dot-iu-cutter v0.5 — Full-Document Trial DESIGN Phase Report

Date: 2026-05-17 · Status: DESIGN ONLY — nothing executed; routes the v0.5 package to GPT review. Predecessor: first controlled production CUT/VERIFY trial CLOSED_PASS (single-IU production write proven, GPT-reviewed closeout). Accepted code commit e93424b5ff7fa5e4b8406131977ce4339cd0856a.

1. What this phase did

Read-only grounding of production (sysid 7611578671664259111, no write) + authored a 6-document design package under knowledge/dev/laws/dieu44-trien-khai/v0.5-full-document-trial-design/:

1. design-master · 2. source-document-and-authority-selection · 3. hien-phap-trial-cut-routing-design · 4. three-existing-cut-documents-merge-and-storage-design · 5. scale-index-and-label-metadata-risk-note · 6. this report.

No production write, no CUT/VERIFY, no dry-run, no code change, no commit, no schema/index/label/vector/alias change. Only read-only SELECT/catalog/docker inspect were issued (no row content/secret exposed).

2. Required analysis — answered (summary; detail in the package)

1. 3 already-cut documents + storage: DIEU-28 (27), DIEU-32 (23), DIEU-35 (36) in public.tac_logical_unit; governance/manifest in cutter_governance.* (currently only the 1 trial IU). All draft/draft_only/canonical-address-v1.
2. Hiến pháp source: absent from corpus; external authoritative source required (OD-2) — 2013 Constitution (11 Chương, 120 Điều) as worked example; not invented/selected.
3. Granularity: root/section/unit × section_type; leaf IU = Khoản/Điểm/đoạn; article-vs-clause-vs-point = OD-3; Chương grammar = OD-4.
4. Next-trial mode: recommend dry-run-first → staged production small-batch; single big-bang not recommended; Hiến pháp not yet (OD-1).
5. Volume estimate: ≈300–500 leaf IUs → ≈5,000–7,500 governance rows at clause/point granularity (≈1,800 at article); +15/IU invariant. Exact = OD-5.
6. Pre-scale index DDL: REQUIRED — concrete unindexed hot paths identified; proposal-only index list; index DDL itself NOT authorized (OD-10).
7. Label/metadata: registry needed before multi-unit labeling; deferred, separately gated (OD-7).
8. Merge/storage: same tac_logical_unit SSOT + cutter_governance, new config-driven doc_code, co-resident with DIEU-28/32/35; no NoSQL/new store; no cross-doc row merge.
9. Manifest strategy: (A) per-IU envelope (recommended, preserves +15 invariant) vs (B) document-level envelope — OD-6.
10. Rollback/forward-compensation (multi-IU): per-IU append-only forward-compensation; no document-wide rollback/delete; resumable via deterministic entry_id; staged checkpoints (OD-11).
11. No-hardcode: no fixed source path, no hardcoded labels, no hardcoded storage destination — all config/derived; only auditable safety constants remain literals.
12. SQL/NoSQL: SQL (tac_logical_unit + cutter_governance) = SSOT; vector/qdrant projection/search only; no NoSQL in write path.
13. Open decisions: OD-1 … OD-12 consolidated in design-master §13.

3. Data-quality finding (flagged, not actioned)

DIEU-32 and DIEU-35 have blank tier (59/86 corpus rows); only DIEU-28 is fully tiered. Recorded as OD-9; no UPDATE proposed (separately-authorized data-quality task).

4. Scale blockers (must clear before any full-document execution)

Index-only DDL · label/metadata registry design · multi-level canonical-address grammar (Chương) · existing-corpus tier normalisation · authoritative Hiến pháp source+ingestion · dry-run-first-at-volume · vector/NoSQL stays projection/search only. None cleared here; all carried forward.

5. Git / hardcode / boundaries

• Git: branch main · HEAD e93424b5ff7fa5e4b8406131977ce4339cd0856a · git status --short -- iu-cutter = clean (0 lines) — no code change, no commit (none expected/authorized).
• No fixed IP/DSN/password/container/vector-collection introduced; no runtime label/key hardcoding; no schema change. SQL / deployed cutter_governance remains SSOT.
• Boundaries honoured: design only · no production writes · no CUT/VERIFY · no deploy/restart · no schema migration · no index DDL · no label registry creation · no vector/NoSQL integration · no alias writes · no code change · no git commit · no self-advance.

6. Next

GPT review of the v0.5 design package + rulings on OD-1 … OD-12. No execution, dry-run, code change, or further phase until a separate GPT review and (where applicable) sovereign authorization. Self-advance PROHIBITED.