@energy8platform/golem 0.7.0 → 0.8.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@energy8platform/golem",
3
- "version": "0.7.0",
3
+ "version": "0.8.0",
4
4
  "description": "AI-first 2D rigging for slot games: rig.json format, PixiJS v8 runtime, MCP tools and a browser editor",
5
5
  "license": "MIT",
6
6
  "repository": {
@@ -54,7 +54,7 @@ The node's origin is the centre of the rig's canvas (`meta.width` × `meta.heigh
54
54
  "win": { "type": "golem", "props": { "rig": "assets/rigs/gem/rig.json", "animation": "win", "loop": false } } }
55
55
  ```
56
56
 
57
- As a symbol state the node plays its `animation` once to the end.
57
+ As a symbol state the node plays its `animation` once to the end. The state is fitted to the cell by the rig's parts at rest, not by its canvas, and keeps that size through its own pop: the rig of the static PNG fills the cell exactly as the PNG does, and `rig_create_symbol`'s `margin` (room for the motion, default 0.15) does not shrink it. `RigSymbol` outside e8 fits the same way.
58
58
 
59
59
  ## Live editing and undo
60
60
 
@@ -17,8 +17,11 @@ PCHIP gives continuous velocity, not continuous acceleration. For a deliberate c
17
17
  rest-to-rest reach, use `ease: "minimum_jerk"`: `10u³−15u⁴+6u⁵`. Key x/y with the same normalized
18
18
  phase to follow a straight path; author a curved spatial path when the action needs it. Put
19
19
  minimum-jerk segment boundaries only at real stops, not every sampled waypoint. For longer
20
- custom trajectories, sample a differentiable function into explicit linear keys at a rate
21
- that passes acceleration/jerk measurements; do not re-ease each sample.
20
+ custom trajectories, sample a differentiable function into keys on an `interpolation: "spline"`
21
+ track: a C2 cubic through them (continuous velocity AND acceleration), periodic across a closed
22
+ loop, at rest at the ends of an open one. It may overshoot between keys, so keep PCHIP for holds
23
+ and ground contacts (two equal keys in a row) — the built-in presets do exactly this. Do not
24
+ re-ease each sample.
22
25
 
23
26
  For idle, t=0 is the approved setup pose unless the established brief says otherwise. Author
24
27
  both endpoint values and velocities across every channel. PCHIP wraps endpoint tangents only
@@ -43,7 +43,10 @@ all mathematically possible motion.
43
43
 
44
44
  `rig_check` adds `pass_key_stop`, `velocity_jump`, `acceleration_jump`, `loop_velocity`,
45
45
  `secondary_frozen`, `draw_order_rule` and `attachment_landmark_jump` to existing geometry
46
- checks. Acceleration findings are informational: PCHIP is C1, not C2. `motion_pop` looks for
46
+ checks. Each line names its subject (a track as `target.id.prop`), animation and key time. To accept
47
+ a finding you have reviewed, pass `ignore: [{ code, subject?, animation? }]`: the reply counts what
48
+ it dropped and names any rule that matched nothing, so a typo does not read as "OK".
49
+ Acceleration findings are informational: PCHIP is C1, not C2; a `spline` track is C2. `motion_pop` looks for
47
50
  an isolated displacement spike relative to neighbouring frames; a smooth fast gesture alone
48
51
  is not evidence of a teleport. Source-atlas bboxes are provenance, not document placement:
49
52
  mark `asset.origin.bboxSpace: "source"` so `patch_offset` does not compare unrelated spaces.