Bots Getting Twice The Gravity

The Bot Gravity Bug

When the input queue held more than one command, simulate() ran twice per tick:

entities got 2x gravity
bots ~2-3x gravity  (vy delta ~-17.6/tick vs ~-8/tick)

Also fixed: a 50-unit stop threshold that friction was using to stop bots dead, and a missing yaw rotation on the bot's direction in _sendInput().

Then the bots got *NavGraph A\ pathfinding, 3-ray feelers, strafe-jump button flags, and a jump probability cut from 2% to 0.3-0.4%**.

The Protocol Bug

BinaryProtocol.decode placed movementState on the velocity component instead of the movementState field — so RemotePlayerRenderer never saw a crouch, slide or walljump flag. One property name, and every remote player stood up straight.

Movement state is now sent without delta compression: 1 byte per entity per snapshot.

Remote Models, Detail By Detail

• ▸Crouchslide model anchor lowered 20 units to match the shorter hitbox, because

the Mixamo slide does not lower the skeleton.

• ▸Crouch walk uses directional animations with speed-proportional playback;

backward uses a negative speedRatio to reverse the clip.

• ▸Crouch idle is the frozen first frame of crouch_walk_forward at

speedRatio 0.

• ▸Split Y offsets: -10 for crouch walk and idle, -20 for crouchslide.
• ▸Hitbox wireframes shrink from the head down, keeping the bottom at feet level

— previously from the bottom, which slid the whole box into the floor.

Remote weapons attach to mixamorig:RightHand with scale compensation for MODEL_SCALE, and plasmagun dual barrels got Z offsets of -4 and 0 — Z being left/right in GLB.

Plasmagun recoilScale 1.0 → 0.4. Riotgun pellet impact lights collapsed into one light at the impact centroid.

Known Issues: none stated.