CLM Glossary Build Kit — Incremental Update Support (v1.2)

Tag: S-2026-06-08-glossary-kit-incremental-updates Type: own-writing Author(s): Paul Miles (AI-assisted) Date of source: 2026-06-08 Date ingested: 2026-06-08 Authority weight: high — the controlled record of the v1.2 incremental-update enhancement (captured as §6 of the QA changes-applied report). Raw file: /_raw_sources/S-2026-06-08-glossary-kit-incremental-updates.md

What it claims

Documents the v1.2 enhancement that lets the CLM Business Glossary Build Kit merge additional batches of source files into the master deterministically, prompted by Paul’s question on how a second batch incrementally adds to the master. Core addition: a Business Concept Key (BCK) — a durable per-concept identity, minted once when a concept is first mastered and reused by every later batch that matches it — added as a column in T-03 and as the first Glossary-Mastering field in T-06 (30 added fields; all 19 Taxonomy Template V0.1 columns still preserved). New guardrail GR-14 requires that, on any batch after the first, the existing master (T-06) is in the clustering comparison scope and new terms are matched against it first, never creating a second row for an existing concept. P-02 gained first-run vs incremental-run modes; incremental matches inherit the existing BCK and route to a new T-03 disposition “Update existing master concept (new source / CR)“. P-05/glossary-mastering reuse an existing BCK to update a concept’s row (adding the source, raising a Change Request if the definition differs materially) or mint a new BCK for new concepts. glossary-qa check 9 verifies BCK integrity. Before the first baseline a second batch is just more rows through Stages 1–5; after publication it is a Stage 7 delta rolled into the next baseline. Guidelines §11, the process-diagram deck (incremental callout looping to Step 2) and README were updated; verification confirmed zero formula errors, preserved V0.1 columns, matching column lists and a re-validated docx.

Notable quotes

“On any batch after the first, always run clustering with the existing master in scope and flag it as an incremental run.” (Guidelines §11, the key operating rule).

What’s speculative vs. asserted

Asserted: the design changes and the verification performed. Conditional: incremental matching still depends on the existing master being supplied to P-02 as comparison scope (now mandated by GR-14); the initial match on each incremental run remains an AI inference subject to the same human review as any clustering output — the BCK only makes a confirmed match persistent.

Topics this feeds

CLM Glossary Acceleration Squad — the v1.2 build kit.

Open questions raised

Whether to add a content-derived hash alongside the BCK for fully automated re-matching (deferred — current design keeps the human-confirmed match as the persistence anchor).