Server Performance

Machine learning sounds heavy, and owners are right to ask whether it will cost them TPS. The engine is built so that it does not.

Will the engine lag my server?

No. The everyday work, recording sales and answering /ah suggest, is light and quick. The heavy work, training the models, is deliberately kept away from your players.

When does the heavy lifting happen?

Only when your server is quiet. The engine learns your calm hours and trains then, and it refuses to start training while the server is under load, waiting until things are running smoothly with few players online. Your peak-hour performance is never the price of a smarter economy.

Where does the maths actually run?

In the plugin's native core, on a background thread, off to the side of the main game loop. Even while it is training, the server keeps ticking. If the server ever dips below a healthy frame rate, training holds off rather than push it further.

Does it use a lot of memory or disk?

No. The models it builds are small, measured in kilobytes, and it keeps a compact history of trades rather than a copy of everything forever. A large, busy server uses more than a tiny one, but it stays modest.

I run a big network. Any surprises?

The busier your server, the more the engine has to learn from, which is a good thing. Its training still waits for quiet windows, and its data collection is buffered, so a rush of sales is written in batches rather than all at once. It scales with your server rather than fighting it.