feat(ontology): preserve temporal imported primary Voice intervals (ADR 0252)

ARCHITECTURE.md

@@ -86,12 +86,8 @@ flowchart LR
 | `commitment_extraction.py` | Pluggable LLM derivation of a customer commitment (promise + deadline) from a post; `Null` default, `ContextualOrchestrator` real impl |
 | `temporal_expressions.py` | Pure Korean relative-time resolver for Global Ask (ADR 0150) |
 | `ask_time_axis.py` | Event-time vs ingestion-time clock choice for that window (ADR 0202) |
-| `ontology.py` | Loads the governed Turtle source tree (`lineageweave-kg.ttl` plus generated fragments), the formal OWL 2/RDFS/SKOS vocabulary for the Knowledge Graph's node/edge types, source taxonomies, and published O*NET linkages (ADR 0004, ADR 0252, ADR 0255, ADR 0256) |
-| `backend/app/occupation_rating_ingestion.py` | Projects authenticated occupation-rating evidence plus persisted source and represented-occupation catalogs (ADR 0258, ADR 0260, ADR 0261) |
-| `frontend/src/components/OccupationRatingProfile.tsx` | Selects an imported source, filters stored occupation titles without ranking, and reads exact Dashboard evidence while preserving absence, uncertainty, and warning semantics (ADR 0259–0262) |
+| `ontology.py` | Loads `docs/ontology/lineageweave-kg.ttl`, the formal OWL 2/RDFS/SKOS vocabulary for the Knowledge Graph's node/edge types (ADR 0004) |
 | `ontology_neighborhood.py` | Bounded typed ontology/provenance neighborhood (ADR 0184); PostgreSQL stays authoritative, OWL subclass is not an instance edge |
-| `occupational_construct_catalog.py` | Official O*NET 31.0 construct catalog sync (ADR 0250); no ratings or invented IRIs |
-| `backend/app/occupational_construct_search.py` | Authorized catalog-label search over assertion-backed constructs (ADR 0257); hidden Posts never mint a hit |
 | `ontology_source_cursor.py` | Opaque HMAC source-window continuation (ADR 0124); keyset pagination, never OFFSET |
 | `period_report.py` | Fit GRM/GPCM on persisted IRT rows, FIPC-select, EAP-score a period (ADR 0003 slice 3; Bock & Mislevy, 1982) |
 | `fixtures.py` | Synthetic demo dataset -- no real data ships in this repo |
@@ -379,6 +375,11 @@ governed atomic Voice with an explicit truth state and an ABAC-visible evidence
 Post. The server creates the normalized PROV-O derivation and assignment in one
 transaction; clients never submit an internal assertion id, and this route
 cannot replace the imported primary Voice.
+Imported-primary changes are retained in `source_post_voice` as non-overlapping
+half-open effective intervals (ADR 0252). The database closes the current row
+and opens the new observed primary at one statement instant; live reads select
+the open interval, cutoff reads select the containing interval, and an ontology
+continuation without an explicit cutoff uses its frozen snapshot instant.
 The bounded ontology response carries a visible Voice assignment's evidence
 Post id alongside its exact-value row. The exact-value table therefore offers
 separate carrying-Post and derivation-evidence actions; hidden evidence removes
@@ -683,8 +684,7 @@ vocabulary (`node_type`, `edge_type`, `entity_relationship_type`,
 `person_side`, `corporate_entity_level`) actually matches what the
 Ontology/Semantic-Layer claim implies.
 
-`docs/ontology/lineageweave-kg.ttl` and its deterministic governed fragments
-are a real OWL 2 / RDFS / SKOS
+`docs/ontology/lineageweave-kg.ttl` is a real OWL 2 / RDFS / SKOS
 ontology in Turtle syntax: classes for `Post`/`Person`/`CorporateEntity`
 (with `OurSidePerson`/`CounterpartyPerson` subclasses), object
 properties for each `edge_type_code` and `entity_relationship_type`
@@ -698,7 +698,7 @@ specification over it, in the same sense W3C's own stack uses "semantic
 layer" (RDFS/OWL as the governed conceptual layer over raw data), not a
 separate BI-metrics product and not a parallel triple store.
 
-`lineageweave/ontology.py` parses the Turtle source tree once with `rdflib`
+`lineageweave/ontology.py` parses the Turtle file once with `rdflib`
 (pure Python, no Rust toolchain, unlike `fast-mlsirm`) and exposes the
 vocabulary as importable IRI constants, so application code has one
 canonical name per class/property instead of re-typing lookup codes as
@@ -715,19 +715,6 @@ enforcement mechanism: a future PR that adds a new `edge_type` or
 `entity_relationship_type` code without updating the ontology fails
 this test, not just a docstring's word.
 
-### Authorized job architecture snapshots
-
-The public SOC/O*NET vocabulary and an employer's job architecture remain
-different graphs. ADR 0263 adds an organization-scoped PostgreSQL source
-boundary for private job-family/job-series snapshots: immutable source
-metadata owns normalized nodes, source-declared broader/narrower edges, and
-optional explicit bindings to a versioned external occupation scheme. An edge
-table preserves multiple-family membership; the importer rejects cycles and
-never derives a parent or binding from a label or code pattern. The snapshot
-is source evidence only. It does not create a person, post, organizational
-unit, competency, score, weight, or ontology assertion, and runtime rows never
-enter repository artifacts.
-
 ## Phase 6c: post content normalization before any LLM/embedding call
 
 The brief's latest revision calls out, explicitly, that a post body mixing

CHANGELOG.md

@@ -9,6 +9,12 @@ All notable changes to this project are documented here. Format follows
 ### Added
 
 - Normalized Voice-of-X composition persistence (ADR 0256): every imported
+- Imported primary Voice changes now retain non-overlapping effective intervals
+  (ADR 0252), including recurring A → B → A values. Live, knowledge-cutoff,
+  and ontology-snapshot reads select the same period contract; PostgreSQL
+  rejects overlapping primary intervals instead of relying on application
+  ordering.
+- Normalized Voice-of-X composition persistence (ADR 0251): every imported
   primary voice is mirrored into `source_post_voice`; each additional voice
   requires its own PROV-O assertion and truth status. Compound lookup codes,
   keyword inference, confidence thresholds, and invented weights remain out of

backend/app/ontology_neighborhood_ingestion.py

@@ -866,6 +866,10 @@ def neighborhood_to_payload(neighborhood: OntologyNeighborhood) -> dict[str, Any
                 "is_primary": assignment.is_primary,
                 "truth_status_code": assignment.truth_status_code,
                 "recorded_at": assignment.recorded_at.isoformat(),
+                "effective_from": assignment.effective_from.isoformat(),
+                "effective_to": assignment.effective_to.isoformat()
+                if assignment.effective_to
+                else None,
                 "provenance_reference": assignment.provenance_reference,
                 "evidence_post_id": assignment.evidence_post_id,
             }
@@ -890,6 +894,7 @@ async def _load_voice_assignments(
         """
         select voice.post_id, voice.voice_type_code, lookup.lookup_label, voice.is_primary,
                voice.truth_status_code, voice.recorded_at,
+               voice.effective_from, voice.effective_to,
                case when evidence.node_id = any($1::uuid[]) then evidence.node_id end
                    as evidence_post_id
           from source_post_voice voice
@@ -903,10 +908,11 @@ async def _load_voice_assignments(
            and evidence.node_type_code = 'node_post'
          where voice.post_id = any($1::uuid[])
            and (voice.is_primary or evidence.node_id = any($1::uuid[]))
-           and (($2::timestamptz is null and voice.effective_to is null)
-                or ($2::timestamptz is not null
-                    and voice.effective_from <= $2
-                    and (voice.effective_to is null or $2 < voice.effective_to)))
+           and voice.effective_from <= coalesce($2::timestamptz, $3::timestamptz)
+           and (
+               voice.effective_to is null
+               or coalesce($2::timestamptz, $3::timestamptz) < voice.effective_to
+           )
            and voice.recorded_at <= $3::timestamptz
          order by voice.post_id, voice.is_primary desc,
                   lookup.display_order, voice.voice_type_code
@@ -931,6 +937,8 @@ async def _load_voice_assignments(
                 is_primary=row["is_primary"],
                 truth_status_code=row["truth_status_code"],
                 recorded_at=row["recorded_at"],
+                effective_from=row["effective_from"],
+                effective_to=row["effective_to"],
                 provenance_reference=(
                     "Evidence-backed additional voice"
                     if not row["is_primary"]