Make the governed RAG evidence path executable end to end

Separate local parsing from model indexing, bind review decisions to immutable manifests, persist vectors behind active profiles, and expose retrieval, chat, evaluation, and document workflows through the React workbench.

Constraint: Live Bailian authentication currently fails for all three configured capabilities

Rejected: Direct upload-to-embedding flow | bypasses local review and manifest binding

Confidence: high

Scope-risk: broad

Directive: Keep private-data deployment blocked until authentication, RBAC, and separate database roles land

Tested: make verify; fresh and replay Docker document smoke; worker recovery smoke; frozen synthetic evaluation; migration 0003-0004 roundtrip

Not-tested: Successful live Bailian calls, OCR, real multi-user authorization

backend/tests/unit/test_document_review_persistence.py

@@ -0,0 +1,80 @@
+from __future__ import annotations
+
+import uuid
+from pathlib import Path
+
+import pytest
+
+from app.core.config import Settings
+from app.persistence.document_review import (
+    _APPROVE_CHUNKS,
+    _APPROVE_VERSION,
+    _ENQUEUE_EMBED_JOB,
+    _LOCK_REVIEW,
+    _REJECT_VERSION,
+    PostgresDocumentReviewRepository,
+)
+from app.persistence.documents import DocumentActor
+
+MANIFEST = "a" * 64
+ACTOR = DocumentActor(
+    subject="synthetic-demo-maintainer",
+    knowledge_base_id=uuid.UUID("10000000-0000-0000-0000-000000000001"),
+    access_scope_id=uuid.UUID("20000000-0000-0000-0000-000000000002"),
+)
+
+
+def _normalized(statement: str) -> str:
+    return " ".join(statement.lower().split())
+
+
+def test_review_lock_and_mutations_are_scope_manifest_and_revision_bound() -> None:
+    lock = _normalized(_LOCK_REVIEW)
+    approve = _normalized(_APPROVE_VERSION)
+    chunks = _normalized(_APPROVE_CHUNKS)
+    reject = _normalized(_REJECT_VERSION)
+
+    assert "document.knowledge_base_id = %s" in lock
+    assert "document.access_scope_id = %s" in lock
+    assert "for update of document, version" in lock
+    assert "review_revision = %s" in approve
+    assert "outbound_manifest_sha256 = %s" in approve
+    assert "review_state = 'local_parsed_pending_cloud_review'" in approve
+    assert "approval_status = 'local_parsed_pending_cloud_review'" in chunks
+    assert "embedding_model = %s" in chunks
+    assert "embedding_dimension = 1024" in chunks
+    assert "review_revision = %s" in reject
+
+
+def test_embedding_job_payload_parameter_has_an_explicit_postgres_type() -> None:
+    normalized = _normalized(_ENQUEUE_EMBED_JOB)
+
+    assert "jsonb_build_object('document_version_id', %s::text)" in normalized
+
+
+def test_decision_validation_fails_closed_before_database_access(tmp_path: Path) -> None:
+    password = tmp_path / "password"
+    password.write_text("synthetic-password", encoding="utf-8")
+    repository = PostgresDocumentReviewRepository(Settings(postgres_password_file=password))
+
+    with pytest.raises(ValueError, match="approval requires"):
+        repository.apply_decision(
+            actor=ACTOR,
+            document_id=uuid.uuid4(),
+            decision="APPROVE",
+            reason_code="SYNTHETIC_REVIEW_APPROVED",
+            expected_revision=0,
+            outbound_manifest_sha256=None,
+            trace_id=uuid.uuid4(),
+        )
+
+    with pytest.raises(ValueError, match="rejection requires"):
+        repository.apply_decision(
+            actor=ACTOR,
+            document_id=uuid.uuid4(),
+            decision="REJECT",
+            reason_code="SYNTHETIC_REVIEW_APPROVED",
+            expected_revision=0,
+            outbound_manifest_sha256=MANIFEST,
+            trace_id=uuid.uuid4(),
+        )