MLSAC AI AntiCheat Plugin v1.0-b42

[•] Feat: implement HologramManager to display dynamic AI probability names

[•] Due to internal issues, we will be moving from https://mlsac.wtf to https://mlsac.net and from github.com/MLSAC/client-side to https://github.com/SoMax1soft/MLSAC. We will disclaim responsibility if you have not updated from the official channels above.

[•] Update AI server endpoint in config.yml

While on the server, run the command /mlsac reinstall
This command will delete the previous config but preserve the value
detection-ai: true/false and your api key, and automatically insert it.

This is required for this update.

[•] Now, the cheat probability in all parameters is displayed as a % instead of 0.1, etc.
[•] Fixed the anti-cheat disabler due to a package bug (versions 1.17, 1.18, and 1.19 now work perfectly)
[•] The /mlsac suspects menu has been completely redesigned for full server-side customization, and many parameters are no longer hardcoded into the plugin itself. Customizing actions for specific keys, etc. is now possible.
[•] Holograms above the player have been rewritten and optimized; they no longer affect PvP (all attacks now hit even if the hologram is low).
[•] Holograms now show the result specifically for the model that produces it: F for FAST and P for PRO.
[•] /mlsac prob now also displays PRO and FAST results separately.
[•] The PRO model is now enabled by default.
[•] Trolling system: if a player receives a certain number of detections within a 25-second window, two things can happen.
The inventory will be moved and the sword will drop very far away. This depends on the configuration.
[•] /mlsac profile nickname - has been slightly redesigned and a clickable link to the player's cheat list has been added. This feature will be available within a day after the update.
[•] Fixed a false FOLIA detection on versions 1.21.
[•] Added helpful comments in the config for easy setup.
[•] FOLIA VERSIONS HOT-FIX (If you are not on folia no need to change)

[•] Full update for the Cheat-List system
[•] Fixed bugs for the /mlsac suspects command

Update Fast model
- Fix some False positives
and GRIM fix
You can use GRIM WITH MLSAC without errors.
Before on some servers grim is blocking but now its tottaly fixed.

[•] Fix problem

[+] Added model name display in alerts ({MODEL} placeholder)
[+] Added per-model "only-alert" mode configuration
[+] Added playerName to inference requests
[+] Added pluginVersion to session creation
[/] Changed config structure: models now have name + only-alert settings
[/] Changed only-alert behavior: buffer no longer accumulates in monitoring mode
[/] Moved pluginVersion from predict request to connect request
[-] Removed global only-alert setting (now per-model)
[+] Added Server Streaming support for AI checks: model responses are now processed as a real-time stream.
[+] Added mask-armor functionality for custom mask items (stats for armor, toughness, and knockback resistance).
[/] Optimized the /mlsac suspects menu: head and data loading is now fully asynchronous, eliminating TPS drops when opening.
[/] Rewrote the SignalR networking layer: migrated to RxJava Observable for more efficient prediction handling.
[-] Fixed a bug where the model name was displayed as "unknown" in logs and alerts.
[-] Adjusted the message displayed when using the /mlsac stop all command.