02 — Seam Mapping to Real Sources

The two committed SB-2 integration seams currently resolve WHERE false. This doc maps each to the exact real source tables/views that should feed it later, with the exact column wiring. No wiring is performed here — this is the build-prep spec for a future ratified CREATE OR REPLACE.

2.0 Why this is a ratified decision, not pure engineering

A non-empty inventory seam makes the substrate begin to see governed objects — i.e. it crosses inert→active for coverage purposes. Per the deferred-substrate build (doc 03 §3.2), wiring these catalogs is a coverage/coupling decision owned by governance, not an engineering default. This doc therefore produces a spec to be ratified, not a change to apply.

2.1 Seam A — v_governance_object_inventory (governed-object inventory)

Current shape (live pg_get_viewdef):

SELECT NULL::text AS object_type, NULL::text AS object_ref,
       false AS requires_owner, false AS born
WHERE false;

Consumers: v_object_owner_gap (inventory LEFT JOIN effective-owner ⇒ owner IS NULL = gap); coverage detectors iu_object_orphan, inventory_gap.

Real-source wiring (proposed, ratified follow-up):

Output col	Source	Derivation
object_type	collection_registry / birth_registry	collection grain → 'collection'; born-row grain → 'entity' (or the collection name as type)
object_ref	collection_registry.collection_name / birth_registry key	collection name, or collection:entity_code for object grain
requires_owner	collection_registry.governance_role	governance_role IN ('governed','locked') ⇒ true; observed/excluded ⇒ false
born	collection_registry.coverage_status / birth_registry	coverage_status='BIRTH_REQUIRED' and a corresponding birth_registry row exists ⇒ true

Two valid grains (pick at ratification):

• Collection grain (recommended for first dry-run): one inventory row per collection_registry row WHERE governance_role IN ('governed','locked') AND coverage_status='BIRTH_REQUIRED'. ≤ ~38 rows. Bounded, auditable.
• Object grain (deferred): join birth_registry (born=true) filtered to governed collections. 1.08M-row source ⇒ must be cursor-tailed (SB-13), never full-swept in a dry-run.

Drift cross-check: meta_catalog.orphan_count > 0 flags collections drifting out of coverage (Σ=161 today) — an inventory_gap / iu_object_orphan input, not part of the inventory row itself.

2.2 Seam B — v_governance_object_containment (containment tree)

Current shape (live):

SELECT NULL::text AS object_type, NULL::text AS object_ref,
       NULL::text AS parent_type, NULL::text AS parent_ref
WHERE false;

Consumer: recursive v_object_effective_owner — climbs containment through is_inherited_anchor=true ancestors (nearest-ancestor, anti-hiding, 64-depth guard). With an empty seam it yields direct owners only.

Real-source wiring (proposed, ratified follow-up) — union of the live containment edges:

Edge family	Source	parent derivation
Pivot hierarchy	pivot_definitions.parent_code (37 rows)	(pivot, code) → (pivot, parent_code)
IU containment	information_unit container ref + iu_relation (60 edges)	container/parent relation rows where relation type = containment
Topic hierarchy	taxonomy.facet_id → taxonomy_facets (58 nodes / 10 facets)	FAC-08 node → parent node / facet
Object→collection	collection_registry / meta_catalog	entity collection:entity_code → 'collection' parent

Caution: the union must be acyclic for the recursive view (it has a 64-depth cycle guard but a true cycle would saturate it). A containment_cycle detector (doc 08 of the hardening pkg) must pre-screen; the dry-run must assert depth ≤ small bound.

2.3 Axis / topic / relation / assignment coverage sources

Coverage target	Real source(s)	Status
Topic (FAC-08 nodes)	taxonomy (status='active', facet_id) + taxonomy_facets (FAC-08) → owner via v_object_effective_owner	sources LIVE (58/10 rows); detector wireable once containment seam fed
Axis surfaces	taxonomy_facets, entity_labels, pivot_definitions, KG edge types	sources LIVE, but comparison target axis_registry is ABSENT
Axis registration	axis_registry (NOT BUILT)	fake/test-only until axis substrate ratified+built
Assignment (zone/confidence/stale)	axis_assignment (NOT BUILT), coverage_rule (NOT BUILT)	fake/test-only
Label reconcile	entity_labels (795,841) ⇄ axis_assignment	half-live; needs axis_assignment
UI-visible-unapproved	v_ui_exposed_objects helper view (NOT BUILT)	fake/test-only

2.4 Which seams remain fake / test-only

1. All axis-layer coverage (axis_unregistered, axis_owner_gap, axis_vocab_unowned, ui_visible_unapproved, wrong_topic/topic_stale) — depend on axis_registry/axis_assignment/coverage_rule/v_axis_surfaces/v_ui_exposed_objects, none of which exist. The axis substrate (hardening pkg docs 04/05) is design-only.
2. Object-grain inventory — wireable in principle but deliberately deferred (1.08M source; cursor-only).

2.5 Exact future wiring (the only DDL a ratified follow-up needs)

Two CREATE OR REPLACE VIEW statements (Seam A, Seam B) — additive, no migration, instantly reversible to the WHERE false form (the SB-2 rollback script already restores the inert version). Not authored here; specified so the future macro is a fill-in-the-blanks, not a design exercise.

-- Seam A (collection-grain, recommended): CREATE OR REPLACE VIEW v_governance_object_inventory AS
--   SELECT 'collection', collection_name,
--          governance_role IN ('governed','locked'),
--          coverage_status='BIRTH_REQUIRED'
--   FROM collection_registry
--   WHERE governance_role IN ('governed','locked','observed');
-- Seam B: CREATE OR REPLACE VIEW v_governance_object_containment AS
--   <UNION of pivot_definitions / iu_relation / taxonomy / collection edges, acyclic-screened>

(Shown as a comment block — prepared, not applied.)

2.6 Seam mapping verdict

CONCRETE. Both seams map to enumerable, live sources with exact column derivations. The inventory seam is wireable at collection grain (≤~38 rows) the moment it is ratified. The containment seam is wireable from 4 live edge families subject to acyclic screening. Everything axis-related is correctly classified fake/test-only pending the unbuilt axis substrate.