:].isdigit():
+                num = int(username[5:])
+                if 1 <= num <= 10: # Assuming max 10 guests for this range
+                    port = 14000 + num
+                else:
+                     logging.warning(f"Guest user number {num} out of range (1-10), using default port {port}.")
             else:
-                port = 7888
-        else:
-            port = 7888
-    
-    uvicorn.run(app, host="0.0.0.0", port=port)
+                logging.info(f"Username '{username}' doesn't match specific rules, using default port {port}.")
+        except Exception as e:
+             logging.error(f"Error determining port from username: {e}. Using default port {port}.", exc_info=True)
+    else:
+         logging.info(f"Not running on Windows, using default port {port}.")
+    # --- End of restored port calculation --- 
+
+    logging.info(f"Final Host={host}, Port={port}")
+
+    try:
+        logging.info(f"Starting Uvicorn server on {host}:{port}")
+        # Use the calculated port and specific host
+        uvicorn.run(app, host=host, port=port)
+        logging.info("Uvicorn server stopped.")
+    except Exception as main_e:
+        logging.error("Error in main execution:", exc_info=True)
+    finally:
+        logging.info("App main() function finished.")
 
 if __name__ == "__main__":
-    # main()
+    main()
 
     # Test log_ootb_request
     log_ootb_request("http://ec2-44-234-43-86.us-west-2.compute.amazonaws.com", "test_request", {"message": "Test message"})

{computer_use_ootb_internal-0.0.116 → computer_use_ootb_internal-0.0.118}/src/computer_use_ootb_internal/guard_service.py

@@ -34,7 +34,7 @@ _LISTEN_HOST = "0.0.0.0" # Listen on all interfaces
 _LISTEN_PORT = 14000 # Port for server to POST commands TO
 # _SHARED_SECRET = "YOUR_SECRET_HERE" # !! REMOVED !! - No secret check implemented now
 # --- End Server POST Configuration ---
-_SERVER_STATUS_REPORT_URL = "http://52.160.105.102:7000/api/guard/status" # URL to POST status back TO
+_SERVER_STATUS_REPORT_URL = "http://52.160.105.102:7000/guard/status" # URL to POST status back TO (Path changed)
 _LOG_FILE = pathlib.Path(os.environ['PROGRAMDATA']) / "OOTBGuardService" / "guard_post_mode.log" # Different log file
 # --- End Configuration ---
 
@@ -427,18 +427,34 @@ class GuardService(win32serviceutil.ServiceFramework):
                      token = win32ts.WTSQueryUserToken(session_id)
                      env = win32profile.CreateEnvironmentBlock(token, False)
                      startup = win32process.STARTUPINFO()
+                     # Simplify startup flags: Run hidden, don't explicitly set desktop
                      startup.dwFlags = win32process.STARTF_USESHOWWINDOW
-                     startup.wShowWindow = win32con.SW_SHOW # Try showing the window
-                     startup.lpDesktop = 'winsta0\\default'
+                     startup.wShowWindow = win32con.SW_HIDE
+                     # startup.lpDesktop = 'winsta0\\default' # Removed
+
                      creation_flags = win32process.CREATE_NEW_CONSOLE | win32process.CREATE_UNICODE_ENVIRONMENT
+                     # Define cwd as user's profile directory if possible
+                     user_profile_dir = win32profile.GetUserProfileDirectory(token)
 
                      hProcess, hThread, dwPid, dwTid = win32process.CreateProcessAsUser(
                          token, self.python_exe, self.ootb_command,
-                         None, None, False, creation_flags, env, None, startup
+                         None, None, False, creation_flags, env, 
+                         user_profile_dir, # Set current directory
+                         startup
                      )
-                     log_info(f"Successfully started OOTB for user '{user}' (PID: {dwPid}).")
-                     started_count += 1
-                     target_users_started.add(user)
+                     log_info(f"CreateProcessAsUser call succeeded for user '{user}' (PID: {dwPid}).")
+
+                     # Add post-start check
+                     time.sleep(1) # Small delay
+                     if psutil.pid_exists(dwPid):
+                         log_info(f"Process PID {dwPid} confirmed to exist shortly after creation.")
+                         started_count += 1
+                         target_users_started.add(user)
+                     else:
+                         log_warning(f"Process PID {dwPid} reported by CreateProcessAsUser does NOT exist shortly after creation. It likely exited immediately.")
+                         users_failed_to_start.add(user)
+                         # Attempt to get exit code? Difficult without waiting.
+
                      win32api.CloseHandle(hProcess)
                      win32api.CloseHandle(hThread)