Attach per-string provenance to assembled text #105

New issue

Closed

opened 2026-07-01 00:54:09 +00:00 by manlycucumber · 1 comment

manlycucumber commented

2026-07-01 00:54:09 +00:00

(Migrated from github.com)

Problem

The platform requires per-string provenance for all text. Assembled output must carry the source and licence of every Latin and English string.

Acceptance criteria

Every assembled string carries source and licence metadata
Assembly fails if any string lacks provenance
Provenance is queryable in the output
Tests verify provenance presence across an office

Notes

Reuses the corpus provenance shape from v0.1.0.

## Problem The platform requires per-string provenance for all text. Assembled output must carry the source and licence of every Latin and English string. ## Acceptance criteria - [ ] Every assembled string carries source and licence metadata - [ ] Assembly fails if any string lacks provenance - [ ] Provenance is queryable in the output - [ ] Tests verify provenance presence across an office ## Notes Reuses the corpus provenance shape from v0.1.0.

manlycucumber referenced this issue

2026-07-17 15:04:35 +00:00

[Epic] Bilingual text assembler #103

manlycucumber commented

2026-08-05 19:53:23 +00:00

Owner

Done in fd26918 and 70f3133, with the locator half of the rule closed later in d073227.

ACs: every assembled string carries source + rights ✓ (TextLeaf); assembly fails if any string lacks provenance ✓ — fail-closed at both the build gate (tools/generator/src/provenance.mjs, thrown from build.mjs) and at assembly (CorpusOfficeTextProviderTest::testProperFailsClosedWhenTheReferenceHasNoLatin); provenance queryable in the output ✓ — carried on each leaf in the office block; tests verify provenance across an office ✓ — tests/Office/CorpusOfficeTextProviderTest.php, plus tests/Golden/GoldenProvenanceLockTest.php gating digest moves.

Note d073227 tightened this after the fact: the gate checked that a text leaf cited a public-domain source but not that the citation carried a :locator, though the schema and comments claimed it did. Build-time guard only — corpus and golden digest unmoved.

Done in `fd26918` and `70f3133`, with the locator half of the rule closed later in `d073227`. ACs: every assembled string carries `source` + `rights` ✓ (`TextLeaf`); assembly fails if any string lacks provenance ✓ — fail-closed at both the build gate (`tools/generator/src/provenance.mjs`, thrown from `build.mjs`) and at assembly (`CorpusOfficeTextProviderTest::testProperFailsClosedWhenTheReferenceHasNoLatin`); provenance queryable in the output ✓ — carried on each leaf in the `office` block; tests verify provenance across an office ✓ — `tests/Office/CorpusOfficeTextProviderTest.php`, plus `tests/Golden/GoldenProvenanceLockTest.php` gating digest moves. Note `d073227` tightened this after the fact: the gate checked that a text leaf cited a public-domain source but not that the citation carried a `:locator`, though the schema and comments claimed it did. Build-time guard only — corpus and golden digest unmoved.

manlycucumber closed this issue

2026-08-05 19:53:23 +00:00

manlycucumber referenced this issue

2026-08-05 19:53:31 +00:00

[Epic] Bilingual text assembler #103