### Added

• Create_GantryShaft can now see a carriage while it is moving.** A gantry

carriage stops being a block the instant it assembles — Create anchors the
contraption at the carriage's own position and removes its blocks from the
world — so every position readback went nil for the entire duration of a
move, which is exactly when a control program needs one. The peripheral now
looks for the contraption entity first and falls back to a block scan, using
Create's own rail-membership test, so positions stay live end to end and are
fractional rather than whole while travelling.

• Create_GantryShaft: getCarriage, returning position, state and the

applicable one of id/error/remaining in a single read. Preferred over the
individual getters in a control loop — every field comes from one
observation, so they cannot disagree the way separate polls straddling a
tick boundary can.

• Create_GantryShaft: getState, distinguishing empty, parked,

moving, stalled and failed. Previously every one of those answered nil
and there was no way to tell "no carriage" from "blocked" from "assembly
failed".

• Create_GantryShaft: isAssembled, isStalled, getRemainingMovement,

getLastAssemblyError, getRailStart, and disassemble. This brings the
gantry to parity with the other contraption controllers — piston, rope
pulley, bearing and elevator pulley all had assembly state and an error
readback already. getRailStart exposes the origin rail indices are
measured from, so waypoints can be anchored in world space instead of
silently shifting when a player extends the rail at the start end.

• Create_GantryShaft now queues gantry_departed, gantry_arrived,

gantry_stalled and gantry_assembly_failed, so a controller can wait on
an event instead of polling. Nothing upstream fires on any of these; the
rail is polled for edges every five ticks, and only on shafts that actually
have a computer attached.

### Fixed

• Create_GantryShaft: hasCarriage and getCarriagePosition were looking

for the wrong block.** The carriage-detection test was inverted — it
required GantryCarriageBlock.FACING to point back at the shaft, when in
Create it points *away* from it. The predicate therefore never matched a
carriage on the queried rail at all; what it did match was a carriage one
block away belonging to a *different* rail two blocks away. Since direction
iteration starts at DOWN, the practical effect was "reports the carriage
below the rail, never its own" — most visible on a stacked gantry, where
rail, carriage and second rail are exactly that geometry. Reported in #24.

• Create_GantryShaft: rail length and index ran past the end of the rail.**

The rail walk accepted any neighbouring shaft on the same *axis*; Create
joins shafts into one rail only on exact facing equality. Two abutting
shafts facing opposite ways are two rails, and sneak-placing against an
existing shaft deliberately gives you the opposite facing — so this was one
keystroke away, not a corner case. Affected getRailLength, getRailIndex
and the origin getCarriagePosition measures from.

• Create_GantryShaft could load, and generate, chunks from a Lua call.**

The rail walk called getBlockState up to 256 times per direction with no
load check, and Level.getBlockState resolves its chunk with
requireChunk = true. A rail pointing into ungenerated terrain would
therefore generate it, synchronously, on the server thread. The walk now
stops at unloaded chunks, as Create's own does.

• Create_GantryShaft documented a redstone behaviour that does not

exist.** isPowered claimed powering a shaft inverts the carriage's
direction of travel and getMovementSpeed claimed to factor that inversion
in. Create's getPinionMovementSpeed reads no redstone state whatsoever. A
powered gantry shaft stops translating its carriage and transmits rotation
into the carriage's output shaft instead — the docs had it backwards from a
real mechanic. Also now documented: the ±0.49 blocks/tick clamp, and that
canAssembleOn answers per-shaft, with single shafts never able to
assemble and the two ends requiring opposite signs.

### Docs

• altitude_sensor.getVerticalSpeed documented a distinction no Lua caller

can observe. The note said "server-side only — returns 0 on the client
side", which describes the block entity: the finite difference is
accumulated in a tick handler that returns early on a client level. But
CC:Tweaked runs Lua on the server, so a peripheral call never sees the
client copy. Reworded to what is actually observable — one tick of lag, and
0 until the sensor has ticked twice after being placed or loaded. Same
wording leak removed from four gas_provider balloon getters ("or 0 if no
server-side balloon" → "or 0 if no balloon is attached"); in a peripheral
call an attached balloon is always the server one.

### Added

• Releases now publish to CurseForge.** The CurseForge listing had been

stuck on 0.4.0 since 2026-05-11 while Modrinth carried through 0.5.1, so
anyone browsing from a launcher's CurseForge index saw a build three
releases old. Note that Create Aeronautics — which supplies the required
simulated dependency — is published only on Modrinth, and still has to
be installed from there.

### Fixed

• Runs on Create: Simulated 1.3.0.** mods.toml has always declared

simulated [1.2.1,), but 1.3.0 renamed getLaserRange to
getRaycastLength and retyped the recovery-compass placer component from
String to UUID, so the jar would not in fact have run against it. Both
are bridged: one jar now covers Simulated 1.2.1 through 1.3.0, and Sable
1.1.x through 2.0.x, with no change to any Lua-visible value.

• mounted_potato_cannon: getAimingVector and getBarrelPos returned

body/local-frame values instead of the world-frame values their docs
promise — the aim came back as a fixed body axis (e.g. {1, 0, 0}) and
the muzzle position carried raw sub-level local coordinates. Both now
project out of the host sub-level (aim rotated into world frame, muzzle
position projected to world frame), and pass through unchanged when the
cannon isn't on a sub-level ([#18], [#22]).

• linked_typewriter: the key event fired before the keypress had been

applied and carried an "alive" flag read from the pre-update state. It now
fires once the key has registered, with true as the third argument.

### Changed

• Startup logs when Create: Simulated's own ComputerCraft registration is

suppressed, alongside Avionics' own registration lines. Avionics replaces
that registration wholesale, and previously nothing in the log said so —
which made reports about missing or misbehaving Simulated peripherals hard
to read.

• Internal: the Create-side peripheral table moved out of

ComputerBehaviourMixin into CreatePeripheralRegistry; the kinetic SCADA
pack now routes entirely through KineticReadback so its two declarations
cannot drift; the three CC integration classes share one add() helper.
No behavioural change.

• CI compiles and tests against the ends of the declared upstream ranges

(Simulated 1.2.1/1.3.0 × Sable 1.1.1/2.0.3), so an upstream rename fails a
build instead of a user's launch. Adds the first tests in the repo.

> ⚠️
> **0.5.0 was withdrawn** — it crashed on launch with every mod configuration and has been removed from Modrinth and GitHub releases. Upgrade from 0.4.0 directly to 0.5.1. Since nobody should have 0.5.0, its full changelog is included below.

> 🛑
> **Breaking change for Lua scripts** (introduced in 0.5.0): the per-block getStressCapacity of the kinetic SCADA pack is renamed to **getStressContribution** (on KineticPeripheral, SimKineticPeripheral, and KineticSource). The old name collided with Create_Stressometer.getStressCapacity, which reports the network-wide total and keeps its name.

## 0.5.1

### Fixed

• 0.5.0 crashed on launch** for everyone: PeripheralComposition

resolved Simulated's portable-engine block-entity holder during mod
construction, before block-entity registries are bound, throwing
"Trying to access unbound value". Suppliers are now parked and resolved
on first lookup. 0.5.0 has been withdrawn from Modrinth and GitHub
releases — upgrade directly from 0.4.0 to 0.5.1.

• Composed peripherals now report their generic type names: portable

engines answer peripheral.hasType(name, "inventory") and are returned
by peripheral.find("inventory"). 0.5.0 composed only the methods, not
the types, so scripts that locate inventories by type still missed them
(follow-up to [#15]).

## 0.5.0 (yanked) — included because the release was withdrawn

### Added

• swivel_bearing: locking-mode control and assembly diagnostics —

isLocked, getLockingMode, setLockingMode (one of locked_always,
locked_default, unlocked_default, unlocked_always), and
getLastAssemblyException (the message from the most recent failed
assembly attempt, or nil). Public API
ink.astrius.create_avionics.api.simulated.SwivelBearingExt —
mixin-supplied interface for downstream addons.

• Kinetic SCADA pack on bare kinetic blocks: a new KineticSource generic

source applies the full pack to every KineticBlockEntity that has no
explicit peripheral (encased shafts, gearboxes, mixers, …). Thanks
@TechTastic ([#6]).

• Generic-source methods now compose onto capability-supplied peripherals:

blocks whose peripheral Avionics supplies via the NeoForge capability
(e.g. portable_engine) also expose matching generic packs instead of
shadowing them.

### Changed

• Breaking**: the per-block getStressCapacity of the kinetic SCADA pack

is renamed to getStressContribution (on KineticPeripheral,
SimKineticPeripheral, and KineticSource). The old name collided with
Create_Stressometer.getStressCapacity, which reports the network-wide
total and keeps its name.

• gas_provider.getGasType(): third-party LiftingGasType implementations

now return their class's simple name instead of "unknown". Stock types
still return "steam" / "default".

• Simulated-side peripherals deduplicated against the originals Simulated

ships: altitude_sensor, gimbal_sensor, and linked_typewriter now
extend their upstream counterparts; standalone duplicates of the
nameplate, modulating-link, directional-link, docking-connector,
nav-table, optical-sensor, and velocity-sensor peripherals were removed
in favor of upstream's, with method parity preserved. Thanks @TechTastic
([#6]).

### Fixed

• Portable engines are inventory peripherals again: the capability-supplied

peripheral no longer drops CC: Tweaked's generic inventory methods
([#15]).

### Docs

• The kinetic SCADA pack is declared once on the KineticScadaSurface

interface; the cct-javadoc fork gained interface scanning
(1.10.0-mainthread.4) so @LuaFunction default methods on interfaces
document correctly.

• StressGauge: use @code instead of @link for non-peripheral types so

references render instead of breaking.

• Full Changelog**: https://github.com/SolAstrius/CreateAvionics/compare/v0.5.0...v0.5.1

## Added

• Create: Offroad support.** New wheel_mount peripheral.
• Script overrides for steering and braking.** setSteering(-1..1) and setBrake(0..1) bypass the redstone reads. Both independent.
• Wheel state reads.** hasTire, getTireRadius, getExtension, getAngularVelocity, getTouchingFriction, isLiftedUp, plus the usual kinetic SCADA surface.

[Full changelog](https://github.com/SolAstrius/CreateAvionics/blob/main/CHANGELOG.md) · [Docs](https://solastrius.github.io/CreateAvionics/)

## Highlights

• gyroscopic_propeller_bearing actually controllable from Lua.** New persistent manual-target override replaces the bearing's automatic gravity-tracking with a script-supplied direction. State is synced to the client and survives save/load. The bearing's 12° cone clamp, redstone power gate, and stabilization-strength gate still apply on top.
• setManualTarget({x,y,z}), clearManualTarget(), getManualTarget()

## Removed

• gyroscopic_propeller_bearing.setTilt and setStrictTilt. These were never callable in 0.3.0 — CC silently dropped them because List<Double> is not a supported @LuaFunction parameter type. Even with the marshalling fixed, they were single-tick step primitives that the bearing's own physics tick overwrites on the next tick. **Use setManualTarget instead.**

[Full changelog](https://github.com/SolAstrius/CreateAvionics/blob/main/CHANGELOG.md) · [Docs](https://solastrius.github.io/CreateAvionics/)

## Highlights

• Proper Create SCADA.** Drop-in replacements for every Create peripheral (Create_CreativeMotor, Create_RotationSpeedController, Create_Speedometer, Create_Stressometer, Create_MechanicalBearing, Create_MechanicalPiston, Create_RopePulley, Create_ElevatorPulley, Create_ElevatorContact, Create_GantryShaft, Create_SequencedGearshift) - Create's verbatim methods plus a uniform topology surface (getSelfId, getSourceId, getSubnetworkAnchorId, getNetworkId, getKind, getSpeed, hasSource, isOverstressed, getStressImpact, getStressCapacity).
• nav_table per-target metadata.** Compass, recovery compass, and map targets now expose lodestone position, recovery death-pos, and map id/bounds.
• Public Java API** (ink.astrius.create_avionics.api.create.*) for downstream addons.

## Breaking - behavior

• gimbal_sensor.getLinearAcceleration now subtracts gravity. A stationary sensor previously reported ~9.8 m/s² on its vertical axis; it now reports zero. Scripts that compensated manually must drop that compensation.

## Breaking - Lua API renames

| Old | New |
|---|---|
| propeller.getKineticSpeed() | getSpeed() |
| propeller_bearing.getKineticSpeed() | getSpeed() |
| propeller_bearing.getStressApplied() | getStressImpact() |
| mounted_potato_cannon.getKineticSpeed() | getSpeed() |
| analog_transmission.getInputSpeed() | getSpeed() |
| analog_transmission.getStressApplied() | getStressImpact() |
| analog_transmission.getOutputStressApplied() | getOutputStressImpact() |

[Full changelog](https://github.com/SolAstrius/CreateAvionics/blob/main/CHANGELOG.md) · [Docs](https://solastrius.github.io/CreateAvionics/)