🤖 Agent: Players rebind a mod's keys on its Options > Mods page (the game's own key rows, saved in horizon-mods.txt), with no WIT type change, so the published mods keep loading. Verified with main merged: clippy, the client, mods and SDK tests, and the in-game check in both modes (24/24).
🤖 Agent: native shader packs' post stage in the game with hot reload; pr_ready, 66 affected harnesses and a paired A/B (no cost without a pack) passed.
🤖 Agent: Horizon's official FPS Counter and Keystrokes as mods in mods/official/, same options and look as the extras (pixel-identical, frame cost within noise). Main merged in (wasmtime 48). Verified: pr_ready, clippy and tests of both mods, horizon-mods tests, both mods' settings pages in game.
A native pack named by HORIZON_SHADER_PACK (a folder with horizon-shaders.toml
or a .hshader in shaderpacks/) is checked, then its post passes run over the
world's own target, the one a shaders/ pack reads: no second target and no
extra pass, and without a pack the frame only tests one Option. The pack is
looked at twice a second; a saved file is checked and rebuilt in place, and a
version that fails is refused in the chat while the last good one keeps
drawing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A new case in ui's widget variant stops every ui mod built before it from
loading: the runtime links a component only when each imported
function's types match exactly ("expected variant of 5 cases, found
4"), so the published FPS Counter and Keystrokes would break. Instead,
the game lists a mod's keys first on its page in Options > Mods, as Key
Binds rows, and a mod with keys gets a page even without ui. No WIT
type changes. docs/api/stability.md said a new case the mod sends isn't
breaking; it is, and now says so.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: Mod API (experimental): render.set-zoom and input.claim-wheel, held mod keys let go when a screen opens, mod keys on mouse buttons. The consume choice is in stability.md. Verified: pr_ready (1300 tests), sandbox tests, in game 22/22 in-frame and on the server thread.
🤖 Agent: Mod API calls the official HUD mods use: render corner panels, text width and frame stats; input mappings and clicks per second; a ui button widget. Main merged in (after #1914's widget refactor). Verified: horizon-mods tests, clippy with and without mods, horizon-client tests (1121 lib), Keystrokes' settings page in game.
A ui page's key widget shows one of the mod's keys as Key Binds does
("Zoom: Z"). A click waits for the next key or mouse button, which
becomes the key; Esc binds none. The game keeps the choice in
horizon-mods.txt (key.<mod>.<name>:key.keyboard.z), applies it when
the mods load and to keys registered later, and lets a held key go when
it is rebound. register-key's key is now the default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: Mod API capability input (experimental): claimed items' use and the mod's keys. Main merged in beside command and render (9 capabilities). Verified: horizon-mods tests (sandbox, API reference), clippy with and without mods, horizon-client tests (1121 lib), earlier pr_ready and in-game runs in both modes.
render.set-zoom divides the field of view by up to 10, eased through the
game's FOV easing or at once, with a turning scale (at least 0.05); only
while playing, and the spyglass's own view wins. The strongest of several
mods' zooms counts, and the built-in zoom keeps working beside it.
input.claim-wheel takes the wheel from the hotbar while playing: its steps
are the mod's wheel events. That is the one input a mod consumes; its keys
never do (docs/api/stability.md). A screen opening now lets the mod's held
keys up, and mod keys bound to mouse buttons (key.mouse.4) are heard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Capability::ALL is 9 with Input; the SDK re-exports command and input;
the API reference is regenerated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: ready. Linux runner support (runner-env/vanilla-data Linux paths, linux-runner.sh, runner_pause.py with Ludovic's lock/idle-15-min rule, heat, game, GPU and pause-file checks), desktop parity host, nightly and parity-main pinned to the Mac. verify and parity passed at this head. The desktop runner stays without the horizon-game label until its trial is identical: that is a runner setting, not in this PR. Desktop full run: 342/343 in 39.3 min (the 1 is a harness timing flake, fixed by #1997).
🤖 Agent: Mod API capability render (experimental): HUD shapes and camera shake. Main merged in beside command (8 capabilities). Verified: horizon-mods tests, clippy with and without mods, earlier pr_ready, harnesses and in-game runs.
Both add a Capability variant: the list holds 8 now. The SDK re-exports
both interfaces; the API reference is regenerated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: thank you kacpherek! /pregen is in, off by default: a server turns it on with horizon-pregen=true in server.properties (server.toml [properties]), single player and LAN with --pregen. With it off, the command tree and /help match vanilla. Tested on a dedicated server through horizon-server (on: 1,369 chunks generated and saved), single player on the server thread and in-frame, and 11 harnesses including command_exec.
nix-build names a dev output's GC root <name>-dev, so the runner's PKG_CONFIG_PATH pointed at links
that didn't exist and every client build failed on libudev; it now uses the store path nix-build
prints. The package temperature spikes for a second under load, so the heat check takes the median
of 5 readings. Restarting the watcher no longer throttles a running job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: Mod API capability `command` and the Predator Missile example, by @sdraeger. Thank you! Main merged in, with `command` kept beside `ui` as a 0.1 preview interface. Verified: pr_ready (2106 workspace tests), and in game in-frame and on the server thread, including survival without cheats refused.
Conflicts kept both sides: the capability list (now 7), the WIT, the
SDK's re-exports and docs, the mods state's fields, the workspace
members and AGENTS.md's macOS --exclude list. The API reference is
regenerated, and docs/api/stability.md lists `command` with the 0.1
preview interfaces.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: verified: OUTPUT_VERSION = 1 and output-versions.txt carry the digest measured on 10 CI preparations. tools/prepare_output.py check passes on v0.2.64's preparation and fails on v0.2.47's. This head's own parity job prepares with its preparer and checks the digest. Merging.
parity-main runs the vanilla harnesses on every push to main, one run at a
time: merges that land during a run wait in one pending run, which checks
every merge since main's last green parity-main (tools/parity_main.py base,
then parity.py --changed). A failure is bisected over the batch's
first-parent merges, running only the failing harnesses, and the breaking
merge's pull request gets a comment.
Game releases ship only a commit both checks and parity-main passed on
(release_train.py; release.yml also starts when parity-main finishes; a
forced release doesn't wait). The per-PR parity job now runs only for a PR
that changes a harness, an oracle, a probe or the runner, or carries
run-parity.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The parity job checks out the PR head, so every branch made before #1964
still selected with the old per-crate rule. The step now runs main's
tools/parity.py and tools/parity_reach.py with --changed --list, restores the
head's files, and runs the named harnesses with the head's parity.py. A
branch that edits either file, or a failure of main's selection, falls back
to the head's own `--changed` run.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ludovic's rule (2026-10-26): runner_pause.py pauses at once on keyboard, mouse or gamepad input
(read from /dev/input, only whether bytes came), on an unlocked session touched in the last
15 min, at 90 °C, on a game or the GPU. Fixes the service start: the link loop's last test
failed the script under set -e.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`--changed` / `--changed-files` now narrow each harness's recorded Rust
sources to those its probes can reach (tools/parity_reach.py): an item-level
reachability walk from the probe modules a harness ran (or the binary it
built) over free items, inherent methods, trait impls and macros, matched by
name so it only over-approximates. Parent modules count through their `mod`
declarations, compared with the merge base. A harness whose probe can't be
placed keeps every source; `--by-crate` turns the narrowing off.
On the last 20 PRs merged on 2026-10-25: 2255 -> 1061 harnesses picked,
2384 -> 1592 harness-minutes. Replaying the week's 61 failed CI parity runs,
every failed harness the old selection picked is still picked; LLVM coverage of
six harnesses' probes executes only kept files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Vanilla has no /pregen, so with it always registered /help and the command
tree differed from vanilla's (the command_exec harness failed on /help).
Horizon's own commands are now added to the game's tree only when the owner
asks: `horizon-pregen=true` in server.properties (`horizon-pregen = true`
under [properties] in server.toml) for a dedicated server, `--pregen` for
the game (single player, LAN). Off, the trees are vanilla's own, unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Compile and validate original exporter bundles during release builds, cache development compilation, and reuse initialized JVMs and collision data during preparation. Keep legacy preparers compatible while current releases need only a Java runtime.
Co-Authored-By: Codex <noreply@openai.com>
The FPS counter and keystrokes extras rewritten as mods on the public API
(render, input, ui, storage), in mods/official/, with the extras' options
on their settings pages and the extras' look: the same boxes, colours and
text positions, pixel for pixel. They are workspace members, so clippy and
pr_ready cover them; a maintainer publishes them in horizon-mods. The
built-in extras stay until the removal PR, which installs these for the
players who had them on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Share the asset index and verification state, stage checked imports while hashing, preserve indexed replacements, and clone independent resource copies where supported.
Co-Authored-By: Codex (GPT-6.1) <noreply@openai.com>
runner_pause.py pauses on any seat0 session (locked too), games, GPU use or a pause file, and is
the only thing that starts the runner; a job under way is throttled, then the runner stops. The
desktop host in parity_hosts.json refuses work while paused and its workers stop on a pause. The
nightly audit stays on the Mac mini. Docs and a unit test of the watcher.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The built-in extras become official mods; these are the calls the FPS
counter and keystrokes overlay need to look and behave the same:
- render.set-corner(corner, w, h, shapes): a box the game places with
its own corner stacking (under the status effects, with the overlays
and hud panels), clear-corner; text-width(text) in the HUD font
(advances for U+0000..U+024F, 6 otherwise); frame-stats(): the game's
FPS counter (fps, min, 1% low).
- input.mapping(name): a movement or click mapping's key label and
whether it is held; clicks-per-second() for attack and use.
- ui: a button widget (200 x 20, select(id, 0)) for settings pages.
- The game keeps its FPS and click counters running while a mod runs,
and refreshes the snapshot every frame for mods with a frame hook.
- The ui test guest gains /corner and /readings, with a sandbox test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Launchers copy a player's prepared resources to a new release while its preparer declares the
same OUTPUT_VERSION (the launcher side is its own pull request). The number is bumped by hand;
CI keeps it honest: the parity job already prepares with every change to the preparer, and
tools/prepare_output.py check now fails when that preparation's digest is not the one
crates/horizon-prepare/output-versions.txt gives the current number, or when the file changed
other than by appending a version. The digest canonicalises what differs between two runs of
the same preparer (JSON key order, creative tab tag order, lambda class addresses, logs, the
sounds): 10 preparations by 5 preparer versions on the three Mac runners give one digest, and a Linux preparation by an older preparer gives the Mac runners' digest for it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ludovic Marie
rdbtCVS