Changelog

All notable changes to AntiXray. Dates are in UTC. The docs always describe the current release.

Server compatibility

Each release declares the range of Hytale server versions it supports in its manifest.json. The server enforces that range itself: outside it the plugin is refused at load with an incompatibility message, rather than half-working against an API it wasn't built for.

AntiXray	Declared range	Loads on
1.3.0 (current)	>=0.6.0 <0.7.0	0.6.0 and any later 0.6.x
1.2.0	>=0.5.6 <0.6.0	0.5.6 and any later 0.5.x
1.1.0, 1.0.0	*	any server version — no check at all

Update the plugin and the server together

Hytale 0.6 changed the packet API AntiXray is built on, so each line only runs on its own server line: 1.2.0 refuses to load on 0.6, and 1.3.0 refuses 0.5. The jar name says which is which (AntiXray-1.3.0-hytale-0.6.jar).

If you're still on 1.0.0 or 1.1.0, upgrade before you update the server. Those releases declare "*", so they will load on 0.6 and misbehave instead of refusing.

1.3.0

Support for Hytale 0.6. This line targets 0.6.x only; servers still on 0.5 stay on 1.2.0.

Fixed for 0.6: fakes sent to sections the client didn't have yet

0.6 streams the world one 32-block section at a time instead of a whole column at once, so "the client has this chunk" no longer means it has every section of it. AntiXray now checks each section is on the client before obfuscating it, as the server's own block updates do, and redoes a section when the client unloads it and streams it back. Without this, fakes sent too early were dropped by the client and the section counted as done — the patchwork field from 1.1 all over again.

Changed: spectate runs on Hytale's native spectator

Spectating puts you in AntiXray's spectator game mode, the server's own Spectator with the hotbar kept for the spectator tools. The server now does what AntiXray used to fake by hand: you're invisible to non-spectators and off the player list, you fly through blocks, you can't be hurt or touch anything, and your body follows the suspect on its own. Gone with the workarounds: parking your body under the suspect (Spectate.FollowYOffset is now unused and deprecated), re-hiding you from every player each second, and the invulnerability hack. The third-person camera now stops short of walls instead of clipping into them. AntiXray still handles what the native mode doesn't: following into other worlds, cycling through suspects only, first person, the HUD and the inventory tools.

Faster: one packet per section

0.6 can change many blocks of a section in a single packet, as the server's own block updates do. Obfuscation, reveals and the debug tools now send one packet per section instead of one per block — thousands of packets per chunk become a handful.

Also in this release

• Inventories are read through 0.6's per-section inventory API, replacing the old one Hytale has marked for removal.
• Send-time mode ported to 0.6's new packet encoding.
• The jar is named after the game line it targets: AntiXray-<version>-hytale-<line>.jar.
• AntiXray shows its icon in the world settings' mods list instead of the ? placeholder.
• The documentation is versioned: pick the release you run from the version menu. The 1.2 docs (Hytale 0.5) stay online.

1.2.0

The honeypot works now. Three separate bugs meant that on a normal world, protection covered a fraction of what it claimed and detection could barely fire at all.

Fixed: obfuscation stopped a few seconds after a player joined

MaxTrapsPerPlayer capped a per-player list of every fake position at 40,000. At the default reach one player's field is roughly 100,000 positions, so the cap was hit almost immediately and new chunks stopped being obfuscated — resuming only in scraps as the client unloaded chunks. This is what "fake ores show up in random chunks, not the ones I'm standing in" was.

The field is no longer stored. Which rock becomes a fake, which ore is masked, where a section's bait sits — all of it is a pure function of the block position and the surrounding terrain, identical for every player, so the break handler recomputes the answer instead of looking it up. Nothing per-player is kept but which sections a client has already been sent. MaxTrapsPerPlayer is now unused and deprecated; it stays in the config only so existing files keep parsing.

Fixed: MaxY left the surface unprotected

The old default of 128 sits below the surface on ordinary worlds. A player standing at y140 had everything at and above eye level untouched. It is now 256. To be clear about what this knob is: it's a CPU limiter, not what keeps the surface clean — every fake requires a fully-buried block, so exposed terrain is skipped regardless. Lower it only on a strictly-underground server.

Fixed: honeypots were nearly impossible to trip

The break handler checked whether the block was still fully enclosed. It cannot be — to break a block you must be able to see its face, so the check rejected essentially every genuine hit. Traps are now recognised by position alone.

New: the ore field has two tiers

The single FakeOrePalette split in two, and that split is what makes detection trustworthy:

	Camouflage	Trap
Config	FakeOrePalette, FakeOreDensity	TrapOrePalette, TrapChancePerSection
Ores	common metals (copper, iron)	valuable ores (gold, mithril, …)
Density	hundreds per chunk	~one per 12 sections
Mining near it	revealed before you reach it	never revealed
Breaking it	nothing	honeypot hit

Real valuables are masked as plain rock and the camouflage field is common-only, so every valuable ore an X-ray user can see is a trap. They cannot tell bait from anything else, because there is nothing else. See Honeypots & Detection.

Changed: honeypot hits are a rate, not a lifetime tally

Hits used to accumulate forever, so an honest miner who tunnelled blind into a trap once a month would eventually flag themselves. Hits are now counted inside HoneypotWindowSeconds (default 30 min), like the mining-rate heuristic already was. What separates a cheater isn't the total — it's how many they find in a short time. HoneypotFlagThreshold drops from 4 to 3 to suit the window.

Also in this release

• Pinned server compatibility. The manifest declares >=0.5.6 <0.6.0 instead of "*", so AntiXray is refused at load on a server it wasn't built for — most immediately Hytale 0.6.0, whose native spectate replaces this plugin's follow-camera.
• One source of truth for the version. The banner, the panel title and the GitHub update comparison all read the version the server parsed from manifest.json, filled in by the build from gradle.properties. A constant in the source could previously drift from the built jar and make the update check compare the wrong number.
• Status tab reports camouflage and trap ids separately (Ore ids resolved: 12 camouflage, 8 trap) — a non-zero camouflage count with zero traps means detection can't fire.
• MaxChunksPerTick raised from 8 to 16, so the field keeps up with a player walking.
• The debug Nearest traps tool reads the recomputed field instead of the removed position list.

1.1.0

• Update check. On startup AntiXray asks GitHub whether a newer release exists. If there is one, a banner is printed at the end of the boot log — where it's actually visible rather than buried mid-startup — and admins are told in chat a few seconds after they join.
• "What's new" popup. Admins get the release notes for the version they're running, shown once per release, with Close (see it again next join) and Don't show until next version (persisted per admin in changelog_seen.json). Re-open it any time from Tools → What's new.
• Both lookups are asynchronous with a 5-second timeout and fail silently, so a server with no outbound internet — or one GitHub rate-limits — sees nothing at all: no errors, no startup delay, and no popup.

Upgrading from 1.0.0

The update check ships in this release, so 1.0.0 servers can't announce it — 1.1.0 is the first version that will tell you about future releases.

1.0.0

Initial release.

Protection

• Per-player packet obfuscation: fully-enclosed real ores are shown as plain rock in each client's view only; the server world is never modified.
• Fake-ore honeypot field: decoy ores scattered through hidden host rock at a configurable density, buried CoverDepth deep so they're never visible to a legitimate player.
• Wildcard block ids: FakeOrePalette and TrackedOres accept "Prefix*" entries, auto-discovering the ore variants actually registered in your world — no hand-maintained id list, and no "the ids don't match my build" problem.
• Protected blocks: a second layer for valuable non-ore blocks — hide the buried real ones, and scatter rare buried decoy chests as honeypots. Container blocks are auto-excluded from hiding, since their model replicates on a channel that can't be intercepted. See Protected Blocks.
• Send-time obfuscation (experimental): Obfuscation.SendTimeMode rewrites each chunk as it is sent, so it arrives already-obfuscated — no correction packets, no brief real-ore flash — using the server's own encoder, cached per section and shared across players. Off by default; falls back to the vanilla chunk on any error so terrain can't break. See Send-time mode.
• Safety rules: exposed faces are never touched, and breaking a block re-sends the true neighbours within RevealRadius so mining never uncovers a fake — or leaves a real ore masked.

Detection

• Honeypot-hit signal: breaking a fully-enclosed fake ore is recorded as a high-confidence X-ray indicator.
• Mining-rate heuristic: tracked-ore breaks within a sliding window.
• Combined, time-decaying suspicion score with admin alerts on flag — nothing is ever auto-punished.

Admin tools

• Single command /antixray (alias /ax) opening a tabbed panel: Suspects (worst-first, per-row spectate), Tools, Status.
• Live spectate: a real server follow-camera attached to a suspect, with first/third person, a spectator HUD, cross-world following, and your own gear stashed and restored automatically.
• Suspect inventory: read what a suspect is carrying and confiscate or destroy individual stacks, with a re-check that refuses the action if the stack moved first.
• Probe tool to discover your server's exact block ids.
• Optional debug tools behind -Dantixray.debug=true, including an X-ray audit view that proves the fake field reached your client without needing an X-ray resource pack.

Performance

• The obfuscation scan reads each chunk's slab into a single local snapshot and runs its occlusion checks against it, instead of many per-block world reads.
• Work is tracked per 32-block section rather than per column, so moving vertically never re-scans chunks the client already has.
• Packets are batched into a single flush per chunk instead of one flush per block.

General

• Per-world toggles, explicit bypass list plus optional permission bypass (ops not auto-exempt).
• English + Portuguese localization; every message editable in the data folder.