ADR-0011: The corpus is the case-study plant's own SDS, ingested as files

Status: Accepted — supersedes ADR-0003 Date: 2026-09-02

Decisions

D-14 The corpus is the case-study plant's actual SDS documents, obtained as files and curated with provenance

ADR-0003's curation machinery is kept unchanged — per-document provenance, GHS 16-section parsing, human-reviewed Thai translation where needed. Only the sourcing changes: the documents come from the plant's own binder rather than from manufacturers' public websites.

D-15 There is no live integration with a plant chemical-management system in prototype scope

The source's claim of เชื่อมโยงกับฐานข้อมูลสารเคมีของโรงงานโดยตรง ("links directly to the plant's chemical database") is recorded as production scope, explicitly out of scope for the prototype.

Context

R-26: the revised poster states twice that the product uses SDS ของโรงงาน — the plant's SDS — and contrasts this against general apps that reference generic internet information. Panel 5 specifies เอกสาร SDS ฉบับจริง, actual SDS documents. R-22 narrows the effort to one case-study automotive parts plant, which makes "the plant's SDS" a concrete, obtainable set rather than an abstraction.

R-25 is the deciding evidence: the named process chemicals include น้ำยาล้างชิ้นงาน, a formulated parts-cleaning solution. No manufacturer-PDF corpus assembled from public sources would contain the SDS for the specific blend that plant buys. Only the plant's own binder has it.

Rejected options

• Plant SDS plus a real system integration — rejected as a schedule risk that is entirely outside the project's control: it presumes a plant IT system, an access owner, and a stable interface, any one of which can sink the timeline alone. Recorded as production scope instead of dropped.
• Keep ADR-0003's manufacturer PDFs — rejected because it would make the poster's central differentiator untrue. An app answering from public manufacturer sheets is the "general internet information" the source defines itself against.

This ruling may not be re-decided

If a change contradicts this ADR: stop and raise it. Do not implement over it. In particular, do not substitute public manufacturer PDFs for a plant document that is slow to obtain — the substitution is invisible in the UI and silently reverts the product to the rejected option.

Consequences

The project now has a hard external dependency: without the plant's SDS binder there is no corpus. This dependency also gates objective 2 of R-24 (studying worker needs), which needs access to the same site's staff. It should be the first thing secured, ahead of any build work.

Coverage

Upstream	Landed in	Evidence	Note
R-22	D-14	the case study makes "the plant's SDS" a concrete set
R-25	D-14	the parts-cleaning mixture is the deciding evidence
R-26	D-14, D-15	both source claims addressed; integration deferred
R-06	D-14	the provenance gap R-06 named is closed by The Plant's binder	re-landed from superseded ADR-0003
R-16	D-14	the chemical list is now whatever the binder holds	re-landed from superseded ADR-0003