XGuardian ML pipeline
Asynchronous demo ingestion and model updates — useful for operators who will label samples, not a magic “ban this demo” button.
XGuardian is the machine-learning side of ImpactGuard. It is a pipeline: demos go in, labels get applied, models retrain, weights eventually reach servers that are allowed to use them. It is not a live overwatch jury and it is not a replacement for the 52 CS2 rule modules.
What operators actually do
- Upload a demo you already believe is clean or dirty.
- Label it. Garbage labels make garbage models.
- Watch the public XGuardian tracker for ingestion and health, not for a courtroom verdict.
- Leave enforcement on confidence-gated modules while models catch up.
How this fits detection
Rule modules still do the nightly work. ML is extra signal for later revisions. If you turn every experimental score into an instant ban, you will regret it. Keep confidence floors in place.
Deeper ops notes live in learning docs.
Questions operators ask
Will uploading a demo ban the player?
No. XGuardian is an asynchronous learning workflow. Samples improve future models and may not create an enforcement event at all.
Where can I see pipeline health?
The public tracker is /ml. Staff with panel access also see training and labeled-sample tools.