FMA skeletal spine — bones as the clamped convergence anchor (Located-primary GUID)

[AdaWorldAPI]

What this is

A new ogar-fma-skeleton crate: the hand-curated FMA skeleton as the
clamped convergence anchor for the splat-native ultrasound arc
(docs/SPLAT-NATIVE-CUSTOMER.md §6 litmus). Operator-directed across the
session; the address model converged through several corrections recorded
below.

The constraint (one, not two): bones must be hardcoded and stable
because they are the cross-modal registration frame that ViT, X-ray, and
ultrasound × Doppler all project onto. Bones are the only rigid bodies in the
human, so the skeleton is not data being fit — it is the boundary
condition of the fit. Clamping the splat-fit to a hand-curated skeleton
makes the convergence well-posed and deterministic (the Class IIa SaMD
requirement).

The address — the [256:256] [container:member] GUID

Each bone resolves to a 16-byte Guid: a uniform stack of [256:256] tiers,
each the same scale-free relation [container : member] (Galaxy:planet,
country:city, bodypart:bone, cm²:mm², residue:atom). Bones are
Located-primary (they ARE the position anchor):

classid  = 0x0A03                  Anatomy : bone
HEEL/HIP = rest-pose 3-axis CRS    HEEL coronal x:y · HIP depth z   (ArcGIS / Cesium-addressable)
LEAF     = familyNode : identity   [bodypart : bone]

• HEEL/HIP make the body a first-class GIS layer (Morton/quadkey), so
  off-the-shelf spatial ops work and ViT / ultrasound × Doppler project onto
  the same coordinate. Two HHTL modes are supported (HhtlMode):
  Located (Cesium, position-preserving — the bone case) and Cascade
  (ontology, the self-speaking path — the soft-tissue case).
• LEAF family node IS the group: identity == 0 is the family node,
  identity ≥ 1 are its bones ("identity attached to family node").
• Each byte is a directly-addressable 256-codebook and a 4×4 Morton tile
  carrying the perturbation pyramid's (exponent, location) (OGAR owns those;
  ndarray owns phase, magnitude).

Consumes D-MORTON (DISCOVERY-MAP §2.1); recorded as D-FMA-SKELETON.

Codebook

New ConceptDomain::Anatomy (0x0A) in ogar-vocab: anatomical_structure
/ skeleton / bone / joint (0x0A01..04), in lockstep across CODEBOOK +
class_ids consts + ALL + builders. Anatomy is public reference, not
PHI — kept out of Health 0x09 so it is not dragged into medcare's
fail-closed Health RBAC set (preserves its Health == 7 invariant). The ~206
bones are cascade-path nodes under the bone concept, not codebook slots.

Relations

No 12+4 edge block — that is the pre-family-node taxonomy and was removed.
With family nodes, relations ARE the addressing: local = shared family
prefix (same_family), cross = a reference to another node's Guid.

Atlas coverage & honesty

~80 nodes: full vertebral column C1–coccyx, skull, sternum, 12 rib pairs,
major appendicular bones per side — structured to extend to the full ~206
without schema change. Rest poses are a proportional T-pose v0 (to be
refined against a real FMA-aligned mesh); FMA ids are cross-referenced only
where confidently known (the address identity is a stable atlas id, never a
fabricated FMA id). HIP is coronal-tile + depth-slice (not yet a single 3D
octree — its odd bits are reserved). The address structure is CODED;
the splat-fit convergence claim is CONJECTURE, gated on the splat-native
arc.

Tests / gates

cargo test -p ogar-fma-skeleton — 24 + 2 doctests green: Morton math,
classid tier, HEEL laterality, HIP depth, unique node keys + leaves,
family-node grouping with attached identity, parent-link partonomy, the
clamped-anchor invariant, unit-quaternion frames, classid stability.
ogar-vocab 94 tests green (drift gates updated for the anatomy mints).
The new crate is cargo clippy-clean; the 12 pre-existing ogar-vocab
clippy warnings are untouched (zero added).

Docs

docs/FMA-SKELETON-CONVERGENCE-ANCHOR.md (the unity, the GUID tier model,
located/cascade, the projection operators) + D-FMA-SKELETON in the
DISCOVERY-MAP.

🤖 Generated with Claude Code

---

Generated by Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.