Fix field state introspection and pack manifold geometry

field/state.py

@@ -4,7 +4,7 @@ FieldState — the complete cognitive field at one moment.
 Invariant: versor_condition(F) < 1e-6 always.
 This is checked at injection and maintained structurally by versor_apply().
 
-FieldState is immutable by design (frozen=True, slots=True).
+FieldState is immutable by design (frozen=True).
 The np.ndarray F is copied and validated at construction — the copy() call
 is the explicit contract boundary. Callers must not retain a mutable
 reference to the array passed in and expect coherence.
@@ -17,7 +17,7 @@ import numpy as np
 _EXPECTED_COMPONENTS = 32
 
 
-@dataclass(frozen=True, slots=True)
+@dataclass(frozen=True)
 class FieldState:
     F: np.ndarray   # shape (32,) float32 — Cl(4,1) multivector on the versor manifold
     node: int = 0   # current node index in the vocabulary manifold
@@ -28,7 +28,6 @@ class FieldState:
         # Enforce copy + dtype + shape at the construction boundary.
         # frozen=True prevents reassignment, but ndarray contents are still
         # mutable via the array object; copy() here is the defence.
-        # slots=True closes __dict__ so no incidental attributes can be added.
         F = np.array(self.F, dtype=np.float32).copy()
         if F.shape != (_EXPECTED_COMPONENTS,):
             raise ValueError(

language_packs/compiler.py

@@ -36,13 +36,30 @@ def _feature_rotor(name: str, salt: str, weight: float) -> np.ndarray:
     return rotor
 
 
+def _domain_features(domain: str) -> list[tuple[str, float]]:
+    """
+    Lift hierarchical semantic domains into a small feature chain.
+
+    A domain like ``logos.illumination.photon`` contributes the trunk
+    (``logos``), then the branch (``logos.illumination``), then the leaf.
+    This reduces accidental hash collisions where unrelated surfaces land
+    close together despite having disjoint semantic structure.
+    """
+    parts = domain.lower().split(".")
+    return [
+        (".".join(parts[: depth + 1]), 0.45 / (depth + 1))
+        for depth in range(len(parts))
+    ]
+
+
 def _entry_to_coordinate(entry: LexicalEntry) -> np.ndarray:
     vec = np.zeros(N_COMPONENTS, dtype=np.float32)
     vec[0] = 1.0
 
     pos = (entry.pos or entry.part_of_speech or "").lower()
     for domain in entry.semantic_domains:
-        vec = geometric_product(vec, _feature_rotor(domain.lower(), "domain", 0.7))
+        for feature, weight in _domain_features(domain):
+            vec = geometric_product(vec, _feature_rotor(feature, "domain", weight))
 
     if pos:
         vec = geometric_product(vec, _feature_rotor(pos, "pos", 0.35))