You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hover a belt and press ALT+B: the mod flood-fills the loop (belts,
undergrounds, splitters; balancers as graphs), prunes feed-on/feed-off
spurs so only the circulating core counts, sets one "read entire belt
(hold)" reader per game segment, clusters the readers at the junctions
where segments meet, and stitches them with a minimum spanning tree of
short player-legal wires (reach from the prototypes; unbridgeable gaps
get GPS-linked power-pole suggestions instead of script-cheated spans).
SHIFT+ALT+B removes exactly what the last trace added; a pyops-sushi
remote interface exposes trace/untrace for tooling and future UI.
Segment rules were derived empirically with isolated-reader probes, not
guessed: segments end at splitters and sideload merges (positional —
the input directly behind a belt continues its line; curves keep their
segment), undergrounds CONTINUE the segment with the buried span
readable, and tier-weaved undergrounds never cross lines because
pairing comes from the engine. Verified against exact item censuses on
a scripted torture loop (5 splitters, bypass arms rejoining by sideload
and by splitter, mixed tiers, a T2 line weaved through a T3 span):
reader counts at the theoretical minimum, one network, all wire spans
player-legal, and wire-vs-physical reconciling up to the documented
limits (splitter internals are circuit-invisible; reads bleed into
adjacent feeder branches).
The measurement (sushi.trace, protocol v8) lands in the app: the sushi
planner shows a one-click "measured N tiles" chip for its loop length.
Trace hotkeys register on the next full game restart (data-stage
prototypes); the control-side registration is guarded so reload_mods
on an older data stage doesn't crash.
0 commit comments