Stage 9 measured the released NPC on both frameworks (6000 map): about 3 ms of
median frame per 100 idle while a player keeps them awake, 0.5 ms once they
sleep with nobody near, and 9 ms per 100 fighting, with each NPC reacting about
every 3 s under Rust's shared AI budget. The warning now says idle twice
(awake and asleep) and uses those figures, the worst of the two rigs rounded up.
The awake-idle overhead over the bare swap is runicnpc-rust#13.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E14m6SuuY6i1vASFeGDBeY
- README: status is stage 9; the installer and egg install it now; API 6; the
swap reports 55/55 npc fields; COMMANDS.md and INSTALL.md are linked; a new
"update surface" table lists every Rust type and member the swap and the
brain depend on, where, and what catches a change (compile, the field list,
or only a running server). The staging drill checks exactly that list.
- tools/RunicNpcTest.cs 0.7.0: rnt.cost measures 100 of RunicNPC's own NPCs
against the same session's empty server, idle (awake) and fighting 20
invulnerable stand-ins, with a second empty baseline at the end and the bar
taken against the lower. `idle` skips the fight, `sentry` makes them stand
still, `asleep` leaves out the keeper.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E14m6SuuY6i1vASFeGDBeY
