mirror of
https://github.com/jhd3197/ServerKit.git
synced 2026-10-28 15:47:50 +09:00
Juan Denis
b739a6b17e
The panel runs async_mode='threading' under a plain threaded gunicorn worker; nothing imports gevent. It was still installed everywhere, and on Python 3.14 (no wheel yet) its source build is what broke installs in issue #148. backend/Dockerfile was unreferenced and still launched the gevent-websocket worker that double-answers the WS handshake; the run.py hint recommended the same worker.
36 lines
1.1 KiB
Python
36 lines
1.1 KiB
Python
import os
|
|
import sys
|
|
from dotenv import load_dotenv
|
|
|
|
# Load .env file before creating app
|
|
load_dotenv()
|
|
|
|
from app import create_app, get_socketio
|
|
|
|
app = create_app()
|
|
socketio = get_socketio()
|
|
|
|
if __name__ == '__main__':
|
|
port = int(os.environ.get('PORT', 5000))
|
|
env = os.environ.get('FLASK_ENV', 'development')
|
|
debug = env == 'development'
|
|
|
|
# Prevent running development server in production
|
|
if env == 'production':
|
|
print("ERROR: Do not run the development server in production!", file=sys.stderr)
|
|
print("Use a production WSGI server like gunicorn instead:", file=sys.stderr)
|
|
print(" gunicorn -w 1 --threads 100 run:app", file=sys.stderr)
|
|
sys.exit(1)
|
|
|
|
# Use SocketIO to run the app (supports WebSocket) - only in development
|
|
# use_reloader with stat polling for WSL compatibility
|
|
socketio.run(
|
|
app,
|
|
host='0.0.0.0',
|
|
port=port,
|
|
debug=debug,
|
|
allow_unsafe_werkzeug=True,
|
|
use_reloader=True,
|
|
reloader_type='stat' # Use polling instead of inotify (for WSL)
|
|
)
|