tools(dnn-i18n): #457 content-type extractor + reimport dry-run (fixture, Option C tooling) - #524

…tractor) (#525)

Closes the "inventory content-types" idle task (ai-01 msg-…141408). Extends
PHASE1-content-audit.md (#461) with a complete localization scope map and feeds
the extractor (PR #524, merged).

5 content-types mapped (Argumentum-app scope only; DNN framework out-of-scope):
- A custom-template hardcoded UI strings (2 keys) — extractor-done
- B @Resources.* keys (8, values DB-only) — extractor-done (keys + DB-only flag)
- C 2 static HTML content pages (fallacies.html ~24K chars + MariagePourTous ~13K,
  standalone FR prose, NOT CSV-driven) — repo-extractable gap, recommend separate
  document-tier translation lane (not the ui.*/res.* string task)
- D 2sxc content items (glossary/FAQ/homepage/App resource VALUES) — DB-only, the bulk
- E DNN page settings — DB-only. Both D+E need portal export (jsboige, gated).

Findings: 26 2sxc apps installed but only Argumentum is custom (25 stock). The live
« Materiel » coquille is NOT in any repo template (negative grep across Argumentum scope)
→ it's DB-only (content-type D), consistent with the "canonical FR lives in SQL" pattern.

No prod mutation, no DB/RDP, no dnn-ui-strings.csv touch. Pure scope-map doc.

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

docs(dnn): #457 document-tier translation workflow (Phase 2-4 design)

The Phase 1 inventory (457-site-content-type-inventory.md) named WHAT to
translate but deferred the HOW for content-type C (the 2 standalone FR
HTML pages, ~37k chars of prose). The string-tier toolchain
(tools/dnn_i18n/, PRs #524/#487) handles ui.*/res.* keys but does not
fit prose. This doc designs the document-tier lane: extraction ->
DatasetUpdater -> re-import.

Scope: DNNPlatform/fallacies/fallacies.html (81 KB, structured charte)
+ MariagePourTous.html (38 KB, prose).

Design:
- Phase 2 CSV schema: segment-oriented (segment_id = file#element-index),
  source_path + selector for deterministic re-import, inner-text fields
  per language. NOT the key->value string-tier dialect.
- Phase 2 DatasetUpdater task shape: document-tier variant vs string-tier
  (prose prompt, char-budget chunking, gpt-5.5, reasoning effort low,
  max_output_tokens >=7000, Enabled=false until GO).
- Phase 1.5 extractor: the net-new gating tool (html.parser stdlib,
  file-specific selectors h2/h3/li vs p/h2/h3, inner-text extract +
  selector store for re-wrap). Nothing translates until this exists.
- Phase 4 re-import: selector-based patch, dry-run verifier (same DoD as
  string-tier round-trip), localized HTML files per language.
- RTL/CJK (ar/fa/zh): reuse PR #359 font stacks at page level
  (dir=rtl/lang), template variant decision for jsboige.
- Sequencing: extractor is critical path.
- 4 open questions for jsboige (inner-text vs outer-HTML, one-template vs
  per-lang, are pages still served, GO to build extractor).

Index README #554 updated to reference the new doc in supporting-docs
table. Pure design/references -- no prod mutation, no extraction
executed, no DatasetUpdater enabled.


@

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>