From cd1ba8b62077b4ebbd1f3054b7c63f668401f05d Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 2 Jul 2026 11:42:49 -0400
Subject: [PATCH 01/33] docs(checks): propose highlight-in-verse options
 (Relates to #277, #278)

---
 .../highlight-in-verse-suggestion.md          | 300 ++++++++++++++++++
 1 file changed, 300 insertions(+)
 create mode 100644 docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md

diff --git a/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md b/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md
new file mode 100644
index 00000000..cea6f941
--- /dev/null
+++ b/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md
@@ -0,0 +1,300 @@
+# Highlight Repeated Words in Verse Text — Design Options & Recommendation
+
+**Status:** Proposal / decision-requested. Docs-only. No implementation yet.
+**Type:** Follow-on to the Repeated Word Check UI (cards
+[fluent-web#277](https://github.com/eten-tech-foundation/fluent-web/issues/277) /
+[fluent-web#278](https://github.com/eten-tech-foundation/fluent-web/issues/278)), which
+shipped the **Checks tab + panel**
+([PR #320](https://github.com/eten-tech-foundation/fluent-web/pull/320),
+[fluent-api #203](https://github.com/eten-tech-foundation/fluent-api/pull/203)).
+**Relates to #277, #278.** The tab/panel design this layers on is
+[`checks-ui-integration-suggestion.md`](checks-ui-integration-suggestion.md) (decisions
+**W1–W12**).
+
+> **Why this is a proposal PR and not just an implementation PR.** The highlight-in-verse work
+> was agreed to in the #305 review, and it _looked_ like a small presentational add-on. On
+> close inspection of the actual drafting editor it is **not** small: it forces a real decision
+> about how editable verse text is rendered. That decision is bigger than one feature and will
+> outlive it, so the team should weigh in **before** implementation rather than have an
+> architecture picked unilaterally on an experimental branch.
+
+---
+
+## 0. The original request (what Joel/Ulf asked for)
+
+During the Repeated Word Check proposal review (PR #305), **joelthe1 (Joel) and Ulf** asked
+that, in addition to _listing_ repeated-word findings in the Checks panel, the repeated words
+be **visually marked in the verse text itself** — an underline or color on the offending words
+where they appear, so the translator sees the problem in place, not only in a side list.
+
+It was agreed to, and **deliberately carved out as a separate follow-on PR** so the tab/panel
+(#277/#278) could land first without being blocked on verse-rendering changes. This document is
+that follow-on, brought forward for a design decision.
+
+The concrete visual target discussed on the product side of this thread: **the only visible
+change should be that the flagged (active) repeated words turn red** — no new chrome, no layout
+shift, the editing experience otherwise identical.
+
+---
+
+## 1. TL;DR / recommendation
+
+- Coloring words _inside the box the translator is typing into_ is the hard part. Today that
+  box is a **plain `<textarea>`**, and a `<textarea>` **physically cannot** color individual
+  words — its contents are one uniform color by definition (see §3).
+- Every way to get colored words into an editable field means **changing how the editable verse
+  is rendered** — from a trivial CSS overlay hack up to replacing the editor with a real
+  text-editing engine. The options differ enormously in risk (§5).
+- Fluent is **translation software for minority languages**, so unusual scripts, combining marks
+  / complex shaping, and eventually right-to-left (RTL) text must be assumed. That requirement
+  is what turns the cheap options from "clever hack" into "fragile in exactly the cases the
+  users live in" (§4).
+- **Recommendation:** treat this as an editor-foundation decision. The recommended path is
+  **Option D — adopt a real text-editor engine (CodeMirror 6 preferred, ProseMirror/TipTap
+  alternative), behind a feature flag** — because it is the only option that is _correct_ for
+  arbitrary scripts and RTL and provides a decoration layer reusable by future checks (spell
+  check, Wildebeest). Because that is a significant change, the cheaper options and the **"don't
+  color anything yet"** option (Option A) are documented fully, and the team is asked to choose
+  (§7).
+- **Nothing in this PR ships behavior.** It is a design record + option analysis only.
+
+---
+
+## 2. Background — what already exists (and works)
+
+The Repeated Word Check pipeline is already in place and is **not** what this proposal changes:
+
+- [`useRepeatedWordsCheck.ts`](../../../src/features/checks/hooks/useRepeatedWordsCheck.ts) —
+  fires the check on each auto-save, keyed on `(chapterAssignmentId, saveCounter)`.
+- [`useResolvedFindings.ts`](../../../src/features/checks/hooks/useResolvedFindings.ts) — the
+  pure three-layer suppression cascade that decides which findings are **active** vs.
+  **inactive** (ignored / "Default Ignore"). It also assigns each finding a stable **ordinal**
+  and an **occurrence key** `"{snt_id}|{repeated_word}|{ordinal}"`.
+- [`ChecksPanel.tsx`](../../../src/features/checks/components/ChecksPanel.tsx) — the side list.
+
+Each finding carries `snt_id` (verse), `repeated_word` (NFC + lowercased pair), `surf` (original
+surface text), `start_position` (character offset **as Greek Room counted it**), and
+`legitimate`. **The highlight feature is a pure consumer of the existing `resolved.active`
+list** — it adds a _rendering/decoration layer_, no new API, no new data.
+
+So the data problem is solved. The **rendering problem** is the entire subject of this doc.
+
+---
+
+## 3. The core constraint — words inside a `<textarea>` cannot be colored
+
+The drafting page renders the target (the text the translator is writing) as a native
+`<textarea>`
+([`DraftingPage.tsx` target column](../../../src/features/bible/components/DraftingPage.tsx)).
+A `<textarea>`:
+
+- renders its value as **one uniform run of text** — there is no way to make the third word red
+  and the rest black. HTML simply does not allow child markup inside a `<textarea>`.
+- is also doing a lot of quiet work that must not be lost: **native caret**, text selection,
+  **IME / dead-key composition** (essential for many target languages), spellcheck,
+  autocapitalize/autocorrect, auto-resize on input
+  ([`useDrafting.ts`](../../../src/features/bible/hooks/useDrafting.ts)), focus/caret restoration
+  when switching verses, and Enter-to-advance.
+
+Therefore **any** colored-word solution requires rendering the editable verse _differently_. The
+question is only **how far** to go, and each step up costs risk.
+
+> Note: the **source** column (the reference Bible) is already a styled `<p>`, so highlighting
+> read-only source text would be easy — but Greek Room checks the **target** (the translator's
+> own draft), so that's not where the findings live. The findings are in the editable box.
+
+---
+
+## 4. Why "translation software" makes this harder than a normal highlighter
+
+Most "highlight words in a textarea" tutorials quietly assume English: fixed-width-ish Latin
+text, one code unit per visible character, left-to-right. Fluent cannot assume any of that.
+
+1. **`start_position` is a character offset from Greek Room, not a DOM offset.** Coloring the
+   right glyphs requires mapping that offset onto the rendered text. If the app's idea of
+   "character N" differs from Greek Room's (JS strings count **UTF-16 code units**, so
+   emoji/rare CJK/some historic scripts are 2 units; combining accents and many Indic/complex
+   scripts render **one visual cluster from several code points**), the highlight lands on the
+   wrong glyphs. The cascade already NFC-normalizes, but position math is a separate hazard.
+2. **Combining marks & complex shaping.** Devanagari, Arabic, Khmer, etc. reshape glyphs in
+   context; a naive "wrap characters `start..end` in a span" can split a grapheme cluster and
+   visibly break the script. (Greek Room ships a Devanagari cost-rules file — these scripts are
+   real target languages here, not hypotheticals.)
+3. **RTL is coming.** The assumed product intent is to eventually support right-to-left
+   languages. Any overlay/mirror technique that hand-positions colored text has to get **bidi**
+   right (mixed LTR/RTL runs, the Unicode bidi algorithm). Hand-rolled positioning is where this
+   breaks first.
+4. **Variable fonts / line-wrapping.** An overlay must match the editor's font metrics, padding,
+   line-height, and wrapping **pixel-for-pixel**, per language, or the colors drift off the
+   glyphs. Minority-language fonts are exactly where metrics surprise you.
+
+**Conclusion:** the cheap techniques are cheap _because_ they assume a benign script. For this
+user base they are fragile in precisely the languages the product exists to serve. This is the
+crux of why the decision is being surfaced before building.
+
+---
+
+## 5. The options (with complexity and danger of each)
+
+Ordered from "do nothing" to "do it properly." Each lists what it is, the real cost, and the
+danger — stated honestly.
+
+### Option A — Do **not** color words yet (keep the panel-only experience)
+
+- **What:** ship nothing here. The Checks tab + panel + notification dot already tell the
+  translator which verses/words repeat. In-text coloring is formally recorded as deferred until
+  the editor foundation is decided.
+- **Complexity:** none.
+- **Danger:** none technically. The only cost is that Joel/Ulf's specific ask ("see it in the
+  text") stays unmet for now. **This is a legitimate choice**, not a cop-out: it avoids
+  committing to an editor architecture under time pressure, and the panel already delivers the
+  core check value.
+- **When this is right:** if the team is not ready to change the drafting editor and would
+  rather not carry a hack.
+
+### Option B — CSS overlay / "backdrop" (transparent `<textarea>` + mirrored colored div)
+
+- **What:** keep the real `<textarea>` for input but make its **text color transparent** (caret
+  still visible); behind it render a `<div>` containing the _same_ text with the active repeated
+  words wrapped in red `<span>`s. Both layers are driven by the **same React state in the same
+  render**, so there is _no_ temporal lag between them (addressing the "two sets of text out of
+  sync while typing" worry — they update in one commit).
+- **Complexity:** moderate. The hard part is not lag; it's **pixel-perfect alignment**: the
+  mirror `<div>` must replicate the textarea's font, size, padding, border, line-height,
+  letter-spacing, wrapping, and **auto-resize** behavior exactly, or the red drifts off the
+  words. Plus the offset→span mapping from §4.
+- **Danger (this is where "translation software" bites):**
+  - Alignment is brittle across **fonts/scripts** — the exact scenario (unusual scripts) where
+    correctness matters most is where mirrors most easily misalign.
+  - **RTL and bidi** overlays are notoriously hard to keep aligned.
+  - **Grapheme-cluster splitting** (§4.2) can visibly corrupt complex scripts even though the
+    underlying text is fine.
+  - Long-term maintenance tax: every editor style change must be mirrored in two places.
+- **Verdict:** cheapest way to get red words _today_, but it degrades in exactly the target
+  scripts. Acceptable only as a knowingly-temporary stopgap with a plan to replace it.
+
+### Option C — Raw `contenteditable` (one text layer, red spans on matches)
+
+- **What:** replace the `<textarea>` with a `contenteditable` element; render the text with red
+  spans directly (one layer, no mirror, no alignment math).
+- **Complexity:** high. `contenteditable` is famously painful: the DOM must be serialized back
+  to plain text on every edit, and re-rendering spans **mid-edit can move the caret** unless
+  selection is carefully saved/restored — which is the "messy while typing" failure mode in a
+  new disguise.
+- **Danger:** caret jumps; **IME/composition** breakage (dead keys, complex input) — a serious
+  regression for many target languages; inconsistent spellcheck; browser-specific quirks; paste
+  sanitization. This amounts to hand-building a text editor, badly.
+- **Verdict:** **not recommended.** More effort than Option B for _more_ risk. Listed for
+  completeness because it's the "obvious" idea that turns out worse than both neighbors.
+
+### Option D — Adopt a real text-editor engine, decorate the ranges (RECOMMENDED)
+
+- **What:** replace the target `<textarea>` with a purpose-built editor engine and apply the red
+  as **decorations** (presentational overlays the engine positions itself — they do **not**
+  mutate the document, so they can't corrupt the saved text). Candidates:
+  - **CodeMirror 6** _(preferred)_ — designed for exactly "editable text + decorations,"
+    excellent for large/plain text, first-class **RTL/bidi**, robust selection/IME handling,
+    tree-shakeable modular packages, a clean `Decoration.mark` API for coloring ranges. Lighter
+    and more line/character-oriented than a rich-document editor — a good fit since verses are
+    plain text, not rich documents.
+  - **ProseMirror / TipTap** — rich-document editor with a decorations API; heavier and
+    document-model-oriented (overkill for plain verse text, but viable). **Note:** the existing
+    `TipTapRenderer` in `features/resources` is a _hand-rolled JSON renderer_, **not** the real
+    TipTap library — adopting TipTap would still be a brand-new dependency.
+- **Complexity:** significant up front. The drafting input's behavior (auto-resize, focus/caret
+  restore across verses, Enter-to-advance, controlled value ↔ engine state, spellcheck) must be
+  ported onto the engine, a dependency added, and the whole drafting flow re-tested.
+- **Danger:** the risk is **scope/effort and a temporary editing-behavior regression during the
+  port**, not correctness of the coloring itself — a mature engine handles offsets, bidi, and
+  grapheme clusters far better than a hand-rolled solution. Mitigated by shipping **behind a
+  feature flag** (see §6) so the current `<textarea>` stays the default until the new editor is
+  proven.
+- **Verdict:** the **correct** long-term foundation. It's the only option that is right for
+  arbitrary scripts + RTL, and its decoration layer is **reusable** for every future check
+  (spell check, Wildebeest) — those will want in-text marks too. A full, correct solution is
+  likely required eventually regardless, so investing here avoids building Option B twice.
+
+### Option comparison
+
+| Option                  | Effort            | In-text red? | Correct for odd scripts / RTL      | Editing-behavior risk         | Reusable for future checks |
+| ----------------------- | ----------------- | ------------ | ---------------------------------- | ----------------------------- | -------------------------- |
+| A — do nothing          | none              | no           | n/a                                | none                          | n/a                        |
+| B — CSS overlay         | moderate          | yes          | **poor** (alignment/bidi/clusters) | low (native textarea kept)    | partly                     |
+| C — raw contenteditable | high              | yes          | poor–medium                        | **high** (caret/IME)          | partly                     |
+| D — real editor engine  | **high up front** | yes          | **good**                           | medium during port, low after | **yes**                    |
+
+---
+
+## 6. Cross-cutting concerns (apply to whichever option is chosen)
+
+- **HL1 — Only active findings are colored.** Reuse `resolved.active` from
+  [`useResolvedFindings`](../../../src/features/checks/hooks/useResolvedFindings.ts). Ignored /
+  "Ignore Everywhere" / "Default Ignore" (`legitimate`) words render **normal** — red means
+  "still flagged," mirroring the notification dot. This keeps the highlight and the panel in
+  lockstep by construction (single source of truth).
+- **HL2 — Refresh on the same trigger as the panel.** Decorations recompute when
+  `resolved.active` changes, which is already driven by `saveCounter` + the suppression cascade.
+  No new trigger.
+- **HL3 — Offset mapping is a shared hazard.** Whichever option, the mapping from Greek Room's
+  `start_position` (+ `repeated_word` length / `surf`) onto rendered ranges must be
+  grapheme-aware and NFC-consistent with the cascade (§4). Spec this explicitly during
+  implementation and unit-test it against combining-mark and surrogate-pair fixtures.
+- **HL4 — Ship behind a feature flag.** This is an editor-touching change; it should default
+  **off** until proven. There is an in-flight **feature-flags** proposal
+  ([`fluent-api/docs/proposals/feature-flags/`](../../../../fluent-api/docs/proposals/feature-flags/feature-flags-suggestion.md));
+  if that lands, use it. If not yet available, a local build-time/config flag is acceptable for
+  the experiment. Either way the current `<textarea>` stays the default path.
+- **HL5 — Read-only `/view` mode.** In read-only stages the verse is a `<p>`, not a textarea —
+  coloring there (if wanted) is trivial and low-risk and can be delivered independently of the
+  editable-field decision. Confirm whether product wants highlighting in `/view` too.
+
+---
+
+## 7. Decision requested
+
+This is **not** a request to approve an implementation; it's a request for the team to pick the
+direction so the eventual implementation PR isn't a surprise.
+
+**Q1 (primary).** Which option should be pursued?
+
+- **A** — defer in-text coloring (panel only) for now.
+- **B** — CSS overlay stopgap, knowingly accepting the odd-script/RTL fragility.
+- **C** — raw contenteditable _(recommended against)_.
+- **D** — adopt a real editor engine (the recommendation), **CodeMirror 6** preferred.
+
+**Q2.** If **D**, is **CodeMirror 6** acceptable as the engine, or does anyone prefer
+**ProseMirror/TipTap**? (CM6 is favored here: verses are plain text, and CM6's bidi/decoration
+story is strong.)
+
+**Q3.** Feature-flag mechanism: gate behind the pending feature-flags work, or a temporary local
+flag until that lands? (HL4)
+
+**Q4.** Scope of coloring: active-only (HL1) confirmed? Any desire to also mark ignored words
+differently (dimmed)? Highlight in read-only `/view` too? (HL5)
+
+**Q5.** Sequencing/appetite: is this next now, or parked behind other priorities? (Option A is
+the honest "park it" answer if appetite is low.)
+
+---
+
+## 8. Decisions (to be filled in from review — this is the record)
+
+> Reviewer-confirmed decisions get an ID (**HV1…**) and a `(name, date, link)` note here, per
+> the team's docs convention, and are then referenced from code comments at the edit sites at
+> implementation time.
+
+- **HV1 — Chosen option:** _pending review._
+- **HV2 — Engine (if Option D):** _pending review._
+- **HV3 — Flag mechanism:** _pending review._
+- **HV4 — Coloring scope (active-only / ignored treatment / `/view`):** _pending review._
+
+---
+
+## 9. Explicitly out of scope
+
+- One-click "drop duplicate" quick-fix (already out of scope in the panel v1).
+- Feeding suppressions back to Greek Room.
+- Any change to fluent-ai or the proxy contract — the highlight is a pure client-side consumer
+  of findings already in hand.
+- Highlighting for _other_ checks (spell check, Wildebeest) — but Option D is favored partly
+  because its decoration layer would serve them later.

From b3d0cc08a2cd533b1ddafc374b1311282455a85c Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 2 Jul 2026 13:08:40 -0400
Subject: [PATCH 02/33] =?UTF-8?q?docs(checks):=20clarify=20/view=20highlig?=
 =?UTF-8?q?ht=20still=20shares=20HL3/=C2=A74=20range-mapping?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../highlight-in-verse-suggestion.md                   | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md b/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md
index cea6f941..560e4513 100644
--- a/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md
+++ b/docs/proposals/repeated-word-check/highlight-in-verse-suggestion.md
@@ -244,9 +244,13 @@ danger — stated honestly.
   ([`fluent-api/docs/proposals/feature-flags/`](../../../../fluent-api/docs/proposals/feature-flags/feature-flags-suggestion.md));
   if that lands, use it. If not yet available, a local build-time/config flag is acceptable for
   the experiment. Either way the current `<textarea>` stays the default path.
-- **HL5 — Read-only `/view` mode.** In read-only stages the verse is a `<p>`, not a textarea —
-  coloring there (if wanted) is trivial and low-risk and can be delivered independently of the
-  editable-field decision. Confirm whether product wants highlighting in `/view` too.
+- **HL5 — Read-only `/view` mode.** In read-only stages the verse is a `<p>`, not a textarea, so
+  `/view` highlighting is **simpler on the editing-behavior axis** (no caret / IME / auto-resize /
+  editor-engine concerns) and is **independent of the editable-field Option A–D decision**. It is
+  **not**, however, independent of the mapping work: it still relies on the **same
+  `start_position`→rendered-range mapping and grapheme/NFC handling described in HL3 and §4**, so
+  it can't be treated as a zero-cost shortcut past the hard part. Confirm whether product wants
+  highlighting in `/view` too.
 
 ---
 

From 11189702c44f02c6dd1b8fd573163588018ede82 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 11 Jun 2026 16:20:03 -0400
Subject: [PATCH 03/33] First pass at proposal.

---
 .../checks-ui-integration-suggestion.md       | 188 +++++++-----------
 1 file changed, 74 insertions(+), 114 deletions(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index 43756f3a..85b2462e 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -1,7 +1,6 @@
 # Repeated Word Check UI — Proposal (Checks Tab, Checks Panel, and Suppression Persistence)
 
-**Status:** Revised after first review round (PR [#305](https://github.com/eten-tech-foundation/fluent-web/pull/305)). Incorporates reviewer feedback — see the change log in §0.
-**Reviewer shortcut:** a condensed, stands-on-its-own review summary lives in [`checks-ui-integration-summary.md`](checks-ui-integration-summary.md).
+**Status:** Draft for review.
 **Scope:** Implements the user-facing half of the Repeated Word Check feature: the **Checks tab** (#277) and **Checks view panel** (#278) on fluent-web, plus the small fluent-api additions both cards imply (suppression persistence). This is a single proposal covering both repos so reviewers see the whole design in one place; **implementation will ship as two PRs** (one per repo), and either may land first (§9.4).
 
 **Related cards:**
@@ -20,21 +19,6 @@ This document's own decisions are numbered **W1–W12** (W = web) to avoid colli
 
 ---
 
-## 0. Change log (first review round)
-
-Revisions made in response to PR [#305](https://github.com/eten-tech-foundation/fluent-web/pull/305) review (reviewers: kaseywright, chadw-eten; product input via joelthe1/Ulf). The cards themselves were edited mid-review; this revision tracks the **current** card text and mocks (re-pulled 2026-06-16).
-
-| Change | Driver | Where |
-| --- | --- | --- |
-| Settings endpoint moves from `GET/PUT /users/settings` to a new **`self` domain**: `GET/PUT /self/settings`, guarded by `authenticateUser` alone. | kaseywright (CHANGES_REQUESTED): clean semantic separation, no redundant `{id}` param, no over-exposure risk, simpler middleware (no `requireSelf`/`requireUserAccess`/`UserPolicy`). | W7, §7.2, §8 |
-| Button labels become **"Ignore Here" / "Ignore Everywhere"**. | joelthe1/Ulf (card #278 comment); chadw updated the #278 mocks to match. | W6, §5, §6.5 |
-| **"Ignore Everywhere" now shows a confirmation dialog**, while remaining reversible via `[Undo ▾]`. | card #278 ("Ignore Always should have a standard confirm dialog"). Kept the three-layer cascade & undo (no explicit directive against it; ambiguity of "undo" with surface/suppress entries at multiple layers is better resolved by the cascade — see §6.4/§6.5); added the dialog. | W6, §5.3, §6.5 |
-| **"Show Ignored" toggle**: now default **OFF** and **not persisted** (resets per session). Ignored items show the ignore-type label (e.g. _"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) + an **"Undo Ignore"** control. | card #278 was revised to specify exactly this; we drop our earlier default-ON/persisted proposal and the `showResolvedChecks` editor-state key. | W5, W6, W11, §5.3, §6.4, §6.6, §7.1 |
-| §5.2 mock/text inconsistencies **S1 (zero-state dot)** and **S2 (language dropdown)** are **resolved**: chadw fixed the zero-state mock and removed the language dropdown from the #278 Checks mocks. | chadw-eten ("Fixed." / "good idea… I have adjusted the mocks."). | §5.2, §12 |
-| New requirement adopted: **highlight repeated words in the verse text** (underline/color). | joelthe1/Ulf (card #278 comment). We will implement it; because it touches the verse-rendering pane (outside the left panel this proposal builds) and warrants its own design pass, it ships as a **separate, follow-on PR** so the tab/panel work can land independently. §5.3 item 4 records the intent; the visual/interaction details are specced just before that PR. | §5.3, §11 |
-
----
-
 ## 1. Background
 
 fluent-api now exposes Greek Room's _Repeated Words_ check at `POST /ai/tools/greek-room/repeated-words` (card #172). The endpoint is a thin authenticated proxy to fluent-ai: it accepts a chapter's verses and returns a `ToolJobResponse[RepeatedWordsResult]` envelope whose `findings[]` identify consecutive repeated words (`{snt_id, repeated_word, surf, start_position, legitimate, severity}`).
@@ -42,22 +26,20 @@ fluent-api now exposes Greek Room's _Repeated Words_ check at `POST /ai/tools/gr
 Cards #277 and #278 define how translators consume those findings in the drafting view:
 
 - **#277:** a **Checks tab** is added to the drafting page's left panel alongside the existing Resources tab, with a **notification dot** on the tab header whenever the current chapter has one or more active flags. The dot is absent at zero flags and clears silently on the next auto-save once all flagged issues are resolved.
-- **#278:** the **Checks panel** lists all repeated-word flags for the current chapter, grouped by verse, each with enough context to locate it, refreshed on every auto-save, with two actions per flagged occurrence: **"Ignore Here"** (user-level, this occurrence) and **"Ignore Everywhere"** (user-level, this word pair, across all the user's projects) — both "stored in Fluent's database." "Ignore Everywhere" carries a standard confirm dialog. A **"Show Ignored"** toggle (default off, not persisted) reveals previously ignored occurrences — dimmed, labeled by ignore type, each with an **"Undo Ignore"** action. A "No issues found" zero state is shown when the chapter is clean.
-
-> The #278 card text and mocks were revised during the first review round (labels, the Show Ignored toggle, the confirm dialog). The wording above reflects the **current** card; the original used "Ignore This Time"/"Ignore Always". See §0.
+- **#278:** the **Checks panel** lists all repeated-word flags for the current chapter, grouped by verse, each with enough context to locate it, refreshed on every auto-save, with two actions per flagged occurrence: **"Ignore This Time"** (user-level, this occurrence) and **"Ignore Always"** (user-level, this word pair, across all the user's projects) — both "stored in Fluent's database." A "No issues found" zero state is shown when the chapter is clean.
 
-Both cards carry mockups (drafting page, Judges 4, Gujarati IRV); §5 transcribes them.
+Both cards carry mockups (drafting page, Judges 4, Gujarati IRV); §5 transcribes them and flags two mock/text inconsistencies for product sign-off.
 
 ### 1.1 What the cards imply beyond fluent-web
 
-The ignore actions require server-side persistence that does not exist yet. #172's text ("the backend dependency for all Repeated Word Check UI work") is broad enough to cover suppression storage — the fluent-api proposal's **D1** deferred caching of _tool runs/findings_, which is a different concern from user suppression preferences (a requirement that did not exist until #277/#278 were authored). Accordingly, suppression persistence ships as an **extension of #172's scope** (decision **W1**), designed here and implemented in the fluent-api PR of this pair, which also amends the approved fluent-api proposal pair (the suggestion + operations docs) to record the scope extension. No new product card is needed.
+The ignore actions require server-side persistence that does not exist yet. #172's text ("the backend dependency for all Repeated Word Check UI work") is broad enough to cover suppression storage — the fluent-api proposal's **D1** deferred caching of _tool runs/findings_, which is a different concern from user suppression preferences (a requirement that did not exist until #277/#278 were authored). Accordingly, suppression persistence ships as an **extension of #172's scope** (decision **W1**), designed here and implemented in the fluent-api PR of this pair. No new product card is needed.
 
 ### 1.2 Repos touched
 
 | Repo | Touched | What changes |
 | --- | --- | --- |
 | **fluent-web** | yes (bulk) | Left-panel tab container, Checks panel feature, check-trigger hook, suppression cascade, editor-state keys. |
-| **fluent-api** | yes (small) | New `user_settings` table + a new `self` domain exposing `GET/PUT /self/settings`; Zod-schema extension of the editor-state `resources` blob (no migration). |
+| **fluent-api** | yes (small) | New `user_settings` table + `GET/PUT /users/settings`; Zod-schema extension of the editor-state `resources` blob (no migration). |
 | **fluent-ai** | no | The check itself is unchanged. |
 | **fluent-platform** | no | No new services or env vars. |
 
@@ -72,7 +54,7 @@ The ignore actions require server-side persistence that does not exist yet. #172
 3. The **notification dot** (#277) on the Checks tab header, plus one proposed deviation: mirroring the dot on the panel-toggle button when the panel is closed (§5.3, sign-off item).
 4. A chapter-wide **check trigger** fired on every successful verse auto-save (W3).
 5. A three-layer **active/inactive cascade** unifying Greek Room's `legitimate` verdicts with user ignores, including undo (W5, W6).
-6. **Persistence:** occurrence-level rules in the existing editor-state JSONB; global word-pair rules in a new `user_settings` table exposed at `GET/PUT /self/settings` (W2, W7).
+6. **Persistence:** occurrence-level rules in the existing editor-state JSONB; global word-pair rules in a new `user_settings` table exposed at `GET/PUT /users/settings` (W2, W7).
 7. **Graceful degradation** when the settings backend half is absent, and inline error surfacing when the check call fails (W8, W9).
 8. Tests on both sides using each repo's established test infrastructure (§10).
 
@@ -98,13 +80,13 @@ Restated conclusions; supporting analysis in the cited sections.
 | **W2** | Hybrid storage: "Ignore This Time" lives in the existing per-chapter editor-state JSONB (schema-key addition, no migration); "Ignore Always" lives in a new `user_settings` table (one row per user, single Zod-typed JSONB column). Findings are filtered **client-side**. | Each rule is stored at exactly the scope it governs. Client-side filtering keeps the AI proxy a pure pass-through (D8/D9). `user_settings` deliberately establishes Fluent's user-global preference store: future settings become schema extensions, not migrations. See §7. |
 | **W3** | Chapter-wide check on every successful verse save: a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)`, where `saveCounter` increments in the verse-save mutation's `onSuccess`. The request sends **all currently drafted verses** of the chapter. No extra coalescing debounce. | The dot needs chapter-wide awareness; per-verse merging is more code for no user-visible gain. The existing 2 s per-verse save debounce already rate-limits; the check is <1 s. KISS — optimize if heavier checks arrive. See §6.2. |
 | **W4** | `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM book code, e.g. `JDG 4:3`), matching the convention already used by the repo smoke tests. Occurrence identity for suppressions = `(snt_id, repeated_word, ordinal)` where ordinal is the index of the finding among same-`repeated_word` findings in the verse, ordered by `start_position`. | Ordinals survive unrelated edits (positions don't); adding/removing an earlier same-pair occurrence conservatively re-flags, the safe failure direction. We count Greek Room's findings, never tokenize text ourselves, so Greek Room's equivalence policy is inherited consistently. See §6.3. |
-| **W5** | Every finding resolves to **active/inactive** via a three-layer cascade — Greek Room verdict (`legitimate`), user-global word-pair rule, occurrence rule — where the **most specific non-silent verdict wins**. Rules are tri-state maps (`absent / 'suppress' / 'surface'`). The dot counts active findings only. Inactive findings are hidden by default and revealed by the **"Show Ignored" toggle** (default OFF, not persisted, per revised #278); when shown they render greyed with the ignore-type label. | One mechanism explains machine-legitimate and user-ignored alike ("pre-ignored by Greek Room"). Specificity, not temporal order, keeps resolution deterministic. Greyed-not-removed (under the toggle) keeps undo discoverable. See §6.4. |
-| **W6** | Active findings show `[Ignore Here] [Ignore Everywhere]` (per revised card). `[Ignore Everywhere]` opens a **standard confirm dialog** before writing the global rule. Inactive findings show one `[Undo ▾]` split button: default click acts at the occurrence layer (delete own rule, or write an occurrence `surface` override); the chevron menu offers explicit global actions with consequence-naming labels. Any global write first purges the user's occurrence rules for that pair **in the current chapter's editor state only**. | Default click never silently edits global state. The card asks for a confirm on the global action because it reads as irreversible; we add the dialog **and** keep the `[Undo ▾]` reversibility — the two are complementary (confirm guards the click; undo guards the mistake). Purge-local prevents the just-clicked panel from appearing to ignore the action; other chapters' specific pronouncements deliberately stand. See §6.5. |
-| **W7** | Settings endpoint lives in a **new `self` domain**: **`GET /self/settings` + `PUT /self/settings`**, session-implicit user (no user in URL), guarded by `authenticateUser` alone, full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/self/settings/self-settings.{route,service,repository,types}.ts`. | Per kaseywright: a `self` domain gives clean semantic separation, avoids the redundant `{userId}` param (and the `requireSelf`/`UserPolicy` middleware that `/users/{id}/…` routes carry), and prevents accidental over-exposure (no `{id}` to misuse). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8. |
-| **W8** | Graceful degradation by feature detection: if `GET /self/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Everywhere]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first. | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9. |
+| **W5** | Every finding resolves to **active/inactive** via a three-layer cascade — Greek Room verdict (`legitimate`), user-global word-pair rule, occurrence rule — where the **most specific non-silent verdict wins**. Rules are tri-state maps (`absent / 'suppress' / 'surface'`). The dot counts active findings only. Inactive findings render greyed with a reason label; a "Show ignored & OK" toggle (default ON, persisted) hides them. | One mechanism explains machine-legitimate and user-ignored alike ("pre-ignored by Greek Room"). Specificity, not temporal order, keeps resolution deterministic. Greyed-not-removed keeps undo discoverable. See §6.4. |
+| **W6** | Active findings show `[Ignore This Time] [Ignore Always]` (per card). Inactive findings show one `[Undo ▾]` split button: default click acts at the occurrence layer (delete own rule, or write an occurrence `surface` override); the chevron menu offers explicit global actions with consequence-naming labels. Any global write first purges the user's occurrence rules for that pair **in the current chapter's editor state only**. No confirmation dialogs (global rules are reversible). | Default click never silently edits global state. Purge-local prevents the just-clicked panel from appearing to ignore the action (occurrence beats global in the cascade); other chapters' specific pronouncements deliberately stand. See §6.5. |
+| **W7** | Settings endpoint mirrors the editor-state idiom one level up: **`GET /users/settings` + `PUT /users/settings`**, session-implicit user (no user in URL), full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/users/settings/user-settings.{route,service,repository,types}.ts`. | `user_settings` is to _user_ what `user_chapter_assignment_editor_state` is to _(user, chapter)_; same shape, same endpoint idiom, same review story ("we did what you already do"). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8. |
+| **W8** | Graceful degradation by feature detection: if `GET /users/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Always]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first. | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9. |
 | **W9** | Check-call failure surfaces as a single inline `text-sm text-red-500` line at the top of the Checks panel ("Checks failed to refresh"); the panel keeps rendering the last successful findings (TanStack keeps `query.data` on refetch failure) and the dot reflects that last-known state. Failures log via the existing `Logger`. No toast/banner/popup. | Matches the drafting page's own inline-status precedent ("Auto-save failed"). Failure mode degrades to "results are one save behind." See §9.2. |
-| **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content and while the settings feature-detection probe (§9.1) is unresolved. | Card scope is the drafting view; empty chapters have nothing to check; the probe is one fast `GET`, and waiting for it means the first findings render is already cascade-correct. Review-stage checks are future work. See §6.2. |
-| **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets, the two buttons, and a "Show Ignored" toggle at the bottom; zero state inside the accordion. Only `activeLeftTab` is persisted in editor state (the "Show Ignored" state resets per session, per revised #278). Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5. |
+| **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content. | Card scope is the drafting view; empty chapters have nothing to check. Review-stage checks are future work. See §6.2. |
+| **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets and the two buttons; zero state inside the accordion. Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5. |
 | **W12** | One proposal document (this file) covering both repos; two implementation PRs (fluent-web, fluent-api), cross-referencing each other and the cards. | Splitting the proposal doubles reviewer overhead for a design whose halves only make sense together. See §9.4. |
 
 ---
@@ -139,28 +121,26 @@ The response envelope is consumed whole (D9): the hook inspects `status` and `re
 
 ### 5.1 What the mocks show
 
-The cards' mockups (drafting page, Judges 4, Gujarati IRV project) define the visual target. The #278 mocks were updated during the first review round (§0); the description below tracks the **current** mocks:
+The cards' mockups (drafting page, Judges 4, Gujarati IRV project) define the visual target:
 
 - **Tab header (#277 mock):** the left panel's current `Resources` heading is replaced by a text-tab row — **"Resources | Checks"** — active tab in blue with a blue underline, inactive tab plain. The **notification dot is a solid blue filled circle immediately right of the "Checks" label**, and the mock shows it while the *Resources* tab is active: the dot is visible from either tab whenever the panel is open.
 - **Checks content (#278 mocks):** below the tab header, a card containing a **collapsible "Repeated Words" accordion section** (expanded by default). The accordion-per-check structure anticipates future sibling checks (the board already holds draft cards for Greek Room Wildebeest and Spell checks) without UI rework. Inside the section, findings are grouped by verse:
   - bold **"Verse N"** heading per verse that has findings (verses without findings get no group);
-  - one row per occurrence: a one-line context snippet showing the repeated pair in context;
-  - two solid-blue buttons side by side beneath each snippet: `[Ignore Here] [Ignore Everywhere]`;
-  - a horizontal separator between verse groups;
-  - a **"Show Ignored" toggle** at the bottom of the active list (off by default). When on, previously ignored occurrences appear below the active ones — dimmed, each labeled with its ignore type (_"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) and showing an **"Undo Ignore"** action in place of the two ignore buttons.
+  - one row per occurrence: a one-line context snippet showing the repeated pair in situ;
+  - two solid-blue buttons side by side beneath each snippet: `[Ignore This Time] [Ignore Always]`;
+  - a horizontal separator between verse groups.
 - **Zero state (#278 mock):** the "Repeated Words" section shows a centered bold **"No issues found"**.
 
-### 5.2 Mock/text inconsistencies — resolved in first review round
+### 5.2 Mock/text inconsistencies — for product sign-off
 
-1. **Dot in the zero state.** The original #278 zero-state mock showed the dot on the Checks tab, contradicting #277's text ("the dot is absent when there are no active flags — no zero state"). We proposed following the text; chadw fixed the mock. **Resolved:** no flags, no dot.
-2. **Language dropdown above the Checks panel.** The original #278 mocks showed a language dropdown above the checks content, with no function for checks. We proposed omitting it; chadw agreed and removed it from the mocks. **Resolved:** no dropdown on the Checks tab.
+1. **Dot in the zero state.** The #278 zero-state mock still shows the dot on the Checks tab, contradicting #277's text ("the dot is absent when there are no active flags — no zero state"). **We follow the text:** no flags, no dot. Confirm.
+2. **Language dropdown above the Checks panel.** Both #278 mocks show the Resources tab's "English" language dropdown above the checks content, with no defined function for checks (the check runs against the chapter's target language; there is nothing to select). We presume this is a carryover artifact of editing a Resources-tab screenshot and **propose omitting the dropdown from the Checks tab**. Confirm.
 
 ### 5.3 Deviations from the card text — for product sign-off
 
-1. **Ignored items grey-and-stay rather than disappear.** The card's "Show Ignored" toggle (revised #278) now matches our approach: ignored occurrences are revealed dimmed/subordinate, sorted after active items within their verse group, labeled with the ignore type, with an "Undo Ignore" control (§6.5). We follow the card: the toggle defaults **off** and does **not** persist across sessions. Greyed items never count toward the dot. (This deviation is now largely card-aligned and is listed for completeness.)
-2. **Machine-"legitimate" findings are shown as inactive items.** Greek Room marks some repetitions `legitimate: true` (intentionally repeated words). The cards do not enumerate them explicitly, but the revised #278 "Show Ignored" mock includes a **"Default Ignore"** ignore-type — which is exactly this case. We render machine-legitimate findings in the same dimmed style with that label, surfaceable per-occurrence like any other ignored item (§6.4). This gives translators visibility into what the checker considered and a recourse when Greek Room is wrong in either direction.
+1. **Ignored items grey-and-stay rather than disappear.** #278 says an ignored occurrence "is removed from the panel." We instead render it greyed/subordinate, sorted after active items within its verse group, labeled with the winning rule (e.g. _"Ignored always: 'the the'"_), with an `[Undo ▾]` control (§6.5). Rationale: removal makes a mis-click irreversible-in-place and makes "Ignore Always" indistinguishable from a fix; the card's intent (flag stops demanding attention) is preserved because greyed items don't count toward the dot. A **"Show ignored & OK" checkbox** (default ON, persisted in editor state) collapses them entirely for translators who want the card-literal view.
+2. **Machine-"legitimate" findings are shown as inactive items.** Greek Room marks some repetitions `legitimate: true` (intentionally repeated words). The cards are silent on them. We render them in the same greyed style with reason _"Marked OK by Greek Room"_, surfaceable per-occurrence like any other inactive item (§6.4). This gives translators visibility into what the checker considered and a recourse when Greek Room is wrong in either direction.
 3. **Dot mirrored on the panel-toggle button.** #277 places the dot on the Checks tab header, which is invisible when the whole left panel is closed (it's behind the `BookText` toggle button in the drafting header). To preserve the card's intent — the translator is notified — we mirror the dot on that toggle button whenever the panel is closed and active flags exist.
-4. **Highlight repeated words in the verse text.** Product (joelthe1/Ulf, via #278 comment) asked that flagged repeated words be highlighted in the verse pane (underline or color). We will add this. Because it touches the verse-rendering pane — outside the left-panel tab/panel this proposal builds — and benefits from its own short design pass, it ships as a **follow-on PR** layered on this one, so the tab/panel work can be reviewed and merged independently. The visual/interaction specifics (underline vs. highlight, active-vs-ignored styling, click-to-locate) are worked out just before that PR.
 
 None of these block implementation; they are listed in §12 for explicit confirmation in PR review.
 
@@ -170,7 +150,7 @@ None of these block implementation; they are listed in §12 for explicit confirm
 
 ### 6.1 File layout
 
-```text
+```
 fluent-web/src/
 ├── components/ui/                          # (existing; shadcn primitives incl. dropdown-menu, checkbox)
 ├── features/
@@ -184,10 +164,9 @@ fluent-web/src/
 │       │   ├── LeftPanel.tsx                # Tab container: header row (Resources | Checks + dot),
 │       │   │                                #   renders ResourcePanel or ChecksPanel
 │       │   ├── ChecksPanel.tsx              # Per-check accordion; verse groups; zero state;
-│       │   │                                #   inline error line; "Show Ignored" toggle
-│       │   └── FindingRow.tsx               # Snippet + actions (active: Ignore Here / Ignore
-│       │                                    #   Everywhere [+ confirm dialog]; inactive: greyed +
-│       │                                    #   ignore-type label + Undo split button)
+│       │   │                                #   inline error line; "Show ignored & OK" toggle
+│       │   └── FindingRow.tsx               # Snippet + actions (active: two buttons; inactive:
+│       │                                    #   greyed + reason label + Undo split button)
 │       ├── hooks/
 │       │   ├── useRepeatedWordsCheck.ts     # useQuery keyed (chapterAssignmentId, saveCounter);
 │       │   │                                #   builds request; returns raw findings
@@ -211,7 +190,7 @@ const [saveCounter, setSaveCounter] = useState(0);
 useQuery({
   queryKey: ['repeated-words', chapterAssignmentId, saveCounter],
   queryFn: () => postRepeatedWordsCheck(buildRequest(projectItem, verses)),
-  enabled: !readOnly && versesWithContent.length > 0 && settingsProbeResolved,
+  enabled: !readOnly && versesWithContent.length > 0,
   // TanStack retains previous data on refetch failure; see §9.2
 });
 ```
@@ -220,8 +199,7 @@ useQuery({
 - **No extra debounce:** the per-verse 2 s save debounce already rate-limits typing bursts; the check is sub-second. If a heavier check joins later, a coalescing debounce can wrap `setSaveCounter` without touching anything else.
 - **Request body** is the full `RepeatedWordsRequest` (D8 shape): `lang_code`/`lang_name` from the project's target language, `project_id`/`project_name` from `projectItem`, and `verses[]` covering **all currently drafted verses** of the chapter (content from the drafting state, not a refetch — what the translator sees is what gets checked).
 - **`snt_id` = `"{bookCode} {chapter}:{verse}"`** with the USFM book code (e.g. `JDG 4:3`), the convention the repo smoke tests already use. Implementation note: verify the field carrying the USFM code on the drafting page's `projectItem` (vs. display name) and thread it into the builder.
-- **Where it runs (W10):** drafting route only; `enabled` is false in the read-only `/view` route (reviewers see no Checks activity in v1), when no verse has content, and while the settings feature-detection probe (§9.1) is unresolved — the probe is a single fast `GET`, and waiting for it means the first findings render is already cascade-correct (the user's global rules are known, present or absent).
-- **Emptied chapters resolve naturally:** any save of emptied text triggers a fresh check whose empty findings clear the panel and dot. (When *every* verse is emptied the query disables instead; the panel and dot then treat findings as empty rather than rendering stale data.)
+- **Where it runs (W10):** drafting route only; `enabled` is false in the read-only `/view` route, when no verse has content, and while settings feature-detection is unresolved for suppression-dependent rendering (§9.1 — the check itself still runs; only the Ignore Always affordance waits).
 
 ### 6.3 Occurrence identity (W4)
 
@@ -229,7 +207,7 @@ A suppression must survive verse edits without our re-parsing text. The key is *
 
 - `ordinal` = index of this finding among findings in the same verse with the same `repeated_word`, ordered by `start_position` ("x of n"). Computed from **Greek Room's findings only** — we never tokenize verse text ourselves, so Greek Room's case/diacritic equivalence policy is inherited and consistent between runs.
 - Ordinals survive unrelated edits (`start_position` does not). Adding or removing an *earlier* same-pair occurrence shifts later ordinals and conservatively re-flags — the safe failure direction.
-- String comparison between a stored rule's `repeated_word` and a fresh finding's: **NFC-normalize both, compare exactly, no case folding of our own.** Unicode case folding is locale-sensitive (the Turkish dotless-ı problem) and Fluent targets minority languages; NFC handles composed/decomposed accent variation. Note that Greek Room already delivers `repeated_word` lowercased ("word word" form, per the fluent-ai schema) — original casing lives only in `surf`, which we display but never compare — so case equivalence is wholly Greek Room's policy, inherited rather than re-implemented.
+- String comparison between a stored rule's `repeated_word` and a fresh finding's: **NFC-normalize both, compare exactly, no case folding.** Unicode case folding is locale-sensitive (the Turkish dotless-ı problem) and Fluent targets minority languages; NFC handles composed/decomposed accent variation. Consequence: a capitalization change re-flags — correct conservative behavior.
 - Documented caveat: a triple repetition ("the the the") yields two overlapping findings (ordinals 1 and 2). Mechanically fine; slightly odd UX; accepted for v1.
 
 ### 6.4 The active/inactive cascade (W5)
@@ -245,25 +223,16 @@ Every finding resolves through three layers; the **most specific non-silent verd
 - Stored rules are **tri-state maps** (`key → 'suppress' | 'surface'`, absent = silent), not bare suppression sets — `surface` is what makes per-occurrence undo of a global rule (or of a Greek Room `legitimate` verdict) possible.
 - A machine-legitimate finding is conceptually "pre-ignored by Greek Room": same inactive state as user-ignored, different actor, same undo affordance.
 - **The notification dot counts cascade-resolved active findings only.**
-- Inactive findings render greyed, sorted after active items within their verse group, each labeled with its ignore type — the labels the revised #278 mock uses: _"Ignore Here"_ (occurrence suppress), _"Ignore Everywhere"_ (global suppress), _"Default Ignore"_ (Greek Room `legitimate`). The **"Show Ignored"** toggle (default OFF, **not** persisted — resets each session, per revised #278) reveals them; with the toggle off, only active findings show.
+- Inactive findings render greyed, sorted after active items within their verse group, each labeled with the winning rule (_"Ignored this time"_, _"Ignored always: 'અને અને'"_, _"Marked OK by Greek Room"_). The **"Show ignored & OK"** checkbox (default ON so the undo affordance is discoverable; persisted in editor state) hides them.
 
 ### 6.5 Action surface (W6)
 
-- **Active finding** (per revised card #278): `[Ignore Here]` → writes occurrence rule `suppress`; `[Ignore Everywhere]` → opens a **confirm dialog** ("Fluent will not flag this word pair again across all your projects. Are you sure?") and, on confirm, writes the global pair rule `suppress`.
-- **Confirm dialog on `[Ignore Everywhere]`** (revised #278): the card asks for it because the action reads as global and irreversible. We honor the request **and** keep the action reversible via `[Undo ▾]` below — the dialog guards the deliberate click; the undo guards a mistake that slips past it. The two are not redundant: confirm is pre-action friction, undo is post-action recourse.
-
-  Why reversibility falls out of the design rather than being an extra feature we chose to bolt on — the implication chain (see also §6.4, which this depends on):
-  1. Greek Room reports **both** polarities: repetitions it flags (suspicious) and repetitions it judges `legitimate` (a *negative* / "don't flag" verdict).
-  2. Once we decide to **show** the legitimate ones (§5.3 item 2 — so the translator can see what the checker waved through), it's natural to let the translator disagree and mark one **illegitimate** (re-surface it).
-  3. Expressing "re-surface this" requires a **positive** record (`surface`), not only the negative (`suppress`) records an ignore-list would hold. So the stores must be tri-state, not suppression-only (§6.4).
-  4. Given positive *and* negative records at more than one scope, a **layered** model is the natural fit: a local (occurrence) record can override a global (word-pair) one, and the override works in **both polarities** (locally surface something suppressed globally, or locally suppress something surfaced).
-  5. With that layering in place, a single bare "undo" is **ambiguous** — at which layer, in which polarity? Resolving that ambiguity (the `[Undo ▾]` split button: default = act at the occurrence layer; chevron = act globally) is what the design *has* to do anyway — and the same mechanism incidentally gives the translator the "oops, I hit the wrong button" recourse for `[Ignore Everywhere]`. Reversibility is therefore inherited, not added.
-
-  We are not requesting that this override the card's "cannot be undone" framing; it is documented here for anyone who wants to follow the reasoning, and the implemented PR is reviewed again before merge.
+- **Active finding** (per card #278): `[Ignore This Time]` → writes occurrence rule `suppress`; `[Ignore Always]` → writes global pair rule `suppress`.
 - **Inactive finding:** one `[Undo ▾]` split button (shadcn `dropdown-menu`, already in the codebase — no hidden gestures):
   - **Default click** acts at the occurrence layer: if inactive via its *own* occurrence rule → delete that entry; if inactive via a global rule or Greek Room's verdict → write an occurrence-level `surface` override (the global rule / machine verdict survives; only this occurrence resurfaces).
   - **Chevron menu** offers the deliberate global actions with consequence-naming labels, e.g. _"Stop ignoring 'the the' everywhere"_ (deletes the global entry).
-- **Global writes purge local rules — current chapter only.** Executing any global (everywhere) suppress or surface first deletes the user's occurrence-level rules for that word pair in the *current chapter assignment's* editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
+- **Global writes purge local rules — current chapter only.** Executing any global (`always`) suppress or surface first deletes the user's occurrence-level rules for that word pair in the *current chapter assignment's* editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
+- **No confirmation dialogs:** global rules are reversible via the chevron (and a future settings page, §11), so `[Ignore Always]` does not nag.
 - **Direct edits resolve silently** (card #278): if the translator fixes the text, the next auto-save's fresh findings simply no longer contain the occurrence — no action needed, panel and dot update.
 
 ### 6.6 Left-panel container and persisted UI state (W11)
@@ -272,7 +241,7 @@ Every finding resolves through three layers; the **most specific non-silent verd
 
 - **`LeftPanel`** owns the tab header row ("Resources | Checks", blue underline on the active tab, blue dot right of "Checks" when active findings exist) and renders either the existing `ResourcePanel` or the new `ChecksPanel` below it. `ResourcePanel` loses only its `<h3>Resources</h3>` heading (the tab row replaces it); its content, language dropdown, and accordion are untouched.
 - The dot's state comes from the cascade-resolved active count (§6.4), which `DraftingPage` computes once and threads to both `LeftPanel` (tab dot) and the header toggle button (mirror dot, §5.3 item 3). The check query runs at `DraftingPage` level — **not** inside `ChecksPanel` — so the dot stays live while the Resources tab (or a closed panel) is showing.
-- **Persisted UI state.** One new key rides the same per-chapter editor-state blob the page already saves (debounced 500 ms, §7.1): `activeLeftTab: 'resources' | 'checks'` (reopen where you left off). The **"Show Ignored"** toggle is deliberately **not** persisted — per revised #278 it resets to off each session — so it lives in component state only. The persisted key is cosmetic; failure to persist degrades to defaults.
+- **Persisted UI state.** Two new keys ride the same per-chapter editor-state blob the page already saves (debounced 500 ms, §7.1): `activeLeftTab: 'resources' | 'checks'` (reopen where you left off) and `showResolvedChecks: boolean` (the "Show ignored & OK" checkbox, default `true`). Both are cosmetic; failure to persist degrades to defaults.
 - The panel's existing resize/drag behavior and 20–40 % width constraints are inherited unchanged — `LeftPanel` lives *inside* the resizable container.
 
 ---
@@ -312,6 +281,7 @@ export const editorStateResourcesSchema = z
     tabStatus: z.boolean(),
     // --- NEW, all optional (backward/forward compatible) ---
     activeLeftTab: z.enum(['resources', 'checks']).optional(),
+    showResolvedChecks: z.boolean().optional(),
     checkOccurrenceRules: z
       .record(z.string(), z.enum(['suppress', 'surface']))
       .optional(),
@@ -319,8 +289,8 @@ export const editorStateResourcesSchema = z
   .nullable();
 ```
 
-- `checkOccurrenceRules` is the tri-state map of §6.4 layer 2. Keys are the occurrence identity `"{snt_id}|{repeated_word}|{ordinal}"` (e.g. `"JDG 4:3|અને અને|1"`); the `|` separator cannot appear in a `snt_id` and `repeated_word` is the final segment-pair, so keys are unambiguous. Values: `'suppress'` ("Ignore Here") or `'surface'` (per-occurrence undo of a global/machine verdict). Absent = silent.
-- fluent-web reads/writes these through the **existing** `GET/PUT /chapter-assignments/:id/editor-state` round-trip it already performs — the same debounced save that persists `activeResource` today carries the new keys. No new endpoint for occurrence rules.
+- `checkOccurrenceRules` is the tri-state map of §6.4 layer 2. Keys are the occurrence identity `"{snt_id}|{repeated_word}|{ordinal}"` (e.g. `"JDG 4:3|અને અને|1"`); the `|` separator cannot appear in a `snt_id` and `repeated_word` is the final segment-pair, so keys are unambiguous. Values: `'suppress'` ("Ignore This Time") or `'surface'` (per-occurrence undo of a global/machine verdict). Absent = silent.
+- fluent-web reads/writes these through the **existing** `GET/PUT /users/chapter-assignments/:id/editor-state` round-trip it already performs — the same debounced save that persists `activeResource` today carries the new keys. No new endpoint for occurrence rules.
 - Write pattern: read-modify-write of the blob the page already holds in memory. The page is the only writer for this `(user, chapterAssignment)` pair in practice (concurrent same-user tabs are a pre-existing last-writer-wins, inherited).
 
 ### 7.2 Global rules — new `user_settings` table
@@ -350,9 +320,9 @@ export const user_settings = pgTable('user_settings', {
 ```
 
 - **One row per user** (`userId` is the primary key), one Zod-typed JSONB `settings` column — the same shape discipline as `user_chapter_assignment_editor_state`, one level up the scope ladder.
-- `checkIgnoredWordPairs` is the §6.4 layer-1 map. Keys are the NFC-normalized `repeated_word` string (the pair, e.g. `"the the"`); values `'suppress'` ("Ignore Everywhere") or `'surface'` (reserved — a global "never let Greek Room auto-OK this pair" is expressible but no v1 UI writes it). Absent = silent.
+- `checkIgnoredWordPairs` is the §6.4 layer-1 map. Keys are the NFC-normalized `repeated_word` string (the pair, e.g. `"the the"`); values `'suppress'` ("Ignore Always") or `'surface'` (reserved — a global "never let Greek Room auto-OK this pair" is expressible but no v1 UI writes it). Absent = silent.
 - Requires **one SQL migration** (new table; follows the numbered-migration convention in `src/db/migrations/`).
-- Exposed via `GET/PUT /self/settings` (§8).
+- Exposed via `GET/PUT /users/settings` (§8).
 
 ### 7.3 Why client-side filtering
 
@@ -366,39 +336,32 @@ The proxy therefore remains untouched by this proposal.
 
 ---
 
-## 8. fluent-api design: `GET/PUT /self/settings` (W7)
-
-The settings endpoint lives in a **new `self` domain**, requested by kaseywright in review. `self` is the authenticated caller acting on their own resources — the route never names a user. This is cleaner than hanging settings off `/users/{userId}/…`:
+## 8. fluent-api design: `GET/PUT /users/settings` (W7)
 
-- **Clean semantic separation** — `self/*` is unambiguously "me," leaving the `users/*` domain for (future) admin-style tooling that operates on *other* users by id.
-- **No redundant `{userId}` param** — the user comes from the session, not the URL (the same source `user-chapter-assignment-editor-state.route.ts` already uses).
-- **No accidental over-exposure** — with no `{id}` in the path there is nothing to mis-scope; contrast `/users/{userId}/projects`, which must add `requireSelf()` to stop one user reading another's data.
-- **Simpler middleware** — just `authenticateUser`. No `requireSelf`, no `requireUserAccess`, no `UserPolicy`.
-
-Otherwise it keeps the editor-state idiom: **full-replace upsert** on PUT, one Zod-typed JSONB blob, last-writer-wins on concurrent writes.
+The settings endpoint deliberately mirrors the editor-state endpoint one level up the scope ladder — `user_settings` is to *user* what `user_chapter_assignment_editor_state` is to *(user, chapterAssignment)*. Same idiom throughout: **session-implicit user** (the authenticated user is taken from context, never from the URL — consistent with `user-chapter-assignment-editor-state.route.ts`, where no `userId` appears in the path), **full-replace upsert** on PUT, one Zod-typed JSONB blob.
 
 ### 8.1 Routes
 
 | Method | Path | Auth | Body | Response |
 | --- | --- | --- | --- | --- |
-| `GET` | `/self/settings` | `authenticateUser` only | — | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
-| `PUT` | `/self/settings` | `authenticateUser` only | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation |
+| `GET` | `/users/settings` | session (any authenticated user) | — | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
+| `PUT` | `/users/settings` | session (any authenticated user) | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation |
 
-- **No permission middleware beyond `authenticateUser`.** Every caller owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id` from context. (Editor-state additionally guards with `CONTENT_UPDATE` + chapter-assignment participation because it is tied to a chapter resource; `self/settings` has no resource to scope, so session identity alone is the right guard — and the `self` prefix makes that explicit.)
+- **No special permission** beyond an authenticated session: every user owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id`. (Editor-state precedent; there is nothing to elevate.)
 - **PUT is full-replace.** The client GETs, merges in memory, PUTs the whole blob — the editor-state write pattern. Concurrent tabs are last-writer-wins, identical to editor-state's accepted behavior; for an ignore-list this worst case is "one ignore from another tab is lost until re-clicked."
-- Rejected alternatives, for the record: `/users/settings` and `/users/{userId}/settings` (the original proposal — kaseywright preferred a dedicated `self` domain over loading another route into `users`, see above), `/users/me/settings` (no `me` convention exists in this API), and per-feature endpoints like `/self/settings/check-ignores` (the blob is one document; sub-paths invite partial-update semantics we don't need).
+- Rejected alternatives, for the record: `/user-settings` (breaks the `domains/users` nesting), `/users/me/settings` (no `me` convention exists in this API), and per-feature endpoints like `/users/settings/check-ignores` (the blob is one document; sub-paths invite partial-update semantics we don't need).
 
 ### 8.2 File layout
 
-```text
-fluent-api/src/domains/self/settings/
-├── self-settings.route.ts        # GET + PUT, OpenAPI-described, session user from context
-├── self-settings.service.ts      # get / upsert, Result-typed like editor-state service
-├── self-settings.repository.ts   # drizzle upsert on userId PK (onConflictDoUpdate)
-└── self-settings.types.ts        # response schema; re-exports UserSettings from db schema
+```
+fluent-api/src/domains/users/settings/
+├── user-settings.route.ts        # GET + PUT, OpenAPI-described, session user from context
+├── user-settings.service.ts      # get / upsert, Result-typed like editor-state service
+├── user-settings.repository.ts   # drizzle upsert on userId PK (onConflictDoUpdate)
+└── user-settings.types.ts        # response schema; re-exports UserSettings from db schema
 ```
 
-Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§7.2), one numbered migration in `src/db/migrations/`, and route registration in `src/app.ts`. (The table is still named `user_settings` — it stores per-user rows; only the *route* surface is `self`. The repository upserts on the `userId` PK using the session user's id.)
+Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§7.2), one numbered migration in `src/db/migrations/`, and route registration in `src/app.ts` next to the existing users routes.
 
 ### 8.3 What fluent-api explicitly does **not** do
 
@@ -412,11 +375,11 @@ Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§
 
 ### 9.1 Feature detection for the settings half (W8)
 
-The two PRs (fluent-web, fluent-api) may land and deploy in either order, so the UI must not assume `GET /self/settings` exists:
+The two PRs (fluent-web, fluent-api) may land and deploy in either order, so the UI must not assume `GET /users/settings` exists:
 
-- On drafting-page mount, `useSuppressions` probes `GET /self/settings` once per session. **`404` ⇒ `globalIgnoresAvailable = false`** for the session (the route doesn't exist yet on this deployment); any 2xx — including `settings: null` — ⇒ available.
-- When unavailable, the **capability is hidden, not disabled**: `[Ignore Everywhere]` is simply not rendered, and the `[Undo ▾]` chevron omits global entries. A dead button is a bug report; an absent one is honest. Occurrence-level ignores (editor-state-backed) work regardless.
-- The check query waits for the probe to resolve before its first run (W10, §6.2) — one fast `GET`, so findings never render ahead of the user's global rules. The probe resolves on any terminal response: 2xx ⇒ available, `404` ⇒ unavailable; a network failure resolves it as unavailable for the session (conservative: occurrence ignores only).
+- On drafting-page mount, `useSuppressions` probes `GET /users/settings` once per session. **`404` ⇒ `globalIgnoresAvailable = false`** for the session (the route doesn't exist yet on this deployment); any 2xx — including `settings: null` — ⇒ available.
+- When unavailable, the **capability is hidden, not disabled**: `[Ignore Always]` is simply not rendered, and the `[Undo ▾]` chevron omits global entries. A dead button is a bug report; an absent one is honest. Occurrence-level ignores (editor-state-backed) work regardless.
+- While the probe is unresolved, active findings render with `[Ignore This Time]` only — the check itself never waits on the probe (§6.2).
 - **Unknown-key tolerance both ways:** `userSettingsSchema` uses `.catch({})` server-side; client-side parsing treats absent/unknown keys as empty. A newer client against an older blob (or vice versa) degrades to "no global rules," never to an error. The same applies to the editor-state blob's new optional keys (§7.1) — old clients ignore them, old rows omit them.
 
 ### 9.2 Check-call failure (W9)
@@ -438,7 +401,7 @@ Ignore actions apply **optimistically** (the finding greys immediately) and roll
 ### 9.4 Rollout / landing order (W12)
 
 - **Two PRs**, one per repo, each referencing this proposal and cards #277/#278 (and #172 for the fluent-api half). Either lands first:
-  - fluent-web first ⇒ probe 404s ⇒ occurrence ignores only, no `[Ignore Everywhere]` (§9.1). Fully usable.
+  - fluent-web first ⇒ probe 404s ⇒ occurrence ignores only, no `[Ignore Always]` (§9.1). Fully usable.
   - fluent-api first ⇒ new table/endpoint sits unused. Harmless.
 - The editor-state schema-key extension (§7.1) is backward/forward compatible by construction (optional keys, JSONB column unchanged), so no coordination is needed there either.
 - No new env vars, services, or fluent-platform changes; the feature follows the existing deploy pipeline of each repo.
@@ -453,19 +416,19 @@ The repo now ships test infrastructure under `src/test/` (`msw/server.ts`, `rend
 
 | Area | Representative cases |
 | --- | --- |
-| `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC composed/decomposed key equivalence. |
+| `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC-equal / case-different comparisons. |
 | `useRepeatedWordsCheck` | Fires on `saveCounter` change; request body shape (snake_case verbatim, `snt_id` format, all drafted verses); `enabled` gating (readOnly, empty chapter); previous data retained on 502 (MSW error handler). |
 | `useSuppressions` | Feature-detect: 404 ⇒ `globalIgnoresAvailable=false`; 200-with-null ⇒ available; occurrence rule round-trip through editor-state blob; global write purges current-chapter occurrence rules for the pair; optimistic rollback on PUT failure. |
-| `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and *not* on error); inline error line; active `[Ignore Here]`/`[Ignore Everywhere]` vs. greyed-with-label + `[Undo ▾]`; "Show Ignored" toggle defaults off and is not persisted; `[Ignore Everywhere]` opens confirm dialog and only writes on confirm; `[Ignore Everywhere]` absent when capability unavailable. |
+| `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and *not* on error); inline error line; active buttons vs. greyed-with-reason + `[Undo ▾]`; "Show ignored & OK" toggle; `[Ignore Always]` absent when capability unavailable. |
 | `LeftPanel` / dot | Dot iff cascade-active > 0, visible from Resources tab; tab switch persistence key written; toggle-button mirror dot when panel closed. |
 
 ### 10.2 fluent-api (Vitest, route-level — same style as the ai-tools and editor-state tests)
 
 | Area | Representative cases |
 | --- | --- |
-| `GET /self/settings` | 401 unauthenticated; `settings: null` for new user (200, not 404); returns saved blob; user isolation (only own row, scoped to session user). |
-| `PUT /self/settings` | Upsert create + replace; 400 on schema violation; full-replace semantics (omitted keys gone); `.catch({})` tolerance on read of unknown-shaped stored blobs. |
-| Editor-state schema extension | Existing editor-state tests still pass unchanged (proves backward compatibility); round-trip of `checkOccurrenceRules` / `activeLeftTab`. |
+| `GET /users/settings` | 401 unauthenticated; `settings: null` for new user (200, not 404); returns saved blob; user isolation (only own row). |
+| `PUT /users/settings` | Upsert create + replace; 400 on schema violation; full-replace semantics (omitted keys gone); `.catch({})` tolerance on read of unknown-shaped stored blobs. |
+| Editor-state schema extension | Existing editor-state tests still pass unchanged (proves backward compatibility); round-trip of `checkOccurrenceRules` / `activeLeftTab` / `showResolvedChecks`. |
 
 Manual verification continues to use the stack via `fluent-platform` compose plus the existing repeated-words smoke script for the proxy half.
 
@@ -474,8 +437,7 @@ Manual verification continues to use the stack via `fluent-platform` compose plu
 ## 11. Future work (explicitly deferred)
 
 - **Sibling checks** (Wildebeest character check, spell check — draft cards exist on the board): each becomes another accordion section in `ChecksPanel` and another wire-type module; the cascade, dot, and suppression stores are check-agnostic by construction (occurrence keys embed the finding identity; global keys can be namespaced per check when a second one arrives).
-- **"Manage ignored words" settings page**: a CRUD view over `userSettings.checkIgnoredWordPairs`. The store and `self/settings` endpoint are already shaped for it.
-- **Highlight repeated words in the verse text** (joelthe1/Ulf, §5.3 item 4): a follow-on PR layered on this one, touching the verse-rendering pane; specced just before implementation.
+- **"Manage ignored words" settings page**: a CRUD view over `userSettings.checkIgnoredWordPairs`. The store and endpoint are already shaped for it.
 - **Checks in review/read-only stages** (peer check, `/view` route): needs product definition of who may ignore what; W10 keeps v1 to drafting.
 - **Async/polling tools**: the envelope is consumed whole (D9), so a `status: "queued"` tool slots in with a polling loop inside `useRepeatedWordsCheck` and zero UI-contract change.
 - **"Drop duplicate" quick-fix and Greek Room feedback loop**: excluded by card #278; the occurrence identity (`snt_id`, pair, ordinal, `start_position` at render time) is sufficient input for both when they're picked up.
@@ -484,25 +446,23 @@ Manual verification continues to use the stack via `fluent-platform` compose plu
 
 ## 12. Sign-off checklist
 
-Items needing explicit confirmation in review (referenced from §5.2/§5.3); none block reading the rest of the design. Status reflects the first review round (§0):
+Items needing explicit confirmation in review (referenced from §5.2/§5.3); none block reading the rest of the design:
 
-| # | Item | Proposed resolution | Status |
-| --- | --- | --- | --- |
-| S1 | Zero-state mock showed the notification dot; #277 text says no flags ⇒ no dot | Follow the text: no dot at zero active flags | ✅ Resolved — chadw fixed the mock |
-| S2 | #278 mocks showed a language dropdown above checks content | Omit it on the Checks tab (no function for checks) | ✅ Resolved — chadw removed it from the mocks |
-| S3 | Card originally said ignored items are "removed from the panel" | Reveal them dimmed under a "Show Ignored" toggle (default off, not persisted) with "Undo Ignore"; dot unaffected | ✅ Largely card-aligned — revised #278 added the toggle; confirm the labels match |
-| S4 | Cards did not enumerate Greek-Room-`legitimate` findings | Show as ignored items labeled **"Default Ignore"** (the revised #278 mock's term), per-occurrence surfaceable | ☑ Confirm "Default Ignore" = machine-legitimate |
-| S5 | Dot is invisible when the whole left panel is closed | Mirror the dot on the panel-toggle button while closed | ☐ Open |
-| S6 | Suppression persistence has no card of its own | Ship as extension of #172's scope (W1) — backend dependency of the UI cards | ☐ Open |
-| S7 | Revised #278 confirm dialog says "Ignore Everywhere" "cannot be undone" | Add the confirm dialog **and** keep `[Undo ▾]` reversibility (§6.5); the two are complementary | ☐ Confirm we may retain undo |
-| S8 | Highlight repeated words in the verse text (joelthe1/Ulf) | We will implement it as a **follow-on PR** layered on this one (verse-pane scope; §5.3 item 4) | ☐ Confirm follow-on-PR sequencing |
+| # | Item | Proposed resolution |
+| --- | --- | --- |
+| S1 | Zero-state mock shows the notification dot; #277 text says no flags ⇒ no dot | Follow the text: no dot at zero active flags |
+| S2 | #278 mocks show the Resources language dropdown above checks content | Omit it on the Checks tab (presumed screenshot carryover; no function for checks) |
+| S3 | Card says ignored items are "removed from the panel" | Grey-and-stay with reason label + undo; "Show ignored & OK" unchecked gives the card-literal view; dot unaffected either way |
+| S4 | Cards silent on Greek-Room-`legitimate` findings | Show as inactive items ("Marked OK by Greek Room"), per-occurrence surfaceable |
+| S5 | Dot is invisible when the whole left panel is closed | Mirror the dot on the panel-toggle button while closed |
+| S6 | Suppression persistence has no card of its own | Ship as extension of #172's scope (W1) — backend dependency of the UI cards |
 
 Engineering-side confirmations sought from reviewers (same spirit as the D-series confirmations in the fluent-api proposal):
 
-- **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and a new **`self` domain** (`GET/PUT /self/settings`) as its endpoint (this outlives the feature). _Endpoint shape requested by kaseywright in the first round; confirming the `self`-domain placement._
+- **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and `GET/PUT /users/settings` as its endpoint (this outlives the feature).
 - **W4** — `(snt_id, repeated_word, ordinal)` occurrence identity and NFC-no-case-folding comparison.
-- **W5/W6** — the tri-state cascade, the confirm-dialog-plus-`[Undo ▾]` behavior on "Ignore Everywhere," and the split-button undo (S7).
+- **W5/W6** — the tri-state cascade and the `[Undo ▾]` split-button behavior.
 
 ---
 
-*Prepared against fluent-web `main` (post `2e936d1`), fluent-api `main` (post `7c7f63d`), and the approved fluent-api proposal (PR #173). Revised 2026-06-16 after the first review round on PR #305. Author: JEdward7777.*
+*Prepared against fluent-web `main` (post `0dcea61`), fluent-api `main` (post `7c7f63d`), and the approved fluent-api proposal (PR #173). Author: JEdward7777.*

From 886ee3abd8da2310ed162eedbfb918ecf11766ef Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 11 Jun 2026 16:49:54 -0400
Subject: [PATCH 04/33] Repeated Word Check UI proposal: decision-sync fixes +
 review summary

---
 .../checks-ui-integration-suggestion.md       | 20 ++++++-----
 .../checks-ui-integration-summary.md          | 36 ++++++++-----------
 2 files changed, 25 insertions(+), 31 deletions(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index 85b2462e..e20754e6 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -1,6 +1,7 @@
 # Repeated Word Check UI — Proposal (Checks Tab, Checks Panel, and Suppression Persistence)
 
 **Status:** Draft for review.
+**Reviewer shortcut:** a condensed, stands-on-its-own review summary lives in [`checks-ui-integration-summary.md`](checks-ui-integration-summary.md).
 **Scope:** Implements the user-facing half of the Repeated Word Check feature: the **Checks tab** (#277) and **Checks view panel** (#278) on fluent-web, plus the small fluent-api additions both cards imply (suppression persistence). This is a single proposal covering both repos so reviewers see the whole design in one place; **implementation will ship as two PRs** (one per repo), and either may land first (§9.4).
 
 **Related cards:**
@@ -32,7 +33,7 @@ Both cards carry mockups (drafting page, Judges 4, Gujarati IRV); §5 transcribe
 
 ### 1.1 What the cards imply beyond fluent-web
 
-The ignore actions require server-side persistence that does not exist yet. #172's text ("the backend dependency for all Repeated Word Check UI work") is broad enough to cover suppression storage — the fluent-api proposal's **D1** deferred caching of _tool runs/findings_, which is a different concern from user suppression preferences (a requirement that did not exist until #277/#278 were authored). Accordingly, suppression persistence ships as an **extension of #172's scope** (decision **W1**), designed here and implemented in the fluent-api PR of this pair. No new product card is needed.
+The ignore actions require server-side persistence that does not exist yet. #172's text ("the backend dependency for all Repeated Word Check UI work") is broad enough to cover suppression storage — the fluent-api proposal's **D1** deferred caching of _tool runs/findings_, which is a different concern from user suppression preferences (a requirement that did not exist until #277/#278 were authored). Accordingly, suppression persistence ships as an **extension of #172's scope** (decision **W1**), designed here and implemented in the fluent-api PR of this pair, which also amends the approved fluent-api proposal pair (the suggestion + operations docs) to record the scope extension. No new product card is needed.
 
 ### 1.2 Repos touched
 
@@ -85,7 +86,7 @@ Restated conclusions; supporting analysis in the cited sections.
 | **W7** | Settings endpoint mirrors the editor-state idiom one level up: **`GET /users/settings` + `PUT /users/settings`**, session-implicit user (no user in URL), full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/users/settings/user-settings.{route,service,repository,types}.ts`. | `user_settings` is to _user_ what `user_chapter_assignment_editor_state` is to _(user, chapter)_; same shape, same endpoint idiom, same review story ("we did what you already do"). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8. |
 | **W8** | Graceful degradation by feature detection: if `GET /users/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Always]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first. | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9. |
 | **W9** | Check-call failure surfaces as a single inline `text-sm text-red-500` line at the top of the Checks panel ("Checks failed to refresh"); the panel keeps rendering the last successful findings (TanStack keeps `query.data` on refetch failure) and the dot reflects that last-known state. Failures log via the existing `Logger`. No toast/banner/popup. | Matches the drafting page's own inline-status precedent ("Auto-save failed"). Failure mode degrades to "results are one save behind." See §9.2. |
-| **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content. | Card scope is the drafting view; empty chapters have nothing to check. Review-stage checks are future work. See §6.2. |
+| **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content and while the settings feature-detection probe (§9.1) is unresolved. | Card scope is the drafting view; empty chapters have nothing to check; the probe is one fast `GET`, and waiting for it means the first findings render is already cascade-correct. Review-stage checks are future work. See §6.2. |
 | **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets and the two buttons; zero state inside the accordion. Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5. |
 | **W12** | One proposal document (this file) covering both repos; two implementation PRs (fluent-web, fluent-api), cross-referencing each other and the cards. | Splitting the proposal doubles reviewer overhead for a design whose halves only make sense together. See §9.4. |
 
@@ -190,7 +191,7 @@ const [saveCounter, setSaveCounter] = useState(0);
 useQuery({
   queryKey: ['repeated-words', chapterAssignmentId, saveCounter],
   queryFn: () => postRepeatedWordsCheck(buildRequest(projectItem, verses)),
-  enabled: !readOnly && versesWithContent.length > 0,
+  enabled: !readOnly && versesWithContent.length > 0 && settingsProbeResolved,
   // TanStack retains previous data on refetch failure; see §9.2
 });
 ```
@@ -199,7 +200,8 @@ useQuery({
 - **No extra debounce:** the per-verse 2 s save debounce already rate-limits typing bursts; the check is sub-second. If a heavier check joins later, a coalescing debounce can wrap `setSaveCounter` without touching anything else.
 - **Request body** is the full `RepeatedWordsRequest` (D8 shape): `lang_code`/`lang_name` from the project's target language, `project_id`/`project_name` from `projectItem`, and `verses[]` covering **all currently drafted verses** of the chapter (content from the drafting state, not a refetch — what the translator sees is what gets checked).
 - **`snt_id` = `"{bookCode} {chapter}:{verse}"`** with the USFM book code (e.g. `JDG 4:3`), the convention the repo smoke tests already use. Implementation note: verify the field carrying the USFM code on the drafting page's `projectItem` (vs. display name) and thread it into the builder.
-- **Where it runs (W10):** drafting route only; `enabled` is false in the read-only `/view` route, when no verse has content, and while settings feature-detection is unresolved for suppression-dependent rendering (§9.1 — the check itself still runs; only the Ignore Always affordance waits).
+- **Where it runs (W10):** drafting route only; `enabled` is false in the read-only `/view` route (reviewers see no Checks activity in v1), when no verse has content, and while the settings feature-detection probe (§9.1) is unresolved — the probe is a single fast `GET`, and waiting for it means the first findings render is already cascade-correct (the user's global rules are known, present or absent).
+- **Emptied chapters resolve naturally:** any save of emptied text triggers a fresh check whose empty findings clear the panel and dot. (When *every* verse is emptied the query disables instead; the panel and dot then treat findings as empty rather than rendering stale data.)
 
 ### 6.3 Occurrence identity (W4)
 
@@ -207,7 +209,7 @@ A suppression must survive verse edits without our re-parsing text. The key is *
 
 - `ordinal` = index of this finding among findings in the same verse with the same `repeated_word`, ordered by `start_position` ("x of n"). Computed from **Greek Room's findings only** — we never tokenize verse text ourselves, so Greek Room's case/diacritic equivalence policy is inherited and consistent between runs.
 - Ordinals survive unrelated edits (`start_position` does not). Adding or removing an *earlier* same-pair occurrence shifts later ordinals and conservatively re-flags — the safe failure direction.
-- String comparison between a stored rule's `repeated_word` and a fresh finding's: **NFC-normalize both, compare exactly, no case folding.** Unicode case folding is locale-sensitive (the Turkish dotless-ı problem) and Fluent targets minority languages; NFC handles composed/decomposed accent variation. Consequence: a capitalization change re-flags — correct conservative behavior.
+- String comparison between a stored rule's `repeated_word` and a fresh finding's: **NFC-normalize both, compare exactly, no case folding of our own.** Unicode case folding is locale-sensitive (the Turkish dotless-ı problem) and Fluent targets minority languages; NFC handles composed/decomposed accent variation. Note that Greek Room already delivers `repeated_word` lowercased ("word word" form, per the fluent-ai schema) — original casing lives only in `surf`, which we display but never compare — so case equivalence is wholly Greek Room's policy, inherited rather than re-implemented.
 - Documented caveat: a triple repetition ("the the the") yields two overlapping findings (ordinals 1 and 2). Mechanically fine; slightly odd UX; accepted for v1.
 
 ### 6.4 The active/inactive cascade (W5)
@@ -290,7 +292,7 @@ export const editorStateResourcesSchema = z
 ```
 
 - `checkOccurrenceRules` is the tri-state map of §6.4 layer 2. Keys are the occurrence identity `"{snt_id}|{repeated_word}|{ordinal}"` (e.g. `"JDG 4:3|અને અને|1"`); the `|` separator cannot appear in a `snt_id` and `repeated_word` is the final segment-pair, so keys are unambiguous. Values: `'suppress'` ("Ignore This Time") or `'surface'` (per-occurrence undo of a global/machine verdict). Absent = silent.
-- fluent-web reads/writes these through the **existing** `GET/PUT /users/chapter-assignments/:id/editor-state` round-trip it already performs — the same debounced save that persists `activeResource` today carries the new keys. No new endpoint for occurrence rules.
+- fluent-web reads/writes these through the **existing** `GET/PUT /chapter-assignments/:id/editor-state` round-trip it already performs — the same debounced save that persists `activeResource` today carries the new keys. No new endpoint for occurrence rules.
 - Write pattern: read-modify-write of the blob the page already holds in memory. The page is the only writer for this `(user, chapterAssignment)` pair in practice (concurrent same-user tabs are a pre-existing last-writer-wins, inherited).
 
 ### 7.2 Global rules — new `user_settings` table
@@ -347,7 +349,7 @@ The settings endpoint deliberately mirrors the editor-state endpoint one level u
 | `GET` | `/users/settings` | session (any authenticated user) | — | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
 | `PUT` | `/users/settings` | session (any authenticated user) | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation |
 
-- **No special permission** beyond an authenticated session: every user owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id`. (Editor-state precedent; there is nothing to elevate.)
+- **No special permission** beyond an authenticated session: every user owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id`. (Editor-state additionally guards with `CONTENT_UPDATE` + chapter-assignment participation because it is tied to a chapter resource; settings have no resource to scope, so session identity alone is the right guard.)
 - **PUT is full-replace.** The client GETs, merges in memory, PUTs the whole blob — the editor-state write pattern. Concurrent tabs are last-writer-wins, identical to editor-state's accepted behavior; for an ignore-list this worst case is "one ignore from another tab is lost until re-clicked."
 - Rejected alternatives, for the record: `/user-settings` (breaks the `domains/users` nesting), `/users/me/settings` (no `me` convention exists in this API), and per-feature endpoints like `/users/settings/check-ignores` (the blob is one document; sub-paths invite partial-update semantics we don't need).
 
@@ -379,7 +381,7 @@ The two PRs (fluent-web, fluent-api) may land and deploy in either order, so the
 
 - On drafting-page mount, `useSuppressions` probes `GET /users/settings` once per session. **`404` ⇒ `globalIgnoresAvailable = false`** for the session (the route doesn't exist yet on this deployment); any 2xx — including `settings: null` — ⇒ available.
 - When unavailable, the **capability is hidden, not disabled**: `[Ignore Always]` is simply not rendered, and the `[Undo ▾]` chevron omits global entries. A dead button is a bug report; an absent one is honest. Occurrence-level ignores (editor-state-backed) work regardless.
-- While the probe is unresolved, active findings render with `[Ignore This Time]` only — the check itself never waits on the probe (§6.2).
+- The check query waits for the probe to resolve before its first run (W10, §6.2) — one fast `GET`, so findings never render ahead of the user's global rules. The probe resolves on any terminal response: 2xx ⇒ available, `404` ⇒ unavailable; a network failure resolves it as unavailable for the session (conservative: occurrence ignores only).
 - **Unknown-key tolerance both ways:** `userSettingsSchema` uses `.catch({})` server-side; client-side parsing treats absent/unknown keys as empty. A newer client against an older blob (or vice versa) degrades to "no global rules," never to an error. The same applies to the editor-state blob's new optional keys (§7.1) — old clients ignore them, old rows omit them.
 
 ### 9.2 Check-call failure (W9)
@@ -416,7 +418,7 @@ The repo now ships test infrastructure under `src/test/` (`msw/server.ts`, `rend
 
 | Area | Representative cases |
 | --- | --- |
-| `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC-equal / case-different comparisons. |
+| `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC composed/decomposed key equivalence. |
 | `useRepeatedWordsCheck` | Fires on `saveCounter` change; request body shape (snake_case verbatim, `snt_id` format, all drafted verses); `enabled` gating (readOnly, empty chapter); previous data retained on 502 (MSW error handler). |
 | `useSuppressions` | Feature-detect: 404 ⇒ `globalIgnoresAvailable=false`; 200-with-null ⇒ available; occurrence rule round-trip through editor-state blob; global write purges current-chapter occurrence rules for the pair; optimistic rollback on PUT failure. |
 | `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and *not* on error); inline error line; active buttons vs. greyed-with-reason + `[Undo ▾]`; "Show ignored & OK" toggle; `[Ignore Always]` absent when capability unavailable. |
diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-summary.md b/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
index 12af7694..720ebd9c 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
@@ -1,49 +1,41 @@
 # Repeated Word Check UI — Review Summary
 
-**Status:** Revised after the first review round (PR #305).
-
-**Purpose:** Reviewer orientation for the proposed Repeated Word Check UI (cards [fluent-web#277](https://github.com/eten-tech-foundation/fluent-web/issues/277) and [fluent-web#278](https://github.com/eten-tech-foundation/fluent-web/issues/278)) plus the small fluent-api persistence additions the cards imply. This summary is intended to stand on its own; the long-form proposal with full analysis lives in the sibling file [`checks-ui-integration-suggestion.md`](checks-ui-integration-suggestion.md) (decisions **W1–W12**, §0–§12; §0 is the change log for this revision). It builds on the approved fluent-api proxy design ([PR #173](https://github.com/eten-tech-foundation/fluent-api/pull/173), decisions **D1–D12**). Ships as **two PRs** — fluent-web (the bulk) plus a small fluent-api PR — and either may land first by design (W8/W12). A **highlight-in-verse** capability requested in review ships as a separate follow-on PR.
+**Purpose:** Reviewer orientation for the proposed Repeated Word Check UI (cards [fluent-web#277](https://github.com/eten-tech-foundation/fluent-web/issues/277) and [fluent-web#278](https://github.com/eten-tech-foundation/fluent-web/issues/278)) plus the small fluent-api persistence additions the cards imply. This summary is intended to stand on its own; the long-form proposal with full analysis lives in the sibling file [`checks-ui-integration-suggestion.md`](checks-ui-integration-suggestion.md) (decisions **W1–W12**, §1–§12). It builds on the approved fluent-api proxy design ([PR #173](https://github.com/eten-tech-foundation/fluent-api/pull/173), decisions **D1–D12**). Ships as **two PRs** — fluent-web (the bulk) plus a small fluent-api PR — and either may land first by design (W8/W12).
 
 ## What's being proposed
 
-Give translators the user-facing half of the Repeated Word Check: a **Checks tab** with a notification dot in the drafting page's left panel (#277) and a **Checks panel** listing repeated-word findings per verse, refreshed on every auto-save, with **"Ignore Here" / "Ignore Everywhere"** actions persisted in Fluent's database (#278). The structure (per-check accordion, check-agnostic suppression cascade) anticipates future checks (Wildebeest, spell check) without rework.
+Give translators the user-facing half of the Repeated Word Check: a **Checks tab** with a notification dot in the drafting page's left panel (#277) and a **Checks panel** listing repeated-word findings per verse, refreshed on every auto-save, with "Ignore This Time" / "Ignore Always" actions persisted in Fluent's database (#278). The structure (per-check accordion, check-agnostic suppression cascade) anticipates future checks (Wildebeest, spell check) without rework.
 
 ## Core decisions for review
 
 1. **Suppression persistence ships as an extension of #172's scope — no new product card** (W1). #172's "backend dependency for all Repeated Word Check UI work" covers it; the fluent-api proposal's D1 deferral was about caching tool results, not user preferences.
 
-2. **Hybrid storage at exactly the scope each rule governs** (W2): "Ignore Here" extends the existing per-`(user, chapterAssignment)` editor-state JSONB (optional Zod keys, **no migration**); "Ignore Everywhere" introduces a new **`user_settings` table** — one row per user, one Zod-typed JSONB column — deliberately establishing Fluent's general user-preference store (future settings become schema keys, not migrations). Findings are filtered **client-side**, keeping the AI proxy the pure pass-through D8/D9 established.
+2. **Hybrid storage at exactly the scope each rule governs** (W2): "Ignore This Time" extends the existing per-`(user, chapterAssignment)` editor-state JSONB (optional Zod keys, **no migration**); "Ignore Always" introduces a new **`user_settings` table** — one row per user, one Zod-typed JSONB column — deliberately establishing Fluent's general user-preference store (future settings become schema keys, not migrations). Findings are filtered **client-side**, keeping the AI proxy the pure pass-through D8/D9 established.
 
-3. **Settings endpoint lives in a new `self` domain** (W7, revised per kaseywright's first-round feedback): `GET/PUT /self/settings`, guarded by `authenticateUser` alone (no `{userId}` in the path, so nothing to misuse and no `requireSelf`/`UserPolicy` middleware), full-replace upsert of one blob, file quartet `domains/self/settings/self-settings.{route,service,repository,types}.ts`. Last-writer-wins concurrency inherited from editor-state and accepted.
+3. **Settings endpoint mirrors the editor-state idiom one level up** (W7): `GET/PUT /users/settings`, session-implicit user (never in the URL), full-replace upsert of one blob, file quartet `domains/users/settings/user-settings.{route,service,repository,types}.ts`. Last-writer-wins concurrency inherited from editor-state and accepted.
 
 4. **Chapter-wide check on every successful verse auto-save** (W3/W4): a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)` sending all drafted verses; `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM, the smoke-test convention); no extra debounce beyond the existing 2 s save debounce. Drafting mode only (W10).
 
-5. **Three-layer active/inactive cascade, most-specific-non-silent-wins** (W5/W6): Greek Room's `legitimate` verdict → user-global word-pair rule → occurrence rule `(snt_id, repeated_word, ordinal)`; rules are tri-state maps (`absent/'suppress'/'surface'`), which is what makes per-occurrence **undo** of a global rule or machine verdict possible. The dot counts active findings only; ignored items are hidden by default and revealed dimmed (with "Default Ignore"/"Ignore Here"/"Ignore Always" reason labels and "Undo Ignore") under a **"Show Ignored" toggle** (default off, not persisted), matching the revised #278.
+5. **Three-layer active/inactive cascade, most-specific-non-silent-wins** (W5/W6): Greek Room's `legitimate` verdict → user-global word-pair rule → occurrence rule `(snt_id, repeated_word, ordinal)`; rules are tri-state maps (`absent/'suppress'/'surface'`), which is what makes per-occurrence **undo** of a global rule or machine verdict possible. The dot counts active findings only; ignored items grey-and-stay with reason labels and an `[Undo ▾]` split button.
 
-6. **Graceful degradation, either-PR-lands-first** (W8/W9): the UI feature-detects `GET /self/settings` (404 ⇒ "Ignore Everywhere" simply not rendered — capability hidden, never a dead control); check failures surface as one inline red line in the panel while the last successful findings stay rendered (TanStack default). No toasts/banners — none exist in this codebase. **"Ignore Everywhere" shows a confirm dialog** (per revised #278) while still being reversible via undo.
+6. **Graceful degradation, either-PR-lands-first** (W8/W9): the UI feature-detects `GET /users/settings` (404 ⇒ "Ignore Always" simply not rendered — capability hidden, never a dead control); check failures surface as one inline red line in the panel while the last successful findings stay rendered (TanStack default). No toasts/banners — none exist in this codebase.
 
 ## Explicitly out of scope (deferred)
 
 "Drop duplicate" quick-fix and Greek Room feedback loop (excluded by card #278); checks in review/read-only stages; a "Manage ignored words" settings page (the store is shaped for it); async/polling mode; any change to fluent-ai or the proxy contract; sibling checks.
 
-## Resolved in the first review round
-
-- **Zero-state dot** and **language dropdown** mock issues are fixed — chadw corrected the zero-state mock and removed the dropdown from the #278 mocks.
-- **Button labels** are now **"Ignore Here" / "Ignore Everywhere"** (joelthe1/Ulf).
-- **Ignored items** are revealed dimmed under a **"Show Ignored" toggle** (default off, not persisted) with "Undo Ignore" — the revised #278 added the toggle, largely aligning with our earlier proposal.
-- **Settings endpoint** moved to a new **`self` domain** (`GET/PUT /self/settings`, `authenticateUser` only) per kaseywright.
-
 ## Areas where input would be most valuable
 
-**Product sign-off (§5.3 / §12 of the proposal):**
+**Product sign-off (mock/card deviations, §5.2–§5.3 / §12 of the proposal):**
 
-1. **"Default Ignore" items.** Greek Room sometimes reports a repeated word but marks it `legitimate: true` — i.e. it detected the repetition but judged it intentional/correct (e.g. a word genuinely repeated for emphasis), so it shouldn't be flagged. We render these as ignored items labeled **"Default Ignore"** (the revised #278 mock's term), still surfaceable per-occurrence in case the checker is wrong. Confirm that "Default Ignore" = "repetition Greek Room found but considers legitimate."
-2. Dot mirrored on the panel-toggle button when the left panel is closed, preserving #277's intent.
-3. **"Ignore Everywhere" confirm dialog + undo together:** the revised #278 dialog says the action "cannot be undone," but we keep the `[Undo ▾]` reversibility (the dialog guards against accidental clicks; undo handles deliberate reversals). Confirm we may retain undo.
-4. **Highlight repeated words in the verse text** (joelthe1/Ulf) ships as a **separate follow-on PR** layered on this one — confirm the sequencing.
+1. Zero-state mock shows the dot but #277's text says no flags ⇒ no dot — we follow the text.
+2. The Resources language dropdown appears above the checks content in the #278 mocks with no function there — we propose omitting it.
+3. Ignored items **grey-and-stay with undo** instead of the card's "removed from the panel" (dot behavior unchanged; a "Show ignored & OK" toggle gives the card-literal view).
+4. Machine-`legitimate` findings shown as inactive items ("Marked OK by Greek Room") — cards are silent on them.
+5. Dot mirrored on the panel-toggle button when the left panel is closed, preserving #277's intent.
 
 **Engineering confirmations:**
 
-1. **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and the new **`self` domain** (`GET/PUT /self/settings`) as its endpoint (this outlives the feature).
+1. **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and `GET/PUT /users/settings` as its endpoint (this outlives the feature).
 2. **W4** — `(snt_id, repeated_word, ordinal)` occurrence identity and NFC-exact (no case folding) key comparison.
-3. **W5/W6** — the tri-state cascade and the confirm-dialog-plus-`[Undo ▾]` behavior.
+3. **W5/W6** — the tri-state cascade and the `[Undo ▾]` split-button behavior.

From 14289b914c0c18485a647186ce85311b7988df9f Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Tue, 16 Jun 2026 16:32:39 -0400
Subject: [PATCH 05/33] Revise Repeated Word Check UI proposal per PR #305
 review
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Move settings endpoint to a new self domain (GET/PUT /self/settings,
  authenticateUser only, no userId in path) per kaseywright
- Rename actions to 'Ignore Here' / 'Ignore Everywhere' (joelthe1/Ulf)
- Follow revised #278: 'Show Ignored' toggle (default off, not persisted),
  'Undo Ignore', Ignore Here/Always/Default Ignore labels
- Add confirm dialog on 'Ignore Everywhere' while keeping undo reversibility
- Drop the showResolvedChecks editor-state key (only activeLeftTab persists)
- Mark S1 (zero-state dot) and S2 (language dropdown) resolved; add S7/S8
- Document highlight-in-verse as a follow-on PR
- Add §0 change log
---
 .../checks-ui-integration-suggestion.md       | 156 +++++++++++-------
 .../checks-ui-integration-summary.md          |  36 ++--
 2 files changed, 115 insertions(+), 77 deletions(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index e20754e6..260b5c24 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -1,6 +1,6 @@
 # Repeated Word Check UI — Proposal (Checks Tab, Checks Panel, and Suppression Persistence)
 
-**Status:** Draft for review.
+**Status:** Revised after first review round (PR [#305](https://github.com/eten-tech-foundation/fluent-web/pull/305)). Incorporates reviewer feedback — see the change log in §0.
 **Reviewer shortcut:** a condensed, stands-on-its-own review summary lives in [`checks-ui-integration-summary.md`](checks-ui-integration-summary.md).
 **Scope:** Implements the user-facing half of the Repeated Word Check feature: the **Checks tab** (#277) and **Checks view panel** (#278) on fluent-web, plus the small fluent-api additions both cards imply (suppression persistence). This is a single proposal covering both repos so reviewers see the whole design in one place; **implementation will ship as two PRs** (one per repo), and either may land first (§9.4).
 
@@ -20,6 +20,21 @@ This document's own decisions are numbered **W1–W12** (W = web) to avoid colli
 
 ---
 
+## 0. Change log (first review round)
+
+Revisions made in response to PR [#305](https://github.com/eten-tech-foundation/fluent-web/pull/305) review (reviewers: kaseywright, chadw-eten; product input via joelthe1/Ulf). The cards themselves were edited mid-review; this revision tracks the **current** card text and mocks (re-pulled 2026-06-16).
+
+| Change | Driver | Where |
+| --- | --- | --- |
+| Settings endpoint moves from `GET/PUT /users/settings` to a new **`self` domain**: `GET/PUT /self/settings`, guarded by `authenticateUser` alone. | kaseywright (CHANGES_REQUESTED): clean semantic separation, no redundant `{id}` param, no over-exposure risk, simpler middleware (no `requireSelf`/`requireUserAccess`/`UserPolicy`). | W7, §7.2, §8 |
+| Button labels become **"Ignore Here" / "Ignore Everywhere"**. | joelthe1/Ulf (card #278 comment); chadw updated the #278 mocks to match. | W6, §5, §6.5 |
+| **"Ignore Everywhere" now shows a confirmation dialog**, while remaining reversible via `[Undo ▾]`. | card #278 ("Ignore Always should have a standard confirm dialog"). Kept the three-layer cascade & undo (no explicit directive against it; ambiguity of "undo" with surface/suppress entries at multiple layers is better resolved by the cascade — see §6.4/§6.5); added the dialog. | W6, §5.3, §6.5 |
+| **"Show Ignored" toggle**: now default **OFF** and **not persisted** (resets per session). Ignored items show the ignore-type label (e.g. _"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) + an **"Undo Ignore"** control. | card #278 was revised to specify exactly this; we drop our earlier default-ON/persisted proposal and the `showResolvedChecks` editor-state key. | W5, W6, W11, §5.3, §6.4, §6.6, §7.1 |
+| §5.2 mock/text inconsistencies **S1 (zero-state dot)** and **S2 (language dropdown)** are **resolved**: chadw fixed the zero-state mock and removed the language dropdown from the #278 Checks mocks. | chadw-eten ("Fixed." / "good idea… I have adjusted the mocks."). | §5.2, §12 |
+| New requirement adopted: **highlight repeated words in the verse text** (underline/color). | joelthe1/Ulf (card #278 comment). We will implement it; because it touches the verse-rendering pane (outside the left panel this proposal builds) and warrants its own design pass, it ships as a **separate, follow-on PR** so the tab/panel work can land independently. §5.3 item 4 records the intent; the visual/interaction details are specced just before that PR. | §5.3, §11 |
+
+---
+
 ## 1. Background
 
 fluent-api now exposes Greek Room's _Repeated Words_ check at `POST /ai/tools/greek-room/repeated-words` (card #172). The endpoint is a thin authenticated proxy to fluent-ai: it accepts a chapter's verses and returns a `ToolJobResponse[RepeatedWordsResult]` envelope whose `findings[]` identify consecutive repeated words (`{snt_id, repeated_word, surf, start_position, legitimate, severity}`).
@@ -27,9 +42,11 @@ fluent-api now exposes Greek Room's _Repeated Words_ check at `POST /ai/tools/gr
 Cards #277 and #278 define how translators consume those findings in the drafting view:
 
 - **#277:** a **Checks tab** is added to the drafting page's left panel alongside the existing Resources tab, with a **notification dot** on the tab header whenever the current chapter has one or more active flags. The dot is absent at zero flags and clears silently on the next auto-save once all flagged issues are resolved.
-- **#278:** the **Checks panel** lists all repeated-word flags for the current chapter, grouped by verse, each with enough context to locate it, refreshed on every auto-save, with two actions per flagged occurrence: **"Ignore This Time"** (user-level, this occurrence) and **"Ignore Always"** (user-level, this word pair, across all the user's projects) — both "stored in Fluent's database." A "No issues found" zero state is shown when the chapter is clean.
+- **#278:** the **Checks panel** lists all repeated-word flags for the current chapter, grouped by verse, each with enough context to locate it, refreshed on every auto-save, with two actions per flagged occurrence: **"Ignore Here"** (user-level, this occurrence) and **"Ignore Everywhere"** (user-level, this word pair, across all the user's projects) — both "stored in Fluent's database." "Ignore Everywhere" carries a standard confirm dialog. A **"Show Ignored"** toggle (default off, not persisted) reveals previously ignored occurrences — dimmed, labeled by ignore type, each with an **"Undo Ignore"** action. A "No issues found" zero state is shown when the chapter is clean.
+
+> The #278 card text and mocks were revised during the first review round (labels, the Show Ignored toggle, the confirm dialog). The wording above reflects the **current** card; the original used "Ignore This Time"/"Ignore Always". See §0.
 
-Both cards carry mockups (drafting page, Judges 4, Gujarati IRV); §5 transcribes them and flags two mock/text inconsistencies for product sign-off.
+Both cards carry mockups (drafting page, Judges 4, Gujarati IRV); §5 transcribes them.
 
 ### 1.1 What the cards imply beyond fluent-web
 
@@ -40,7 +57,7 @@ The ignore actions require server-side persistence that does not exist yet. #172
 | Repo | Touched | What changes |
 | --- | --- | --- |
 | **fluent-web** | yes (bulk) | Left-panel tab container, Checks panel feature, check-trigger hook, suppression cascade, editor-state keys. |
-| **fluent-api** | yes (small) | New `user_settings` table + `GET/PUT /users/settings`; Zod-schema extension of the editor-state `resources` blob (no migration). |
+| **fluent-api** | yes (small) | New `user_settings` table + a new `self` domain exposing `GET/PUT /self/settings`; Zod-schema extension of the editor-state `resources` blob (no migration). |
 | **fluent-ai** | no | The check itself is unchanged. |
 | **fluent-platform** | no | No new services or env vars. |
 
@@ -55,7 +72,7 @@ The ignore actions require server-side persistence that does not exist yet. #172
 3. The **notification dot** (#277) on the Checks tab header, plus one proposed deviation: mirroring the dot on the panel-toggle button when the panel is closed (§5.3, sign-off item).
 4. A chapter-wide **check trigger** fired on every successful verse auto-save (W3).
 5. A three-layer **active/inactive cascade** unifying Greek Room's `legitimate` verdicts with user ignores, including undo (W5, W6).
-6. **Persistence:** occurrence-level rules in the existing editor-state JSONB; global word-pair rules in a new `user_settings` table exposed at `GET/PUT /users/settings` (W2, W7).
+6. **Persistence:** occurrence-level rules in the existing editor-state JSONB; global word-pair rules in a new `user_settings` table exposed at `GET/PUT /self/settings` (W2, W7).
 7. **Graceful degradation** when the settings backend half is absent, and inline error surfacing when the check call fails (W8, W9).
 8. Tests on both sides using each repo's established test infrastructure (§10).
 
@@ -81,13 +98,13 @@ Restated conclusions; supporting analysis in the cited sections.
 | **W2** | Hybrid storage: "Ignore This Time" lives in the existing per-chapter editor-state JSONB (schema-key addition, no migration); "Ignore Always" lives in a new `user_settings` table (one row per user, single Zod-typed JSONB column). Findings are filtered **client-side**. | Each rule is stored at exactly the scope it governs. Client-side filtering keeps the AI proxy a pure pass-through (D8/D9). `user_settings` deliberately establishes Fluent's user-global preference store: future settings become schema extensions, not migrations. See §7. |
 | **W3** | Chapter-wide check on every successful verse save: a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)`, where `saveCounter` increments in the verse-save mutation's `onSuccess`. The request sends **all currently drafted verses** of the chapter. No extra coalescing debounce. | The dot needs chapter-wide awareness; per-verse merging is more code for no user-visible gain. The existing 2 s per-verse save debounce already rate-limits; the check is <1 s. KISS — optimize if heavier checks arrive. See §6.2. |
 | **W4** | `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM book code, e.g. `JDG 4:3`), matching the convention already used by the repo smoke tests. Occurrence identity for suppressions = `(snt_id, repeated_word, ordinal)` where ordinal is the index of the finding among same-`repeated_word` findings in the verse, ordered by `start_position`. | Ordinals survive unrelated edits (positions don't); adding/removing an earlier same-pair occurrence conservatively re-flags, the safe failure direction. We count Greek Room's findings, never tokenize text ourselves, so Greek Room's equivalence policy is inherited consistently. See §6.3. |
-| **W5** | Every finding resolves to **active/inactive** via a three-layer cascade — Greek Room verdict (`legitimate`), user-global word-pair rule, occurrence rule — where the **most specific non-silent verdict wins**. Rules are tri-state maps (`absent / 'suppress' / 'surface'`). The dot counts active findings only. Inactive findings render greyed with a reason label; a "Show ignored & OK" toggle (default ON, persisted) hides them. | One mechanism explains machine-legitimate and user-ignored alike ("pre-ignored by Greek Room"). Specificity, not temporal order, keeps resolution deterministic. Greyed-not-removed keeps undo discoverable. See §6.4. |
-| **W6** | Active findings show `[Ignore This Time] [Ignore Always]` (per card). Inactive findings show one `[Undo ▾]` split button: default click acts at the occurrence layer (delete own rule, or write an occurrence `surface` override); the chevron menu offers explicit global actions with consequence-naming labels. Any global write first purges the user's occurrence rules for that pair **in the current chapter's editor state only**. No confirmation dialogs (global rules are reversible). | Default click never silently edits global state. Purge-local prevents the just-clicked panel from appearing to ignore the action (occurrence beats global in the cascade); other chapters' specific pronouncements deliberately stand. See §6.5. |
-| **W7** | Settings endpoint mirrors the editor-state idiom one level up: **`GET /users/settings` + `PUT /users/settings`**, session-implicit user (no user in URL), full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/users/settings/user-settings.{route,service,repository,types}.ts`. | `user_settings` is to _user_ what `user_chapter_assignment_editor_state` is to _(user, chapter)_; same shape, same endpoint idiom, same review story ("we did what you already do"). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8. |
-| **W8** | Graceful degradation by feature detection: if `GET /users/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Always]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first. | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9. |
+| **W5** | Every finding resolves to **active/inactive** via a three-layer cascade — Greek Room verdict (`legitimate`), user-global word-pair rule, occurrence rule — where the **most specific non-silent verdict wins**. Rules are tri-state maps (`absent / 'suppress' / 'surface'`). The dot counts active findings only. Inactive findings are hidden by default and revealed by the **"Show Ignored" toggle** (default OFF, not persisted, per revised #278); when shown they render greyed with the ignore-type label. | One mechanism explains machine-legitimate and user-ignored alike ("pre-ignored by Greek Room"). Specificity, not temporal order, keeps resolution deterministic. Greyed-not-removed (under the toggle) keeps undo discoverable. See §6.4. |
+| **W6** | Active findings show `[Ignore Here] [Ignore Everywhere]` (per revised card). `[Ignore Everywhere]` opens a **standard confirm dialog** before writing the global rule. Inactive findings show one `[Undo ▾]` split button: default click acts at the occurrence layer (delete own rule, or write an occurrence `surface` override); the chevron menu offers explicit global actions with consequence-naming labels. Any global write first purges the user's occurrence rules for that pair **in the current chapter's editor state only**. | Default click never silently edits global state. The card asks for a confirm on the global action because it reads as irreversible; we add the dialog **and** keep the `[Undo ▾]` reversibility — the two are complementary (confirm guards the click; undo guards the mistake). Purge-local prevents the just-clicked panel from appearing to ignore the action; other chapters' specific pronouncements deliberately stand. See §6.5. |
+| **W7** | Settings endpoint lives in a **new `self` domain**: **`GET /self/settings` + `PUT /self/settings`**, session-implicit user (no user in URL), guarded by `authenticateUser` alone, full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/self/settings/self-settings.{route,service,repository,types}.ts`. | Per kaseywright: a `self` domain gives clean semantic separation, avoids the redundant `{userId}` param (and the `requireSelf`/`UserPolicy` middleware that `/users/{id}/…` routes carry), and prevents accidental over-exposure (no `{id}` to misuse). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8. |
+| **W8** | Graceful degradation by feature detection: if `GET /self/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Everywhere]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first. | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9. |
 | **W9** | Check-call failure surfaces as a single inline `text-sm text-red-500` line at the top of the Checks panel ("Checks failed to refresh"); the panel keeps rendering the last successful findings (TanStack keeps `query.data` on refetch failure) and the dot reflects that last-known state. Failures log via the existing `Logger`. No toast/banner/popup. | Matches the drafting page's own inline-status precedent ("Auto-save failed"). Failure mode degrades to "results are one save behind." See §9.2. |
 | **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content and while the settings feature-detection probe (§9.1) is unresolved. | Card scope is the drafting view; empty chapters have nothing to check; the probe is one fast `GET`, and waiting for it means the first findings render is already cascade-correct. Review-stage checks are future work. See §6.2. |
-| **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets and the two buttons; zero state inside the accordion. Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5. |
+| **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets, the two buttons, and a "Show Ignored" toggle at the bottom; zero state inside the accordion. Only `activeLeftTab` is persisted in editor state (the "Show Ignored" state resets per session, per revised #278). Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5. |
 | **W12** | One proposal document (this file) covering both repos; two implementation PRs (fluent-web, fluent-api), cross-referencing each other and the cards. | Splitting the proposal doubles reviewer overhead for a design whose halves only make sense together. See §9.4. |
 
 ---
@@ -122,26 +139,28 @@ The response envelope is consumed whole (D9): the hook inspects `status` and `re
 
 ### 5.1 What the mocks show
 
-The cards' mockups (drafting page, Judges 4, Gujarati IRV project) define the visual target:
+The cards' mockups (drafting page, Judges 4, Gujarati IRV project) define the visual target. The #278 mocks were updated during the first review round (§0); the description below tracks the **current** mocks:
 
 - **Tab header (#277 mock):** the left panel's current `Resources` heading is replaced by a text-tab row — **"Resources | Checks"** — active tab in blue with a blue underline, inactive tab plain. The **notification dot is a solid blue filled circle immediately right of the "Checks" label**, and the mock shows it while the *Resources* tab is active: the dot is visible from either tab whenever the panel is open.
 - **Checks content (#278 mocks):** below the tab header, a card containing a **collapsible "Repeated Words" accordion section** (expanded by default). The accordion-per-check structure anticipates future sibling checks (the board already holds draft cards for Greek Room Wildebeest and Spell checks) without UI rework. Inside the section, findings are grouped by verse:
   - bold **"Verse N"** heading per verse that has findings (verses without findings get no group);
   - one row per occurrence: a one-line context snippet showing the repeated pair in situ;
-  - two solid-blue buttons side by side beneath each snippet: `[Ignore This Time] [Ignore Always]`;
-  - a horizontal separator between verse groups.
+  - two solid-blue buttons side by side beneath each snippet: `[Ignore Here] [Ignore Everywhere]`;
+  - a horizontal separator between verse groups;
+  - a **"Show Ignored" toggle** at the bottom of the active list (off by default). When on, previously ignored occurrences appear below the active ones — dimmed, each labeled with its ignore type (_"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) and showing an **"Undo Ignore"** action in place of the two ignore buttons.
 - **Zero state (#278 mock):** the "Repeated Words" section shows a centered bold **"No issues found"**.
 
-### 5.2 Mock/text inconsistencies — for product sign-off
+### 5.2 Mock/text inconsistencies — resolved in first review round
 
-1. **Dot in the zero state.** The #278 zero-state mock still shows the dot on the Checks tab, contradicting #277's text ("the dot is absent when there are no active flags — no zero state"). **We follow the text:** no flags, no dot. Confirm.
-2. **Language dropdown above the Checks panel.** Both #278 mocks show the Resources tab's "English" language dropdown above the checks content, with no defined function for checks (the check runs against the chapter's target language; there is nothing to select). We presume this is a carryover artifact of editing a Resources-tab screenshot and **propose omitting the dropdown from the Checks tab**. Confirm.
+1. **Dot in the zero state.** The original #278 zero-state mock showed the dot on the Checks tab, contradicting #277's text ("the dot is absent when there are no active flags — no zero state"). We proposed following the text; chadw fixed the mock. **Resolved:** no flags, no dot.
+2. **Language dropdown above the Checks panel.** The original #278 mocks showed a language dropdown above the checks content, with no function for checks. We proposed omitting it; chadw agreed and removed it from the mocks. **Resolved:** no dropdown on the Checks tab.
 
 ### 5.3 Deviations from the card text — for product sign-off
 
-1. **Ignored items grey-and-stay rather than disappear.** #278 says an ignored occurrence "is removed from the panel." We instead render it greyed/subordinate, sorted after active items within its verse group, labeled with the winning rule (e.g. _"Ignored always: 'the the'"_), with an `[Undo ▾]` control (§6.5). Rationale: removal makes a mis-click irreversible-in-place and makes "Ignore Always" indistinguishable from a fix; the card's intent (flag stops demanding attention) is preserved because greyed items don't count toward the dot. A **"Show ignored & OK" checkbox** (default ON, persisted in editor state) collapses them entirely for translators who want the card-literal view.
-2. **Machine-"legitimate" findings are shown as inactive items.** Greek Room marks some repetitions `legitimate: true` (intentionally repeated words). The cards are silent on them. We render them in the same greyed style with reason _"Marked OK by Greek Room"_, surfaceable per-occurrence like any other inactive item (§6.4). This gives translators visibility into what the checker considered and a recourse when Greek Room is wrong in either direction.
+1. **Ignored items grey-and-stay rather than disappear.** The card's "Show Ignored" toggle (revised #278) now matches our approach: ignored occurrences are revealed dimmed/subordinate, sorted after active items within their verse group, labeled with the ignore type, with an "Undo Ignore" control (§6.5). We follow the card: the toggle defaults **off** and does **not** persist across sessions. Greyed items never count toward the dot. (This deviation is now largely card-aligned and is listed for completeness.)
+2. **Machine-"legitimate" findings are shown as inactive items.** Greek Room marks some repetitions `legitimate: true` (intentionally repeated words). The cards do not enumerate them explicitly, but the revised #278 "Show Ignored" mock includes a **"Default Ignore"** ignore-type — which is exactly this case. We render machine-legitimate findings in the same dimmed style with that label, surfaceable per-occurrence like any other ignored item (§6.4). This gives translators visibility into what the checker considered and a recourse when Greek Room is wrong in either direction.
 3. **Dot mirrored on the panel-toggle button.** #277 places the dot on the Checks tab header, which is invisible when the whole left panel is closed (it's behind the `BookText` toggle button in the drafting header). To preserve the card's intent — the translator is notified — we mirror the dot on that toggle button whenever the panel is closed and active flags exist.
+4. **Highlight repeated words in the verse text.** Product (joelthe1/Ulf, via #278 comment) asked that flagged repeated words be highlighted in the verse pane (underline or color). We will add this. Because it touches the verse-rendering pane — outside the left-panel tab/panel this proposal builds — and benefits from its own short design pass, it ships as a **follow-on PR** layered on this one, so the tab/panel work can be reviewed and merged independently. The visual/interaction specifics (underline vs. highlight, active-vs-ignored styling, click-to-locate) are worked out just before that PR.
 
 None of these block implementation; they are listed in §12 for explicit confirmation in PR review.
 
@@ -165,9 +184,10 @@ fluent-web/src/
 │       │   ├── LeftPanel.tsx                # Tab container: header row (Resources | Checks + dot),
 │       │   │                                #   renders ResourcePanel or ChecksPanel
 │       │   ├── ChecksPanel.tsx              # Per-check accordion; verse groups; zero state;
-│       │   │                                #   inline error line; "Show ignored & OK" toggle
-│       │   └── FindingRow.tsx               # Snippet + actions (active: two buttons; inactive:
-│       │                                    #   greyed + reason label + Undo split button)
+│       │   │                                #   inline error line; "Show Ignored" toggle
+│       │   └── FindingRow.tsx               # Snippet + actions (active: Ignore Here / Ignore
+│       │                                    #   Everywhere [+ confirm dialog]; inactive: greyed +
+│       │                                    #   ignore-type label + Undo split button)
 │       ├── hooks/
 │       │   ├── useRepeatedWordsCheck.ts     # useQuery keyed (chapterAssignmentId, saveCounter);
 │       │   │                                #   builds request; returns raw findings
@@ -225,16 +245,17 @@ Every finding resolves through three layers; the **most specific non-silent verd
 - Stored rules are **tri-state maps** (`key → 'suppress' | 'surface'`, absent = silent), not bare suppression sets — `surface` is what makes per-occurrence undo of a global rule (or of a Greek Room `legitimate` verdict) possible.
 - A machine-legitimate finding is conceptually "pre-ignored by Greek Room": same inactive state as user-ignored, different actor, same undo affordance.
 - **The notification dot counts cascade-resolved active findings only.**
-- Inactive findings render greyed, sorted after active items within their verse group, each labeled with the winning rule (_"Ignored this time"_, _"Ignored always: 'અને અને'"_, _"Marked OK by Greek Room"_). The **"Show ignored & OK"** checkbox (default ON so the undo affordance is discoverable; persisted in editor state) hides them.
+- Inactive findings render greyed, sorted after active items within their verse group, each labeled with its ignore type — the labels the revised #278 mock uses: _"Ignore Here"_ (occurrence suppress), _"Ignore Everywhere"_ (global suppress), _"Default Ignore"_ (Greek Room `legitimate`). The **"Show Ignored"** toggle (default OFF, **not** persisted — resets each session, per revised #278) reveals them; with the toggle off, only active findings show.
 
 ### 6.5 Action surface (W6)
 
-- **Active finding** (per card #278): `[Ignore This Time]` → writes occurrence rule `suppress`; `[Ignore Always]` → writes global pair rule `suppress`.
+- **Active finding** (per revised card #278): `[Ignore Here]` → writes occurrence rule `suppress`; `[Ignore Everywhere]` → opens a **confirm dialog** ("Fluent will not flag this word pair again across all your projects. Are you sure?") and, on confirm, writes the global pair rule `suppress`.
+- **Confirm dialog on `[Ignore Everywhere]`** (revised #278): the card asks for it because the action reads as global and irreversible. We honor the request **and** keep the action reversible via `[Undo ▾]` below — the dialog guards the deliberate click; the undo guards a mistake that slips past it. The two are not redundant: confirm is pre-action friction, undo is post-action recourse. (We considered dropping undo to match a strict "cannot be undone" reading, but the cascade already supports clean reversal at the right scope, and there is no explicit directive to remove it; the final PR is reviewed again before merge.)
 - **Inactive finding:** one `[Undo ▾]` split button (shadcn `dropdown-menu`, already in the codebase — no hidden gestures):
   - **Default click** acts at the occurrence layer: if inactive via its *own* occurrence rule → delete that entry; if inactive via a global rule or Greek Room's verdict → write an occurrence-level `surface` override (the global rule / machine verdict survives; only this occurrence resurfaces).
   - **Chevron menu** offers the deliberate global actions with consequence-naming labels, e.g. _"Stop ignoring 'the the' everywhere"_ (deletes the global entry).
-- **Global writes purge local rules — current chapter only.** Executing any global (`always`) suppress or surface first deletes the user's occurrence-level rules for that word pair in the *current chapter assignment's* editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
-- **No confirmation dialogs:** global rules are reversible via the chevron (and a future settings page, §11), so `[Ignore Always]` does not nag.
+- **Global writes purge local rules — current chapter only.** Executing any global (everywhere) suppress or surface first deletes the user's occurrence-level rules for that word pair in the *current chapter assignment's* editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
+- **Why undo despite the "cannot be undone" wording.** The card's confirm-dialog copy ("cannot be undone") is product framing for the *commitment* of a global ignore, not a storage constraint. With the three-layer cascade, "undo" is well-defined at each scope (delete the global entry, or `surface`-override one occurrence), so we surface it rather than stranding the translator. This is the one place we keep more capability than the literal card; flagged in §12 for sign-off.
 - **Direct edits resolve silently** (card #278): if the translator fixes the text, the next auto-save's fresh findings simply no longer contain the occurrence — no action needed, panel and dot update.
 
 ### 6.6 Left-panel container and persisted UI state (W11)
@@ -243,7 +264,7 @@ Every finding resolves through three layers; the **most specific non-silent verd
 
 - **`LeftPanel`** owns the tab header row ("Resources | Checks", blue underline on the active tab, blue dot right of "Checks" when active findings exist) and renders either the existing `ResourcePanel` or the new `ChecksPanel` below it. `ResourcePanel` loses only its `<h3>Resources</h3>` heading (the tab row replaces it); its content, language dropdown, and accordion are untouched.
 - The dot's state comes from the cascade-resolved active count (§6.4), which `DraftingPage` computes once and threads to both `LeftPanel` (tab dot) and the header toggle button (mirror dot, §5.3 item 3). The check query runs at `DraftingPage` level — **not** inside `ChecksPanel` — so the dot stays live while the Resources tab (or a closed panel) is showing.
-- **Persisted UI state.** Two new keys ride the same per-chapter editor-state blob the page already saves (debounced 500 ms, §7.1): `activeLeftTab: 'resources' | 'checks'` (reopen where you left off) and `showResolvedChecks: boolean` (the "Show ignored & OK" checkbox, default `true`). Both are cosmetic; failure to persist degrades to defaults.
+- **Persisted UI state.** One new key rides the same per-chapter editor-state blob the page already saves (debounced 500 ms, §7.1): `activeLeftTab: 'resources' | 'checks'` (reopen where you left off). The **"Show Ignored"** toggle is deliberately **not** persisted — per revised #278 it resets to off each session — so it lives in component state only. The persisted key is cosmetic; failure to persist degrades to defaults.
 - The panel's existing resize/drag behavior and 20–40 % width constraints are inherited unchanged — `LeftPanel` lives *inside* the resizable container.
 
 ---
@@ -283,7 +304,6 @@ export const editorStateResourcesSchema = z
     tabStatus: z.boolean(),
     // --- NEW, all optional (backward/forward compatible) ---
     activeLeftTab: z.enum(['resources', 'checks']).optional(),
-    showResolvedChecks: z.boolean().optional(),
     checkOccurrenceRules: z
       .record(z.string(), z.enum(['suppress', 'surface']))
       .optional(),
@@ -291,7 +311,7 @@ export const editorStateResourcesSchema = z
   .nullable();
 ```
 
-- `checkOccurrenceRules` is the tri-state map of §6.4 layer 2. Keys are the occurrence identity `"{snt_id}|{repeated_word}|{ordinal}"` (e.g. `"JDG 4:3|અને અને|1"`); the `|` separator cannot appear in a `snt_id` and `repeated_word` is the final segment-pair, so keys are unambiguous. Values: `'suppress'` ("Ignore This Time") or `'surface'` (per-occurrence undo of a global/machine verdict). Absent = silent.
+- `checkOccurrenceRules` is the tri-state map of §6.4 layer 2. Keys are the occurrence identity `"{snt_id}|{repeated_word}|{ordinal}"` (e.g. `"JDG 4:3|અને અને|1"`); the `|` separator cannot appear in a `snt_id` and `repeated_word` is the final segment-pair, so keys are unambiguous. Values: `'suppress'` ("Ignore Here") or `'surface'` (per-occurrence undo of a global/machine verdict). Absent = silent.
 - fluent-web reads/writes these through the **existing** `GET/PUT /chapter-assignments/:id/editor-state` round-trip it already performs — the same debounced save that persists `activeResource` today carries the new keys. No new endpoint for occurrence rules.
 - Write pattern: read-modify-write of the blob the page already holds in memory. The page is the only writer for this `(user, chapterAssignment)` pair in practice (concurrent same-user tabs are a pre-existing last-writer-wins, inherited).
 
@@ -322,9 +342,9 @@ export const user_settings = pgTable('user_settings', {
 ```
 
 - **One row per user** (`userId` is the primary key), one Zod-typed JSONB `settings` column — the same shape discipline as `user_chapter_assignment_editor_state`, one level up the scope ladder.
-- `checkIgnoredWordPairs` is the §6.4 layer-1 map. Keys are the NFC-normalized `repeated_word` string (the pair, e.g. `"the the"`); values `'suppress'` ("Ignore Always") or `'surface'` (reserved — a global "never let Greek Room auto-OK this pair" is expressible but no v1 UI writes it). Absent = silent.
+- `checkIgnoredWordPairs` is the §6.4 layer-1 map. Keys are the NFC-normalized `repeated_word` string (the pair, e.g. `"the the"`); values `'suppress'` ("Ignore Everywhere") or `'surface'` (reserved — a global "never let Greek Room auto-OK this pair" is expressible but no v1 UI writes it). Absent = silent.
 - Requires **one SQL migration** (new table; follows the numbered-migration convention in `src/db/migrations/`).
-- Exposed via `GET/PUT /users/settings` (§8).
+- Exposed via `GET/PUT /self/settings` (§8).
 
 ### 7.3 Why client-side filtering
 
@@ -338,32 +358,39 @@ The proxy therefore remains untouched by this proposal.
 
 ---
 
-## 8. fluent-api design: `GET/PUT /users/settings` (W7)
+## 8. fluent-api design: `GET/PUT /self/settings` (W7)
+
+The settings endpoint lives in a **new `self` domain**, requested by kaseywright in review. `self` is the authenticated caller acting on their own resources — the route never names a user. This is cleaner than hanging settings off `/users/{userId}/…`:
 
-The settings endpoint deliberately mirrors the editor-state endpoint one level up the scope ladder — `user_settings` is to *user* what `user_chapter_assignment_editor_state` is to *(user, chapterAssignment)*. Same idiom throughout: **session-implicit user** (the authenticated user is taken from context, never from the URL — consistent with `user-chapter-assignment-editor-state.route.ts`, where no `userId` appears in the path), **full-replace upsert** on PUT, one Zod-typed JSONB blob.
+- **Clean semantic separation** — `self/*` is unambiguously "me," leaving the `users/*` domain for (future) admin-style tooling that operates on *other* users by id.
+- **No redundant `{userId}` param** — the user comes from the session, not the URL (the same source `user-chapter-assignment-editor-state.route.ts` already uses).
+- **No accidental over-exposure** — with no `{id}` in the path there is nothing to mis-scope; contrast `/users/{userId}/projects`, which must add `requireSelf()` to stop one user reading another's data.
+- **Simpler middleware** — just `authenticateUser`. No `requireSelf`, no `requireUserAccess`, no `UserPolicy`.
+
+Otherwise it keeps the editor-state idiom: **full-replace upsert** on PUT, one Zod-typed JSONB blob, last-writer-wins on concurrent writes.
 
 ### 8.1 Routes
 
 | Method | Path | Auth | Body | Response |
 | --- | --- | --- | --- | --- |
-| `GET` | `/users/settings` | session (any authenticated user) | — | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
-| `PUT` | `/users/settings` | session (any authenticated user) | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation |
+| `GET` | `/self/settings` | `authenticateUser` only | — | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
+| `PUT` | `/self/settings` | `authenticateUser` only | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation |
 
-- **No special permission** beyond an authenticated session: every user owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id`. (Editor-state additionally guards with `CONTENT_UPDATE` + chapter-assignment participation because it is tied to a chapter resource; settings have no resource to scope, so session identity alone is the right guard.)
+- **No permission middleware beyond `authenticateUser`.** Every caller owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id` from context. (Editor-state additionally guards with `CONTENT_UPDATE` + chapter-assignment participation because it is tied to a chapter resource; `self/settings` has no resource to scope, so session identity alone is the right guard — and the `self` prefix makes that explicit.)
 - **PUT is full-replace.** The client GETs, merges in memory, PUTs the whole blob — the editor-state write pattern. Concurrent tabs are last-writer-wins, identical to editor-state's accepted behavior; for an ignore-list this worst case is "one ignore from another tab is lost until re-clicked."
-- Rejected alternatives, for the record: `/user-settings` (breaks the `domains/users` nesting), `/users/me/settings` (no `me` convention exists in this API), and per-feature endpoints like `/users/settings/check-ignores` (the blob is one document; sub-paths invite partial-update semantics we don't need).
+- Rejected alternatives, for the record: `/users/settings` and `/users/{userId}/settings` (the original proposal — kaseywright preferred a dedicated `self` domain over loading another route into `users`, see above), `/users/me/settings` (no `me` convention exists in this API), and per-feature endpoints like `/self/settings/check-ignores` (the blob is one document; sub-paths invite partial-update semantics we don't need).
 
 ### 8.2 File layout
 
 ```
-fluent-api/src/domains/users/settings/
-├── user-settings.route.ts        # GET + PUT, OpenAPI-described, session user from context
-├── user-settings.service.ts      # get / upsert, Result-typed like editor-state service
-├── user-settings.repository.ts   # drizzle upsert on userId PK (onConflictDoUpdate)
-└── user-settings.types.ts        # response schema; re-exports UserSettings from db schema
+fluent-api/src/domains/self/settings/
+├── self-settings.route.ts        # GET + PUT, OpenAPI-described, session user from context
+├── self-settings.service.ts      # get / upsert, Result-typed like editor-state service
+├── self-settings.repository.ts   # drizzle upsert on userId PK (onConflictDoUpdate)
+└── self-settings.types.ts        # response schema; re-exports UserSettings from db schema
 ```
 
-Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§7.2), one numbered migration in `src/db/migrations/`, and route registration in `src/app.ts` next to the existing users routes.
+Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§7.2), one numbered migration in `src/db/migrations/`, and route registration in `src/app.ts`. (The table is still named `user_settings` — it stores per-user rows; only the *route* surface is `self`. The repository upserts on the `userId` PK using the session user's id.)
 
 ### 8.3 What fluent-api explicitly does **not** do
 
@@ -377,10 +404,10 @@ Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§
 
 ### 9.1 Feature detection for the settings half (W8)
 
-The two PRs (fluent-web, fluent-api) may land and deploy in either order, so the UI must not assume `GET /users/settings` exists:
+The two PRs (fluent-web, fluent-api) may land and deploy in either order, so the UI must not assume `GET /self/settings` exists:
 
-- On drafting-page mount, `useSuppressions` probes `GET /users/settings` once per session. **`404` ⇒ `globalIgnoresAvailable = false`** for the session (the route doesn't exist yet on this deployment); any 2xx — including `settings: null` — ⇒ available.
-- When unavailable, the **capability is hidden, not disabled**: `[Ignore Always]` is simply not rendered, and the `[Undo ▾]` chevron omits global entries. A dead button is a bug report; an absent one is honest. Occurrence-level ignores (editor-state-backed) work regardless.
+- On drafting-page mount, `useSuppressions` probes `GET /self/settings` once per session. **`404` ⇒ `globalIgnoresAvailable = false`** for the session (the route doesn't exist yet on this deployment); any 2xx — including `settings: null` — ⇒ available.
+- When unavailable, the **capability is hidden, not disabled**: `[Ignore Everywhere]` is simply not rendered, and the `[Undo ▾]` chevron omits global entries. A dead button is a bug report; an absent one is honest. Occurrence-level ignores (editor-state-backed) work regardless.
 - The check query waits for the probe to resolve before its first run (W10, §6.2) — one fast `GET`, so findings never render ahead of the user's global rules. The probe resolves on any terminal response: 2xx ⇒ available, `404` ⇒ unavailable; a network failure resolves it as unavailable for the session (conservative: occurrence ignores only).
 - **Unknown-key tolerance both ways:** `userSettingsSchema` uses `.catch({})` server-side; client-side parsing treats absent/unknown keys as empty. A newer client against an older blob (or vice versa) degrades to "no global rules," never to an error. The same applies to the editor-state blob's new optional keys (§7.1) — old clients ignore them, old rows omit them.
 
@@ -403,7 +430,7 @@ Ignore actions apply **optimistically** (the finding greys immediately) and roll
 ### 9.4 Rollout / landing order (W12)
 
 - **Two PRs**, one per repo, each referencing this proposal and cards #277/#278 (and #172 for the fluent-api half). Either lands first:
-  - fluent-web first ⇒ probe 404s ⇒ occurrence ignores only, no `[Ignore Always]` (§9.1). Fully usable.
+  - fluent-web first ⇒ probe 404s ⇒ occurrence ignores only, no `[Ignore Everywhere]` (§9.1). Fully usable.
   - fluent-api first ⇒ new table/endpoint sits unused. Harmless.
 - The editor-state schema-key extension (§7.1) is backward/forward compatible by construction (optional keys, JSONB column unchanged), so no coordination is needed there either.
 - No new env vars, services, or fluent-platform changes; the feature follows the existing deploy pipeline of each repo.
@@ -421,16 +448,16 @@ The repo now ships test infrastructure under `src/test/` (`msw/server.ts`, `rend
 | `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC composed/decomposed key equivalence. |
 | `useRepeatedWordsCheck` | Fires on `saveCounter` change; request body shape (snake_case verbatim, `snt_id` format, all drafted verses); `enabled` gating (readOnly, empty chapter); previous data retained on 502 (MSW error handler). |
 | `useSuppressions` | Feature-detect: 404 ⇒ `globalIgnoresAvailable=false`; 200-with-null ⇒ available; occurrence rule round-trip through editor-state blob; global write purges current-chapter occurrence rules for the pair; optimistic rollback on PUT failure. |
-| `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and *not* on error); inline error line; active buttons vs. greyed-with-reason + `[Undo ▾]`; "Show ignored & OK" toggle; `[Ignore Always]` absent when capability unavailable. |
+| `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and *not* on error); inline error line; active `[Ignore Here]`/`[Ignore Everywhere]` vs. greyed-with-label + `[Undo ▾]`; "Show Ignored" toggle defaults off and is not persisted; `[Ignore Everywhere]` opens confirm dialog and only writes on confirm; `[Ignore Everywhere]` absent when capability unavailable. |
 | `LeftPanel` / dot | Dot iff cascade-active > 0, visible from Resources tab; tab switch persistence key written; toggle-button mirror dot when panel closed. |
 
 ### 10.2 fluent-api (Vitest, route-level — same style as the ai-tools and editor-state tests)
 
 | Area | Representative cases |
 | --- | --- |
-| `GET /users/settings` | 401 unauthenticated; `settings: null` for new user (200, not 404); returns saved blob; user isolation (only own row). |
-| `PUT /users/settings` | Upsert create + replace; 400 on schema violation; full-replace semantics (omitted keys gone); `.catch({})` tolerance on read of unknown-shaped stored blobs. |
-| Editor-state schema extension | Existing editor-state tests still pass unchanged (proves backward compatibility); round-trip of `checkOccurrenceRules` / `activeLeftTab` / `showResolvedChecks`. |
+| `GET /self/settings` | 401 unauthenticated; `settings: null` for new user (200, not 404); returns saved blob; user isolation (only own row, scoped to session user). |
+| `PUT /self/settings` | Upsert create + replace; 400 on schema violation; full-replace semantics (omitted keys gone); `.catch({})` tolerance on read of unknown-shaped stored blobs. |
+| Editor-state schema extension | Existing editor-state tests still pass unchanged (proves backward compatibility); round-trip of `checkOccurrenceRules` / `activeLeftTab`. |
 
 Manual verification continues to use the stack via `fluent-platform` compose plus the existing repeated-words smoke script for the proxy half.
 
@@ -439,7 +466,8 @@ Manual verification continues to use the stack via `fluent-platform` compose plu
 ## 11. Future work (explicitly deferred)
 
 - **Sibling checks** (Wildebeest character check, spell check — draft cards exist on the board): each becomes another accordion section in `ChecksPanel` and another wire-type module; the cascade, dot, and suppression stores are check-agnostic by construction (occurrence keys embed the finding identity; global keys can be namespaced per check when a second one arrives).
-- **"Manage ignored words" settings page**: a CRUD view over `userSettings.checkIgnoredWordPairs`. The store and endpoint are already shaped for it.
+- **"Manage ignored words" settings page**: a CRUD view over `userSettings.checkIgnoredWordPairs`. The store and `self/settings` endpoint are already shaped for it.
+- **Highlight repeated words in the verse text** (joelthe1/Ulf, §5.3 item 4): a follow-on PR layered on this one, touching the verse-rendering pane; specced just before implementation.
 - **Checks in review/read-only stages** (peer check, `/view` route): needs product definition of who may ignore what; W10 keeps v1 to drafting.
 - **Async/polling tools**: the envelope is consumed whole (D9), so a `status: "queued"` tool slots in with a polling loop inside `useRepeatedWordsCheck` and zero UI-contract change.
 - **"Drop duplicate" quick-fix and Greek Room feedback loop**: excluded by card #278; the occurrence identity (`snt_id`, pair, ordinal, `start_position` at render time) is sufficient input for both when they're picked up.
@@ -448,23 +476,25 @@ Manual verification continues to use the stack via `fluent-platform` compose plu
 
 ## 12. Sign-off checklist
 
-Items needing explicit confirmation in review (referenced from §5.2/§5.3); none block reading the rest of the design:
+Items needing explicit confirmation in review (referenced from §5.2/§5.3); none block reading the rest of the design. Status reflects the first review round (§0):
 
-| # | Item | Proposed resolution |
-| --- | --- | --- |
-| S1 | Zero-state mock shows the notification dot; #277 text says no flags ⇒ no dot | Follow the text: no dot at zero active flags |
-| S2 | #278 mocks show the Resources language dropdown above checks content | Omit it on the Checks tab (presumed screenshot carryover; no function for checks) |
-| S3 | Card says ignored items are "removed from the panel" | Grey-and-stay with reason label + undo; "Show ignored & OK" unchecked gives the card-literal view; dot unaffected either way |
-| S4 | Cards silent on Greek-Room-`legitimate` findings | Show as inactive items ("Marked OK by Greek Room"), per-occurrence surfaceable |
-| S5 | Dot is invisible when the whole left panel is closed | Mirror the dot on the panel-toggle button while closed |
-| S6 | Suppression persistence has no card of its own | Ship as extension of #172's scope (W1) — backend dependency of the UI cards |
+| # | Item | Proposed resolution | Status |
+| --- | --- | --- | --- |
+| S1 | Zero-state mock showed the notification dot; #277 text says no flags ⇒ no dot | Follow the text: no dot at zero active flags | ✅ Resolved — chadw fixed the mock |
+| S2 | #278 mocks showed a language dropdown above checks content | Omit it on the Checks tab (no function for checks) | ✅ Resolved — chadw removed it from the mocks |
+| S3 | Card originally said ignored items are "removed from the panel" | Reveal them dimmed under a "Show Ignored" toggle (default off, not persisted) with "Undo Ignore"; dot unaffected | ✅ Largely card-aligned — revised #278 added the toggle; confirm the labels match |
+| S4 | Cards did not enumerate Greek-Room-`legitimate` findings | Show as ignored items labeled **"Default Ignore"** (the revised #278 mock's term), per-occurrence surfaceable | ☑ Confirm "Default Ignore" = machine-legitimate |
+| S5 | Dot is invisible when the whole left panel is closed | Mirror the dot on the panel-toggle button while closed | ☐ Open |
+| S6 | Suppression persistence has no card of its own | Ship as extension of #172's scope (W1) — backend dependency of the UI cards | ☐ Open |
+| S7 | Revised #278 confirm dialog says "Ignore Everywhere" "cannot be undone" | Add the confirm dialog **and** keep `[Undo ▾]` reversibility (§6.5); the two are complementary | ☐ Confirm we may retain undo |
+| S8 | Highlight repeated words in the verse text (joelthe1/Ulf) | We will implement it as a **follow-on PR** layered on this one (verse-pane scope; §5.3 item 4) | ☐ Confirm follow-on-PR sequencing |
 
 Engineering-side confirmations sought from reviewers (same spirit as the D-series confirmations in the fluent-api proposal):
 
-- **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and `GET/PUT /users/settings` as its endpoint (this outlives the feature).
+- **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and a new **`self` domain** (`GET/PUT /self/settings`) as its endpoint (this outlives the feature). _Endpoint shape requested by kaseywright in the first round; confirming the `self`-domain placement._
 - **W4** — `(snt_id, repeated_word, ordinal)` occurrence identity and NFC-no-case-folding comparison.
-- **W5/W6** — the tri-state cascade and the `[Undo ▾]` split-button behavior.
+- **W5/W6** — the tri-state cascade, the confirm-dialog-plus-`[Undo ▾]` behavior on "Ignore Everywhere," and the split-button undo (S7).
 
 ---
 
-*Prepared against fluent-web `main` (post `0dcea61`), fluent-api `main` (post `7c7f63d`), and the approved fluent-api proposal (PR #173). Author: JEdward7777.*
+*Prepared against fluent-web `main` (post `2e936d1`), fluent-api `main` (post `7c7f63d`), and the approved fluent-api proposal (PR #173). Revised 2026-06-16 after the first review round on PR #305. Author: JEdward7777.*
diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-summary.md b/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
index 720ebd9c..12af7694 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
@@ -1,41 +1,49 @@
 # Repeated Word Check UI — Review Summary
 
-**Purpose:** Reviewer orientation for the proposed Repeated Word Check UI (cards [fluent-web#277](https://github.com/eten-tech-foundation/fluent-web/issues/277) and [fluent-web#278](https://github.com/eten-tech-foundation/fluent-web/issues/278)) plus the small fluent-api persistence additions the cards imply. This summary is intended to stand on its own; the long-form proposal with full analysis lives in the sibling file [`checks-ui-integration-suggestion.md`](checks-ui-integration-suggestion.md) (decisions **W1–W12**, §1–§12). It builds on the approved fluent-api proxy design ([PR #173](https://github.com/eten-tech-foundation/fluent-api/pull/173), decisions **D1–D12**). Ships as **two PRs** — fluent-web (the bulk) plus a small fluent-api PR — and either may land first by design (W8/W12).
+**Status:** Revised after the first review round (PR #305).
+
+**Purpose:** Reviewer orientation for the proposed Repeated Word Check UI (cards [fluent-web#277](https://github.com/eten-tech-foundation/fluent-web/issues/277) and [fluent-web#278](https://github.com/eten-tech-foundation/fluent-web/issues/278)) plus the small fluent-api persistence additions the cards imply. This summary is intended to stand on its own; the long-form proposal with full analysis lives in the sibling file [`checks-ui-integration-suggestion.md`](checks-ui-integration-suggestion.md) (decisions **W1–W12**, §0–§12; §0 is the change log for this revision). It builds on the approved fluent-api proxy design ([PR #173](https://github.com/eten-tech-foundation/fluent-api/pull/173), decisions **D1–D12**). Ships as **two PRs** — fluent-web (the bulk) plus a small fluent-api PR — and either may land first by design (W8/W12). A **highlight-in-verse** capability requested in review ships as a separate follow-on PR.
 
 ## What's being proposed
 
-Give translators the user-facing half of the Repeated Word Check: a **Checks tab** with a notification dot in the drafting page's left panel (#277) and a **Checks panel** listing repeated-word findings per verse, refreshed on every auto-save, with "Ignore This Time" / "Ignore Always" actions persisted in Fluent's database (#278). The structure (per-check accordion, check-agnostic suppression cascade) anticipates future checks (Wildebeest, spell check) without rework.
+Give translators the user-facing half of the Repeated Word Check: a **Checks tab** with a notification dot in the drafting page's left panel (#277) and a **Checks panel** listing repeated-word findings per verse, refreshed on every auto-save, with **"Ignore Here" / "Ignore Everywhere"** actions persisted in Fluent's database (#278). The structure (per-check accordion, check-agnostic suppression cascade) anticipates future checks (Wildebeest, spell check) without rework.
 
 ## Core decisions for review
 
 1. **Suppression persistence ships as an extension of #172's scope — no new product card** (W1). #172's "backend dependency for all Repeated Word Check UI work" covers it; the fluent-api proposal's D1 deferral was about caching tool results, not user preferences.
 
-2. **Hybrid storage at exactly the scope each rule governs** (W2): "Ignore This Time" extends the existing per-`(user, chapterAssignment)` editor-state JSONB (optional Zod keys, **no migration**); "Ignore Always" introduces a new **`user_settings` table** — one row per user, one Zod-typed JSONB column — deliberately establishing Fluent's general user-preference store (future settings become schema keys, not migrations). Findings are filtered **client-side**, keeping the AI proxy the pure pass-through D8/D9 established.
+2. **Hybrid storage at exactly the scope each rule governs** (W2): "Ignore Here" extends the existing per-`(user, chapterAssignment)` editor-state JSONB (optional Zod keys, **no migration**); "Ignore Everywhere" introduces a new **`user_settings` table** — one row per user, one Zod-typed JSONB column — deliberately establishing Fluent's general user-preference store (future settings become schema keys, not migrations). Findings are filtered **client-side**, keeping the AI proxy the pure pass-through D8/D9 established.
 
-3. **Settings endpoint mirrors the editor-state idiom one level up** (W7): `GET/PUT /users/settings`, session-implicit user (never in the URL), full-replace upsert of one blob, file quartet `domains/users/settings/user-settings.{route,service,repository,types}.ts`. Last-writer-wins concurrency inherited from editor-state and accepted.
+3. **Settings endpoint lives in a new `self` domain** (W7, revised per kaseywright's first-round feedback): `GET/PUT /self/settings`, guarded by `authenticateUser` alone (no `{userId}` in the path, so nothing to misuse and no `requireSelf`/`UserPolicy` middleware), full-replace upsert of one blob, file quartet `domains/self/settings/self-settings.{route,service,repository,types}.ts`. Last-writer-wins concurrency inherited from editor-state and accepted.
 
 4. **Chapter-wide check on every successful verse auto-save** (W3/W4): a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)` sending all drafted verses; `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM, the smoke-test convention); no extra debounce beyond the existing 2 s save debounce. Drafting mode only (W10).
 
-5. **Three-layer active/inactive cascade, most-specific-non-silent-wins** (W5/W6): Greek Room's `legitimate` verdict → user-global word-pair rule → occurrence rule `(snt_id, repeated_word, ordinal)`; rules are tri-state maps (`absent/'suppress'/'surface'`), which is what makes per-occurrence **undo** of a global rule or machine verdict possible. The dot counts active findings only; ignored items grey-and-stay with reason labels and an `[Undo ▾]` split button.
+5. **Three-layer active/inactive cascade, most-specific-non-silent-wins** (W5/W6): Greek Room's `legitimate` verdict → user-global word-pair rule → occurrence rule `(snt_id, repeated_word, ordinal)`; rules are tri-state maps (`absent/'suppress'/'surface'`), which is what makes per-occurrence **undo** of a global rule or machine verdict possible. The dot counts active findings only; ignored items are hidden by default and revealed dimmed (with "Default Ignore"/"Ignore Here"/"Ignore Always" reason labels and "Undo Ignore") under a **"Show Ignored" toggle** (default off, not persisted), matching the revised #278.
 
-6. **Graceful degradation, either-PR-lands-first** (W8/W9): the UI feature-detects `GET /users/settings` (404 ⇒ "Ignore Always" simply not rendered — capability hidden, never a dead control); check failures surface as one inline red line in the panel while the last successful findings stay rendered (TanStack default). No toasts/banners — none exist in this codebase.
+6. **Graceful degradation, either-PR-lands-first** (W8/W9): the UI feature-detects `GET /self/settings` (404 ⇒ "Ignore Everywhere" simply not rendered — capability hidden, never a dead control); check failures surface as one inline red line in the panel while the last successful findings stay rendered (TanStack default). No toasts/banners — none exist in this codebase. **"Ignore Everywhere" shows a confirm dialog** (per revised #278) while still being reversible via undo.
 
 ## Explicitly out of scope (deferred)
 
 "Drop duplicate" quick-fix and Greek Room feedback loop (excluded by card #278); checks in review/read-only stages; a "Manage ignored words" settings page (the store is shaped for it); async/polling mode; any change to fluent-ai or the proxy contract; sibling checks.
 
+## Resolved in the first review round
+
+- **Zero-state dot** and **language dropdown** mock issues are fixed — chadw corrected the zero-state mock and removed the dropdown from the #278 mocks.
+- **Button labels** are now **"Ignore Here" / "Ignore Everywhere"** (joelthe1/Ulf).
+- **Ignored items** are revealed dimmed under a **"Show Ignored" toggle** (default off, not persisted) with "Undo Ignore" — the revised #278 added the toggle, largely aligning with our earlier proposal.
+- **Settings endpoint** moved to a new **`self` domain** (`GET/PUT /self/settings`, `authenticateUser` only) per kaseywright.
+
 ## Areas where input would be most valuable
 
-**Product sign-off (mock/card deviations, §5.2–§5.3 / §12 of the proposal):**
+**Product sign-off (§5.3 / §12 of the proposal):**
 
-1. Zero-state mock shows the dot but #277's text says no flags ⇒ no dot — we follow the text.
-2. The Resources language dropdown appears above the checks content in the #278 mocks with no function there — we propose omitting it.
-3. Ignored items **grey-and-stay with undo** instead of the card's "removed from the panel" (dot behavior unchanged; a "Show ignored & OK" toggle gives the card-literal view).
-4. Machine-`legitimate` findings shown as inactive items ("Marked OK by Greek Room") — cards are silent on them.
-5. Dot mirrored on the panel-toggle button when the left panel is closed, preserving #277's intent.
+1. **"Default Ignore" items.** Greek Room sometimes reports a repeated word but marks it `legitimate: true` — i.e. it detected the repetition but judged it intentional/correct (e.g. a word genuinely repeated for emphasis), so it shouldn't be flagged. We render these as ignored items labeled **"Default Ignore"** (the revised #278 mock's term), still surfaceable per-occurrence in case the checker is wrong. Confirm that "Default Ignore" = "repetition Greek Room found but considers legitimate."
+2. Dot mirrored on the panel-toggle button when the left panel is closed, preserving #277's intent.
+3. **"Ignore Everywhere" confirm dialog + undo together:** the revised #278 dialog says the action "cannot be undone," but we keep the `[Undo ▾]` reversibility (the dialog guards against accidental clicks; undo handles deliberate reversals). Confirm we may retain undo.
+4. **Highlight repeated words in the verse text** (joelthe1/Ulf) ships as a **separate follow-on PR** layered on this one — confirm the sequencing.
 
 **Engineering confirmations:**
 
-1. **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and `GET/PUT /users/settings` as its endpoint (this outlives the feature).
+1. **W2/W7** — blessing `user_settings` as Fluent's general user-preference store and the new **`self` domain** (`GET/PUT /self/settings`) as its endpoint (this outlives the feature).
 2. **W4** — `(snt_id, repeated_word, ordinal)` occurrence identity and NFC-exact (no case folding) key comparison.
-3. **W5/W6** — the tri-state cascade and the `[Undo ▾]` split-button behavior.
+3. **W5/W6** — the tri-state cascade and the confirm-dialog-plus-`[Undo ▾]` behavior.

From 2f46757e6a1b9eb56592ba28074d0db276be9b96 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Tue, 16 Jun 2026 16:53:00 -0400
Subject: [PATCH 06/33] Address CodeRabbit nits: 'in situ' -> 'in context'; add
 'text' specifier to directory-tree code fences

(mermaid fence already carried its specifier)
---
 .../repeated-word-check/checks-ui-integration-suggestion.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index 260b5c24..7c58be17 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -144,7 +144,7 @@ The cards' mockups (drafting page, Judges 4, Gujarati IRV project) define the vi
 - **Tab header (#277 mock):** the left panel's current `Resources` heading is replaced by a text-tab row — **"Resources | Checks"** — active tab in blue with a blue underline, inactive tab plain. The **notification dot is a solid blue filled circle immediately right of the "Checks" label**, and the mock shows it while the *Resources* tab is active: the dot is visible from either tab whenever the panel is open.
 - **Checks content (#278 mocks):** below the tab header, a card containing a **collapsible "Repeated Words" accordion section** (expanded by default). The accordion-per-check structure anticipates future sibling checks (the board already holds draft cards for Greek Room Wildebeest and Spell checks) without UI rework. Inside the section, findings are grouped by verse:
   - bold **"Verse N"** heading per verse that has findings (verses without findings get no group);
-  - one row per occurrence: a one-line context snippet showing the repeated pair in situ;
+  - one row per occurrence: a one-line context snippet showing the repeated pair in context;
   - two solid-blue buttons side by side beneath each snippet: `[Ignore Here] [Ignore Everywhere]`;
   - a horizontal separator between verse groups;
   - a **"Show Ignored" toggle** at the bottom of the active list (off by default). When on, previously ignored occurrences appear below the active ones — dimmed, each labeled with its ignore type (_"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) and showing an **"Undo Ignore"** action in place of the two ignore buttons.
@@ -170,7 +170,7 @@ None of these block implementation; they are listed in §12 for explicit confirm
 
 ### 6.1 File layout
 
-```
+```text
 fluent-web/src/
 ├── components/ui/                          # (existing; shadcn primitives incl. dropdown-menu, checkbox)
 ├── features/
@@ -382,7 +382,7 @@ Otherwise it keeps the editor-state idiom: **full-replace upsert** on PUT, one Z
 
 ### 8.2 File layout
 
-```
+```text
 fluent-api/src/domains/self/settings/
 ├── self-settings.route.ts        # GET + PUT, OpenAPI-described, session user from context
 ├── self-settings.service.ts      # get / upsert, Result-typed like editor-state service

From 181258e00dfcfd96de75c2e26eb0538be002b173 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Tue, 16 Jun 2026 17:11:46 -0400
Subject: [PATCH 07/33] docs(checks-ui): expand S6.5 with the reversibility
 implication chain

Document why [Undo] reversibility falls out of the layered cascade design
rather than being an added feature, and fold out the now-redundant
'why undo despite cannot be undone' bullet. Framed as documentation for
anyone following the reasoning, not a request to override the card's
'cannot be undone' wording (S7 remains a confirm-only sign-off item).
---
 .../checks-ui-integration-suggestion.md              | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index 7c58be17..43756f3a 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -250,12 +250,20 @@ Every finding resolves through three layers; the **most specific non-silent verd
 ### 6.5 Action surface (W6)
 
 - **Active finding** (per revised card #278): `[Ignore Here]` → writes occurrence rule `suppress`; `[Ignore Everywhere]` → opens a **confirm dialog** ("Fluent will not flag this word pair again across all your projects. Are you sure?") and, on confirm, writes the global pair rule `suppress`.
-- **Confirm dialog on `[Ignore Everywhere]`** (revised #278): the card asks for it because the action reads as global and irreversible. We honor the request **and** keep the action reversible via `[Undo ▾]` below — the dialog guards the deliberate click; the undo guards a mistake that slips past it. The two are not redundant: confirm is pre-action friction, undo is post-action recourse. (We considered dropping undo to match a strict "cannot be undone" reading, but the cascade already supports clean reversal at the right scope, and there is no explicit directive to remove it; the final PR is reviewed again before merge.)
+- **Confirm dialog on `[Ignore Everywhere]`** (revised #278): the card asks for it because the action reads as global and irreversible. We honor the request **and** keep the action reversible via `[Undo ▾]` below — the dialog guards the deliberate click; the undo guards a mistake that slips past it. The two are not redundant: confirm is pre-action friction, undo is post-action recourse.
+
+  Why reversibility falls out of the design rather than being an extra feature we chose to bolt on — the implication chain (see also §6.4, which this depends on):
+  1. Greek Room reports **both** polarities: repetitions it flags (suspicious) and repetitions it judges `legitimate` (a *negative* / "don't flag" verdict).
+  2. Once we decide to **show** the legitimate ones (§5.3 item 2 — so the translator can see what the checker waved through), it's natural to let the translator disagree and mark one **illegitimate** (re-surface it).
+  3. Expressing "re-surface this" requires a **positive** record (`surface`), not only the negative (`suppress`) records an ignore-list would hold. So the stores must be tri-state, not suppression-only (§6.4).
+  4. Given positive *and* negative records at more than one scope, a **layered** model is the natural fit: a local (occurrence) record can override a global (word-pair) one, and the override works in **both polarities** (locally surface something suppressed globally, or locally suppress something surfaced).
+  5. With that layering in place, a single bare "undo" is **ambiguous** — at which layer, in which polarity? Resolving that ambiguity (the `[Undo ▾]` split button: default = act at the occurrence layer; chevron = act globally) is what the design *has* to do anyway — and the same mechanism incidentally gives the translator the "oops, I hit the wrong button" recourse for `[Ignore Everywhere]`. Reversibility is therefore inherited, not added.
+
+  We are not requesting that this override the card's "cannot be undone" framing; it is documented here for anyone who wants to follow the reasoning, and the implemented PR is reviewed again before merge.
 - **Inactive finding:** one `[Undo ▾]` split button (shadcn `dropdown-menu`, already in the codebase — no hidden gestures):
   - **Default click** acts at the occurrence layer: if inactive via its *own* occurrence rule → delete that entry; if inactive via a global rule or Greek Room's verdict → write an occurrence-level `surface` override (the global rule / machine verdict survives; only this occurrence resurfaces).
   - **Chevron menu** offers the deliberate global actions with consequence-naming labels, e.g. _"Stop ignoring 'the the' everywhere"_ (deletes the global entry).
 - **Global writes purge local rules — current chapter only.** Executing any global (everywhere) suppress or surface first deletes the user's occurrence-level rules for that word pair in the *current chapter assignment's* editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
-- **Why undo despite the "cannot be undone" wording.** The card's confirm-dialog copy ("cannot be undone") is product framing for the *commitment* of a global ignore, not a storage constraint. With the three-layer cascade, "undo" is well-defined at each scope (delete the global entry, or `surface`-override one occurrence), so we surface it rather than stranding the translator. This is the one place we keep more capability than the literal card; flagged in §12 for sign-off.
 - **Direct edits resolve silently** (card #278): if the translator fixes the text, the next auto-save's fresh findings simply no longer contain the occurrence — no action needed, panel and dot update.
 
 ### 6.6 Left-panel container and persisted UI state (W11)

From e0e1a3b5c2dfeae60be06aaaaebc0e599eb0a86a Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 18 Jun 2026 14:07:05 -0400
Subject: [PATCH 08/33] feat(checks): add Repeated Word Check UI foundations
 (#277, #278)

Implements the unblocked portions of the Checks UI per the proposal
(docs/proposals/repeated-word-check, decisions W1-W12, file layout S6.1):

- features/checks/checks.types.ts: wire types verbatim snake_case per
  fluent-api D8 (lang_code, snt_id, repeated_word, start_position, surf,
  legitimate, severity) plus the full ToolJobResponse envelope; UI-derived
  types in camelCase (RuleVerdict, OccurrenceRules, ResolvedFinding, ...).
- features/checks/hooks/useRepeatedWordsCheck.ts: TanStack useQuery keyed
  (chapterAssignmentId, saveCounter), POST /ai/tools/greek-room/repeated-words
  with credentials:'include', consumes the full envelope and branches on
  status. snt_id = '{bookCode} {chapter}:{verse}' using projectItem.bookCode
  (USFM code). Follows the features/bible/hooks/useBibleTarget.ts fetch pattern.
- features/checks/hooks/useResolvedFindings.ts (+ tests): pure three-layer
  active/inactive cascade (W5/W6, most-specific-non-silent-wins) with NFC-exact
  comparison (no case folding) and zero-based per-verse ordinals. 24 tests cover
  the precedence matrix, ordinals incl. 'the the the', and NFC equivalence.
- features/checks/components/LeftPanel.tsx (+ tests): Resources | Checks tab
  container with content-agnostic slots and a notification dot beside 'Checks'
  when active findings > 0.
- resources/components/ResourcePanel.tsx: remove the <h3>Resources</h3> header
  (the tab row is now the header).
- bible/hooks/useResourceStatePersistence.ts: persist activeLeftTab.

Held for sign-off (not included): S5 toggle-button dot, S7 confirm-dialog +
[Undo], ChecksPanel/FindingRow, useSuppressions, DraftingPage wiring, and the
fluent-api /self/settings persistence half.

Relates to #277, #278
---
 .../hooks/useResourceStatePersistence.ts      |  10 +
 src/features/checks/checks.types.ts           | 159 ++++++++++++
 .../checks/components/LeftPanel.test.tsx      |  61 +++++
 src/features/checks/components/LeftPanel.tsx  | 104 ++++++++
 .../checks/hooks/useRepeatedWordsCheck.ts     | 136 ++++++++++
 .../checks/hooks/useResolvedFindings.test.ts  | 232 ++++++++++++++++++
 .../checks/hooks/useResolvedFindings.ts       | 198 +++++++++++++++
 .../resources/components/ResourcePanel.tsx    |   5 +-
 8 files changed, 902 insertions(+), 3 deletions(-)
 create mode 100644 src/features/checks/checks.types.ts
 create mode 100644 src/features/checks/components/LeftPanel.test.tsx
 create mode 100644 src/features/checks/components/LeftPanel.tsx
 create mode 100644 src/features/checks/hooks/useRepeatedWordsCheck.ts
 create mode 100644 src/features/checks/hooks/useResolvedFindings.test.ts
 create mode 100644 src/features/checks/hooks/useResolvedFindings.ts

diff --git a/src/features/bible/hooks/useResourceStatePersistence.ts b/src/features/bible/hooks/useResourceStatePersistence.ts
index 54691a8c..4694cda6 100644
--- a/src/features/bible/hooks/useResourceStatePersistence.ts
+++ b/src/features/bible/hooks/useResourceStatePersistence.ts
@@ -3,10 +3,19 @@ import { useMutation, useQuery, useQueryClient } from '@tanstack/react-query';
 import { config } from '@/lib/config';
 import { Logger } from '@/lib/services/logger';
 
+/**
+ * Which tab of the drafting-page left panel is showing (Resources | Checks).
+ * Persisted in the editor-state blob so the panel reopens where the translator
+ * left off (Repeated Word Check, W11/§6.6). Optional + backward-compatible:
+ * old rows simply omit it and the panel defaults to 'resources'.
+ */
+export type LeftTab = 'resources' | 'checks';
+
 export interface FetchResourceState {
   activeResource: string;
   languageCode: string;
   tabStatus: boolean;
+  activeLeftTab?: LeftTab;
 }
 
 interface ResourceState {
@@ -17,6 +26,7 @@ interface ResourceState {
     activeResource: string;
     languageCode: string;
     tabStatus: boolean;
+    activeLeftTab?: LeftTab;
   };
 }
 
diff --git a/src/features/checks/checks.types.ts b/src/features/checks/checks.types.ts
new file mode 100644
index 00000000..f969eae9
--- /dev/null
+++ b/src/features/checks/checks.types.ts
@@ -0,0 +1,159 @@
+/**
+ * Type definitions for the Checks feature (Repeated Word Check, cards #277/#278).
+ *
+ * This module deliberately mixes two naming conventions:
+ *
+ *  1. **Wire types** (the `*Request` / `*Response` / `*Finding` / envelope shapes)
+ *     mirror the fluent-ai contract **verbatim in snake_case** (`lang_code`,
+ *     `snt_id`, `repeated_word`, `start_position`, …). This is the same
+ *     intentional, contained exception fluent-api made — see fluent-api decision
+ *     **D8** and the proposal §6.1 note. Do **not** "normalize" these to
+ *     camelCase: renaming silently breaks the verbatim pass-through contract with
+ *     fluent-ai (the request body is forwarded unchanged through fluent-api).
+ *     Approved in review:
+ *     https://github.com/eten-tech-foundation/fluent-api/pull/173#discussion_r3343677813
+ *
+ *  2. **UI-derived types** (resolved findings, rules, cascade verdicts) use the
+ *     repo's normal **camelCase** convention. The snake_case exception is scoped
+ *     strictly to the wire types above.
+ */
+
+// ---------------------------------------------------------------------------
+// Wire types — snake_case, verbatim fluent-ai contract (fluent-api D8).
+// Mirror of fluent-api `src/domains/ai-tools/ai-tools.types.ts` and
+// `src/lib/services/fluent-ai/fluent-ai.types.ts`.
+// ---------------------------------------------------------------------------
+
+/** Job lifecycle status from the fluent-ai `ToolJobResponse` envelope. */
+export type JobStatus = 'queued' | 'running' | 'completed' | 'failed' | 'cancelled';
+
+/** One verse sent for checking. `snt_id` is the verse identifier (see W4). */
+export interface VerseInput {
+  snt_id: string;
+  text: string;
+}
+
+/**
+ * Request body for `POST /ai/tools/greek-room/repeated-words`.
+ * Forwarded verbatim by fluent-api to fluent-ai (D8). snake_case is intentional.
+ */
+export interface RepeatedWordsRequest {
+  lang_code: string;
+  lang_name: string;
+  project_id: string | number;
+  project_name: string;
+  verses: VerseInput[];
+}
+
+/** A single consecutive-repeated-word finding from Greek Room. */
+export interface RepeatedWordsFinding {
+  /** Verse identifier, e.g. `"JDG 4:3"` (USFM book code; see W4). */
+  snt_id: string;
+  /** The repeated pair, NFC + lowercased by Greek Room, e.g. `"the the"`. */
+  repeated_word: string;
+  /** The original surface text as it appeared (preserves casing); display only. */
+  surf: string;
+  /** Character offset of the finding within the verse text. */
+  start_position: number;
+  /** Greek Room's verdict: `true` ⇒ intentional/correct (machine "Default Ignore"). */
+  legitimate: boolean;
+  /** Upstream numeric severity (e.g. 0.1 legitimate, 0.5 suspicious). */
+  severity: number;
+}
+
+export interface RepeatedWordsSummary {
+  total_findings: number;
+  legitimate_count: number;
+  verse_count: number;
+}
+
+/** The `result` payload when `status === "completed"`. */
+export interface RepeatedWordsResult {
+  lang_code: string;
+  provider: string;
+  check: string;
+  findings: RepeatedWordsFinding[];
+  summary: RepeatedWordsSummary;
+}
+
+/** fluent-ai's error shape inside the envelope. */
+export interface ToolJobError {
+  code: string;
+  message: string;
+  details?: unknown;
+}
+
+/**
+ * The full `ToolJobResponse` envelope, passed through fluent-api unchanged (D9).
+ * The hook consumes the whole envelope and branches on `status`, so a future
+ * async/polling tool (`status: "queued"`) drops in without reshaping the UI.
+ */
+export interface ToolJobResponse<TResult> {
+  job_id: string;
+  tool: string;
+  status: JobStatus;
+  result: TResult | null;
+  error: ToolJobError | null;
+  created_at: string;
+  completed_at: string | null;
+}
+
+/** Convenience alias for the repeated-words envelope. */
+export type RepeatedWordsResponse = ToolJobResponse<RepeatedWordsResult>;
+
+// ---------------------------------------------------------------------------
+// UI-derived types — camelCase (the repo's normal convention).
+// ---------------------------------------------------------------------------
+
+/**
+ * A tri-state suppression rule (W5/W6). `'suppress'` hides a finding,
+ * `'surface'` forces it active (per-occurrence undo of a global rule or of a
+ * Greek Room `legitimate` verdict). Absence of a key = silent (no opinion).
+ */
+export type RuleVerdict = 'suppress' | 'surface';
+
+/**
+ * Occurrence-level rules (cascade layer 2), scoped to one chapter assignment.
+ * Keyed by occurrence identity `"{snt_id}|{repeated_word}|{ordinal}"`.
+ * Persisted in the editor-state JSONB blob (`checkOccurrenceRules`).
+ */
+export type OccurrenceRules = Record<string, RuleVerdict>;
+
+/**
+ * User-global word-pair rules (cascade layer 1), across all the user's projects.
+ * Keyed by the NFC-normalized `repeated_word` pair (e.g. `"the the"`).
+ * Persisted in `user_settings.checkIgnoredWordPairs`.
+ */
+export type GlobalRules = Record<string, RuleVerdict>;
+
+/** Why a finding is currently inactive (drives the dimmed ignore-type label). */
+export type InactiveReason =
+  | 'occurrence' // user "Ignore Here"
+  | 'global' // user "Ignore Everywhere"
+  | 'legitimate'; // Greek Room "Default Ignore"
+
+/**
+ * A finding after the three-layer cascade has been applied (§6.4).
+ * Carries the original wire finding plus the derived identity and verdict.
+ */
+export interface ResolvedFinding {
+  /** The raw finding from Greek Room (snake_case wire shape preserved). */
+  finding: RepeatedWordsFinding;
+  /**
+   * Index of this finding among same-`repeated_word` findings in the same
+   * verse, ordered by `start_position` (the "x of n" ordinal, W4). Zero-based.
+   */
+  ordinal: number;
+  /** Stable occurrence identity key: `"{snt_id}|{repeated_word}|{ordinal}"`. */
+  occurrenceKey: string;
+  /** Whether the finding is currently active (counts toward the dot). */
+  isActive: boolean;
+  /** Present only when `isActive === false`: which layer suppressed it. */
+  inactiveReason?: InactiveReason;
+}
+
+/** Output of the cascade: findings split into active and inactive buckets. */
+export interface ResolvedFindings {
+  active: ResolvedFinding[];
+  inactive: ResolvedFinding[];
+}
diff --git a/src/features/checks/components/LeftPanel.test.tsx b/src/features/checks/components/LeftPanel.test.tsx
new file mode 100644
index 00000000..07fd26ca
--- /dev/null
+++ b/src/features/checks/components/LeftPanel.test.tsx
@@ -0,0 +1,61 @@
+import { fireEvent, render, screen } from '@testing-library/react';
+import { describe, expect, it, vi } from 'vitest';
+
+import { LeftPanel } from './LeftPanel';
+
+const baseProps = {
+  activeTab: 'resources' as const,
+  onTabChange: vi.fn(),
+  activeFindingsCount: 0,
+  resourcesContent: <div>resources-body</div>,
+  checksContent: <div>checks-body</div>,
+};
+
+describe('LeftPanel', () => {
+  it('renders both tabs', () => {
+    render(<LeftPanel {...baseProps} />);
+    expect(screen.getByRole('tab', { name: /Resources/ })).toBeInTheDocument();
+    expect(screen.getByRole('tab', { name: /Checks/ })).toBeInTheDocument();
+  });
+
+  it('shows the active tab body and hides the inactive one', () => {
+    const { rerender } = render(<LeftPanel {...baseProps} activeTab='resources' />);
+    expect(screen.getByText('resources-body')).toBeInTheDocument();
+    expect(screen.queryByText('checks-body')).not.toBeInTheDocument();
+
+    rerender(<LeftPanel {...baseProps} activeTab='checks' />);
+    expect(screen.getByText('checks-body')).toBeInTheDocument();
+    expect(screen.queryByText('resources-body')).not.toBeInTheDocument();
+  });
+
+  it('marks the active tab via aria-selected', () => {
+    render(<LeftPanel {...baseProps} activeTab='checks' />);
+    expect(screen.getByRole('tab', { name: /Checks/ })).toHaveAttribute('aria-selected', 'true');
+    expect(screen.getByRole('tab', { name: /Resources/ })).toHaveAttribute(
+      'aria-selected',
+      'false'
+    );
+  });
+
+  it('calls onTabChange when a tab is clicked', () => {
+    const onTabChange = vi.fn();
+    render(<LeftPanel {...baseProps} activeTab='resources' onTabChange={onTabChange} />);
+    fireEvent.click(screen.getByRole('tab', { name: /Checks/ }));
+    expect(onTabChange).toHaveBeenCalledWith('checks');
+  });
+
+  it('does not render the notification dot when there are no active findings', () => {
+    render(<LeftPanel {...baseProps} activeFindingsCount={0} />);
+    expect(screen.queryByTestId('checks-notification-dot')).not.toBeInTheDocument();
+  });
+
+  it('renders the dot (visible from the Resources tab) when active findings exist', () => {
+    render(<LeftPanel {...baseProps} activeFindingsCount={3} activeTab='resources' />);
+    expect(screen.getByTestId('checks-notification-dot')).toBeInTheDocument();
+  });
+
+  it('renders the dot on the Checks tab when active findings exist', () => {
+    render(<LeftPanel {...baseProps} activeFindingsCount={1} activeTab='checks' />);
+    expect(screen.getByTestId('checks-notification-dot')).toBeInTheDocument();
+  });
+});
diff --git a/src/features/checks/components/LeftPanel.tsx b/src/features/checks/components/LeftPanel.tsx
new file mode 100644
index 00000000..f3e79a20
--- /dev/null
+++ b/src/features/checks/components/LeftPanel.tsx
@@ -0,0 +1,104 @@
+import { type ReactNode } from 'react';
+
+import { cn } from '@/lib/utils';
+
+/** Which tab of the drafting-page left panel is showing (persisted, W11). */
+export type LeftTab = 'resources' | 'checks';
+
+export interface LeftPanelProps {
+  /** Controlled active tab; lifted to `DraftingPage` so it can be persisted
+   *  in the editor-state blob as `activeLeftTab` (W11, §6.6). */
+  activeTab: LeftTab;
+  /** Called when the translator switches tabs; the parent persists the value. */
+  onTabChange: (tab: LeftTab) => void;
+  /** Cascade-resolved active-finding count (§6.4). The notification dot shows
+   *  iff this is > 0; the dot is visible from either tab. */
+  activeFindingsCount: number;
+  /** Body for the Resources tab (the existing `ResourcePanel`). */
+  resourcesContent: ReactNode;
+  /** Body for the Checks tab (the `ChecksPanel`, composed by the parent). */
+  checksContent: ReactNode;
+}
+
+interface TabButtonProps {
+  label: string;
+  isActive: boolean;
+  onClick: () => void;
+  showDot?: boolean;
+}
+
+const TabButton: React.FC<TabButtonProps> = ({ label, isActive, onClick, showDot = false }) => (
+  <button
+    aria-selected={isActive}
+    className={cn(
+      'flex items-center gap-1.5 border-b-2 px-1 pb-2 text-base font-semibold transition-colors',
+      isActive
+        ? 'border-blue-600 text-blue-600'
+        : 'text-muted-foreground hover:text-foreground border-transparent'
+    )}
+    role='tab'
+    type='button'
+    onClick={onClick}
+  >
+    {label}
+    {showDot && (
+      <span
+        aria-label='Active checks present'
+        className='inline-block h-2 w-2 shrink-0 rounded-full bg-blue-600'
+        data-testid='checks-notification-dot'
+      />
+    )}
+  </button>
+);
+
+/**
+ * Tabbed left-panel container for the drafting page (W11, §6.6).
+ *
+ * Owns the text-tab header row ("Resources | Checks", blue underline on the
+ * active tab, a blue notification dot right of "Checks" when there are active
+ * findings) and renders the active tab's body below it. The dot's state comes
+ * from the cascade-resolved active count computed at `DraftingPage` level and
+ * threaded in via {@link LeftPanelProps.activeFindingsCount}, so it stays live
+ * while the Resources tab is showing.
+ *
+ * `ResourcePanel` lost its own "Resources" heading; this tab row is now the
+ * panel's header. The component is deliberately content-agnostic (slots for
+ * each tab body) so it can be unit-tested in isolation and so the real
+ * `ChecksPanel`/`ResourcePanel` are composed by the parent.
+ */
+export const LeftPanel: React.FC<LeftPanelProps> = ({
+  activeTab,
+  onTabChange,
+  activeFindingsCount,
+  resourcesContent,
+  checksContent,
+}) => {
+  const hasActiveFindings = activeFindingsCount > 0;
+
+  return (
+    <div className='flex h-full flex-col'>
+      <div aria-label='Left panel' className='flex items-center gap-6 px-1 pt-4' role='tablist'>
+        <TabButton
+          isActive={activeTab === 'resources'}
+          label='Resources'
+          onClick={() => onTabChange('resources')}
+        />
+        {/* The notification dot sits to the right of the "Checks" label only,
+            and is visible whatever tab is active (mock shows it while the
+            Resources tab is active too) — §5.1, W11. */}
+        <TabButton
+          isActive={activeTab === 'checks'}
+          label='Checks'
+          showDot={hasActiveFindings}
+          onClick={() => onTabChange('checks')}
+        />
+      </div>
+
+      <div className='min-h-0 flex-1'>
+        {activeTab === 'resources' ? resourcesContent : checksContent}
+      </div>
+    </div>
+  );
+};
+
+export default LeftPanel;
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.ts b/src/features/checks/hooks/useRepeatedWordsCheck.ts
new file mode 100644
index 00000000..1a3146f0
--- /dev/null
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.ts
@@ -0,0 +1,136 @@
+import { useQuery } from '@tanstack/react-query';
+
+import { config } from '@/lib/config';
+import { Logger } from '@/lib/services/logger';
+import { type ProjectItem } from '@/lib/types';
+
+import {
+  type RepeatedWordsRequest,
+  type RepeatedWordsResponse,
+  type VerseInput,
+} from '../checks.types';
+
+/**
+ * A drafted verse to feed into the check. `verseNumber` is the canonical verse
+ * number; `content` is what the translator currently sees (drafting state, not
+ * a refetch — what they see is what gets checked, per §6.2).
+ */
+export interface CheckVerseInput {
+  verseNumber: number;
+  content: string;
+}
+
+/**
+ * Build the verse identifier per **W4**: `"{bookCode} {chapter}:{verse}"` using
+ * the **USFM book code** (e.g. `"JDG 4:3"`).
+ *
+ * Implementation note (W4 resolved): `ProjectItem.bookCode` carries the USFM
+ * code, whereas `ProjectItem.book` is the human display name. The drafting
+ * page's editor-state persistence happens to store the display name under a key
+ * it (confusingly) also calls `bookCode`; we deliberately use the real USFM
+ * `projectItem.bookCode` here so `snt_id` matches the smoke-test convention.
+ */
+export const buildSntId = (bookCode: string, chapterNumber: number, verseNumber: number): string =>
+  `${bookCode} ${chapterNumber}:${verseNumber}`;
+
+/**
+ * Assemble the snake_case `RepeatedWordsRequest` (D8 shape) from the project
+ * context and the currently drafted verses. Only verses with non-empty content
+ * are sent.
+ */
+export const buildRepeatedWordsRequest = (
+  projectItem: ProjectItem,
+  verses: CheckVerseInput[]
+): RepeatedWordsRequest => {
+  const verseInputs: VerseInput[] = verses
+    .filter(v => v.content.trim() !== '')
+    .map(v => ({
+      snt_id: buildSntId(projectItem.bookCode, projectItem.chapterNumber, v.verseNumber),
+      text: v.content,
+    }));
+
+  return {
+    lang_code: projectItem.targetLanguage,
+    lang_name: projectItem.targetLanguage,
+    project_id: projectItem.projectUnitId,
+    project_name: projectItem.projectName,
+    verses: verseInputs,
+  };
+};
+
+/** POST the request to the fluent-api proxy and return the full envelope (D9). */
+export const postRepeatedWordsCheck = async (
+  request: RepeatedWordsRequest
+): Promise<RepeatedWordsResponse> => {
+  const res = await fetch(`${config.api.url}/ai/tools/greek-room/repeated-words`, {
+    method: 'POST',
+    credentials: 'include',
+    headers: {
+      'Content-Type': 'application/json',
+    },
+    body: JSON.stringify(request),
+  });
+
+  if (!res.ok) {
+    throw new Error('Failed to run repeated-words check');
+  }
+
+  return (await res.json()) as RepeatedWordsResponse;
+};
+
+export interface UseRepeatedWordsCheckParams {
+  projectItem: ProjectItem;
+  /** All currently drafted verses of the chapter (content from drafting state). */
+  verses: CheckVerseInput[];
+  /**
+   * Increments on every successful verse auto-save; part of the query key so
+   * the check re-fires exactly on the auto-save event (W3, card #172).
+   */
+  saveCounter: number;
+  /**
+   * When false the query is disabled — read-only `/view` route, no verse has
+   * content, or any other gating the caller wants to apply (W10).
+   */
+  enabled?: boolean;
+}
+
+/**
+ * Chapter-wide Repeated Word Check query (W3/W4/W10).
+ *
+ * Keyed on `(chapterAssignmentId, saveCounter)` so it re-runs on each
+ * successful auto-save. Consumes the full `ToolJobResponse` envelope and
+ * branches on `status`:
+ *  - `completed` ⇒ the synchronous case today; findings are in `result`.
+ *  - `queued` / `running` ⇒ a future async tool; the shape is identical, so a
+ *    `refetchInterval` can be added later without reshaping callers (D3/D9).
+ *  - `failed` / `cancelled` ⇒ fluent-api surfaces these as HTTP 502 errors, so
+ *    they arrive here as `query.isError` (handled by the panel's inline error,
+ *    W9). TanStack retains the last successful `data` on refetch failure.
+ *
+ * Follows the fetch pattern in `features/bible/hooks/useBibleTarget.ts`.
+ */
+export const useRepeatedWordsCheck = ({
+  projectItem,
+  verses,
+  saveCounter,
+  enabled = true,
+}: UseRepeatedWordsCheckParams) => {
+  const hasContent = verses.some(v => v.content.trim() !== '');
+
+  return useQuery<RepeatedWordsResponse>({
+    queryKey: ['repeated-words', projectItem.chapterAssignmentId, saveCounter],
+    queryFn: async () => {
+      const request = buildRepeatedWordsRequest(projectItem, verses);
+      try {
+        return await postRepeatedWordsCheck(request);
+      } catch (error) {
+        Logger.logException(error, { context: 'Repeated-words check failed' });
+        throw error;
+      }
+    },
+    enabled: enabled && hasContent,
+    // Keep the last successful findings visible on refetch failure (W9); the
+    // panel renders an inline error line above them.
+    retry: false,
+  });
+};
diff --git a/src/features/checks/hooks/useResolvedFindings.test.ts b/src/features/checks/hooks/useResolvedFindings.test.ts
new file mode 100644
index 00000000..511e2240
--- /dev/null
+++ b/src/features/checks/hooks/useResolvedFindings.test.ts
@@ -0,0 +1,232 @@
+import { describe, expect, it } from 'vitest';
+
+import { type GlobalRules, type OccurrenceRules, type RepeatedWordsFinding } from '../checks.types';
+
+import { buildOccurrenceKey, normalizePair, resolveFindings } from './useResolvedFindings';
+
+/**
+ * Factory for a Greek Room finding with sensible defaults. `surf` defaults to a
+ * Title-cased echo of `repeated_word` to prove it is never used for comparison.
+ */
+const makeFinding = (over: Partial<RepeatedWordsFinding> = {}): RepeatedWordsFinding => ({
+  snt_id: 'JDG 4:3',
+  repeated_word: 'the the',
+  surf: 'The the',
+  start_position: 0,
+  legitimate: false,
+  severity: 0.5,
+  ...over,
+});
+
+const NO_OCCURRENCE: OccurrenceRules = {};
+const NO_GLOBAL: GlobalRules = {};
+
+describe('normalizePair', () => {
+  it('NFC-normalizes without case folding', () => {
+    // Decomposed "é" (e + combining acute) vs. composed "é".
+    const decomposed = 'cafe\u0301 cafe\u0301';
+    const composed = 'caf\u00e9 caf\u00e9';
+    expect(normalizePair(decomposed)).toBe(normalizePair(composed));
+    // Case is preserved (no folding).
+    expect(normalizePair('The The')).toBe('The The');
+    expect(normalizePair('The The')).not.toBe(normalizePair('the the'));
+  });
+});
+
+describe('buildOccurrenceKey', () => {
+  it('formats as "{snt_id}|{repeated_word}|{ordinal}" with NFC-normalized pair', () => {
+    expect(buildOccurrenceKey('JDG 4:3', 'the the', 0)).toBe('JDG 4:3|the the|0');
+  });
+
+  it('produces equal keys for composed/decomposed accent variants', () => {
+    const decomposed = buildOccurrenceKey('GEN 1:1', 'cafe\u0301 cafe\u0301', 1);
+    const composed = buildOccurrenceKey('GEN 1:1', 'caf\u00e9 caf\u00e9', 1);
+    expect(decomposed).toBe(composed);
+  });
+});
+
+describe('resolveFindings — Greek Room layer (layer 0)', () => {
+  it('suspicious findings are active', () => {
+    const result = resolveFindings([makeFinding({ legitimate: false })], NO_OCCURRENCE, NO_GLOBAL);
+    expect(result.active).toHaveLength(1);
+    expect(result.inactive).toHaveLength(0);
+  });
+
+  it('legitimate findings are inactive with reason "legitimate"', () => {
+    const result = resolveFindings([makeFinding({ legitimate: true })], NO_OCCURRENCE, NO_GLOBAL);
+    expect(result.active).toHaveLength(0);
+    expect(result.inactive).toHaveLength(1);
+    expect(result.inactive[0].inactiveReason).toBe('legitimate');
+  });
+});
+
+describe('resolveFindings — global layer (layer 1)', () => {
+  it('global suppress hides an otherwise-suspicious finding (reason "global")', () => {
+    const global: GlobalRules = { 'the the': 'suppress' };
+    const result = resolveFindings([makeFinding()], NO_OCCURRENCE, global);
+    expect(result.active).toHaveLength(0);
+    expect(result.inactive[0].inactiveReason).toBe('global');
+  });
+
+  it('global surface re-activates a Greek-Room-legitimate finding', () => {
+    const global: GlobalRules = { 'the the': 'surface' };
+    const result = resolveFindings([makeFinding({ legitimate: true })], NO_OCCURRENCE, global);
+    expect(result.active).toHaveLength(1);
+    expect(result.inactive).toHaveLength(0);
+  });
+
+  it('global rule matches via NFC normalization', () => {
+    // Stored rule composed; finding decomposed.
+    const global: GlobalRules = { [normalizePair('caf\u00e9 caf\u00e9')]: 'suppress' };
+    const finding = makeFinding({ repeated_word: 'cafe\u0301 cafe\u0301' });
+    const result = resolveFindings([finding], NO_OCCURRENCE, global);
+    expect(result.inactive).toHaveLength(1);
+    expect(result.inactive[0].inactiveReason).toBe('global');
+  });
+});
+
+describe('resolveFindings — occurrence layer (layer 2) precedence', () => {
+  it('occurrence suppress beats a silent global (reason "occurrence")', () => {
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|0': 'suppress' };
+    const result = resolveFindings([makeFinding()], occ, NO_GLOBAL);
+    expect(result.inactive[0].inactiveReason).toBe('occurrence');
+  });
+
+  it('occurrence surface overrides a global suppress (most specific wins)', () => {
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|0': 'surface' };
+    const global: GlobalRules = { 'the the': 'suppress' };
+    const result = resolveFindings([makeFinding()], occ, global);
+    expect(result.active).toHaveLength(1);
+    expect(result.inactive).toHaveLength(0);
+  });
+
+  it('occurrence suppress overrides a global surface (most specific wins)', () => {
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|0': 'suppress' };
+    const global: GlobalRules = { 'the the': 'surface' };
+    const result = resolveFindings([makeFinding({ legitimate: true })], occ, global);
+    expect(result.inactive[0].inactiveReason).toBe('occurrence');
+  });
+
+  it('occurrence surface overrides a Greek-Room-legitimate verdict', () => {
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|0': 'surface' };
+    const result = resolveFindings([makeFinding({ legitimate: true })], occ, NO_GLOBAL);
+    expect(result.active).toHaveLength(1);
+  });
+});
+
+describe('resolveFindings — full precedence matrix', () => {
+  // most-specific non-silent verdict wins; layer 0 is the Greek Room verdict.
+  interface Case {
+    occ?: 'suppress' | 'surface';
+    global?: 'suppress' | 'surface';
+    legitimate: boolean;
+    expectActive: boolean;
+    reason?: 'occurrence' | 'global' | 'legitimate';
+  }
+
+  const cases: Case[] = [
+    // occurrence dominates regardless of lower layers
+    {
+      occ: 'suppress',
+      global: 'surface',
+      legitimate: false,
+      expectActive: false,
+      reason: 'occurrence',
+    },
+    { occ: 'surface', global: 'suppress', legitimate: true, expectActive: true },
+    // global decides when occurrence is silent
+    { global: 'suppress', legitimate: false, expectActive: false, reason: 'global' },
+    { global: 'surface', legitimate: true, expectActive: true },
+    // greek room decides when both upper layers silent
+    { legitimate: true, expectActive: false, reason: 'legitimate' },
+    { legitimate: false, expectActive: true },
+  ];
+
+  it.each(cases)(
+    'occ=$occ global=$global legitimate=$legitimate -> active=$expectActive',
+    ({ occ, global, legitimate, expectActive, reason }: Case) => {
+      const occRules: OccurrenceRules = occ ? { 'JDG 4:3|the the|0': occ } : {};
+      const globalRules: GlobalRules = global ? { 'the the': global } : {};
+      const result = resolveFindings([makeFinding({ legitimate })], occRules, globalRules);
+
+      if (expectActive) {
+        expect(result.active).toHaveLength(1);
+        expect(result.inactive).toHaveLength(0);
+      } else {
+        expect(result.inactive).toHaveLength(1);
+        expect(result.inactive[0].inactiveReason).toBe(reason);
+      }
+    }
+  );
+});
+
+describe('resolveFindings — ordinal assignment', () => {
+  it('numbers same-pair findings in a verse by start_position (zero-based)', () => {
+    // Provided out of order; ordinals must follow start_position, not input order.
+    const findings = [
+      makeFinding({ start_position: 40 }),
+      makeFinding({ start_position: 10 }),
+      makeFinding({ start_position: 25 }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
+    const byStart = new Map([...result.active].map(r => [r.finding.start_position, r.ordinal]));
+    expect(byStart.get(10)).toBe(0);
+    expect(byStart.get(25)).toBe(1);
+    expect(byStart.get(40)).toBe(2);
+  });
+
+  it('keeps ordinals independent per repeated_word and per verse', () => {
+    const findings = [
+      makeFinding({ snt_id: 'JDG 4:3', repeated_word: 'the the', start_position: 5 }),
+      makeFinding({ snt_id: 'JDG 4:3', repeated_word: 'and and', start_position: 20 }),
+      makeFinding({ snt_id: 'JDG 4:4', repeated_word: 'the the', start_position: 0 }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
+    // Each is the first (ordinal 0) of its own (verse, pair) group.
+    expect(result.active.every(r => r.ordinal === 0)).toBe(true);
+    expect(result.active.map(r => r.occurrenceKey).sort()).toEqual(
+      ['JDG 4:3|and and|0', 'JDG 4:3|the the|0', 'JDG 4:4|the the|0'].sort()
+    );
+  });
+
+  it('handles the "the the the" overlap: two findings, ordinals 0 and 1', () => {
+    // A triple repetition yields two overlapping consecutive-pair findings.
+    const findings = [
+      makeFinding({ start_position: 0, surf: 'the the' }),
+      makeFinding({ start_position: 4, surf: 'the the' }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
+    const ordinals = result.active.map(r => r.ordinal).sort();
+    expect(ordinals).toEqual([0, 1]);
+    expect(result.active.map(r => r.occurrenceKey).sort()).toEqual(
+      ['JDG 4:3|the the|0', 'JDG 4:3|the the|1'].sort()
+    );
+  });
+
+  it('allows independently suppressing one occurrence of an overlap', () => {
+    const findings = [makeFinding({ start_position: 0 }), makeFinding({ start_position: 4 })];
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|1': 'suppress' };
+    const result = resolveFindings(findings, occ, NO_GLOBAL);
+    expect(result.active).toHaveLength(1);
+    expect(result.active[0].ordinal).toBe(0);
+    expect(result.inactive).toHaveLength(1);
+    expect(result.inactive[0].ordinal).toBe(1);
+    expect(result.inactive[0].inactiveReason).toBe('occurrence');
+  });
+});
+
+describe('resolveFindings — output shape', () => {
+  it('preserves the original finding reference and omits inactiveReason when active', () => {
+    const finding = makeFinding();
+    const result = resolveFindings([finding], NO_OCCURRENCE, NO_GLOBAL);
+    expect(result.active[0].finding).toBe(finding);
+    expect(result.active[0]).not.toHaveProperty('inactiveReason');
+    expect(result.active[0].occurrenceKey).toBe('JDG 4:3|the the|0');
+  });
+
+  it('returns empty buckets for no findings', () => {
+    const result = resolveFindings([], NO_OCCURRENCE, NO_GLOBAL);
+    expect(result.active).toEqual([]);
+    expect(result.inactive).toEqual([]);
+  });
+});
diff --git a/src/features/checks/hooks/useResolvedFindings.ts b/src/features/checks/hooks/useResolvedFindings.ts
new file mode 100644
index 00000000..c4edffb6
--- /dev/null
+++ b/src/features/checks/hooks/useResolvedFindings.ts
@@ -0,0 +1,198 @@
+import { useMemo } from 'react';
+
+import {
+  type GlobalRules,
+  type InactiveReason,
+  type OccurrenceRules,
+  type RepeatedWordsFinding,
+  type ResolvedFinding,
+  type ResolvedFindings,
+  type RuleVerdict,
+} from '../checks.types';
+
+/**
+ * Pure three-layer suppression cascade (W5/W6, §6.4).
+ *
+ * Every finding resolves to active/inactive through three layers; the
+ * **most-specific non-silent verdict wins** (specificity, not temporal order,
+ * so the result is deterministic and reproducible from stored state):
+ *
+ *   | Layer | Scope                              | Verdicts                          |
+ *   | ----- | ---------------------------------- | --------------------------------- |
+ *   | 2     | occurrence (this chapter)          | silent / 'suppress' / 'surface'   |
+ *   | 1     | word pair (all the user's projects)| silent / 'suppress' / 'surface'   |
+ *   | 0     | Greek Room verdict (`legitimate`)  | active / inactive (always present)|
+ *
+ * Layer 2 (occurrence) is more specific than layer 1 (global), which is more
+ * specific than layer 0 (Greek Room). A `'surface'` rule at any layer forces
+ * the finding active (per-occurrence undo of a global rule or of a Greek Room
+ * `legitimate` verdict); a `'suppress'` rule forces it inactive; absence of a
+ * rule (silent) defers to the next-less-specific layer.
+ *
+ * This module is intentionally a pure function (wrapped in a memo hook) with no
+ * I/O — it is exhaustively unit-tested in `useResolvedFindings.test.ts`.
+ */
+
+/**
+ * NFC-normalize a string for comparison/keying. We normalize but deliberately
+ * do **not** case-fold: Unicode case folding is locale-sensitive (the Turkish
+ * dotless-ı problem) and Fluent targets minority languages. Greek Room already
+ * delivers `repeated_word` lowercased, so case equivalence is wholly its policy,
+ * inherited here rather than re-implemented (W4, §6.3).
+ */
+export const normalizePair = (repeatedWord: string): string => repeatedWord.normalize('NFC');
+
+/**
+ * Build the stable occurrence-identity key `"{snt_id}|{repeated_word}|{ordinal}"`.
+ * The `repeated_word` segment is NFC-normalized so a stored rule matches a fresh
+ * finding regardless of composed/decomposed accent representation. The `|`
+ * separator cannot appear in a `snt_id`, and `repeated_word` is the final
+ * segment-pair, so keys are unambiguous (§7.1).
+ */
+export const buildOccurrenceKey = (sntId: string, repeatedWord: string, ordinal: number): string =>
+  `${sntId}|${normalizePair(repeatedWord)}|${ordinal}`;
+
+/**
+ * Assign each finding its ordinal: the index among findings in the **same
+ * verse** with the **same (NFC-normalized) `repeated_word`**, ordered by
+ * `start_position` ("x of n", zero-based). Computed from Greek Room's findings
+ * only — we never tokenize verse text ourselves (W4, §6.3).
+ *
+ * Caveat (§6.3): a triple repetition ("the the the") yields two overlapping
+ * findings, which receive ordinals 0 and 1. Mechanically fine; accepted for v1.
+ */
+const assignOrdinals = (
+  findings: RepeatedWordsFinding[]
+): Array<{ finding: RepeatedWordsFinding; ordinal: number }> => {
+  // Counter per (snt_id, normalized repeated_word). We must process each group
+  // in start_position order so ordinals are stable against unrelated edits.
+  const groups = new Map<string, RepeatedWordsFinding[]>();
+  for (const finding of findings) {
+    const groupKey = `${finding.snt_id}|${normalizePair(finding.repeated_word)}`;
+    const bucket = groups.get(groupKey);
+    if (bucket) {
+      bucket.push(finding);
+    } else {
+      groups.set(groupKey, [finding]);
+    }
+  }
+
+  // Map each finding (by reference identity) to its assigned ordinal.
+  const ordinalOf = new Map<RepeatedWordsFinding, number>();
+  for (const bucket of groups.values()) {
+    const ordered = [...bucket].sort((a, b) => a.start_position - b.start_position);
+    ordered.forEach((finding, index) => ordinalOf.set(finding, index));
+  }
+
+  // Preserve the caller's original finding order in the output.
+  return findings.map(finding => ({
+    finding,
+    ordinal: ordinalOf.get(finding) ?? 0,
+  }));
+};
+
+/**
+ * Resolve one finding through the cascade. Returns whether it is active and, if
+ * not, which layer suppressed it.
+ */
+/**
+ * Look up a verdict in a tri-state rule map. The map's value type is
+ * `RuleVerdict`, but a missing key yields `undefined` at runtime; this helper
+ * makes that `| undefined` explicit (the tsconfig has no `noUncheckedIndexed
+ * AccessProperty`, so a bare index access would over-narrow).
+ */
+const lookupVerdict = (
+  rules: OccurrenceRules | GlobalRules,
+  key: string
+): RuleVerdict | undefined => (key in rules ? rules[key] : undefined);
+
+const resolveOne = (
+  finding: RepeatedWordsFinding,
+  occurrenceKey: string,
+  occurrenceRules: OccurrenceRules,
+  globalRules: GlobalRules
+): { isActive: boolean; inactiveReason?: InactiveReason } => {
+  // Layer 2 — occurrence rule (most specific). A present verdict wins outright:
+  // 'suppress' => inactive (reason 'occurrence'); 'surface' => active.
+  const occurrenceVerdict = lookupVerdict(occurrenceRules, occurrenceKey);
+  if (occurrenceVerdict !== undefined) {
+    return occurrenceVerdict === 'suppress'
+      ? { isActive: false, inactiveReason: 'occurrence' }
+      : { isActive: true };
+  }
+
+  // Layer 1 — user-global word-pair rule.
+  const globalVerdict = lookupVerdict(globalRules, normalizePair(finding.repeated_word));
+  if (globalVerdict !== undefined) {
+    return globalVerdict === 'suppress'
+      ? { isActive: false, inactiveReason: 'global' }
+      : { isActive: true };
+  }
+
+  // Layer 0 — Greek Room verdict.
+  if (finding.legitimate) {
+    return { isActive: false, inactiveReason: 'legitimate' };
+  }
+  return { isActive: true };
+};
+
+/**
+ * Resolve a list of raw findings into `{ active[], inactive[] }` given the
+ * occurrence- and global-rule maps. Pure; safe to call outside React.
+ */
+export const resolveFindings = (
+  findings: RepeatedWordsFinding[],
+  occurrenceRules: OccurrenceRules,
+  globalRules: GlobalRules
+): ResolvedFindings => {
+  const withOrdinals = assignOrdinals(findings);
+
+  const active: ResolvedFinding[] = [];
+  const inactive: ResolvedFinding[] = [];
+
+  for (const { finding, ordinal } of withOrdinals) {
+    const occurrenceKey = buildOccurrenceKey(finding.snt_id, finding.repeated_word, ordinal);
+    const { isActive, inactiveReason } = resolveOne(
+      finding,
+      occurrenceKey,
+      occurrenceRules,
+      globalRules
+    );
+
+    const resolved: ResolvedFinding = {
+      finding,
+      ordinal,
+      occurrenceKey,
+      isActive,
+      ...(isActive ? {} : { inactiveReason }),
+    };
+
+    if (isActive) {
+      active.push(resolved);
+    } else {
+      inactive.push(resolved);
+    }
+  }
+
+  return { active, inactive };
+};
+
+export interface UseResolvedFindingsParams {
+  findings: RepeatedWordsFinding[];
+  occurrenceRules: OccurrenceRules;
+  globalRules: GlobalRules;
+}
+
+/**
+ * React hook wrapper around {@link resolveFindings}, memoized on its inputs.
+ * The notification dot counts `active.length`.
+ */
+export const useResolvedFindings = ({
+  findings,
+  occurrenceRules,
+  globalRules,
+}: UseResolvedFindingsParams): ResolvedFindings =>
+  useMemo(
+    () => resolveFindings(findings, occurrenceRules, globalRules),
+    [findings, occurrenceRules, globalRules]
+  );
diff --git a/src/features/resources/components/ResourcePanel.tsx b/src/features/resources/components/ResourcePanel.tsx
index 929a1aa5..8868fe9b 100644
--- a/src/features/resources/components/ResourcePanel.tsx
+++ b/src/features/resources/components/ResourcePanel.tsx
@@ -410,10 +410,9 @@ export const ResourcePanel: React.FC<ResourcePanelProps> = ({
 
   return (
     <aside className='bg-background flex h-full flex-col'>
+      {/* The "Resources" heading was removed: the shared LeftPanel tab row
+          ("Resources | Checks") now serves as this panel's header (W11, §6.6). */}
       <div className='bg-background top-0 py-4'>
-        <div className='flex items-center gap-2'>
-          <h3 className='text-foreground text-xl font-semibold'>Resources</h3>
-        </div>
         <ResourceChipRow
           resourceNames={resourceNames}
           selectedResource={selectedResource}

From 1a0a4d2cbfde8bb4ed101c363a25d2d67370b40b Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 18 Jun 2026 14:15:22 -0400
Subject: [PATCH 09/33] =?UTF-8?q?docs(checks-ui):=20add=20=C2=A76.1.1=20im?=
 =?UTF-8?q?plementation-status=20checklist?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Track per-deliverable build progress on the implementation branch. The
proposal branch / PR #305 deliberately preserves the as-reviewed,
all-unchecked state; this checklist advances on jel-checks-ui-impl as
files land. Marks the committed foundations done and the held items
(DraftingPage wiring, ChecksPanel/FindingRow, useSuppressions, the
fluent-api /self/settings backend) as pending.
---
 .../checks-ui-integration-progress.md         |   0
 .../checks-ui-integration-suggestion.md       | 159 +++++++++---------
 2 files changed, 78 insertions(+), 81 deletions(-)
 create mode 100644 docs/proposals/repeated-word-check/checks-ui-integration-progress.md

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-progress.md b/docs/proposals/repeated-word-check/checks-ui-integration-progress.md
new file mode 100644
index 00000000..e69de29b
diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index 43756f3a..af497233 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -24,14 +24,14 @@ This document's own decisions are numbered **W1–W12** (W = web) to avoid colli
 
 Revisions made in response to PR [#305](https://github.com/eten-tech-foundation/fluent-web/pull/305) review (reviewers: kaseywright, chadw-eten; product input via joelthe1/Ulf). The cards themselves were edited mid-review; this revision tracks the **current** card text and mocks (re-pulled 2026-06-16).
 
-| Change | Driver | Where |
-| --- | --- | --- |
-| Settings endpoint moves from `GET/PUT /users/settings` to a new **`self` domain**: `GET/PUT /self/settings`, guarded by `authenticateUser` alone. | kaseywright (CHANGES_REQUESTED): clean semantic separation, no redundant `{id}` param, no over-exposure risk, simpler middleware (no `requireSelf`/`requireUserAccess`/`UserPolicy`). | W7, §7.2, §8 |
-| Button labels become **"Ignore Here" / "Ignore Everywhere"**. | joelthe1/Ulf (card #278 comment); chadw updated the #278 mocks to match. | W6, §5, §6.5 |
-| **"Ignore Everywhere" now shows a confirmation dialog**, while remaining reversible via `[Undo ▾]`. | card #278 ("Ignore Always should have a standard confirm dialog"). Kept the three-layer cascade & undo (no explicit directive against it; ambiguity of "undo" with surface/suppress entries at multiple layers is better resolved by the cascade — see §6.4/§6.5); added the dialog. | W6, §5.3, §6.5 |
-| **"Show Ignored" toggle**: now default **OFF** and **not persisted** (resets per session). Ignored items show the ignore-type label (e.g. _"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) + an **"Undo Ignore"** control. | card #278 was revised to specify exactly this; we drop our earlier default-ON/persisted proposal and the `showResolvedChecks` editor-state key. | W5, W6, W11, §5.3, §6.4, §6.6, §7.1 |
-| §5.2 mock/text inconsistencies **S1 (zero-state dot)** and **S2 (language dropdown)** are **resolved**: chadw fixed the zero-state mock and removed the language dropdown from the #278 Checks mocks. | chadw-eten ("Fixed." / "good idea… I have adjusted the mocks."). | §5.2, §12 |
-| New requirement adopted: **highlight repeated words in the verse text** (underline/color). | joelthe1/Ulf (card #278 comment). We will implement it; because it touches the verse-rendering pane (outside the left panel this proposal builds) and warrants its own design pass, it ships as a **separate, follow-on PR** so the tab/panel work can land independently. §5.3 item 4 records the intent; the visual/interaction details are specced just before that PR. | §5.3, §11 |
+| Change                                                                                                                                                                                                                                | Driver                                                                                                                                                                                                                                                                                                                                                                     | Where                               |
+| ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------- |
+| Settings endpoint moves from `GET/PUT /users/settings` to a new **`self` domain**: `GET/PUT /self/settings`, guarded by `authenticateUser` alone.                                                                                     | kaseywright (CHANGES_REQUESTED): clean semantic separation, no redundant `{id}` param, no over-exposure risk, simpler middleware (no `requireSelf`/`requireUserAccess`/`UserPolicy`).                                                                                                                                                                                      | W7, §7.2, §8                        |
+| Button labels become **"Ignore Here" / "Ignore Everywhere"**.                                                                                                                                                                         | joelthe1/Ulf (card #278 comment); chadw updated the #278 mocks to match.                                                                                                                                                                                                                                                                                                   | W6, §5, §6.5                        |
+| **"Ignore Everywhere" now shows a confirmation dialog**, while remaining reversible via `[Undo ▾]`.                                                                                                                                   | card #278 ("Ignore Always should have a standard confirm dialog"). Kept the three-layer cascade & undo (no explicit directive against it; ambiguity of "undo" with surface/suppress entries at multiple layers is better resolved by the cascade — see §6.4/§6.5); added the dialog.                                                                                       | W6, §5.3, §6.5                      |
+| **"Show Ignored" toggle**: now default **OFF** and **not persisted** (resets per session). Ignored items show the ignore-type label (e.g. _"Ignore Here"_, _"Ignore Everywhere"_, _"Default Ignore"_) + an **"Undo Ignore"** control. | card #278 was revised to specify exactly this; we drop our earlier default-ON/persisted proposal and the `showResolvedChecks` editor-state key.                                                                                                                                                                                                                            | W5, W6, W11, §5.3, §6.4, §6.6, §7.1 |
+| §5.2 mock/text inconsistencies **S1 (zero-state dot)** and **S2 (language dropdown)** are **resolved**: chadw fixed the zero-state mock and removed the language dropdown from the #278 Checks mocks.                                 | chadw-eten ("Fixed." / "good idea… I have adjusted the mocks.").                                                                                                                                                                                                                                                                                                           | §5.2, §12                           |
+| New requirement adopted: **highlight repeated words in the verse text** (underline/color).                                                                                                                                            | joelthe1/Ulf (card #278 comment). We will implement it; because it touches the verse-rendering pane (outside the left panel this proposal builds) and warrants its own design pass, it ships as a **separate, follow-on PR** so the tab/panel work can land independently. §5.3 item 4 records the intent; the visual/interaction details are specced just before that PR. | §5.3, §11                           |
 
 ---
 
@@ -54,12 +54,12 @@ The ignore actions require server-side persistence that does not exist yet. #172
 
 ### 1.2 Repos touched
 
-| Repo | Touched | What changes |
-| --- | --- | --- |
-| **fluent-web** | yes (bulk) | Left-panel tab container, Checks panel feature, check-trigger hook, suppression cascade, editor-state keys. |
-| **fluent-api** | yes (small) | New `user_settings` table + a new `self` domain exposing `GET/PUT /self/settings`; Zod-schema extension of the editor-state `resources` blob (no migration). |
-| **fluent-ai** | no | The check itself is unchanged. |
-| **fluent-platform** | no | No new services or env vars. |
+| Repo                | Touched     | What changes                                                                                                                                                 |
+| ------------------- | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| **fluent-web**      | yes (bulk)  | Left-panel tab container, Checks panel feature, check-trigger hook, suppression cascade, editor-state keys.                                                  |
+| **fluent-api**      | yes (small) | New `user_settings` table + a new `self` domain exposing `GET/PUT /self/settings`; Zod-schema extension of the editor-state `resources` blob (no migration). |
+| **fluent-ai**       | no          | The check itself is unchanged.                                                                                                                               |
+| **fluent-platform** | no          | No new services or env vars.                                                                                                                                 |
 
 ---
 
@@ -92,20 +92,20 @@ The ignore actions require server-side persistence that does not exist yet. #172
 
 Restated conclusions; supporting analysis in the cited sections.
 
-| # | Decision | Short rationale |
-| --- | --- | --- |
-| **W1** | Suppression persistence ships as an extension of #172's scope; no new product card. | #172's wording covers backend dependencies of the UI work; D1's deferral was about caching tool results, not user preferences. See §1.1. |
-| **W2** | Hybrid storage: "Ignore This Time" lives in the existing per-chapter editor-state JSONB (schema-key addition, no migration); "Ignore Always" lives in a new `user_settings` table (one row per user, single Zod-typed JSONB column). Findings are filtered **client-side**. | Each rule is stored at exactly the scope it governs. Client-side filtering keeps the AI proxy a pure pass-through (D8/D9). `user_settings` deliberately establishes Fluent's user-global preference store: future settings become schema extensions, not migrations. See §7. |
-| **W3** | Chapter-wide check on every successful verse save: a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)`, where `saveCounter` increments in the verse-save mutation's `onSuccess`. The request sends **all currently drafted verses** of the chapter. No extra coalescing debounce. | The dot needs chapter-wide awareness; per-verse merging is more code for no user-visible gain. The existing 2 s per-verse save debounce already rate-limits; the check is <1 s. KISS — optimize if heavier checks arrive. See §6.2. |
-| **W4** | `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM book code, e.g. `JDG 4:3`), matching the convention already used by the repo smoke tests. Occurrence identity for suppressions = `(snt_id, repeated_word, ordinal)` where ordinal is the index of the finding among same-`repeated_word` findings in the verse, ordered by `start_position`. | Ordinals survive unrelated edits (positions don't); adding/removing an earlier same-pair occurrence conservatively re-flags, the safe failure direction. We count Greek Room's findings, never tokenize text ourselves, so Greek Room's equivalence policy is inherited consistently. See §6.3. |
-| **W5** | Every finding resolves to **active/inactive** via a three-layer cascade — Greek Room verdict (`legitimate`), user-global word-pair rule, occurrence rule — where the **most specific non-silent verdict wins**. Rules are tri-state maps (`absent / 'suppress' / 'surface'`). The dot counts active findings only. Inactive findings are hidden by default and revealed by the **"Show Ignored" toggle** (default OFF, not persisted, per revised #278); when shown they render greyed with the ignore-type label. | One mechanism explains machine-legitimate and user-ignored alike ("pre-ignored by Greek Room"). Specificity, not temporal order, keeps resolution deterministic. Greyed-not-removed (under the toggle) keeps undo discoverable. See §6.4. |
-| **W6** | Active findings show `[Ignore Here] [Ignore Everywhere]` (per revised card). `[Ignore Everywhere]` opens a **standard confirm dialog** before writing the global rule. Inactive findings show one `[Undo ▾]` split button: default click acts at the occurrence layer (delete own rule, or write an occurrence `surface` override); the chevron menu offers explicit global actions with consequence-naming labels. Any global write first purges the user's occurrence rules for that pair **in the current chapter's editor state only**. | Default click never silently edits global state. The card asks for a confirm on the global action because it reads as irreversible; we add the dialog **and** keep the `[Undo ▾]` reversibility — the two are complementary (confirm guards the click; undo guards the mistake). Purge-local prevents the just-clicked panel from appearing to ignore the action; other chapters' specific pronouncements deliberately stand. See §6.5. |
-| **W7** | Settings endpoint lives in a **new `self` domain**: **`GET /self/settings` + `PUT /self/settings`**, session-implicit user (no user in URL), guarded by `authenticateUser` alone, full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/self/settings/self-settings.{route,service,repository,types}.ts`. | Per kaseywright: a `self` domain gives clean semantic separation, avoids the redundant `{userId}` param (and the `requireSelf`/`UserPolicy` middleware that `/users/{id}/…` routes carry), and prevents accidental over-exposure (no `{id}` to misuse). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8. |
-| **W8** | Graceful degradation by feature detection: if `GET /self/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Everywhere]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first. | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9. |
-| **W9** | Check-call failure surfaces as a single inline `text-sm text-red-500` line at the top of the Checks panel ("Checks failed to refresh"); the panel keeps rendering the last successful findings (TanStack keeps `query.data` on refetch failure) and the dot reflects that last-known state. Failures log via the existing `Logger`. No toast/banner/popup. | Matches the drafting page's own inline-status precedent ("Auto-save failed"). Failure mode degrades to "results are one save behind." See §9.2. |
-| **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content and while the settings feature-detection probe (§9.1) is unresolved. | Card scope is the drafting view; empty chapters have nothing to check; the probe is one fast `GET`, and waiting for it means the first findings render is already cascade-correct. Review-stage checks are future work. See §6.2. |
-| **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets, the two buttons, and a "Show Ignored" toggle at the bottom; zero state inside the accordion. Only `activeLeftTab` is persisted in editor state (the "Show Ignored" state resets per session, per revised #278). Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5. |
-| **W12** | One proposal document (this file) covering both repos; two implementation PRs (fluent-web, fluent-api), cross-referencing each other and the cards. | Splitting the proposal doubles reviewer overhead for a design whose halves only make sense together. See §9.4. |
+| #       | Decision                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 | Short rationale                                                                                                                                                                                                                                                                                                                                                                                                                         |
+| ------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| **W1**  | Suppression persistence ships as an extension of #172's scope; no new product card.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      | #172's wording covers backend dependencies of the UI work; D1's deferral was about caching tool results, not user preferences. See §1.1.                                                                                                                                                                                                                                                                                                |
+| **W2**  | Hybrid storage: "Ignore This Time" lives in the existing per-chapter editor-state JSONB (schema-key addition, no migration); "Ignore Always" lives in a new `user_settings` table (one row per user, single Zod-typed JSONB column). Findings are filtered **client-side**.                                                                                                                                                                                                                                                                                                              | Each rule is stored at exactly the scope it governs. Client-side filtering keeps the AI proxy a pure pass-through (D8/D9). `user_settings` deliberately establishes Fluent's user-global preference store: future settings become schema extensions, not migrations. See §7.                                                                                                                                                            |
+| **W3**  | Chapter-wide check on every successful verse save: a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)`, where `saveCounter` increments in the verse-save mutation's `onSuccess`. The request sends **all currently drafted verses** of the chapter. No extra coalescing debounce.                                                                                                                                                                                                                                                                                        | The dot needs chapter-wide awareness; per-verse merging is more code for no user-visible gain. The existing 2 s per-verse save debounce already rate-limits; the check is <1 s. KISS — optimize if heavier checks arrive. See §6.2.                                                                                                                                                                                                     |
+| **W4**  | `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM book code, e.g. `JDG 4:3`), matching the convention already used by the repo smoke tests. Occurrence identity for suppressions = `(snt_id, repeated_word, ordinal)` where ordinal is the index of the finding among same-`repeated_word` findings in the verse, ordered by `start_position`.                                                                                                                                                                                                                                           | Ordinals survive unrelated edits (positions don't); adding/removing an earlier same-pair occurrence conservatively re-flags, the safe failure direction. We count Greek Room's findings, never tokenize text ourselves, so Greek Room's equivalence policy is inherited consistently. See §6.3.                                                                                                                                         |
+| **W5**  | Every finding resolves to **active/inactive** via a three-layer cascade — Greek Room verdict (`legitimate`), user-global word-pair rule, occurrence rule — where the **most specific non-silent verdict wins**. Rules are tri-state maps (`absent / 'suppress' / 'surface'`). The dot counts active findings only. Inactive findings are hidden by default and revealed by the **"Show Ignored" toggle** (default OFF, not persisted, per revised #278); when shown they render greyed with the ignore-type label.                                                                       | One mechanism explains machine-legitimate and user-ignored alike ("pre-ignored by Greek Room"). Specificity, not temporal order, keeps resolution deterministic. Greyed-not-removed (under the toggle) keeps undo discoverable. See §6.4.                                                                                                                                                                                               |
+| **W6**  | Active findings show `[Ignore Here] [Ignore Everywhere]` (per revised card). `[Ignore Everywhere]` opens a **standard confirm dialog** before writing the global rule. Inactive findings show one `[Undo ▾]` split button: default click acts at the occurrence layer (delete own rule, or write an occurrence `surface` override); the chevron menu offers explicit global actions with consequence-naming labels. Any global write first purges the user's occurrence rules for that pair **in the current chapter's editor state only**.                                              | Default click never silently edits global state. The card asks for a confirm on the global action because it reads as irreversible; we add the dialog **and** keep the `[Undo ▾]` reversibility — the two are complementary (confirm guards the click; undo guards the mistake). Purge-local prevents the just-clicked panel from appearing to ignore the action; other chapters' specific pronouncements deliberately stand. See §6.5. |
+| **W7**  | Settings endpoint lives in a **new `self` domain**: **`GET /self/settings` + `PUT /self/settings`**, session-implicit user (no user in URL), guarded by `authenticateUser` alone, full-replace upsert of one Zod-typed JSONB blob. File quartet `domains/self/settings/self-settings.{route,service,repository,types}.ts`.                                                                                                                                                                                                                                                               | Per kaseywright: a `self` domain gives clean semantic separation, avoids the redundant `{userId}` param (and the `requireSelf`/`UserPolicy` middleware that `/users/{id}/…` routes carry), and prevents accidental over-exposure (no `{id}` to misuse). Last-writer-wins on concurrent tabs is inherited from editor-state and accepted. See §8.                                                                                        |
+| **W8**  | Graceful degradation by feature detection: if `GET /self/settings` 404s, the session records `globalIgnoresAvailable = false` and the `[Ignore Everywhere]` button plus global menu entries are **not rendered** (capability hidden, never a dead control). Unknown/absent JSONB keys parse as empty on both sides. Either repo's PR can land first.                                                                                                                                                                                                                                     | The UI must not assume the backend half exists. An invisible capability is honest; a dead button is a bug report. See §9.                                                                                                                                                                                                                                                                                                               |
+| **W9**  | Check-call failure surfaces as a single inline `text-sm text-red-500` line at the top of the Checks panel ("Checks failed to refresh"); the panel keeps rendering the last successful findings (TanStack keeps `query.data` on refetch failure) and the dot reflects that last-known state. Failures log via the existing `Logger`. No toast/banner/popup.                                                                                                                                                                                                                               | Matches the drafting page's own inline-status precedent ("Auto-save failed"). Failure mode degrades to "results are one save behind." See §9.2.                                                                                                                                                                                                                                                                                         |
+| **W10** | The check runs in drafting mode only — not in the read-only `/view` route — and is skipped (`enabled: false`) when no verse has content and while the settings feature-detection probe (§9.1) is unresolved.                                                                                                                                                                                                                                                                                                                                                                             | Card scope is the drafting view; empty chapters have nothing to check; the probe is one fast `GET`, and waiting for it means the first findings render is already cascade-correct. Review-stage checks are future work. See §6.2.                                                                                                                                                                                                       |
+| **W11** | Left-panel architecture follows the mocks: a text-tab header row ("Resources \| Checks", blue underline active state, blue dot after "Checks"), dot visible from either tab; Checks content = per-check accordion sections with verse-grouped snippets, the two buttons, and a "Show Ignored" toggle at the bottom; zero state inside the accordion. Only `activeLeftTab` is persisted in editor state (the "Show Ignored" state resets per session, per revised #278). Proposed deviation for sign-off: when the whole panel is closed, the dot is mirrored on the panel-toggle button. | Mock-faithful where the mocks speak; the toggle-button dot preserves #277's intent (translator is notified) when the panel is hidden. See §5.                                                                                                                                                                                                                                                                                           |
+| **W12** | One proposal document (this file) covering both repos; two implementation PRs (fluent-web, fluent-api), cross-referencing each other and the cards.                                                                                                                                                                                                                                                                                                                                                                                                                                      | Splitting the proposal doubles reviewer overhead for a design whose halves only make sense together. See §9.4.                                                                                                                                                                                                                                                                                                                          |
 
 ---
 
@@ -141,7 +141,7 @@ The response envelope is consumed whole (D9): the hook inspects `status` and `re
 
 The cards' mockups (drafting page, Judges 4, Gujarati IRV project) define the visual target. The #278 mocks were updated during the first review round (§0); the description below tracks the **current** mocks:
 
-- **Tab header (#277 mock):** the left panel's current `Resources` heading is replaced by a text-tab row — **"Resources | Checks"** — active tab in blue with a blue underline, inactive tab plain. The **notification dot is a solid blue filled circle immediately right of the "Checks" label**, and the mock shows it while the *Resources* tab is active: the dot is visible from either tab whenever the panel is open.
+- **Tab header (#277 mock):** the left panel's current `Resources` heading is replaced by a text-tab row — **"Resources | Checks"** — active tab in blue with a blue underline, inactive tab plain. The **notification dot is a solid blue filled circle immediately right of the "Checks" label**, and the mock shows it while the _Resources_ tab is active: the dot is visible from either tab whenever the panel is open.
 - **Checks content (#278 mocks):** below the tab header, a card containing a **collapsible "Repeated Words" accordion section** (expanded by default). The accordion-per-check structure anticipates future sibling checks (the board already holds draft cards for Greek Room Wildebeest and Spell checks) without UI rework. Inside the section, findings are grouped by verse:
   - bold **"Verse N"** heading per verse that has findings (verses without findings get no group);
   - one row per occurrence: a one-line context snippet showing the repeated pair in context;
@@ -221,14 +221,14 @@ useQuery({
 - **Request body** is the full `RepeatedWordsRequest` (D8 shape): `lang_code`/`lang_name` from the project's target language, `project_id`/`project_name` from `projectItem`, and `verses[]` covering **all currently drafted verses** of the chapter (content from the drafting state, not a refetch — what the translator sees is what gets checked).
 - **`snt_id` = `"{bookCode} {chapter}:{verse}"`** with the USFM book code (e.g. `JDG 4:3`), the convention the repo smoke tests already use. Implementation note: verify the field carrying the USFM code on the drafting page's `projectItem` (vs. display name) and thread it into the builder.
 - **Where it runs (W10):** drafting route only; `enabled` is false in the read-only `/view` route (reviewers see no Checks activity in v1), when no verse has content, and while the settings feature-detection probe (§9.1) is unresolved — the probe is a single fast `GET`, and waiting for it means the first findings render is already cascade-correct (the user's global rules are known, present or absent).
-- **Emptied chapters resolve naturally:** any save of emptied text triggers a fresh check whose empty findings clear the panel and dot. (When *every* verse is emptied the query disables instead; the panel and dot then treat findings as empty rather than rendering stale data.)
+- **Emptied chapters resolve naturally:** any save of emptied text triggers a fresh check whose empty findings clear the panel and dot. (When _every_ verse is emptied the query disables instead; the panel and dot then treat findings as empty rather than rendering stale data.)
 
 ### 6.3 Occurrence identity (W4)
 
 A suppression must survive verse edits without our re-parsing text. The key is **`(snt_id, repeated_word, ordinal)`**:
 
 - `ordinal` = index of this finding among findings in the same verse with the same `repeated_word`, ordered by `start_position` ("x of n"). Computed from **Greek Room's findings only** — we never tokenize verse text ourselves, so Greek Room's case/diacritic equivalence policy is inherited and consistent between runs.
-- Ordinals survive unrelated edits (`start_position` does not). Adding or removing an *earlier* same-pair occurrence shifts later ordinals and conservatively re-flags — the safe failure direction.
+- Ordinals survive unrelated edits (`start_position` does not). Adding or removing an _earlier_ same-pair occurrence shifts later ordinals and conservatively re-flags — the safe failure direction.
 - String comparison between a stored rule's `repeated_word` and a fresh finding's: **NFC-normalize both, compare exactly, no case folding of our own.** Unicode case folding is locale-sensitive (the Turkish dotless-ı problem) and Fluent targets minority languages; NFC handles composed/decomposed accent variation. Note that Greek Room already delivers `repeated_word` lowercased ("word word" form, per the fluent-ai schema) — original casing lives only in `surf`, which we display but never compare — so case equivalence is wholly Greek Room's policy, inherited rather than re-implemented.
 - Documented caveat: a triple repetition ("the the the") yields two overlapping findings (ordinals 1 and 2). Mechanically fine; slightly odd UX; accepted for v1.
 
@@ -236,11 +236,11 @@ A suppression must survive verse edits without our re-parsing text. The key is *
 
 Every finding resolves through three layers; the **most specific non-silent verdict wins** (specificity, not temporal order — deterministic and reproducible from stored state):
 
-| Layer | Scope | Verdicts |
-| --- | --- | --- |
-| 0 — Greek Room | this finding | `active` (suspicious) / `inactive` (`legitimate: true`) — always present |
-| 1 — User global | word pair, all the user's projects | _silent_ / `suppress` / `surface` |
-| 2 — Occurrence | `(snt_id, repeated_word, ordinal)`, this chapter assignment | _silent_ / `suppress` / `surface` |
+| Layer           | Scope                                                       | Verdicts                                                                 |
+| --------------- | ----------------------------------------------------------- | ------------------------------------------------------------------------ |
+| 0 — Greek Room  | this finding                                                | `active` (suspicious) / `inactive` (`legitimate: true`) — always present |
+| 1 — User global | word pair, all the user's projects                          | _silent_ / `suppress` / `surface`                                        |
+| 2 — Occurrence  | `(snt_id, repeated_word, ordinal)`, this chapter assignment | _silent_ / `suppress` / `surface`                                        |
 
 - Stored rules are **tri-state maps** (`key → 'suppress' | 'surface'`, absent = silent), not bare suppression sets — `surface` is what makes per-occurrence undo of a global rule (or of a Greek Room `legitimate` verdict) possible.
 - A machine-legitimate finding is conceptually "pre-ignored by Greek Room": same inactive state as user-ignored, different actor, same undo affordance.
@@ -253,17 +253,18 @@ Every finding resolves through three layers; the **most specific non-silent verd
 - **Confirm dialog on `[Ignore Everywhere]`** (revised #278): the card asks for it because the action reads as global and irreversible. We honor the request **and** keep the action reversible via `[Undo ▾]` below — the dialog guards the deliberate click; the undo guards a mistake that slips past it. The two are not redundant: confirm is pre-action friction, undo is post-action recourse.
 
   Why reversibility falls out of the design rather than being an extra feature we chose to bolt on — the implication chain (see also §6.4, which this depends on):
-  1. Greek Room reports **both** polarities: repetitions it flags (suspicious) and repetitions it judges `legitimate` (a *negative* / "don't flag" verdict).
+  1. Greek Room reports **both** polarities: repetitions it flags (suspicious) and repetitions it judges `legitimate` (a _negative_ / "don't flag" verdict).
   2. Once we decide to **show** the legitimate ones (§5.3 item 2 — so the translator can see what the checker waved through), it's natural to let the translator disagree and mark one **illegitimate** (re-surface it).
   3. Expressing "re-surface this" requires a **positive** record (`surface`), not only the negative (`suppress`) records an ignore-list would hold. So the stores must be tri-state, not suppression-only (§6.4).
-  4. Given positive *and* negative records at more than one scope, a **layered** model is the natural fit: a local (occurrence) record can override a global (word-pair) one, and the override works in **both polarities** (locally surface something suppressed globally, or locally suppress something surfaced).
-  5. With that layering in place, a single bare "undo" is **ambiguous** — at which layer, in which polarity? Resolving that ambiguity (the `[Undo ▾]` split button: default = act at the occurrence layer; chevron = act globally) is what the design *has* to do anyway — and the same mechanism incidentally gives the translator the "oops, I hit the wrong button" recourse for `[Ignore Everywhere]`. Reversibility is therefore inherited, not added.
+  4. Given positive _and_ negative records at more than one scope, a **layered** model is the natural fit: a local (occurrence) record can override a global (word-pair) one, and the override works in **both polarities** (locally surface something suppressed globally, or locally suppress something surfaced).
+  5. With that layering in place, a single bare "undo" is **ambiguous** — at which layer, in which polarity? Resolving that ambiguity (the `[Undo ▾]` split button: default = act at the occurrence layer; chevron = act globally) is what the design _has_ to do anyway — and the same mechanism incidentally gives the translator the "oops, I hit the wrong button" recourse for `[Ignore Everywhere]`. Reversibility is therefore inherited, not added.
 
   We are not requesting that this override the card's "cannot be undone" framing; it is documented here for anyone who wants to follow the reasoning, and the implemented PR is reviewed again before merge.
+
 - **Inactive finding:** one `[Undo ▾]` split button (shadcn `dropdown-menu`, already in the codebase — no hidden gestures):
-  - **Default click** acts at the occurrence layer: if inactive via its *own* occurrence rule → delete that entry; if inactive via a global rule or Greek Room's verdict → write an occurrence-level `surface` override (the global rule / machine verdict survives; only this occurrence resurfaces).
+  - **Default click** acts at the occurrence layer: if inactive via its _own_ occurrence rule → delete that entry; if inactive via a global rule or Greek Room's verdict → write an occurrence-level `surface` override (the global rule / machine verdict survives; only this occurrence resurfaces).
   - **Chevron menu** offers the deliberate global actions with consequence-naming labels, e.g. _"Stop ignoring 'the the' everywhere"_ (deletes the global entry).
-- **Global writes purge local rules — current chapter only.** Executing any global (everywhere) suppress or surface first deletes the user's occurrence-level rules for that word pair in the *current chapter assignment's* editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
+- **Global writes purge local rules — current chapter only.** Executing any global (everywhere) suppress or surface first deletes the user's occurrence-level rules for that word pair in the _current chapter assignment's_ editor state, then writes the global rule. This is purely UI coherence: without it, the just-clicked panel would appear to ignore the action (occurrence beats global in the cascade). Occurrence rules in **other** chapters/projects deliberately stand — they were specific pronouncements, and we don't revert decisions from afar. Client-side operation; no special endpoint semantics.
 - **Direct edits resolve silently** (card #278): if the translator fixes the text, the next auto-save's fresh findings simply no longer contain the occurrence — no action needed, panel and dot update.
 
 ### 6.6 Left-panel container and persisted UI state (W11)
@@ -273,7 +274,7 @@ Every finding resolves through three layers; the **most specific non-silent verd
 - **`LeftPanel`** owns the tab header row ("Resources | Checks", blue underline on the active tab, blue dot right of "Checks" when active findings exist) and renders either the existing `ResourcePanel` or the new `ChecksPanel` below it. `ResourcePanel` loses only its `<h3>Resources</h3>` heading (the tab row replaces it); its content, language dropdown, and accordion are untouched.
 - The dot's state comes from the cascade-resolved active count (§6.4), which `DraftingPage` computes once and threads to both `LeftPanel` (tab dot) and the header toggle button (mirror dot, §5.3 item 3). The check query runs at `DraftingPage` level — **not** inside `ChecksPanel` — so the dot stays live while the Resources tab (or a closed panel) is showing.
 - **Persisted UI state.** One new key rides the same per-chapter editor-state blob the page already saves (debounced 500 ms, §7.1): `activeLeftTab: 'resources' | 'checks'` (reopen where you left off). The **"Show Ignored"** toggle is deliberately **not** persisted — per revised #278 it resets to off each session — so it lives in component state only. The persisted key is cosmetic; failure to persist degrades to defaults.
-- The panel's existing resize/drag behavior and 20–40 % width constraints are inherited unchanged — `LeftPanel` lives *inside* the resizable container.
+- The panel's existing resize/drag behavior and 20–40 % width constraints are inherited unchanged — `LeftPanel` lives _inside_ the resizable container.
 
 ---
 
@@ -299,7 +300,7 @@ export const editorStateResourcesSchema = z
   .nullable();
 ```
 
-"Ignore This Time" is scoped to *this user's view of this chapter assignment* — precisely this table's grain. The fluent-api PR extends the schema with **optional** keys (old rows parse unchanged; **no SQL migration** — the column is already JSONB):
+"Ignore This Time" is scoped to _this user's view of this chapter assignment_ — precisely this table's grain. The fluent-api PR extends the schema with **optional** keys (old rows parse unchanged; **no SQL migration** — the column is already JSONB):
 
 ```ts
 export const editorStateResourcesSchema = z
@@ -312,9 +313,7 @@ export const editorStateResourcesSchema = z
     tabStatus: z.boolean(),
     // --- NEW, all optional (backward/forward compatible) ---
     activeLeftTab: z.enum(['resources', 'checks']).optional(),
-    checkOccurrenceRules: z
-      .record(z.string(), z.enum(['suppress', 'surface']))
-      .optional(),
+    checkOccurrenceRules: z.record(z.string(), z.enum(['suppress', 'surface'])).optional(),
   })
   .nullable();
 ```
@@ -331,9 +330,7 @@ export const editorStateResourcesSchema = z
 // fluent-api/src/db/schema.ts — NEW
 export const userSettingsSchema = z
   .object({
-    checkIgnoredWordPairs: z
-      .record(z.string(), z.enum(['suppress', 'surface']))
-      .optional(),
+    checkIgnoredWordPairs: z.record(z.string(), z.enum(['suppress', 'surface'])).optional(),
   })
   .catch({}); // unknown/old shapes parse as empty, never throw (W8)
 
@@ -359,7 +356,7 @@ export const user_settings = pgTable('user_settings', {
 The alternative — teaching the AI proxy to subtract suppressed findings server-side — was rejected:
 
 1. It breaks the proxy's reviewed role as a **verbatim pass-through** (D8: contract mirrored exactly; D9: envelope passthrough). Filtering injects Fluent domain state into an AI-contract endpoint.
-2. The cascade needs *all* findings anyway to render inactive items with reason labels (§5.3 item 1) — a server that pre-filters would have to return them annotated regardless, which is the client cascade with extra steps.
+2. The cascade needs _all_ findings anyway to render inactive items with reason labels (§5.3 item 1) — a server that pre-filters would have to return them annotated regardless, which is the client cascade with extra steps.
 3. Suppression maps are small (a translator's ignore list, not a corpus) and already at the client's fingertips via blobs it loads for other reasons.
 
 The proxy therefore remains untouched by this proposal.
@@ -370,7 +367,7 @@ The proxy therefore remains untouched by this proposal.
 
 The settings endpoint lives in a **new `self` domain**, requested by kaseywright in review. `self` is the authenticated caller acting on their own resources — the route never names a user. This is cleaner than hanging settings off `/users/{userId}/…`:
 
-- **Clean semantic separation** — `self/*` is unambiguously "me," leaving the `users/*` domain for (future) admin-style tooling that operates on *other* users by id.
+- **Clean semantic separation** — `self/*` is unambiguously "me," leaving the `users/*` domain for (future) admin-style tooling that operates on _other_ users by id.
 - **No redundant `{userId}` param** — the user comes from the session, not the URL (the same source `user-chapter-assignment-editor-state.route.ts` already uses).
 - **No accidental over-exposure** — with no `{id}` in the path there is nothing to mis-scope; contrast `/users/{userId}/projects`, which must add `requireSelf()` to stop one user reading another's data.
 - **Simpler middleware** — just `authenticateUser`. No `requireSelf`, no `requireUserAccess`, no `UserPolicy`.
@@ -379,10 +376,10 @@ Otherwise it keeps the editor-state idiom: **full-replace upsert** on PUT, one Z
 
 ### 8.1 Routes
 
-| Method | Path | Auth | Body | Response |
-| --- | --- | --- | --- | --- |
-| `GET` | `/self/settings` | `authenticateUser` only | — | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
-| `PUT` | `/self/settings` | `authenticateUser` only | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation |
+| Method | Path             | Auth                    | Body                         | Response                                                                                                                                                            |
+| ------ | ---------------- | ----------------------- | ---------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `GET`  | `/self/settings` | `authenticateUser` only | —                            | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
+| `PUT`  | `/self/settings` | `authenticateUser` only | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation                                                                                                                        |
 
 - **No permission middleware beyond `authenticateUser`.** Every caller owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id` from context. (Editor-state additionally guards with `CONTENT_UPDATE` + chapter-assignment participation because it is tied to a chapter resource; `self/settings` has no resource to scope, so session identity alone is the right guard — and the `self` prefix makes that explicit.)
 - **PUT is full-replace.** The client GETs, merges in memory, PUTs the whole blob — the editor-state write pattern. Concurrent tabs are last-writer-wins, identical to editor-state's accepted behavior; for an ignore-list this worst case is "one ignore from another tab is lost until re-clicked."
@@ -398,7 +395,7 @@ fluent-api/src/domains/self/settings/
 └── self-settings.types.ts        # response schema; re-exports UserSettings from db schema
 ```
 
-Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§7.2), one numbered migration in `src/db/migrations/`, and route registration in `src/app.ts`. (The table is still named `user_settings` — it stores per-user rows; only the *route* surface is `self`. The repository upserts on the `userId` PK using the session user's id.)
+Plus: the `user_settings` table + `userSettingsSchema` in `src/db/schema.ts` (§7.2), one numbered migration in `src/db/migrations/`, and route registration in `src/app.ts`. (The table is still named `user_settings` — it stores per-user rows; only the _route_ surface is `self`. The repository upserts on the `userId` PK using the session user's id.)
 
 ### 8.3 What fluent-api explicitly does **not** do
 
@@ -426,14 +423,14 @@ When the repeated-words query errors (network, 502 `AI_SERVICE_UNAVAILABLE` / `A
 - A single inline line renders at the top of the Checks panel: `Checks failed to refresh` in `text-sm text-red-500` — the drafting page's own status idiom (it already shows "Auto-save failed" inline in the header). **No toast, banner, or popup**; none exist in this codebase and a transient check hiccup doesn't warrant introducing one.
 - TanStack Query retains the last successful `data` on refetch failure, so the panel **keeps rendering the last-known findings** below the error line, and the dot reflects that last-known state. The failure mode is "results are one save behind," not "results vanish."
 - The error is logged via the existing `Logger` for diagnosis; the next successful auto-save naturally retries (new `saveCounter` key).
-- A failure on the *initial* load (no previous data) renders the error line over an empty section — not the "No issues found" zero state, which must never appear on error.
+- A failure on the _initial_ load (no previous data) renders the error line over an empty section — not the "No issues found" zero state, which must never appear on error.
 
 ### 9.3 Suppression-write failure
 
 Ignore actions apply **optimistically** (the finding greys immediately) and roll back on write failure:
 
 - Occurrence rules ride the debounced editor-state save; a failed save leaves the in-memory rule intact and retries with the next editor-state write (the page already tolerates editor-state save failures this way).
-- Global rules PUT immediately; on failure the optimistic update is reverted (the finding returns to active) and the action can be re-clicked. No queued retry for v1 — the user sees the flag come back, which *is* the failure notification.
+- Global rules PUT immediately; on failure the optimistic update is reverted (the finding returns to active) and the action can be re-clicked. No queued retry for v1 — the user sees the flag come back, which _is_ the failure notification.
 
 ### 9.4 Rollout / landing order (W12)
 
@@ -451,21 +448,21 @@ Ignore actions apply **optimistically** (the finding greys immediately) and roll
 
 The repo now ships test infrastructure under `src/test/` (`msw/server.ts`, `render.tsx`, `setup.ts`); all UI tests use it — MSW intercepts at the network boundary, so the hooks under test exercise their real fetch paths.
 
-| Area | Representative cases |
-| --- | --- |
-| `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC composed/decomposed key equivalence. |
-| `useRepeatedWordsCheck` | Fires on `saveCounter` change; request body shape (snake_case verbatim, `snt_id` format, all drafted verses); `enabled` gating (readOnly, empty chapter); previous data retained on 502 (MSW error handler). |
-| `useSuppressions` | Feature-detect: 404 ⇒ `globalIgnoresAvailable=false`; 200-with-null ⇒ available; occurrence rule round-trip through editor-state blob; global write purges current-chapter occurrence rules for the pair; optimistic rollback on PUT failure. |
-| `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and *not* on error); inline error line; active `[Ignore Here]`/`[Ignore Everywhere]` vs. greyed-with-label + `[Undo ▾]`; "Show Ignored" toggle defaults off and is not persisted; `[Ignore Everywhere]` opens confirm dialog and only writes on confirm; `[Ignore Everywhere]` absent when capability unavailable. |
-| `LeftPanel` / dot | Dot iff cascade-active > 0, visible from Resources tab; tab switch persistence key written; toggle-button mirror dot when panel closed. |
+| Area                         | Representative cases                                                                                                                                                                                                                                                                                                                                        |
+| ---------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `useResolvedFindings` (pure) | Cascade precedence table: machine-legitimate vs. global vs. occurrence, `suppress`/`surface`/silent at each layer, most-specific-wins; ordinal assignment incl. "the the the" overlap; NFC composed/decomposed key equivalence.                                                                                                                             |
+| `useRepeatedWordsCheck`      | Fires on `saveCounter` change; request body shape (snake_case verbatim, `snt_id` format, all drafted verses); `enabled` gating (readOnly, empty chapter); previous data retained on 502 (MSW error handler).                                                                                                                                                |
+| `useSuppressions`            | Feature-detect: 404 ⇒ `globalIgnoresAvailable=false`; 200-with-null ⇒ available; occurrence rule round-trip through editor-state blob; global write purges current-chapter occurrence rules for the pair; optimistic rollback on PUT failure.                                                                                                               |
+| `ChecksPanel` / `FindingRow` | Verse grouping + separators; zero state (and _not_ on error); inline error line; active `[Ignore Here]`/`[Ignore Everywhere]` vs. greyed-with-label + `[Undo ▾]`; "Show Ignored" toggle defaults off and is not persisted; `[Ignore Everywhere]` opens confirm dialog and only writes on confirm; `[Ignore Everywhere]` absent when capability unavailable. |
+| `LeftPanel` / dot            | Dot iff cascade-active > 0, visible from Resources tab; tab switch persistence key written; toggle-button mirror dot when panel closed.                                                                                                                                                                                                                     |
 
 ### 10.2 fluent-api (Vitest, route-level — same style as the ai-tools and editor-state tests)
 
-| Area | Representative cases |
-| --- | --- |
-| `GET /self/settings` | 401 unauthenticated; `settings: null` for new user (200, not 404); returns saved blob; user isolation (only own row, scoped to session user). |
-| `PUT /self/settings` | Upsert create + replace; 400 on schema violation; full-replace semantics (omitted keys gone); `.catch({})` tolerance on read of unknown-shaped stored blobs. |
-| Editor-state schema extension | Existing editor-state tests still pass unchanged (proves backward compatibility); round-trip of `checkOccurrenceRules` / `activeLeftTab`. |
+| Area                          | Representative cases                                                                                                                                         |
+| ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| `GET /self/settings`          | 401 unauthenticated; `settings: null` for new user (200, not 404); returns saved blob; user isolation (only own row, scoped to session user).                |
+| `PUT /self/settings`          | Upsert create + replace; 400 on schema violation; full-replace semantics (omitted keys gone); `.catch({})` tolerance on read of unknown-shaped stored blobs. |
+| Editor-state schema extension | Existing editor-state tests still pass unchanged (proves backward compatibility); round-trip of `checkOccurrenceRules` / `activeLeftTab`.                    |
 
 Manual verification continues to use the stack via `fluent-platform` compose plus the existing repeated-words smoke script for the proxy half.
 
@@ -486,16 +483,16 @@ Manual verification continues to use the stack via `fluent-platform` compose plu
 
 Items needing explicit confirmation in review (referenced from §5.2/§5.3); none block reading the rest of the design. Status reflects the first review round (§0):
 
-| # | Item | Proposed resolution | Status |
-| --- | --- | --- | --- |
-| S1 | Zero-state mock showed the notification dot; #277 text says no flags ⇒ no dot | Follow the text: no dot at zero active flags | ✅ Resolved — chadw fixed the mock |
-| S2 | #278 mocks showed a language dropdown above checks content | Omit it on the Checks tab (no function for checks) | ✅ Resolved — chadw removed it from the mocks |
-| S3 | Card originally said ignored items are "removed from the panel" | Reveal them dimmed under a "Show Ignored" toggle (default off, not persisted) with "Undo Ignore"; dot unaffected | ✅ Largely card-aligned — revised #278 added the toggle; confirm the labels match |
-| S4 | Cards did not enumerate Greek-Room-`legitimate` findings | Show as ignored items labeled **"Default Ignore"** (the revised #278 mock's term), per-occurrence surfaceable | ☑ Confirm "Default Ignore" = machine-legitimate |
-| S5 | Dot is invisible when the whole left panel is closed | Mirror the dot on the panel-toggle button while closed | ☐ Open |
-| S6 | Suppression persistence has no card of its own | Ship as extension of #172's scope (W1) — backend dependency of the UI cards | ☐ Open |
-| S7 | Revised #278 confirm dialog says "Ignore Everywhere" "cannot be undone" | Add the confirm dialog **and** keep `[Undo ▾]` reversibility (§6.5); the two are complementary | ☐ Confirm we may retain undo |
-| S8 | Highlight repeated words in the verse text (joelthe1/Ulf) | We will implement it as a **follow-on PR** layered on this one (verse-pane scope; §5.3 item 4) | ☐ Confirm follow-on-PR sequencing |
+| #   | Item                                                                          | Proposed resolution                                                                                              | Status                                                                            |
+| --- | ----------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------- |
+| S1  | Zero-state mock showed the notification dot; #277 text says no flags ⇒ no dot | Follow the text: no dot at zero active flags                                                                     | ✅ Resolved — chadw fixed the mock                                                |
+| S2  | #278 mocks showed a language dropdown above checks content                    | Omit it on the Checks tab (no function for checks)                                                               | ✅ Resolved — chadw removed it from the mocks                                     |
+| S3  | Card originally said ignored items are "removed from the panel"               | Reveal them dimmed under a "Show Ignored" toggle (default off, not persisted) with "Undo Ignore"; dot unaffected | ✅ Largely card-aligned — revised #278 added the toggle; confirm the labels match |
+| S4  | Cards did not enumerate Greek-Room-`legitimate` findings                      | Show as ignored items labeled **"Default Ignore"** (the revised #278 mock's term), per-occurrence surfaceable    | ☑ Confirm "Default Ignore" = machine-legitimate                                   |
+| S5  | Dot is invisible when the whole left panel is closed                          | Mirror the dot on the panel-toggle button while closed                                                           | ☐ Open                                                                            |
+| S6  | Suppression persistence has no card of its own                                | Ship as extension of #172's scope (W1) — backend dependency of the UI cards                                      | ☐ Open                                                                            |
+| S7  | Revised #278 confirm dialog says "Ignore Everywhere" "cannot be undone"       | Add the confirm dialog **and** keep `[Undo ▾]` reversibility (§6.5); the two are complementary                   | ☐ Confirm we may retain undo                                                      |
+| S8  | Highlight repeated words in the verse text (joelthe1/Ulf)                     | We will implement it as a **follow-on PR** layered on this one (verse-pane scope; §5.3 item 4)                   | ☐ Confirm follow-on-PR sequencing                                                 |
 
 Engineering-side confirmations sought from reviewers (same spirit as the D-series confirmations in the fluent-api proposal):
 
@@ -505,4 +502,4 @@ Engineering-side confirmations sought from reviewers (same spirit as the D-serie
 
 ---
 
-*Prepared against fluent-web `main` (post `2e936d1`), fluent-api `main` (post `7c7f63d`), and the approved fluent-api proposal (PR #173). Revised 2026-06-16 after the first review round on PR #305. Author: JEdward7777.*
+_Prepared against fluent-web `main` (post `2e936d1`), fluent-api `main` (post `7c7f63d`), and the approved fluent-api proposal (PR #173). Revised 2026-06-16 after the first review round on PR #305. Author: JEdward7777._

From b6a4813be4da740cb790520811dace4a3a09458c Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 18 Jun 2026 15:06:57 -0400
Subject: [PATCH 10/33] docs(checks-ui): remove standalone progress checklist

Implementation progress is now tracked privately in the execution
harness (self-notes/), not in a doc under the team repo. The proposal
docs stay focused on the design.
---
 .../repeated-word-check/checks-ui-integration-progress.md         | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 delete mode 100644 docs/proposals/repeated-word-check/checks-ui-integration-progress.md

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-progress.md b/docs/proposals/repeated-word-check/checks-ui-integration-progress.md
deleted file mode 100644
index e69de29b..00000000

From aee3f7d5a37ae74b17cd520228ca96c641236309 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 21 Jun 2026 09:55:45 -0400
Subject: [PATCH 11/33] feat(checks): add useSuppressions hook (Phase 2)

- Owns global word-pair ignores via /self/settings (session probe +
  optimistic PUT + rollback; full-object-replace body matching the
  Phase 1 contract); occurrence rules handled through an injected
  editor-state writer so useResourceStatePersistence stays the single
  writer of the editor-state blob.
- Extend useResourceStatePersistence state shapes with optional
  checkOccurrenceRules.
- 16 new tests (probe, pure helpers, occurrence round-trip, global
  optimistic+purge+rollback); lint 0/0, typecheck + prettier clean.
- Drive-by: MultiSelectFilter ||->?? to clear a pre-existing lint warning.

Relates to #277, #278
---
 .../hooks/useResourceStatePersistence.ts      |  12 +
 .../checks/hooks/useSuppressions.test.ts      | 261 +++++++++++++
 src/features/checks/hooks/useSuppressions.ts  | 366 ++++++++++++++++++
 3 files changed, 639 insertions(+)
 create mode 100644 src/features/checks/hooks/useSuppressions.test.ts
 create mode 100644 src/features/checks/hooks/useSuppressions.ts

diff --git a/src/features/bible/hooks/useResourceStatePersistence.ts b/src/features/bible/hooks/useResourceStatePersistence.ts
index 4694cda6..01bc076d 100644
--- a/src/features/bible/hooks/useResourceStatePersistence.ts
+++ b/src/features/bible/hooks/useResourceStatePersistence.ts
@@ -1,5 +1,6 @@
 import { useMutation, useQuery, useQueryClient } from '@tanstack/react-query';
 
+import { type OccurrenceRules } from '@/features/checks/checks.types';
 import { config } from '@/lib/config';
 import { Logger } from '@/lib/services/logger';
 
@@ -16,6 +17,16 @@ export interface FetchResourceState {
   languageCode: string;
   tabStatus: boolean;
   activeLeftTab?: LeftTab;
+  /**
+   * Occurrence-level suppression rules for the Repeated Word Check, scoped to
+   * this chapter assignment (cascade layer 2, §7.1). Keyed by occurrence
+   * identity `"{snt_id}|{repeated_word}|{ordinal}"`. Optional +
+   * backward-compatible: old rows omit it and `useSuppressions` treats it as
+   * empty. This blob has a **single writer** (`useSaveResourceState`); the
+   * Checks feature reads/updates the map through that one writer rather than
+   * opening a second PUT path that would clobber the other keys.
+   */
+  checkOccurrenceRules?: OccurrenceRules;
 }
 
 interface ResourceState {
@@ -27,6 +38,7 @@ interface ResourceState {
     languageCode: string;
     tabStatus: boolean;
     activeLeftTab?: LeftTab;
+    checkOccurrenceRules?: OccurrenceRules;
   };
 }
 
diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
new file mode 100644
index 00000000..0e93ab5e
--- /dev/null
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -0,0 +1,261 @@
+import { http, HttpResponse } from 'msw';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+import { config } from '@/lib/config';
+import { server } from '@/test/msw/server';
+import { act, renderHook, waitFor } from '@/test/render';
+
+import { type OccurrenceRules, type ResolvedFinding } from '../checks.types';
+
+import {
+  deleteRule,
+  purgeLocalForPair,
+  setRule,
+  useSuppressions,
+  type UseSuppressionsParams,
+} from './useSuppressions';
+
+const SETTINGS_URL = `${config.api.url}/self/settings`;
+
+/** Register a GET /self/settings handler returning the given body + status. */
+const mockGet = (body: Record<string, unknown>, status = 200) => {
+  server.use(http.get(SETTINGS_URL, () => HttpResponse.json(body, { status })));
+};
+
+/** Register a PUT /self/settings handler; resolves the captured body. */
+const mockPut = (status = 200) => {
+  const captured: { body: unknown } = { body: undefined };
+  server.use(
+    http.put(SETTINGS_URL, async ({ request }) => {
+      captured.body = await request.json();
+      return HttpResponse.json({ settings: {}, updatedAt: null }, { status });
+    })
+  );
+  return captured;
+};
+
+/** Render the hook with an injected occurrence store we can assert against. */
+const setup = (
+  occurrenceRules: OccurrenceRules = {},
+  saveOccurrenceRules: UseSuppressionsParams['saveOccurrenceRules'] = vi.fn()
+) => {
+  const result = renderHook((props: UseSuppressionsParams) => useSuppressions(props), {
+    initialProps: { occurrenceRules, saveOccurrenceRules },
+  });
+  return { ...result, saveOccurrenceRules };
+};
+
+const makeResolved = (over: Partial<ResolvedFinding> = {}): ResolvedFinding => ({
+  finding: {
+    snt_id: 'JDG 4:3',
+    repeated_word: 'the the',
+    surf: 'the the',
+    start_position: 0,
+    legitimate: false,
+    severity: 0.5,
+  },
+  ordinal: 0,
+  occurrenceKey: 'JDG 4:3|the the|0',
+  isActive: false,
+  inactiveReason: 'occurrence',
+  ...over,
+});
+
+beforeEach(() => {
+  // Default: route present, empty settings, so probe resolves "available".
+  mockGet({ settings: null, updatedAt: null });
+});
+
+afterEach(() => {
+  vi.restoreAllMocks();
+});
+
+// ---------------------------------------------------------------------------
+// Pure helpers
+// ---------------------------------------------------------------------------
+
+describe('setRule / deleteRule', () => {
+  it('setRule returns a new map with the key set, leaving the original untouched', () => {
+    const original: OccurrenceRules = { a: 'suppress' };
+    const next = setRule(original, 'b', 'surface');
+    expect(next).toEqual({ a: 'suppress', b: 'surface' });
+    expect(original).toEqual({ a: 'suppress' });
+  });
+
+  it('deleteRule removes the key in a copy; no-op when absent', () => {
+    expect(deleteRule({ a: 'suppress', b: 'surface' }, 'a')).toEqual({ b: 'surface' });
+    expect(deleteRule({ a: 'suppress' }, 'missing')).toEqual({ a: 'suppress' });
+  });
+});
+
+describe('purgeLocalForPair', () => {
+  it('drops every occurrence rule whose pair segment matches (NFC), keeps others', () => {
+    const rules: OccurrenceRules = {
+      'JDG 4:3|the the|0': 'suppress',
+      'JDG 4:5|the the|0': 'surface',
+      'JDG 4:3|and and|0': 'suppress',
+    };
+    const { rules: next, changed } = purgeLocalForPair(rules, 'the the');
+    expect(changed).toBe(true);
+    expect(next).toEqual({ 'JDG 4:3|and and|0': 'suppress' });
+  });
+
+  it('matches composed vs decomposed accents', () => {
+    const rules: OccurrenceRules = { ['GEN 1:1|cafe\u0301 cafe\u0301|0']: 'suppress' };
+    const { rules: next, changed } = purgeLocalForPair(rules, 'caf\u00e9 caf\u00e9');
+    expect(changed).toBe(true);
+    expect(next).toEqual({});
+  });
+
+  it('returns the same instance + changed=false when nothing matches', () => {
+    const rules: OccurrenceRules = { 'JDG 4:3|and and|0': 'suppress' };
+    const out = purgeLocalForPair(rules, 'the the');
+    expect(out.changed).toBe(false);
+    expect(out.rules).toBe(rules);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Feature-detect probe (§9.1, W8)
+// ---------------------------------------------------------------------------
+
+describe('useSuppressions — feature-detect probe', () => {
+  it('200 with settings:null ⇒ available, empty global rules, probe resolved', async () => {
+    mockGet({ settings: null, updatedAt: null });
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.globalIgnoresAvailable).toBe(true);
+    expect(result.current.globalRules).toEqual({});
+  });
+
+  it('200 with rules ⇒ available and parses checkIgnoredWordPairs', async () => {
+    mockGet({
+      settings: { checkIgnoredWordPairs: { 'the the': 'suppress' } },
+      updatedAt: '2026-06-18T00:00:00Z',
+    });
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.globalIgnoresAvailable).toBe(true);
+    expect(result.current.globalRules).toEqual({ 'the the': 'suppress' });
+  });
+
+  it('404 ⇒ unavailable (route not deployed), probe still resolves', async () => {
+    mockGet({ message: 'Not Found' }, 404);
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.globalIgnoresAvailable).toBe(false);
+    expect(result.current.globalRules).toEqual({});
+  });
+
+  it('network failure ⇒ conservatively unavailable, probe resolves', async () => {
+    server.use(http.get(SETTINGS_URL, () => HttpResponse.error()));
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.globalIgnoresAvailable).toBe(false);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Occurrence actions (write through the single editor-state writer, §7.1)
+// ---------------------------------------------------------------------------
+
+describe('useSuppressions — occurrence actions', () => {
+  it('ignoreHere writes a "suppress" verdict for the key, preserving siblings', async () => {
+    const save = vi.fn();
+    const { result } = setup({ 'JDG 4:3|and and|0': 'suppress' }, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreHere('JDG 4:3|the the|0'));
+    expect(save).toHaveBeenCalledWith({
+      'JDG 4:3|and and|0': 'suppress',
+      'JDG 4:3|the the|0': 'suppress',
+    });
+  });
+
+  it('undoOccurrence DELETES the rule when the finding is hidden by its own occurrence rule', async () => {
+    const save = vi.fn();
+    const { result } = setup({ 'JDG 4:3|the the|0': 'suppress' }, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.undoOccurrence(makeResolved({ inactiveReason: 'occurrence' })));
+    expect(save).toHaveBeenCalledWith({});
+  });
+
+  it('undoOccurrence writes "surface" when the finding is hidden by a GLOBAL rule', async () => {
+    const save = vi.fn();
+    const { result } = setup({}, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.undoOccurrence(makeResolved({ inactiveReason: 'global' })));
+    expect(save).toHaveBeenCalledWith({ 'JDG 4:3|the the|0': 'surface' });
+  });
+
+  it('undoOccurrence writes "surface" when the finding is hidden by Greek Room (legitimate)', async () => {
+    const save = vi.fn();
+    const { result } = setup({}, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.undoOccurrence(makeResolved({ inactiveReason: 'legitimate' })));
+    expect(save).toHaveBeenCalledWith({ 'JDG 4:3|the the|0': 'surface' });
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Global actions (optimistic + rollback; purge current-chapter local, §6.5/§9.3)
+// ---------------------------------------------------------------------------
+
+describe('useSuppressions — global actions', () => {
+  it('ignoreEverywhere optimistically suppresses, PUTs the full blob, and purges current-chapter occurrence rules for the pair', async () => {
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    const put = mockPut();
+    const save = vi.fn();
+    const { result } = setup(
+      { 'JDG 4:3|the the|0': 'suppress', 'JDG 4:3|and and|0': 'suppress' },
+      save
+    );
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreEverywhere('the the'));
+
+    // Optimistic: the global map updates immediately.
+    expect(result.current.globalRules).toEqual({ 'the the': 'suppress' });
+    // Purge-local: the current-chapter occurrence rule for "the the" is dropped,
+    // the unrelated "and and" rule stands.
+    expect(save).toHaveBeenCalledWith({ 'JDG 4:3|and and|0': 'suppress' });
+    // Full-replace PUT carries the whole blob under settings.checkIgnoredWordPairs.
+    await waitFor(() =>
+      expect(put.body).toEqual({ settings: { checkIgnoredWordPairs: { 'the the': 'suppress' } } })
+    );
+  });
+
+  it('ignoreEverywhere rolls back the optimistic update on PUT failure', async () => {
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    mockPut(500);
+    const { result } = setup({}, vi.fn());
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreEverywhere('the the'));
+    // Briefly optimistic, then reverts to the prior (empty) state.
+    await waitFor(() => expect(result.current.globalRules).toEqual({}));
+  });
+
+  it('stopIgnoringEverywhere deletes the global entry and PUTs the reduced blob', async () => {
+    mockGet({
+      settings: { checkIgnoredWordPairs: { 'the the': 'suppress', 'and and': 'suppress' } },
+      updatedAt: null,
+    });
+    const put = mockPut();
+    const { result } = setup({}, vi.fn());
+    await waitFor(() =>
+      expect(result.current.globalRules).toEqual({ 'the the': 'suppress', 'and and': 'suppress' })
+    );
+
+    act(() => result.current.stopIgnoringEverywhere('the the'));
+    expect(result.current.globalRules).toEqual({ 'and and': 'suppress' });
+    await waitFor(() =>
+      expect(put.body).toEqual({
+        settings: { checkIgnoredWordPairs: { 'and and': 'suppress' } },
+      })
+    );
+  });
+});
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
new file mode 100644
index 00000000..dd532eb3
--- /dev/null
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -0,0 +1,366 @@
+import { useCallback, useEffect, useMemo, useRef, useState } from 'react';
+
+import { config } from '@/lib/config';
+import { Logger } from '@/lib/services/logger';
+
+import {
+  type GlobalRules,
+  type OccurrenceRules,
+  type ResolvedFinding,
+  type RuleVerdict,
+} from '../checks.types';
+
+import { normalizePair } from './useResolvedFindings';
+
+/**
+ * `useSuppressions` — owns the read/write of the two suppression stores that
+ * feed the cascade in `useResolvedFindings` (Repeated Word Check, phase 2;
+ * proposal §6.5, §7, §9.1, §9.3, W6/W8).
+ *
+ * Two stores, two scopes:
+ *
+ *  - **Occurrence rules** (cascade layer 2) live in the editor-state JSONB blob
+ *    (`checkOccurrenceRules`), scoped to one `(user, chapterAssignment)`. There
+ *    is **one writer** for that blob — `useResourceStatePersistence`'s save
+ *    mutation — so this hook does not open a second PUT path. The page injects
+ *    the current map plus a single `saveOccurrenceRules` callback; this hook
+ *    does the read-modify-write and hands the merged map back through it (§7.1,
+ *    "one writer for the editor-state blob").
+ *
+ *  - **Global word-pair rules** (cascade layer 1) live in
+ *    `user_settings.checkIgnoredWordPairs` behind `GET/PUT /self/settings`. This
+ *    hook owns that half end to end: the once-per-session feature-detect probe
+ *    (§9.1), optimistic writes with rollback (§9.3), and the full-replace PUT
+ *    (§8.1).
+ *
+ * The action callbacks (`ignoreHere`, `ignoreEverywhere`, `undoOccurrence`,
+ * `stopIgnoringEverywhere`) are what `FindingRow` (phase 3) calls. The confirm
+ * dialog for "Ignore Everywhere" lives in the UI (phase 3); it calls
+ * `ignoreEverywhere` only on confirm.
+ */
+
+// ---------------------------------------------------------------------------
+// /self/settings wire shape (Fluent-domain → camelCase, conventions §B).
+// ---------------------------------------------------------------------------
+
+/** The user-settings blob behind `/self/settings`. Tolerant of absent keys. */
+interface SelfSettings {
+  /** Global word-pair rules, keyed by NFC-normalized `repeated_word`. */
+  checkIgnoredWordPairs?: GlobalRules;
+}
+
+/** GET response: the server returns `{ settings }`, possibly `null` (W8). */
+interface SelfSettingsResponse {
+  settings: SelfSettings | null;
+}
+
+const SELF_SETTINGS_URL = `${config.api.url}/self/settings`;
+
+/**
+ * Result of the once-per-session probe. `available` distinguishes "the route
+ * exists" (any 2xx, including `settings: null`) from "not deployed / network
+ * down" (404 or failure) so phase 3 can hide the global-ignore controls
+ * without ever rendering a dead button (W8).
+ */
+interface ProbeResult {
+  available: boolean;
+  rules: GlobalRules;
+}
+
+/** Pull the global rules out of a settings blob, tolerating any odd shape. */
+const extractGlobalRules = (settings: SelfSettings | null): GlobalRules => {
+  const pairs = settings?.checkIgnoredWordPairs;
+  if (pairs && typeof pairs === 'object') {
+    return pairs;
+  }
+  return {};
+};
+
+/**
+ * Probe `GET /self/settings` once. Any 2xx ⇒ available (parse rules, tolerate
+ * `null`); 404 ⇒ route not deployed yet ⇒ unavailable; network/parse failure ⇒
+ * conservatively unavailable for the session. Never throws.
+ */
+const probeSelfSettings = async (): Promise<ProbeResult> => {
+  try {
+    const res = await fetch(SELF_SETTINGS_URL, {
+      method: 'GET',
+      credentials: 'include',
+      headers: { 'Content-Type': 'application/json' },
+    });
+    if (res.status === 404) {
+      return { available: false, rules: {} };
+    }
+    if (!res.ok) {
+      // Other non-2xx (401/500/…): treat as unavailable for the session.
+      return { available: false, rules: {} };
+    }
+    const body = (await res.json()) as SelfSettingsResponse | null;
+    return { available: true, rules: extractGlobalRules(body?.settings ?? null) };
+  } catch (error) {
+    Logger.logException(error, { context: 'Repeated-words self/settings probe failed' });
+    return { available: false, rules: {} };
+  }
+};
+
+/**
+ * PUT the full settings blob (full-replace, §8.1). The caller supplies the
+ * complete `checkIgnoredWordPairs` map it wants persisted; this hook never
+ * sends a partial patch.
+ */
+const putSelfSettings = async (rules: GlobalRules): Promise<void> => {
+  const res = await fetch(SELF_SETTINGS_URL, {
+    method: 'PUT',
+    credentials: 'include',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({ settings: { checkIgnoredWordPairs: rules } }),
+  });
+  if (!res.ok) {
+    throw new Error('Failed to save global ignore rules');
+  }
+};
+
+// ---------------------------------------------------------------------------
+// Pure map helpers (exported for unit tests).
+// ---------------------------------------------------------------------------
+
+/** Set `key → verdict` in a copy of `rules`. */
+export const setRule = <T extends Record<string, RuleVerdict>>(
+  rules: T,
+  key: string,
+  verdict: RuleVerdict
+): T => ({ ...rules, [key]: verdict });
+
+/** Delete `key` from a copy of `rules` (no-op if absent). */
+export const deleteRule = <T extends Record<string, RuleVerdict>>(rules: T, key: string): T => {
+  if (!(key in rules)) {
+    return { ...rules };
+  }
+  const next = { ...rules };
+  delete next[key];
+  return next;
+};
+
+/**
+ * Remove every occurrence rule that targets `repeatedWord` (NFC-normalized) in
+ * the current chapter's occurrence map — used by "purge-local on global write"
+ * (§6.5). The occurrence key is `"{snt_id}|{repeated_word}|{ordinal}"`, so we
+ * match the normalized middle segment. Returns a new map (or the same instance
+ * if nothing matched, to avoid needless writes).
+ */
+export const purgeLocalForPair = (
+  occurrenceRules: OccurrenceRules,
+  repeatedWord: string
+): { rules: OccurrenceRules; changed: boolean } => {
+  const target = normalizePair(repeatedWord);
+  let changed = false;
+  const next: OccurrenceRules = {};
+  for (const [key, verdict] of Object.entries(occurrenceRules)) {
+    const segments = key.split('|');
+    // key = snt_id | repeated_word | ordinal  (snt_id has no '|')
+    const pair = segments.length >= 3 ? segments[1] : undefined;
+    if (pair !== undefined && normalizePair(pair) === target) {
+      changed = true;
+      continue; // drop it
+    }
+    next[key] = verdict;
+  }
+  return changed ? { rules: next, changed } : { rules: occurrenceRules, changed };
+};
+
+// ---------------------------------------------------------------------------
+// Hook
+// ---------------------------------------------------------------------------
+
+export interface UseSuppressionsParams {
+  /**
+   * Current occurrence-rule map from the editor-state blob (read through the
+   * single editor-state writer; defaults to empty for old rows). The page owns
+   * this state; this hook never fetches it.
+   */
+  occurrenceRules: OccurrenceRules;
+  /**
+   * The **single writer** for the occurrence map. Given the next full map, it
+   * persists it via the existing debounced editor-state save
+   * (`useResourceStatePersistence`). This hook does the read-modify-write and
+   * never opens its own editor-state PUT (the "one writer" rule, §7.1).
+   */
+  saveOccurrenceRules: (next: OccurrenceRules) => void;
+}
+
+export interface UseSuppressionsResult {
+  /** Layer-2 map → `useResolvedFindings`. */
+  occurrenceRules: OccurrenceRules;
+  /** Layer-1 map → `useResolvedFindings`. */
+  globalRules: GlobalRules;
+  /** False until the probe resolves, and on 404/failure (W8). Phase 3 hides
+   * `[Ignore Everywhere]` and omits global undo when this is false. */
+  globalIgnoresAvailable: boolean;
+  /** Flips true on any terminal probe response. Phase 4 gates the check
+   * query's `enabled` on this so findings never render ahead of global rules
+   * (W10/§9.1). */
+  settingsProbeResolved: boolean;
+
+  // Actions called by FindingRow (phase 3) -----------------------------------
+
+  /** "Ignore Here" — suppress this single occurrence (layer 2). */
+  ignoreHere: (occurrenceKey: string) => void;
+  /** "Ignore Everywhere" — suppress this word pair across the user's projects
+   * (layer 1) and purge any current-chapter occurrence rules for the pair.
+   * Optimistic; rolls back on PUT failure. Caller (phase 3) shows the confirm
+   * dialog first. */
+  ignoreEverywhere: (repeatedWord: string) => void;
+  /** Default `[Undo ▾]` click (§6.5): undo whatever layer is currently hiding
+   * this finding — delete the occurrence rule if it's the user's own, else
+   * write an occurrence-level `'surface'` override so the global rule / Greek
+   * Room verdict survives but this one occurrence re-surfaces. */
+  undoOccurrence: (resolved: ResolvedFinding) => void;
+  /** Chevron action: stop ignoring this pair everywhere — delete the global
+   * entry (and purge current-chapter occurrence rules for it). Optimistic;
+   * rolls back on PUT failure. */
+  stopIgnoringEverywhere: (repeatedWord: string) => void;
+}
+
+/**
+ * @see the module doc above. The occurrence half is injected (single writer);
+ * the global half is owned here (probe + optimistic PUT/rollback).
+ */
+export const useSuppressions = ({
+  occurrenceRules,
+  saveOccurrenceRules,
+}: UseSuppressionsParams): UseSuppressionsResult => {
+  // --- Global half: probe once per session, then hold the map locally so we
+  // can apply optimistic updates and roll them back without a query refetch. ---
+  const [globalRules, setGlobalRules] = useState<GlobalRules>({});
+  const [globalIgnoresAvailable, setGlobalIgnoresAvailable] = useState(false);
+  const [settingsProbeResolved, setSettingsProbeResolved] = useState(false);
+
+  // Guard against double-invocation (StrictMode) and state updates after
+  // unmount; the probe runs exactly once per mounted session.
+  const probeStarted = useRef(false);
+  const mounted = useRef(true);
+  useEffect(() => {
+    mounted.current = true;
+    return () => {
+      mounted.current = false;
+    };
+  }, []);
+
+  useEffect(() => {
+    if (probeStarted.current) {
+      return;
+    }
+    probeStarted.current = true;
+    void probeSelfSettings().then(({ available, rules }) => {
+      if (!mounted.current) {
+        return;
+      }
+      setGlobalIgnoresAvailable(available);
+      setGlobalRules(available ? rules : {});
+      setSettingsProbeResolved(true);
+    });
+  }, []);
+
+  // Latest occurrence map for read-modify-write inside callbacks without
+  // making the callbacks change identity on every keystroke.
+  const occurrenceRef = useRef(occurrenceRules);
+  occurrenceRef.current = occurrenceRules;
+
+  // --- Occurrence actions (write through the single editor-state writer) ---
+
+  const ignoreHere = useCallback(
+    (occurrenceKey: string) => {
+      saveOccurrenceRules(setRule(occurrenceRef.current, occurrenceKey, 'suppress'));
+    },
+    [saveOccurrenceRules]
+  );
+
+  const undoOccurrence = useCallback(
+    (resolved: ResolvedFinding) => {
+      const { occurrenceKey, inactiveReason } = resolved;
+      if (inactiveReason === 'occurrence') {
+        // The user's own occurrence rule is what's hiding it → remove it.
+        saveOccurrenceRules(deleteRule(occurrenceRef.current, occurrenceKey));
+      } else {
+        // A global rule or Greek Room `legitimate` is hiding it → re-surface
+        // just this occurrence; the broader verdict stands.
+        saveOccurrenceRules(setRule(occurrenceRef.current, occurrenceKey, 'surface'));
+      }
+    },
+    [saveOccurrenceRules]
+  );
+
+  // --- Global actions (optimistic + rollback; purge current-chapter local) ---
+
+  /** Apply a current-chapter purge for `pair` through the single writer. */
+  const purgeCurrentChapter = useCallback(
+    (repeatedWord: string) => {
+      const { rules, changed } = purgeLocalForPair(occurrenceRef.current, repeatedWord);
+      if (changed) {
+        saveOccurrenceRules(rules);
+      }
+    },
+    [saveOccurrenceRules]
+  );
+
+  const writeGlobal = useCallback(
+    (repeatedWord: string, mutate: (rules: GlobalRules) => GlobalRules) => {
+      const pair = normalizePair(repeatedWord);
+      const previous = globalRules;
+      const next = mutate(previous);
+      // Optimistic: the finding greys/un-greys immediately.
+      setGlobalRules(next);
+      // Purge current-chapter occurrence rules for the pair so the just-clicked
+      // panel doesn't appear to ignore the action (occurrence beats global).
+      purgeCurrentChapter(pair);
+      void putSelfSettings(next).catch(error => {
+        Logger.logException(error, {
+          context: 'Repeated-words global ignore write failed; rolling back',
+        });
+        if (mounted.current) {
+          // Rollback: the flag returns to its prior state; re-click allowed.
+          // (No queued retry for v1 — the visible flip is the failure notice.)
+          setGlobalRules(previous);
+        }
+      });
+    },
+    [globalRules, purgeCurrentChapter]
+  );
+
+  const ignoreEverywhere = useCallback(
+    (repeatedWord: string) => {
+      writeGlobal(repeatedWord, rules => setRule(rules, normalizePair(repeatedWord), 'suppress'));
+    },
+    [writeGlobal]
+  );
+
+  const stopIgnoringEverywhere = useCallback(
+    (repeatedWord: string) => {
+      writeGlobal(repeatedWord, rules => deleteRule(rules, normalizePair(repeatedWord)));
+    },
+    [writeGlobal]
+  );
+
+  return useMemo(
+    () => ({
+      occurrenceRules,
+      globalRules,
+      globalIgnoresAvailable,
+      settingsProbeResolved,
+      ignoreHere,
+      ignoreEverywhere,
+      undoOccurrence,
+      stopIgnoringEverywhere,
+    }),
+    [
+      occurrenceRules,
+      globalRules,
+      globalIgnoresAvailable,
+      settingsProbeResolved,
+      ignoreHere,
+      ignoreEverywhere,
+      undoOccurrence,
+      stopIgnoringEverywhere,
+    ]
+  );
+};

From 5b055e717a7afba3af5d002b1ec13b7a40622ea8 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 21 Jun 2026 10:25:57 -0400
Subject: [PATCH 12/33] feat(checks): add ChecksPanel and FindingRow components
 (Phase 3)

Presentational Checks-tab body for the Repeated Word Check (cards #277/#278):

- FindingRow: active rows show [Ignore Here]/[Ignore Everywhere] (the latter
  gated on globalIgnoresAvailable and opening an S7 confirm dialog, firing only
  on confirm); inactive rows render dimmed with the ignore-type label
  (Ignore Here / Ignore Always / Default Ignore) and an [Undo \u25be] split button
  whose chevron offers the deliberate global stop-ignoring action.
- ChecksPanel: per-check accordion (Repeated Words, expanded by default),
  findings grouped by verse with 'Verse N' headings (robust snt_id parse with
  raw-id fallback) and separators between groups, active-before-inactive,
  session-local 'Show Ignored' toggle (default off, not persisted), centered
  'No issues found' zero state, and an inline 'Checks failed to refresh' error
  line that never shows the zero state on error.

Components are presentational (no query/hook ownership) so Phase 4 can compose
them. 19 new tests; full suite 104 green, lint 0/0, typecheck + prettier clean.
---
 .../checks/components/ChecksPanel.test.tsx    | 142 ++++++++++++++
 .../checks/components/ChecksPanel.tsx         | 178 ++++++++++++++++++
 .../checks/components/FindingRow.test.tsx     | 143 ++++++++++++++
 src/features/checks/components/FindingRow.tsx | 174 +++++++++++++++++
 4 files changed, 637 insertions(+)
 create mode 100644 src/features/checks/components/ChecksPanel.test.tsx
 create mode 100644 src/features/checks/components/ChecksPanel.tsx
 create mode 100644 src/features/checks/components/FindingRow.test.tsx
 create mode 100644 src/features/checks/components/FindingRow.tsx

diff --git a/src/features/checks/components/ChecksPanel.test.tsx b/src/features/checks/components/ChecksPanel.test.tsx
new file mode 100644
index 00000000..66d82326
--- /dev/null
+++ b/src/features/checks/components/ChecksPanel.test.tsx
@@ -0,0 +1,142 @@
+import { describe, expect, it, vi } from 'vitest';
+
+import { renderWithProviders, screen } from '@/test/render';
+
+import { type ResolvedFinding, type ResolvedFindings } from '../checks.types';
+
+import { ChecksPanel } from './ChecksPanel';
+
+const active = (sntId: string, word: string, ordinal = 0): ResolvedFinding => ({
+  finding: {
+    snt_id: sntId,
+    repeated_word: word,
+    surf: `${word} (in ${sntId})`,
+    start_position: 0,
+    legitimate: false,
+    severity: 0.5,
+  },
+  ordinal,
+  occurrenceKey: `${sntId}|${word}|${ordinal}`,
+  isActive: true,
+});
+
+const inactive = (
+  sntId: string,
+  word: string,
+  reason: ResolvedFinding['inactiveReason'],
+  ordinal = 0
+): ResolvedFinding => ({
+  ...active(sntId, word, ordinal),
+  isActive: false,
+  inactiveReason: reason,
+});
+
+const handlers = () => ({
+  onIgnoreHere: vi.fn(),
+  onIgnoreEverywhere: vi.fn(),
+  onUndo: vi.fn(),
+  onStopIgnoringEverywhere: vi.fn(),
+});
+
+const renderPanel = (resolved: ResolvedFindings, over: Partial<{ isError: boolean }> = {}) =>
+  renderWithProviders(
+    <ChecksPanel
+      globalIgnoresAvailable
+      isError={over.isError ?? false}
+      resolved={resolved}
+      {...handlers()}
+    />
+  );
+
+describe('ChecksPanel — grouping & zero state', () => {
+  it('groups active findings by verse with a "Verse N" heading per findings-bearing verse', () => {
+    renderPanel({
+      active: [active('JDG 4:1', 'the the'), active('JDG 4:3', 'and and')],
+      inactive: [],
+    });
+    expect(screen.getByRole('heading', { name: 'Verse 1' })).toBeInTheDocument();
+    expect(screen.getByRole('heading', { name: 'Verse 3' })).toBeInTheDocument();
+    // Verse 2 had no findings -> no heading.
+    expect(screen.queryByRole('heading', { name: 'Verse 2' })).not.toBeInTheDocument();
+  });
+
+  it('renders a separator between verse groups (one fewer than the group count)', () => {
+    const { container } = renderPanel({
+      active: [active('JDG 4:1', 'the the'), active('JDG 4:3', 'and and')],
+      inactive: [],
+    });
+    expect(container.querySelectorAll('[data-slot="separator"]')).toHaveLength(1);
+  });
+
+  it('shows "No issues found" when there are no active findings', () => {
+    renderPanel({ active: [], inactive: [] });
+    expect(screen.getByText('No issues found')).toBeInTheDocument();
+  });
+
+  it('falls back to the raw snt_id when it does not parse to a verse number', () => {
+    renderPanel({ active: [active('weird-id', 'the the')], inactive: [] });
+    expect(screen.getByRole('heading', { name: 'weird-id' })).toBeInTheDocument();
+  });
+});
+
+describe('ChecksPanel — error line', () => {
+  it('renders the inline error line when isError', () => {
+    renderPanel({ active: [active('JDG 4:1', 'the the')], inactive: [] }, { isError: true });
+    expect(screen.getByRole('alert')).toHaveTextContent('Checks failed to refresh');
+  });
+
+  it('does NOT show the zero state on error (error over empty section instead)', () => {
+    renderPanel({ active: [], inactive: [] }, { isError: true });
+    expect(screen.getByRole('alert')).toHaveTextContent('Checks failed to refresh');
+    expect(screen.queryByText('No issues found')).not.toBeInTheDocument();
+  });
+
+  it('keeps rendering last-known findings below the error line', () => {
+    renderPanel({ active: [active('JDG 4:1', 'the the')], inactive: [] }, { isError: true });
+    expect(screen.getByRole('alert')).toBeInTheDocument();
+    expect(screen.getByRole('heading', { name: 'Verse 1' })).toBeInTheDocument();
+  });
+});
+
+describe('ChecksPanel — Show Ignored toggle', () => {
+  it('hides inactive findings by default (toggle off) and reveals them when toggled on', async () => {
+    const resolved: ResolvedFindings = {
+      active: [active('JDG 4:1', 'the the')],
+      inactive: [inactive('JDG 4:1', 'and and', 'occurrence')],
+    };
+    const { user } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+
+    // Default OFF: the inactive snippet is not shown.
+    expect(screen.queryByText('and and (in JDG 4:1)')).not.toBeInTheDocument();
+
+    await user.click(screen.getByRole('checkbox', { name: 'Show Ignored' }));
+    expect(screen.getByText('and and (in JDG 4:1)')).toBeInTheDocument();
+    expect(screen.getByTestId('inactive-label')).toHaveTextContent('Ignore Here');
+  });
+
+  it('does not render the toggle when there are no inactive findings', () => {
+    renderPanel({ active: [active('JDG 4:1', 'the the')], inactive: [] });
+    expect(screen.queryByRole('checkbox', { name: 'Show Ignored' })).not.toBeInTheDocument();
+  });
+
+  it('resets the toggle (not persisted) across remounts', async () => {
+    const resolved: ResolvedFindings = {
+      active: [],
+      inactive: [inactive('JDG 4:1', 'and and', 'global')],
+    };
+    const { user, unmount } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+    await user.click(screen.getByRole('checkbox', { name: 'Show Ignored' }));
+    expect(screen.getByText('and and (in JDG 4:1)')).toBeInTheDocument();
+    unmount();
+
+    // Fresh mount = fresh session: toggle is OFF again.
+    renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+    expect(screen.queryByText('and and (in JDG 4:1)')).not.toBeInTheDocument();
+  });
+});
diff --git a/src/features/checks/components/ChecksPanel.tsx b/src/features/checks/components/ChecksPanel.tsx
new file mode 100644
index 00000000..7f780245
--- /dev/null
+++ b/src/features/checks/components/ChecksPanel.tsx
@@ -0,0 +1,178 @@
+import { useState } from 'react';
+
+import {
+  Accordion,
+  AccordionContent,
+  AccordionItem,
+  AccordionTrigger,
+} from '@/components/ui/accordion';
+import { Checkbox } from '@/components/ui/checkbox';
+import { Label } from '@/components/ui/label';
+import { Separator } from '@/components/ui/separator';
+
+import { type ResolvedFinding, type ResolvedFindings } from '../checks.types';
+
+import { FindingRow } from './FindingRow';
+
+export interface ChecksPanelProps {
+  /** Cascade output: `{ active[], inactive[] }` from `useResolvedFindings`. */
+  resolved: ResolvedFindings;
+  /** The check query errored — show the inline red line (W9). */
+  isError: boolean;
+  /** Hide the global "Ignore Everywhere" affordance when unavailable (W8). */
+  globalIgnoresAvailable: boolean;
+  /** Active: suppress this occurrence ("Ignore Here"). */
+  onIgnoreHere: (occurrenceKey: string) => void;
+  /** Active: suppress this word pair globally (called only after confirm, S7). */
+  onIgnoreEverywhere: (repeatedWord: string) => void;
+  /** Inactive `[Undo ▾]` default click. */
+  onUndo: (resolved: ResolvedFinding) => void;
+  /** Inactive `[Undo ▾]` chevron: stop ignoring this pair everywhere. */
+  onStopIgnoringEverywhere: (repeatedWord: string) => void;
+}
+
+/**
+ * Derive the "Verse N" group heading from a `snt_id` (`"{book} {ch}:{verse}"`,
+ * e.g. `"JDG 4:3"` → "Verse 3"). Robust: take the segment after the last `:`;
+ * if it isn't a number (malformed `snt_id`), fall back to the raw `snt_id` so
+ * the heading is never blank.
+ */
+const verseHeading = (sntId: string): string => {
+  const afterColon = sntId.includes(':') ? sntId.slice(sntId.lastIndexOf(':') + 1).trim() : '';
+  return /^\d+$/.test(afterColon) ? `Verse ${afterColon}` : sntId;
+};
+
+interface VerseGroup {
+  sntId: string;
+  heading: string;
+  rows: ResolvedFinding[];
+}
+
+/**
+ * Group resolved findings by verse (`snt_id`), preserving first-seen verse order
+ * and, within each verse, active findings before inactive ones. Only verses that
+ * actually have rows to show produce a group (§5.1, §6.4).
+ */
+const groupByVerse = (
+  active: ResolvedFinding[],
+  inactive: ResolvedFinding[],
+  showIgnored: boolean
+): VerseGroup[] => {
+  const order: string[] = [];
+  const byVerse = new Map<string, { active: ResolvedFinding[]; inactive: ResolvedFinding[] }>();
+
+  const bucket = (sntId: string) => {
+    let entry = byVerse.get(sntId);
+    if (!entry) {
+      entry = { active: [], inactive: [] };
+      byVerse.set(sntId, entry);
+      order.push(sntId);
+    }
+    return entry;
+  };
+
+  for (const f of active) bucket(f.finding.snt_id).active.push(f);
+  if (showIgnored) {
+    for (const f of inactive) bucket(f.finding.snt_id).inactive.push(f);
+  }
+
+  return order
+    .map(sntId => {
+      const entry = byVerse.get(sntId) ?? { active: [], inactive: [] };
+      return {
+        sntId,
+        heading: verseHeading(sntId),
+        rows: [...entry.active, ...entry.inactive],
+      };
+    })
+    .filter(group => group.rows.length > 0);
+};
+
+/**
+ * The Checks-tab body (§5.1, §6.4–§6.6).
+ *
+ * A per-check accordion (only "Repeated Words" is wired; the structure
+ * anticipates sibling checks) containing verse-grouped findings, a bottom
+ * "Show Ignored" toggle, a zero state, and an inline error line. Purely
+ * presentational — it receives the cascade-resolved buckets and the suppression
+ * callbacks; it owns neither the check query nor the suppression hook (Phase 4
+ * composes them), and only the session-local "Show Ignored" toggle is local
+ * state (deliberately **not** persisted, per revised #278).
+ */
+export const ChecksPanel: React.FC<ChecksPanelProps> = ({
+  resolved,
+  isError,
+  globalIgnoresAvailable,
+  onIgnoreHere,
+  onIgnoreEverywhere,
+  onUndo,
+  onStopIgnoringEverywhere,
+}) => {
+  const [showIgnored, setShowIgnored] = useState(false);
+
+  const { active, inactive } = resolved;
+  const groups = groupByVerse(active, inactive, showIgnored);
+  // The zero state must never show on error (§9.2 — an error line over an empty
+  // section instead) and only when there is genuinely nothing to display: no
+  // active findings AND nothing currently revealed by the "Show Ignored" toggle.
+  const showZeroState = !isError && groups.length === 0;
+  const hasIgnored = inactive.length > 0;
+
+  return (
+    <div className='px-1 py-3'>
+      {isError && (
+        <p className='mb-2 text-sm text-red-500' role='alert'>
+          Checks failed to refresh
+        </p>
+      )}
+
+      <div className='rounded-lg border p-2'>
+        <Accordion collapsible defaultValue='repeated-words' type='single'>
+          <AccordionItem className='border-b-0' value='repeated-words'>
+            <AccordionTrigger className='py-2 text-base'>Repeated Words</AccordionTrigger>
+            <AccordionContent>
+              {showZeroState ? (
+                <p className='py-4 text-center text-sm font-bold'>No issues found</p>
+              ) : (
+                <div>
+                  {groups.map((group, index) => (
+                    <div key={group.sntId}>
+                      {index > 0 && <Separator className='my-2' />}
+                      <h4 className='text-sm font-bold'>{group.heading}</h4>
+                      {group.rows.map(row => (
+                        <FindingRow
+                          key={row.occurrenceKey}
+                          globalIgnoresAvailable={globalIgnoresAvailable}
+                          resolved={row}
+                          onIgnoreEverywhere={onIgnoreEverywhere}
+                          onIgnoreHere={onIgnoreHere}
+                          onStopIgnoringEverywhere={onStopIgnoringEverywhere}
+                          onUndo={onUndo}
+                        />
+                      ))}
+                    </div>
+                  ))}
+                </div>
+              )}
+
+              {hasIgnored && (
+                <div className='mt-3 flex items-center gap-2 border-t pt-3'>
+                  <Checkbox
+                    checked={showIgnored}
+                    id='checks-show-ignored'
+                    onCheckedChange={value => setShowIgnored(value === true)}
+                  />
+                  <Label className='text-sm font-normal' htmlFor='checks-show-ignored'>
+                    Show Ignored
+                  </Label>
+                </div>
+              )}
+            </AccordionContent>
+          </AccordionItem>
+        </Accordion>
+      </div>
+    </div>
+  );
+};
+
+export default ChecksPanel;
diff --git a/src/features/checks/components/FindingRow.test.tsx b/src/features/checks/components/FindingRow.test.tsx
new file mode 100644
index 00000000..bb7a2222
--- /dev/null
+++ b/src/features/checks/components/FindingRow.test.tsx
@@ -0,0 +1,143 @@
+import { describe, expect, it, vi } from 'vitest';
+
+import { renderWithProviders, screen, within } from '@/test/render';
+
+import { type ResolvedFinding } from '../checks.types';
+
+import { FindingRow } from './FindingRow';
+
+const makeResolved = (over: Partial<ResolvedFinding> = {}): ResolvedFinding => ({
+  finding: {
+    snt_id: 'JDG 4:3',
+    repeated_word: 'the the',
+    surf: 'and The the made',
+    start_position: 4,
+    legitimate: false,
+    severity: 0.5,
+  },
+  ordinal: 0,
+  occurrenceKey: 'JDG 4:3|the the|0',
+  isActive: true,
+  ...over,
+});
+
+const handlers = () => ({
+  onIgnoreHere: vi.fn(),
+  onIgnoreEverywhere: vi.fn(),
+  onUndo: vi.fn(),
+  onStopIgnoringEverywhere: vi.fn(),
+});
+
+describe('FindingRow — active', () => {
+  it('renders the surface snippet and both ignore buttons', () => {
+    renderWithProviders(
+      <FindingRow globalIgnoresAvailable resolved={makeResolved()} {...handlers()} />
+    );
+    expect(screen.getByText('and The the made')).toBeInTheDocument();
+    expect(screen.getByRole('button', { name: 'Ignore Here' })).toBeInTheDocument();
+    expect(screen.getByRole('button', { name: 'Ignore Everywhere' })).toBeInTheDocument();
+  });
+
+  it('fires onIgnoreHere with the occurrence key', async () => {
+    const h = handlers();
+    const { user } = renderWithProviders(
+      <FindingRow globalIgnoresAvailable resolved={makeResolved()} {...h} />
+    );
+    await user.click(screen.getByRole('button', { name: 'Ignore Here' }));
+    expect(h.onIgnoreHere).toHaveBeenCalledWith('JDG 4:3|the the|0');
+  });
+
+  it('hides Ignore Everywhere when global ignores are unavailable', () => {
+    renderWithProviders(
+      <FindingRow globalIgnoresAvailable={false} resolved={makeResolved()} {...handlers()} />
+    );
+    expect(screen.queryByRole('button', { name: 'Ignore Everywhere' })).not.toBeInTheDocument();
+  });
+
+  it('Ignore Everywhere opens a confirm dialog and only fires on confirm', async () => {
+    const h = handlers();
+    const { user } = renderWithProviders(
+      <FindingRow globalIgnoresAvailable resolved={makeResolved()} {...h} />
+    );
+
+    await user.click(screen.getByRole('button', { name: 'Ignore Everywhere' }));
+    // Dialog is open; the action has NOT fired yet.
+    expect(screen.getByRole('dialog')).toBeInTheDocument();
+    expect(h.onIgnoreEverywhere).not.toHaveBeenCalled();
+
+    // Confirm via the dialog's own "Ignore Everywhere" button.
+    const dialog = screen.getByRole('dialog');
+    await user.click(within(dialog).getByRole('button', { name: 'Ignore Everywhere' }));
+    expect(h.onIgnoreEverywhere).toHaveBeenCalledWith('the the');
+  });
+
+  it('Cancel in the confirm dialog does not fire onIgnoreEverywhere', async () => {
+    const h = handlers();
+    const { user } = renderWithProviders(
+      <FindingRow globalIgnoresAvailable resolved={makeResolved()} {...h} />
+    );
+    await user.click(screen.getByRole('button', { name: 'Ignore Everywhere' }));
+    await user.click(screen.getByRole('button', { name: 'Cancel' }));
+    expect(h.onIgnoreEverywhere).not.toHaveBeenCalled();
+  });
+});
+
+describe('FindingRow — inactive', () => {
+  it('renders the correct ignore-type label for each inactive reason', () => {
+    const cases: Array<[ResolvedFinding['inactiveReason'], string]> = [
+      ['occurrence', 'Ignore Here'],
+      ['global', 'Ignore Always'],
+      ['legitimate', 'Default Ignore'],
+    ];
+    for (const [reason, label] of cases) {
+      const { unmount } = renderWithProviders(
+        <FindingRow
+          globalIgnoresAvailable
+          resolved={makeResolved({ isActive: false, inactiveReason: reason })}
+          {...handlers()}
+        />
+      );
+      expect(screen.getByTestId('inactive-label')).toHaveTextContent(label);
+      // No active ignore buttons on an inactive row.
+      expect(screen.queryByRole('button', { name: 'Ignore Here' })).not.toBeInTheDocument();
+      unmount();
+    }
+  });
+
+  it('Undo Ignore default click fires onUndo with the resolved finding', async () => {
+    const h = handlers();
+    const resolved = makeResolved({ isActive: false, inactiveReason: 'occurrence' });
+    const { user } = renderWithProviders(
+      <FindingRow globalIgnoresAvailable resolved={resolved} {...h} />
+    );
+    await user.click(screen.getByRole('button', { name: 'Undo Ignore' }));
+    expect(h.onUndo).toHaveBeenCalledWith(resolved);
+  });
+
+  it('chevron menu fires onStopIgnoringEverywhere when global ignores are available', async () => {
+    const h = handlers();
+    const { user } = renderWithProviders(
+      <FindingRow
+        globalIgnoresAvailable
+        resolved={makeResolved({ isActive: false, inactiveReason: 'global' })}
+        {...h}
+      />
+    );
+    await user.click(screen.getByRole('button', { name: 'More undo options' }));
+    await user.click(screen.getByRole('menuitem', { name: /Stop ignoring/ }));
+    expect(h.onStopIgnoringEverywhere).toHaveBeenCalledWith('the the');
+  });
+
+  it('omits the chevron (global entry) when global ignores are unavailable', () => {
+    renderWithProviders(
+      <FindingRow
+        globalIgnoresAvailable={false}
+        resolved={makeResolved({ isActive: false, inactiveReason: 'occurrence' })}
+        {...handlers()}
+      />
+    );
+    expect(screen.queryByRole('button', { name: 'More undo options' })).not.toBeInTheDocument();
+    // The plain Undo button is still present.
+    expect(screen.getByRole('button', { name: 'Undo Ignore' })).toBeInTheDocument();
+  });
+});
diff --git a/src/features/checks/components/FindingRow.tsx b/src/features/checks/components/FindingRow.tsx
new file mode 100644
index 00000000..fd1fe80d
--- /dev/null
+++ b/src/features/checks/components/FindingRow.tsx
@@ -0,0 +1,174 @@
+import { useState } from 'react';
+
+import { ChevronDown } from 'lucide-react';
+
+import { Button } from '@/components/ui/button';
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from '@/components/ui/dialog';
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from '@/components/ui/dropdown-menu';
+import { cn } from '@/lib/utils';
+
+import { type InactiveReason, type ResolvedFinding } from '../checks.types';
+
+/**
+ * Human-readable ignore-type label for an inactive finding (§6.4, revised #278).
+ *
+ * Note the deliberate label mismatch: the **active** button is "Ignore
+ * Everywhere", but a finding inactivated by a global rule shows "Ignore Always"
+ * — that is the term the #278 mock uses for the inactive label. Don't unify the
+ * two (conventions §G / phase-03 gotchas).
+ */
+const INACTIVE_LABEL: Record<InactiveReason, string> = {
+  occurrence: 'Ignore Here',
+  global: 'Ignore Always',
+  legitimate: 'Default Ignore',
+};
+
+export interface FindingRowProps {
+  /** The cascade-resolved finding to render. */
+  resolved: ResolvedFinding;
+  /** Hide the global "Ignore Everywhere" affordance when the backend half is
+   *  absent (feature detection, W8). */
+  globalIgnoresAvailable: boolean;
+  /** Active: suppress just this occurrence ("Ignore Here"). */
+  onIgnoreHere: (occurrenceKey: string) => void;
+  /** Active: suppress this word pair globally — called **only after** the user
+   *  confirms the dialog (S7). */
+  onIgnoreEverywhere: (repeatedWord: string) => void;
+  /** Inactive `[Undo ▾]` default click: reverse at the occurrence layer. */
+  onUndo: (resolved: ResolvedFinding) => void;
+  /** Inactive `[Undo ▾]` chevron: deliberately stop ignoring this pair
+   *  everywhere (only offered when `globalIgnoresAvailable`). */
+  onStopIgnoringEverywhere: (repeatedWord: string) => void;
+}
+
+/**
+ * One repeated-word finding row (§5.1, §6.4–§6.5).
+ *
+ * - **Active** findings show the context snippet plus two solid-blue buttons,
+ *   `[Ignore Here]` and (when available) `[Ignore Everywhere]`. The latter opens
+ *   a confirm dialog (S7) and only fires `onIgnoreEverywhere` on confirm, while
+ *   the resulting inactive row stays reversible via `[Undo ▾]` — the two are
+ *   complementary (confirm guards the click; undo guards the mistake).
+ * - **Inactive** findings render greyed/dimmed with their ignore-type label and
+ *   one `[Undo ▾]` split button: the main button reverses at the occurrence
+ *   layer; the chevron offers the deliberate global "stop ignoring everywhere".
+ *
+ * Presentational only: it owns the confirm-dialog open state (local to the row)
+ * but never the suppression hook or the check query.
+ */
+export const FindingRow: React.FC<FindingRowProps> = ({
+  resolved,
+  globalIgnoresAvailable,
+  onIgnoreHere,
+  onIgnoreEverywhere,
+  onUndo,
+  onStopIgnoringEverywhere,
+}) => {
+  const [confirmOpen, setConfirmOpen] = useState(false);
+  const { finding, isActive, inactiveReason, occurrenceKey } = resolved;
+  // Display the original surface text (preserves casing); never compare on it —
+  // all matching already happened in the cascade (conventions §C).
+  const snippet = finding.surf;
+
+  if (isActive) {
+    return (
+      <div className='py-2' data-testid='finding-row'>
+        <p className='text-sm'>{snippet}</p>
+        <div className='mt-2 flex flex-wrap gap-2'>
+          <Button size='sm' type='button' onClick={() => onIgnoreHere(occurrenceKey)}>
+            Ignore Here
+          </Button>
+          {globalIgnoresAvailable && (
+            <Button size='sm' type='button' onClick={() => setConfirmOpen(true)}>
+              Ignore Everywhere
+            </Button>
+          )}
+        </div>
+
+        <Dialog open={confirmOpen} onOpenChange={setConfirmOpen}>
+          <DialogContent>
+            <DialogHeader>
+              <DialogTitle>Ignore everywhere?</DialogTitle>
+              <DialogDescription>
+                Fluent will not flag this word pair again across all your projects. Are you sure?
+              </DialogDescription>
+            </DialogHeader>
+            <DialogFooter>
+              <Button type='button' variant='outline' onClick={() => setConfirmOpen(false)}>
+                Cancel
+              </Button>
+              <Button
+                type='button'
+                onClick={() => {
+                  setConfirmOpen(false);
+                  onIgnoreEverywhere(finding.repeated_word);
+                }}
+              >
+                Ignore Everywhere
+              </Button>
+            </DialogFooter>
+          </DialogContent>
+        </Dialog>
+      </div>
+    );
+  }
+
+  // Inactive — greyed, ignore-type label, [Undo ▾] split button.
+  const label = inactiveReason ? INACTIVE_LABEL[inactiveReason] : 'Ignored';
+
+  return (
+    <div className='py-2 opacity-50' data-testid='finding-row'>
+      <p className='text-sm'>{snippet}</p>
+      <div className='mt-2 flex flex-wrap items-center gap-2'>
+        <span className='text-muted-foreground text-xs font-medium' data-testid='inactive-label'>
+          {label}
+        </span>
+        <div className='flex'>
+          <Button
+            className={cn(globalIgnoresAvailable && 'rounded-r-none')}
+            size='sm'
+            type='button'
+            variant='outline'
+            onClick={() => onUndo(resolved)}
+          >
+            Undo Ignore
+          </Button>
+          {globalIgnoresAvailable && (
+            <DropdownMenu>
+              <DropdownMenuTrigger asChild>
+                <Button
+                  aria-label='More undo options'
+                  className='rounded-l-none border-l-0 px-2'
+                  size='sm'
+                  type='button'
+                  variant='outline'
+                >
+                  <ChevronDown />
+                </Button>
+              </DropdownMenuTrigger>
+              <DropdownMenuContent align='end'>
+                <DropdownMenuItem onClick={() => onStopIgnoringEverywhere(finding.repeated_word)}>
+                  Stop ignoring “{finding.repeated_word}” everywhere
+                </DropdownMenuItem>
+              </DropdownMenuContent>
+            </DropdownMenu>
+          )}
+        </div>
+      </div>
+    </div>
+  );
+};
+
+export default FindingRow;

From 69c11bc6a1194536c584de227a4e1f4a2593f2c6 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 21 Jun 2026 13:01:31 -0400
Subject: [PATCH 13/33] feat(checks): wire Repeated Word Check into
 DraftingPage (Phase 4)

Wire the chapter-wide Repeated Word Check into the drafting/view page:

- Host LeftPanel (Resources | Checks) in place of the bare ResourcePanel,
  feeding ChecksPanel the resolved findings and ignore/undo actions.
- Run useRepeatedWordsCheck keyed on a saveCounter that bumps on each
  successful verse auto-save (W3, #172); gate via enabled = !readOnly &&
  hasContent && settingsProbeResolved (W10, sec 9.1).
- Apply the three-layer cascade with useResolvedFindings + useSuppressions;
  occurrence rules ride the existing debounced editor-state save (one
  writer, sec 7.1).
- Thread the active-finding count to two notification dots: the Checks tab
  dot and an S5 mirror dot on the closed-panel BookText toggle (#277).
- Persist + hydrate activeLeftTab and checkOccurrenceRules in the
  editor-state blob (W11, sec 6.6).

Adds useRepeatedWordsCheck.test.ts (builders + hook gating/refire/error).
A full DraftingPage render test is intentionally omitted: the component is
integration-heavy (router/store/loader) and the wiring behavior is already
unit-covered by the checks hooks/components; the page-level contract is
verified by the manual smoke pass.

Relates to #277, #278, #172
---
 .../bible/components/DraftingHeader.tsx       |  17 +-
 .../components/DraftingResourceSidebar.tsx    |  57 +++--
 .../bible/components/DraftingUI.test.tsx      |  26 ++
 src/features/bible/components/DraftingUI.tsx  | 110 ++++++++-
 .../hooks/useRepeatedWordsCheck.test.ts       | 233 ++++++++++++++++++
 5 files changed, 423 insertions(+), 20 deletions(-)
 create mode 100644 src/features/checks/hooks/useRepeatedWordsCheck.test.ts

diff --git a/src/features/bible/components/DraftingHeader.tsx b/src/features/bible/components/DraftingHeader.tsx
index 48eb6454..a452f4c0 100644
--- a/src/features/bible/components/DraftingHeader.tsx
+++ b/src/features/bible/components/DraftingHeader.tsx
@@ -23,6 +23,10 @@ interface DraftingHeaderProps {
   isComplete: boolean;
   isDraft: boolean;
   buttonText: string | undefined;
+  /** Cascade-resolved active-finding count (Repeated Word Check, §6.4). Drives
+   *  the S5 mirror notification dot on the resource-toggle button while the
+   *  left panel is closed (#277). */
+  activeFindingsCount: number;
   onBack: () => void;
   onToggleResources: () => void;
   onSubmit: () => Promise<void>;
@@ -39,6 +43,7 @@ export const DraftingHeader: React.FC<DraftingHeaderProps> = ({
   isComplete,
   isDraft,
   buttonText,
+  activeFindingsCount,
   onBack,
   onToggleResources,
   onSubmit,
@@ -100,11 +105,21 @@ export const DraftingHeader: React.FC<DraftingHeaderProps> = ({
                 <TooltipTrigger asChild>
                   <Button
                     aria-pressed={showResources}
-                    className='bg-primary flex cursor-pointer items-center gap-2'
+                    className='bg-primary relative flex cursor-pointer items-center gap-2'
                     type='button'
                     onClick={onToggleResources}
                   >
                     <BookText color='#ffffff' />
+                    {/* S5: mirror the active-checks notification dot on the
+                        panel-toggle button while the panel is closed, so the
+                        translator is still notified (#277). */}
+                    {!showResources && activeFindingsCount > 0 && (
+                      <span
+                        aria-label='Active checks present'
+                        className='absolute -top-1 -right-1 inline-block h-2.5 w-2.5 rounded-full bg-blue-600 ring-2 ring-white'
+                        data-testid='checks-toggle-notification-dot'
+                      />
+                    )}
                   </Button>
                 </TooltipTrigger>
                 <TooltipContent
diff --git a/src/features/bible/components/DraftingResourceSidebar.tsx b/src/features/bible/components/DraftingResourceSidebar.tsx
index b76bceec..1e958799 100644
--- a/src/features/bible/components/DraftingResourceSidebar.tsx
+++ b/src/features/bible/components/DraftingResourceSidebar.tsx
@@ -1,7 +1,9 @@
-import React, { useCallback, useEffect, useRef, useState } from 'react';
+import React, { type ReactNode, useCallback, useEffect, useRef, useState } from 'react';
 
 import { GripVertical } from 'lucide-react';
 
+import { type LeftTab } from '@/features/bible/hooks/useResourceStatePersistence';
+import { LeftPanel } from '@/features/checks/components/LeftPanel';
 import { ResourcePanel } from '@/features/resources/components/ResourcePanel';
 import { type BibleVerse } from '@/features/resources/hooks/hooks';
 import { type ProjectItem, type ResourceName } from '@/lib/types';
@@ -21,6 +23,15 @@ interface DraftingResourceSidebarProps {
   setBibleVerses: (verses: BibleVerse[]) => void;
   setCurrentLanguage: (lang: string) => void;
   setCurrentResource: (res: ResourceName) => void;
+  /** Controlled active left-panel tab (Resources | Checks), persisted by the
+   *  parent in the editor-state blob (Repeated Word Check, W11/§6.6). */
+  activeLeftTab: LeftTab;
+  /** Called when the translator switches tabs; the parent persists it. */
+  onTabChange: (tab: LeftTab) => void;
+  /** Cascade-resolved active-finding count; drives the tab notification dot. */
+  activeFindingsCount: number;
+  /** Body for the Checks tab (the `ChecksPanel`, composed by the parent). */
+  checksContent: ReactNode;
 }
 
 export const DraftingResourceSidebar: React.FC<DraftingResourceSidebarProps> = ({
@@ -38,6 +49,10 @@ export const DraftingResourceSidebar: React.FC<DraftingResourceSidebarProps> = (
   setBibleVerses,
   setCurrentLanguage,
   setCurrentResource,
+  activeLeftTab,
+  onTabChange,
+  activeFindingsCount,
+  checksContent,
 }) => {
   const [resourcePanelWidth, setResourcePanelWidth] = useState(25); // percentage
   const isDraggingRef = useRef(false);
@@ -101,22 +116,30 @@ export const DraftingResourceSidebar: React.FC<DraftingResourceSidebarProps> = (
   return (
     <>
       <div className='min-w-0 shrink-0 overflow-hidden' style={{ width: `${resourcePanelWidth}%` }}>
-        <ResourcePanel
-          activeVerseId={resourceVerseId}
-          bibleResourceName={setBibleTabLabel}
-          initialLanguage={currentLanguage}
-          initialResource={currentResource}
-          openResourceBiblePanel={setOpenResourcePanel}
-          registerClearBible={fn => {
-            clearBibleRef.current = fn;
-          }}
-          resourceNames={resourceNames}
-          selectPanel={panel => setSelectedPanel(panel as 1 | 2)}
-          sourceData={projectItem}
-          onBibleLoadingChange={setBibleContentLoading}
-          onBibleVersesChange={setBibleVerses}
-          onLanguageChange={setCurrentLanguage}
-          onResourceChange={setCurrentResource}
+        <LeftPanel
+          activeFindingsCount={activeFindingsCount}
+          activeTab={activeLeftTab}
+          checksContent={checksContent}
+          resourcesContent={
+            <ResourcePanel
+              activeVerseId={resourceVerseId}
+              bibleResourceName={setBibleTabLabel}
+              initialLanguage={currentLanguage}
+              initialResource={currentResource}
+              openResourceBiblePanel={setOpenResourcePanel}
+              registerClearBible={fn => {
+                clearBibleRef.current = fn;
+              }}
+              resourceNames={resourceNames}
+              selectPanel={panel => setSelectedPanel(panel as 1 | 2)}
+              sourceData={projectItem}
+              onBibleLoadingChange={setBibleContentLoading}
+              onBibleVersesChange={setBibleVerses}
+              onLanguageChange={setCurrentLanguage}
+              onResourceChange={setCurrentResource}
+            />
+          }
+          onTabChange={onTabChange}
         />
       </div>
       {/* Drag handle */}
diff --git a/src/features/bible/components/DraftingUI.test.tsx b/src/features/bible/components/DraftingUI.test.tsx
index 37a70ee5..1da97403 100644
--- a/src/features/bible/components/DraftingUI.test.tsx
+++ b/src/features/bible/components/DraftingUI.test.tsx
@@ -64,6 +64,32 @@ vi.mock('@/features/bible/hooks/usePericope', () => ({
   usePericope: (props: unknown) => mockUsePericope(props) as unknown,
 }));
 
+// Mock the Repeated Word Check hooks (Phase 4). These wrap TanStack Query /
+// fetch; this suite exercises drafting/pericope/resource behavior, not the
+// check itself (which is unit-tested in `useRepeatedWordsCheck.test.ts` and the
+// checks-feature component tests). Mocking them keeps DraftingUI renderable
+// without a QueryClientProvider and with no network I/O.
+vi.mock('@/features/checks/hooks/useSuppressions', () => ({
+  useSuppressions: () => ({
+    occurrenceRules: {},
+    globalRules: {},
+    globalIgnoresAvailable: false,
+    settingsProbeResolved: true,
+    ignoreHere: vi.fn(),
+    ignoreEverywhere: vi.fn(),
+    undoOccurrence: vi.fn(),
+    stopIgnoringEverywhere: vi.fn(),
+  }),
+}));
+
+vi.mock('@/features/checks/hooks/useRepeatedWordsCheck', () => ({
+  useRepeatedWordsCheck: () => ({ data: undefined, isError: false }),
+}));
+
+vi.mock('@/features/checks/hooks/useResolvedFindings', () => ({
+  useResolvedFindings: () => ({ active: [], inactive: [] }),
+}));
+
 // Mock ResourcePanel
 vi.mock('@/features/resources/components/ResourcePanel', () => ({
   ResourcePanel: ({
diff --git a/src/features/bible/components/DraftingUI.tsx b/src/features/bible/components/DraftingUI.tsx
index f973ea30..ebdea4ca 100644
--- a/src/features/bible/components/DraftingUI.tsx
+++ b/src/features/bible/components/DraftingUI.tsx
@@ -11,9 +11,15 @@ import { useChapterPresence } from '@/features/bible/hooks/useChapterPresence';
 import { useDrafting } from '@/features/bible/hooks/useDrafting';
 import { usePericope } from '@/features/bible/hooks/usePericope';
 import {
+  type LeftTab,
   useResourceState,
   useSaveResourceState,
 } from '@/features/bible/hooks/useResourceStatePersistence';
+import { type OccurrenceRules } from '@/features/checks/checks.types';
+import { ChecksPanel } from '@/features/checks/components/ChecksPanel';
+import { useRepeatedWordsCheck } from '@/features/checks/hooks/useRepeatedWordsCheck';
+import { useResolvedFindings } from '@/features/checks/hooks/useResolvedFindings';
+import { useSuppressions } from '@/features/checks/hooks/useSuppressions';
 import { type BibleVerse } from '@/features/resources/hooks/hooks';
 import { Logger } from '@/lib/services/logger';
 import {
@@ -64,6 +70,17 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
   const [bibleVerses, setBibleVerses] = useState<BibleVerse[]>([]);
   const [bibleContentLoading, setBibleContentLoading] = useState(false);
 
+  // Which left-panel tab is showing (Resources | Checks). Persisted in the
+  // editor-state blob as `activeLeftTab` (W11, §6.6).
+  const [activeLeftTab, setActiveLeftTab] = useState<LeftTab>('resources');
+  // Occurrence-level suppression rules for the Repeated Word Check (cascade
+  // layer 2). Held here so they ride the existing debounced editor-state save
+  // — the single writer for the blob (§7.1).
+  const [occurrenceRules, setOccurrenceRules] = useState<OccurrenceRules>({});
+  // Increments on every successful verse auto-save; part of the check query key
+  // so the check re-fires exactly on the auto-save event (W3, card #172).
+  const [saveCounter, setSaveCounter] = useState(0);
+
   const containerRef = useRef<HTMLDivElement>(null);
   const clearBibleRef = useRef<(() => void) | null>(null);
 
@@ -78,6 +95,8 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
     activeResource: string;
     languageCode: string;
     tabStatus: boolean;
+    activeLeftTab: LeftTab;
+    checkOccurrenceRules: OccurrenceRules;
   } | null>(null);
   const saveTimeoutRef = useRef<NodeJS.Timeout | null>(null);
 
@@ -119,6 +138,11 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
           assignedUserId: userdetail.id,
         },
       });
+
+      // Bump on the successful auto-save event so the Repeated Word Check
+      // re-fires (W3, card #172). `useAddTranslatedVerse` doesn't forward
+      // mutate-time `onSuccess`, so we bump after the awaited resolve.
+      setSaveCounter(c => c + 1);
     },
     [addVerseMutation, projectItem.projectUnitId, sourceVerses, userdetail]
   );
@@ -170,6 +194,54 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
     revealNextVerse,
   });
 
+  // --- Repeated Word Check wiring (Phase 4, §6.2/§6.6, W3/W10/W11) ----------
+
+  // The single writer for the occurrence-rule map: `useSuppressions` does the
+  // read-modify-write and hands the next full map back here; updating state
+  // makes it ride the existing debounced editor-state save (one writer, §7.1).
+  const saveOccurrenceRules = useCallback((next: OccurrenceRules) => {
+    setOccurrenceRules(next);
+  }, []);
+
+  const {
+    occurrenceRules: liveOccurrenceRules,
+    globalRules,
+    globalIgnoresAvailable,
+    settingsProbeResolved,
+    ignoreHere,
+    ignoreEverywhere,
+    undoOccurrence,
+    stopIgnoringEverywhere,
+  } = useSuppressions({ occurrenceRules, saveOccurrenceRules });
+
+  // What the translator currently sees is what gets checked (§6.2) — feed the
+  // live drafting verses, not a refetch.
+  const checkVerses = verses.map(v => ({ verseNumber: v.verseNumber, content: v.content }));
+  const hasContent = checkVerses.some(v => v.content.trim() !== '');
+
+  const checkQuery = useRepeatedWordsCheck({
+    projectItem,
+    verses: checkVerses,
+    saveCounter,
+    // Wait for the settings probe so the first render is cascade-correct, and
+    // never run in read-only `/view` or when the chapter is empty (W10/§9.1).
+    enabled: !readOnly && hasContent && settingsProbeResolved,
+  });
+
+  const resolved = useResolvedFindings({
+    findings: checkQuery.data?.result?.findings ?? [],
+    occurrenceRules: liveOccurrenceRules,
+    globalRules,
+  });
+
+  // Active-finding count drives both notification dots; computed once here and
+  // threaded to the tab and the closed-panel toggle button (S5, §6.4).
+  const activeFindingsCount = resolved.active.length;
+
+  const handleTabChange = useCallback((tab: LeftTab) => {
+    setActiveLeftTab(tab);
+  }, []);
+
   const handleBack = useCallback(() => {
     clearCurrentProjectItem();
 
@@ -186,12 +258,19 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
     if (!isFetched || isInitializedRef.current) return;
 
     if (savedResourceState) {
-      const { languageCode, tabStatus } = savedResourceState;
+      const { languageCode, tabStatus, activeLeftTab: savedTab } = savedResourceState;
+      const savedOccurrenceRules = savedResourceState.checkOccurrenceRules ?? {};
 
       if (typeof tabStatus === 'boolean') {
         setShowResources(tabStatus);
       }
 
+      if (savedTab === 'resources' || savedTab === 'checks') {
+        setActiveLeftTab(savedTab);
+      }
+
+      setOccurrenceRules(savedOccurrenceRules);
+
       setCurrentLanguage(languageCode || projectItem.sourceLangCode);
 
       lastSavedStateRef.current = {
@@ -201,6 +280,8 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
         activeResource: RESOURCE_NAMES[0].id,
         languageCode: languageCode || projectItem.sourceLangCode,
         tabStatus: typeof tabStatus === 'boolean' ? tabStatus : false,
+        activeLeftTab: savedTab === 'checks' ? 'checks' : 'resources',
+        checkOccurrenceRules: savedOccurrenceRules,
       };
     } else {
       setCurrentLanguage(projectItem.sourceLangCode);
@@ -212,6 +293,8 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
         activeResource: RESOURCE_NAMES[0].id,
         languageCode: projectItem.sourceLangCode,
         tabStatus: false,
+        activeLeftTab: 'resources',
+        checkOccurrenceRules: {},
       };
     }
 
@@ -236,6 +319,8 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
       activeResource: currentResource.id,
       languageCode: currentLanguage || projectItem.sourceLangCode,
       tabStatus: showResources,
+      activeLeftTab,
+      checkOccurrenceRules: occurrenceRules,
     };
 
     if (lastSavedStateRef.current) {
@@ -245,7 +330,11 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
         lastSavedStateRef.current.verseNumber !== currentState.verseNumber ||
         lastSavedStateRef.current.activeResource !== currentState.activeResource ||
         lastSavedStateRef.current.languageCode !== currentState.languageCode ||
-        lastSavedStateRef.current.tabStatus !== currentState.tabStatus;
+        lastSavedStateRef.current.tabStatus !== currentState.tabStatus ||
+        lastSavedStateRef.current.activeLeftTab !== currentState.activeLeftTab ||
+        // Occurrence rules are replaced by-reference on every write (the hook
+        // returns a fresh map), so an identity check is sufficient and cheap.
+        lastSavedStateRef.current.checkOccurrenceRules !== currentState.checkOccurrenceRules;
 
       if (!hasChanged) return;
     }
@@ -271,6 +360,8 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
     currentResource.id,
     currentLanguage,
     showResources,
+    activeLeftTab,
+    occurrenceRules,
     activeVerseId,
     projectItem.chapterAssignmentId,
     projectItem.book,
@@ -406,6 +497,7 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
   return (
     <div className='flex h-full flex-col overflow-hidden'>
       <DraftingHeader
+        activeFindingsCount={activeFindingsCount}
         buttonText={buttonText}
         hasAnyError={hasAnyError}
         isAnythingSaving={isAnythingSaving}
@@ -424,6 +516,19 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
       <div ref={containerRef} className='flex h-full overflow-hidden'>
         {showResources && isInitializedRef.current && (
           <DraftingResourceSidebar
+            activeFindingsCount={activeFindingsCount}
+            activeLeftTab={activeLeftTab}
+            checksContent={
+              <ChecksPanel
+                globalIgnoresAvailable={globalIgnoresAvailable}
+                isError={checkQuery.isError}
+                resolved={resolved}
+                onIgnoreEverywhere={ignoreEverywhere}
+                onIgnoreHere={ignoreHere}
+                onStopIgnoringEverywhere={stopIgnoringEverywhere}
+                onUndo={undoOccurrence}
+              />
+            }
             clearBibleRef={clearBibleRef}
             containerRef={containerRef}
             currentLanguage={currentLanguage}
@@ -438,6 +543,7 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
             setCurrentResource={setCurrentResource}
             setOpenResourcePanel={setOpenResourcePanel}
             setSelectedPanel={setSelectedPanel}
+            onTabChange={handleTabChange}
           />
         )}
 
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
new file mode 100644
index 00000000..518b1c16
--- /dev/null
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -0,0 +1,233 @@
+import { createElement, type ReactNode } from 'react';
+
+import { QueryClientProvider } from '@tanstack/react-query';
+import { http, HttpResponse } from 'msw';
+import { afterEach, describe, expect, it, vi } from 'vitest';
+
+import { config } from '@/lib/config';
+import { type ProjectItem } from '@/lib/types';
+import { server } from '@/test/msw/server';
+import { createTestQueryClient, renderHook, waitFor } from '@/test/render';
+
+import { type RepeatedWordsResponse } from '../checks.types';
+
+import {
+  buildRepeatedWordsRequest,
+  buildSntId,
+  type CheckVerseInput,
+  useRepeatedWordsCheck,
+} from './useRepeatedWordsCheck';
+
+const CHECK_URL = `${config.api.url}/ai/tools/greek-room/repeated-words`;
+
+/**
+ * A QueryClientProvider wrapper pinned to one client for the test's lifetime,
+ * so a `rerender(...)` doesn't swap in a fresh client and reset the cache.
+ * (`@/test/render`'s `renderHook` is the raw Testing Library one — no providers.)
+ * This file is `.test.ts` (no JSX), so the element is built with `createElement`.
+ */
+const makeWrapper = () => {
+  const queryClient = createTestQueryClient();
+  return ({ children }: { children: ReactNode }) =>
+    createElement(QueryClientProvider, { client: queryClient }, children);
+};
+
+/** A ProjectItem with the fields the check reads; the rest are filler. */
+const makeProjectItem = (over: Partial<ProjectItem> = {}): ProjectItem => ({
+  chapterAssignmentId: 42,
+  projectId: 100,
+  projectName: 'Test Project',
+  projectUnitId: 7,
+  bibleId: 1,
+  bibleName: 'Test Bible',
+  targetLanguage: 'spa',
+  bookId: 7,
+  // `book` is the human display name; the check must NOT use it for snt_id.
+  book: 'Judges',
+  chapterStatus: 'draft',
+  chapterNumber: 4,
+  totalVerses: 24,
+  completedVerses: 0,
+  submittedTime: null,
+  // `bookCode` is the USFM code the check builds snt_id from (W4).
+  bookCode: 'JDG',
+  sourceLangCode: 'eng',
+  ...over,
+});
+
+/** A completed-envelope response with the given findings count. */
+const makeResponse = (findingsCount = 0): RepeatedWordsResponse => ({
+  job_id: 'job-1',
+  tool: 'greek-room/repeated-words',
+  status: 'completed',
+  result: {
+    lang_code: 'spa',
+    provider: 'greek-room',
+    check: 'repeated-words',
+    findings: Array.from({ length: findingsCount }, (_, i) => ({
+      snt_id: 'JDG 4:3',
+      repeated_word: 'the the',
+      surf: 'the the',
+      start_position: i,
+      legitimate: false,
+      severity: 0.5,
+    })),
+    summary: { total_findings: findingsCount, legitimate_count: 0, verse_count: 1 },
+  },
+  error: null,
+  created_at: '2026-06-21T00:00:00Z',
+  completed_at: '2026-06-21T00:00:01Z',
+});
+
+/** Register a POST handler that captures the request body and returns a body. */
+const mockPost = (response: RepeatedWordsResponse = makeResponse(0), status = 200) => {
+  const calls: { count: number; lastBody: unknown } = { count: 0, lastBody: undefined };
+  server.use(
+    http.post(CHECK_URL, async ({ request }) => {
+      calls.count += 1;
+      calls.lastBody = await request.json();
+      return HttpResponse.json(response, { status });
+    })
+  );
+  return calls;
+};
+
+afterEach(() => {
+  vi.restoreAllMocks();
+});
+
+// ---------------------------------------------------------------------------
+// Pure builders
+// ---------------------------------------------------------------------------
+
+describe('buildSntId', () => {
+  it('formats "{bookCode} {chapter}:{verse}" with the USFM book code', () => {
+    expect(buildSntId('JDG', 4, 3)).toBe('JDG 4:3');
+  });
+});
+
+describe('buildRepeatedWordsRequest', () => {
+  it('uses projectItem.bookCode (USFM) not projectItem.book for snt_id', () => {
+    const request = buildRepeatedWordsRequest(makeProjectItem(), [
+      { verseNumber: 3, content: 'the the cat' },
+    ]);
+    expect(request.verses[0].snt_id).toBe('JDG 4:3');
+  });
+
+  it('drops empty/whitespace-only verses', () => {
+    const verses: CheckVerseInput[] = [
+      { verseNumber: 1, content: '   ' },
+      { verseNumber: 2, content: '' },
+      { verseNumber: 3, content: 'real content' },
+    ];
+    const request = buildRepeatedWordsRequest(makeProjectItem(), verses);
+    expect(request.verses).toHaveLength(1);
+    expect(request.verses[0]).toEqual({ snt_id: 'JDG 4:3', text: 'real content' });
+  });
+
+  it('carries the snake_case project context (D8 wire shape)', () => {
+    const request = buildRepeatedWordsRequest(makeProjectItem(), [
+      { verseNumber: 3, content: 'x' },
+    ]);
+    expect(request).toMatchObject({
+      lang_code: 'spa',
+      lang_name: 'spa',
+      project_id: 7,
+      project_name: 'Test Project',
+    });
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Hook
+// ---------------------------------------------------------------------------
+
+describe('useRepeatedWordsCheck — enabled gating', () => {
+  it('does not fire when disabled', async () => {
+    const calls = mockPost();
+    renderHook(
+      () =>
+        useRepeatedWordsCheck({
+          projectItem: makeProjectItem(),
+          verses: [{ verseNumber: 3, content: 'the the' }],
+          saveCounter: 0,
+          enabled: false,
+        }),
+      { wrapper: makeWrapper() }
+    );
+    // Give any rogue request a tick to land.
+    await new Promise(r => setTimeout(r, 50));
+    expect(calls.count).toBe(0);
+  });
+
+  it('does not fire when no verse has content (even if enabled)', async () => {
+    const calls = mockPost();
+    renderHook(
+      () =>
+        useRepeatedWordsCheck({
+          projectItem: makeProjectItem(),
+          verses: [{ verseNumber: 3, content: '   ' }],
+          saveCounter: 0,
+          enabled: true,
+        }),
+      { wrapper: makeWrapper() }
+    );
+    await new Promise(r => setTimeout(r, 50));
+    expect(calls.count).toBe(0);
+  });
+
+  it('fires and returns the completed envelope when enabled with content', async () => {
+    mockPost(makeResponse(2));
+    const { result } = renderHook(
+      () =>
+        useRepeatedWordsCheck({
+          projectItem: makeProjectItem(),
+          verses: [{ verseNumber: 3, content: 'the the' }],
+          saveCounter: 0,
+          enabled: true,
+        }),
+      { wrapper: makeWrapper() }
+    );
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.status).toBe('completed');
+    expect(result.current.data?.result?.findings).toHaveLength(2);
+  });
+});
+
+describe('useRepeatedWordsCheck — saveCounter re-fires', () => {
+  it('re-runs the check when saveCounter changes', async () => {
+    const calls = mockPost(makeResponse(1));
+    const { result, rerender } = renderHook(
+      (saveCounter: number) =>
+        useRepeatedWordsCheck({
+          projectItem: makeProjectItem(),
+          verses: [{ verseNumber: 3, content: 'the the' }],
+          saveCounter,
+          enabled: true,
+        }),
+      { initialProps: 0, wrapper: makeWrapper() }
+    );
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(calls.count).toBe(1);
+
+    rerender(1);
+    await waitFor(() => expect(calls.count).toBe(2));
+  });
+});
+
+describe('useRepeatedWordsCheck — error handling', () => {
+  it('surfaces a non-2xx as query.isError (no retry)', async () => {
+    mockPost(makeResponse(0), 502);
+    const { result } = renderHook(
+      () =>
+        useRepeatedWordsCheck({
+          projectItem: makeProjectItem(),
+          verses: [{ verseNumber: 3, content: 'the the' }],
+          saveCounter: 0,
+          enabled: true,
+        }),
+      { wrapper: makeWrapper() }
+    );
+    await waitFor(() => expect(result.current.isError).toBe(true));
+  });
+});

From 9fa34515d4579c39be13379b5b1e4f64f854a186 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Tue, 23 Jun 2026 16:17:57 -0400
Subject: [PATCH 14/33] fix(checks): send contract-correct lang_code +
 project_id to the AI check
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Two web→AI request-shape fixes found during the Phase-4 manual smoke of the
Repeated Word Check:

- lang_code: send the ISO 639-3 code (e.g. "eng"), not the language display
  name ("English"). greek-room keys its legitimate-duplicate whitelist on the
  ISO code, so sending the display name silently disabled all layer-0
  (legitimate) suppression — duplicates that should be auto-ignored (e.g.
  "truly truly") were flagged. Adds targetLanguageCode to ProjectItem (optional,
  with a "<unknown>" fallback) and UserChapterAssignment (required); lang_name
  still carries the human name.
- project_id: coerce to String(). fluent-ai declares project_id: str under
  Pydantic v2 strict (no int→str coercion), so a numeric value was rejected 422
  and surfaced by fluent-api as 502.

Tests: new regression cases (distinct code/name, ISO-code, empty/missing-code
fallback) + factory updates. typecheck + lint clean, 116/116 green.

Relates to #277, #278
---
 .../hooks/useRepeatedWordsCheck.test.ts       | 48 +++++++++++++++++--
 .../checks/hooks/useRepeatedWordsCheck.ts     | 21 +++++++-
 .../dashboard/user/UserHomePage.test.tsx      |  1 +
 src/lib/types.ts                              | 12 +++++
 4 files changed, 77 insertions(+), 5 deletions(-)

diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
index 518b1c16..2a974e37 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -40,7 +40,12 @@ const makeProjectItem = (over: Partial<ProjectItem> = {}): ProjectItem => ({
   projectUnitId: 7,
   bibleId: 1,
   bibleName: 'Test Bible',
-  targetLanguage: 'spa',
+  // `targetLanguage` is the human display NAME; the check must send it as
+  // `lang_name`, NOT as `lang_code` (greek-room keys its legitimate-duplicate
+  // whitelist on the ISO code). See phase-04 manual smoke (BUG #2, 2026-06-23).
+  targetLanguage: 'Spanish',
+  // `targetLanguageCode` is the ISO 639-3 code the check must send as `lang_code`.
+  targetLanguageCode: 'spa',
   bookId: 7,
   // `book` is the human display name; the check must NOT use it for snt_id.
   book: 'Judges',
@@ -130,12 +135,49 @@ describe('buildRepeatedWordsRequest', () => {
       { verseNumber: 3, content: 'x' },
     ]);
     expect(request).toMatchObject({
+      // lang_code is the ISO 639-3 CODE; lang_name is the human display NAME.
       lang_code: 'spa',
-      lang_name: 'spa',
-      project_id: 7,
+      lang_name: 'Spanish',
+      // project_id is sent as a STRING: fluent-ai declares it `str` (strict), so
+      // a numeric value is rejected 422 → 502. See the phase-04 manual smoke.
+      project_id: '7',
       project_name: 'Test Project',
     });
   });
+
+  it('sends the ISO code as lang_code and the display name as lang_name (BUG #2 regression)', () => {
+    // Greek-room keys its legitimate-duplicate whitelist on the ISO 639-3 code
+    // (e.g. "eng" recognises "truly truly"); sending the display name ("English")
+    // silently disables legitimate-duplicate suppression. lang_code MUST be the
+    // code, lang_name the name. See phase-04 manual smoke (BUG #2, 2026-06-23).
+    const request = buildRepeatedWordsRequest(
+      makeProjectItem({ targetLanguage: 'English', targetLanguageCode: 'eng' }),
+      [{ verseNumber: 3, content: 'truly truly' }]
+    );
+    expect(request.lang_code).toBe('eng');
+    expect(request.lang_name).toBe('English');
+    // Guard against a regression that sends the name as the code.
+    expect(request.lang_code).not.toBe('English');
+  });
+
+  it('falls back to "<unknown>" when targetLanguageCode is missing (no crash)', () => {
+    const request = buildRepeatedWordsRequest(
+      makeProjectItem({ targetLanguage: 'English', targetLanguageCode: undefined }),
+      [{ verseNumber: 3, content: 'truly truly' }]
+    );
+    // A missing code must NOT produce undefined/empty on the wire; greek-room
+    // just won't match any legitimate whitelist for an unknown code.
+    expect(request.lang_code).toBe('<unknown>');
+    expect(request.lang_name).toBe('English');
+  });
+
+  it('sends project_id as a string (fluent-ai requires str, not int)', () => {
+    const request = buildRepeatedWordsRequest(makeProjectItem({ projectUnitId: 123 }), [
+      { verseNumber: 3, content: 'x' },
+    ]);
+    expect(typeof request.project_id).toBe('string');
+    expect(request.project_id).toBe('123');
+  });
 });
 
 // ---------------------------------------------------------------------------
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.ts b/src/features/checks/hooks/useRepeatedWordsCheck.ts
index 1a3146f0..6ee4f655 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.ts
@@ -49,10 +49,27 @@ export const buildRepeatedWordsRequest = (
       text: v.content,
     }));
 
+  // `lang_code` MUST be the ISO 639-3 code (e.g. "eng"): greek-room keys its
+  // legitimate-duplicate whitelist on the code, so sending the display name
+  // ("English") silently disables legitimate-duplicate suppression. `lang_name`
+  // carries the human name. See phase-04 manual smoke (BUG #2, 2026-06-23).
+  //
+  // Defensive fallback: if the API omits the code (undefined) OR sends it empty
+  // (fluent-api defaults a null ISO column to ""), send "<unknown>" rather than
+  // an empty/undefined value. greek-room simply won't match any legitimate
+  // whitelist for an unknown code (duplicates are still flagged) — a safe
+  // degradation that keeps the check running instead of crashing.
+  const langCode = projectItem.targetLanguageCode?.trim() ? projectItem.targetLanguageCode : null;
+
   return {
-    lang_code: projectItem.targetLanguage,
+    lang_code: langCode ?? '<unknown>',
     lang_name: projectItem.targetLanguage,
-    project_id: projectItem.projectUnitId,
+    // fluent-ai declares `project_id: str` (Pydantic v2, strict — it does NOT
+    // coerce an int to a str), so a numeric `projectUnitId` is rejected with a
+    // 422 that fluent-api surfaces as a 502. fluent-api's own schema is the
+    // tolerant side (accepts string | number); we are the compliant side and
+    // send the contract-correct string. See phase-04 manual smoke (2026-06-23).
+    project_id: String(projectItem.projectUnitId),
     project_name: projectItem.projectName,
     verses: verseInputs,
   };
diff --git a/src/features/dashboard/user/UserHomePage.test.tsx b/src/features/dashboard/user/UserHomePage.test.tsx
index 2345a37f..62d6d2f8 100644
--- a/src/features/dashboard/user/UserHomePage.test.tsx
+++ b/src/features/dashboard/user/UserHomePage.test.tsx
@@ -37,6 +37,7 @@ const assignment = (overrides: Partial<UserChapterAssignment>): UserChapterAssig
   bibleName: 'Test Bible',
   chapterStatus: 'draft',
   targetLanguage: 'Spanish',
+  targetLanguageCode: 'spa',
   sourceLangCode: 'eng',
   bookCode: 'GEN',
   bookId: 1,
diff --git a/src/lib/types.ts b/src/lib/types.ts
index 9a2dbccc..76aa3f8b 100644
--- a/src/lib/types.ts
+++ b/src/lib/types.ts
@@ -114,7 +114,16 @@ export interface ProjectItem {
   projectUnitId: number;
   bibleId: number;
   bibleName: string;
+  /** Human-readable target language display NAME, e.g. "English". */
   targetLanguage: string;
+  /**
+   * ISO 639-3 target language CODE, e.g. "eng". Sent as the check's `lang_code`
+   * — greek-room keys its legitimate-duplicate whitelist on this code, so the
+   * display name must NOT be used here. See phase-04 manual smoke (BUG #2).
+   * Optional: older API responses may omit it; the check falls back to
+   * "<unknown>" rather than crashing.
+   */
+  targetLanguageCode?: string;
   bookId: number;
   book: string;
   chapterStatus: string;
@@ -247,7 +256,10 @@ export interface UserChapterAssignment {
   bibleId: number;
   bibleName: string;
   chapterStatus: string;
+  /** Human-readable target language display NAME, e.g. "English". */
   targetLanguage: string;
+  /** ISO 639-3 target language CODE, e.g. "eng" (sent as the check's lang_code). */
+  targetLanguageCode: string;
   sourceLangCode: string;
   bookCode: string;
   bookId: number;

From 67dbbfe484cd781734df19429d90d66be27d37b2 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Tue, 23 Jun 2026 17:14:24 -0400
Subject: [PATCH 15/33] fix(checks): apply toggle principle to suppression
 actions (BUG #3/#4)

BUG #3: ignoreHere on a surfaced legitimate finding was stacking
'suppress' over 'surface', requiring two undos. Now deletes the
'surface' override instead, letting the lower layer re-suppress.

BUG #4: stopIgnoringEverywhere chevron on a legitimate-only finding
was a no-op (no global rule to delete). Now writes a global 'surface'
to override the legitimate verdict for the pair everywhere.

Also applied the toggle principle symmetrically to ignoreEverywhere
(deletes a 'surface' global instead of stacking 'suppress').

Added 3 regression tests (19 total in useSuppressions). 119/119 green.
---
 .../checks/hooks/useSuppressions.test.ts      | 53 +++++++++++++++++++
 src/features/checks/hooks/useSuppressions.ts  | 36 ++++++++++---
 2 files changed, 83 insertions(+), 6 deletions(-)

diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 0e93ab5e..76d11fd7 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -172,6 +172,20 @@ describe('useSuppressions — occurrence actions', () => {
     });
   });
 
+  it('BUG #3 regression: ignoreHere DELETES a "surface" override instead of stacking "suppress"', async () => {
+    // Scenario: a legitimate finding was surfaced (undo), then the user clicks
+    // Ignore Here again. The 'surface' override should be removed so the finding
+    // falls back to the lower-layer verdict (legitimate), not overwritten with
+    // 'suppress' which would change the inactiveReason and require two undos.
+    const save = vi.fn();
+    const { result } = setup({ 'JDG 4:3|the the|0': 'surface' }, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreHere('JDG 4:3|the the|0'));
+    // Should delete the 'surface' key, not write 'suppress' over it.
+    expect(save).toHaveBeenCalledWith({});
+  });
+
   it('undoOccurrence DELETES the rule when the finding is hidden by its own occurrence rule', async () => {
     const save = vi.fn();
     const { result } = setup({ 'JDG 4:3|the the|0': 'suppress' }, save);
@@ -258,4 +272,43 @@ describe('useSuppressions — global actions', () => {
       })
     );
   });
+
+  it('BUG #4 regression: stopIgnoringEverywhere writes "surface" when no global suppress exists (e.g. legitimate-only finding)', async () => {
+    // Scenario: a finding is only suppressed by Greek Room legitimate (no global
+    // rule). The chevron "Stop ignoring everywhere" should write a global 'surface'
+    // to override the legitimate verdict for this pair across all projects.
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    const put = mockPut();
+    const { result } = setup({}, vi.fn());
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.stopIgnoringEverywhere('truly truly'));
+    expect(result.current.globalRules).toEqual({ 'truly truly': 'surface' });
+    await waitFor(() =>
+      expect(put.body).toEqual({
+        settings: { checkIgnoredWordPairs: { 'truly truly': 'surface' } },
+      })
+    );
+  });
+
+  it('ignoreEverywhere DELETES a "surface" global override instead of stacking "suppress" (toggle principle)', async () => {
+    // Scenario: a pair was previously globally surfaced (e.g. to override
+    // legitimate). Clicking "Ignore Everywhere" should just remove the 'surface'
+    // override, letting the lower layer (legitimate) re-suppress naturally.
+    mockGet({
+      settings: { checkIgnoredWordPairs: { 'truly truly': 'surface' } },
+      updatedAt: null,
+    });
+    const put = mockPut();
+    const { result } = setup({}, vi.fn());
+    await waitFor(() => expect(result.current.globalRules).toEqual({ 'truly truly': 'surface' }));
+
+    act(() => result.current.ignoreEverywhere('truly truly'));
+    expect(result.current.globalRules).toEqual({});
+    await waitFor(() =>
+      expect(put.body).toEqual({
+        settings: { checkIgnoredWordPairs: {} },
+      })
+    );
+  });
 });
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index dd532eb3..5816838c 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -215,9 +215,11 @@ export interface UseSuppressionsResult {
    * write an occurrence-level `'surface'` override so the global rule / Greek
    * Room verdict survives but this one occurrence re-surfaces. */
   undoOccurrence: (resolved: ResolvedFinding) => void;
-  /** Chevron action: stop ignoring this pair everywhere — delete the global
-   * entry (and purge current-chapter occurrence rules for it). Optimistic;
-   * rolls back on PUT failure. */
+  /** Chevron action: stop ignoring this pair everywhere. If a global
+   * `'suppress'` exists, deletes it; otherwise writes a global `'surface'`
+   * override (e.g. to override Greek Room `legitimate` for the pair globally).
+   * Purges current-chapter occurrence rules for the pair. Optimistic; rolls
+   * back on PUT failure. */
   stopIgnoringEverywhere: (repeatedWord: string) => void;
 }
 
@@ -270,7 +272,16 @@ export const useSuppressions = ({
 
   const ignoreHere = useCallback(
     (occurrenceKey: string) => {
-      saveOccurrenceRules(setRule(occurrenceRef.current, occurrenceKey, 'suppress'));
+      const currentVerdict = occurrenceRef.current[occurrenceKey] as RuleVerdict | undefined;
+      if (currentVerdict === 'surface') {
+        // The finding is only active because of a 'surface' override (e.g. the
+        // user previously undid a legitimate or global suppression). To re-ignore,
+        // just remove the override — the lower layer will re-suppress naturally.
+        // This avoids stacking 'suppress' over 'surface' (BUG #3).
+        saveOccurrenceRules(deleteRule(occurrenceRef.current, occurrenceKey));
+      } else {
+        saveOccurrenceRules(setRule(occurrenceRef.current, occurrenceKey, 'suppress'));
+      }
     },
     [saveOccurrenceRules]
   );
@@ -329,14 +340,27 @@ export const useSuppressions = ({
 
   const ignoreEverywhere = useCallback(
     (repeatedWord: string) => {
-      writeGlobal(repeatedWord, rules => setRule(rules, normalizePair(repeatedWord), 'suppress'));
+      const pair = normalizePair(repeatedWord);
+      writeGlobal(repeatedWord, rules => {
+        const current = rules[pair] as RuleVerdict | undefined;
+        // Toggle principle: if there's a 'surface' override, just remove it
+        // (the lower layer or default will re-suppress). Otherwise write 'suppress'.
+        return current === 'surface' ? deleteRule(rules, pair) : setRule(rules, pair, 'suppress');
+      });
     },
     [writeGlobal]
   );
 
   const stopIgnoringEverywhere = useCallback(
     (repeatedWord: string) => {
-      writeGlobal(repeatedWord, rules => deleteRule(rules, normalizePair(repeatedWord)));
+      const pair = normalizePair(repeatedWord);
+      writeGlobal(repeatedWord, rules => {
+        const current = rules[pair] as RuleVerdict | undefined;
+        // Toggle principle: if there's a 'suppress' rule, just remove it.
+        // Otherwise write 'surface' to override a lower-layer verdict (e.g.
+        // Greek Room `legitimate`) globally for this pair (BUG #4).
+        return current === 'suppress' ? deleteRule(rules, pair) : setRule(rules, pair, 'surface');
+      });
     },
     [writeGlobal]
   );

From d840aee3cfa549bb2ae3bc6dce2069f5eb7ab047 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Tue, 23 Jun 2026 17:50:15 -0400
Subject: [PATCH 16/33] feat(checks): align repeated-word panel with new mocks

- Move ignored findings into a dedicated section below the Show Ignored
  toggle (previously interleaved above it)
- Replace Show Ignored checkbox with a Switch (@radix-ui/react-switch)
- Inactive FindingRow: label left / Undo right per mock

Visual/markup only; suppression cascade and handlers untouched.
Undo split-button (delta #3) intentionally left as-is.
---
 package.json                                  |   1 +
 pnpm-lock.yaml                                | 183 ++++++++++++++++++
 src/components/ui/switch.tsx                  |  27 +++
 .../checks/components/ChecksPanel.test.tsx    |  29 ++-
 .../checks/components/ChecksPanel.tsx         | 134 +++++++------
 src/features/checks/components/FindingRow.tsx |   2 +-
 6 files changed, 310 insertions(+), 66 deletions(-)
 create mode 100644 src/components/ui/switch.tsx

diff --git a/package.json b/package.json
index d9ca7632..75656486 100644
--- a/package.json
+++ b/package.json
@@ -38,6 +38,7 @@
     "@radix-ui/react-select": "^2.2.6",
     "@radix-ui/react-separator": "^1.1.7",
     "@radix-ui/react-slot": "^1.2.2",
+    "@radix-ui/react-switch": "^1.3.1",
     "@radix-ui/react-tooltip": "^1.2.8",
     "@radix-ui/react-visually-hidden": "^1.2.3",
     "@tanstack/react-query": "^5.84.1",
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml
index ddb97cd0..6139410d 100644
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -42,6 +42,9 @@ importers:
       '@radix-ui/react-slot':
         specifier: ^1.2.2
         version: 1.2.4(@types/react@18.3.28)(react@18.3.1)
+      '@radix-ui/react-switch':
+        specifier: ^1.3.1
+        version: 1.3.1(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)
       '@radix-ui/react-tooltip':
         specifier: ^1.2.8
         version: 1.2.8(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)
@@ -799,6 +802,9 @@ packages:
   '@radix-ui/primitive@1.1.3':
     resolution: {integrity: sha512-JTF99U/6XIjCBo0wqkU5sK10glYe27MRRsfwoiq5zzOEZLHU3A3KCMa5X/azekYRCJ0HlwI0crAXS/5dEHTzDg==}
 
+  '@radix-ui/primitive@1.1.4':
+    resolution: {integrity: sha512-7AdCK9PQyiljKoBDbN8OuctCbd/esdwZPQ8RtOE3SsyQtUpiPb+ND75q0jEhC1m1ecBI0MFNeLJvwIh9iKHRcQ==}
+
   '@radix-ui/react-accordion@1.2.12':
     resolution: {integrity: sha512-T4nygeh9YE9dLRPhAHSeOZi7HBXo+0kYIPJXayZfvWOWA0+n3dESrZbjfDPUABkUNym6Hd+f2IR113To8D2GPA==}
     peerDependencies:
@@ -873,6 +879,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-compose-refs@1.1.3':
+    resolution: {integrity: sha512-rYOP8OMnuuPMQF1uhPVlGNcCDlkokKqGFE3JcxFViIkAXP7EvFWUliJAstrapypaBLJNHbZL6jGhbVDGTwmVhA==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-context@1.1.2':
     resolution: {integrity: sha512-jCi/QKUM2r1Ju5a3J64TH2A5SpKAgh0LpknyqdQ4m6DCV0xJ2HG1xARRwNGPQfi1SLdLWZ1OJz6F4OMBBNiGJA==}
     peerDependencies:
@@ -882,6 +897,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-context@1.1.4':
+    resolution: {integrity: sha512-QwH4PO5urrbO+FaGd5Aglg+YJgWTyyuZ3g/6mKvsqraLkglDdckw9JafgL5McL5VEJ6EPNduPaT3ZE9BttDAqg==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-dialog@1.1.15':
     resolution: {integrity: sha512-TCglVRtzlffRNxRMEyR36DGBLJpeusFcgMVD9PZEzAKnUs1lKCgX5u9BmC2Yg+LL9MgZDugFFs1Vl+Jp4t/PGw==}
     peerDependencies:
@@ -1065,6 +1089,19 @@ packages:
       '@types/react-dom':
         optional: true
 
+  '@radix-ui/react-primitive@2.1.6':
+    resolution: {integrity: sha512-wetd0QI77DbvrPpTAvH1SqOxsYF2wZe5TNxqwOd5Ty4XDpV3dpV0s8K/1MGMJBeY5o7lg8ub5VIt1Ub+yVen6g==}
+    peerDependencies:
+      '@types/react': '*'
+      '@types/react-dom': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+      react-dom: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+      '@types/react-dom':
+        optional: true
+
   '@radix-ui/react-roving-focus@1.1.11':
     resolution: {integrity: sha512-7A6S9jSgm/S+7MdtNDSb+IU859vQqJ/QAtcYQcfFC6W8RS4IxIZDldLR0xqCFZ6DCyrQLjLPsxtTNch5jVA4lA==}
     peerDependencies:
@@ -1122,6 +1159,28 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-slot@1.3.0':
+    resolution: {integrity: sha512-MojKku4U/miO8Av4Dkb+ctMAQx7JmY96LmtDQlAarCRtd7rN52QCSzBF+XAvr5S6coSVj9HEPBgHAHKEJVk/WA==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
+  '@radix-ui/react-switch@1.3.1':
+    resolution: {integrity: sha512-55bQtCnOB0BohomSHi6qvQXpJEEqUGDm6hRrM0Bph5OXwhSegqkd8IqgBAQkM1IlgUlWZIxpxRcpOEfRIgimyw==}
+    peerDependencies:
+      '@types/react': '*'
+      '@types/react-dom': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+      react-dom: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+      '@types/react-dom':
+        optional: true
+
   '@radix-ui/react-tooltip@1.2.8':
     resolution: {integrity: sha512-tY7sVt1yL9ozIxvmbtN5qtmH2krXcBCfjEiCgKGLqunJHvgvZG2Pcl2oQ3kbcZARb1BGEHdkLzcYGO8ynVlieg==}
     peerDependencies:
@@ -1153,6 +1212,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-use-controllable-state@1.2.3':
+    resolution: {integrity: sha512-PLzC90MS+ReootmjC597dvopoelpZ8Q61HJkDXZSExitIq7PL55vHNnesAHwguHK0aPfBnpdNzQtv1uliaqQrA==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-use-effect-event@0.0.2':
     resolution: {integrity: sha512-Qp8WbZOBe+blgpuUT+lw2xheLP8q0oatc9UpmiemEICxGvFLYmHm9QowVZGHtJlGbS6A6yJ3iViad/2cVjnOiA==}
     peerDependencies:
@@ -1162,6 +1230,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-use-effect-event@0.0.3':
+    resolution: {integrity: sha512-6c8ZqvPTWILEKnyVkP53EGRCcpnJiKTC21sS/6R1GF5xKyHJJWQEPfkqlcgUkdRQivd6tb23abUwe4ngWmY0JA==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-use-escape-keydown@1.1.1':
     resolution: {integrity: sha512-Il0+boE7w/XebUHyBjroE+DbByORGR9KKmITzbR7MyQ4akpORYP/ZmbhAr0DG7RmmBqoOnZdy2QlvajJ2QA59g==}
     peerDependencies:
@@ -1180,6 +1257,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-use-layout-effect@1.1.2':
+    resolution: {integrity: sha512-jrBWOxZITuGcnjRCM2t2U5ZPkCLxD+Ym6DjfssS5haTj2iiak/DOb64JeN6OdLfLgptb6/e2kKR+ZuTrGoZTPA==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-use-previous@1.1.1':
     resolution: {integrity: sha512-2dHfToCj/pzca2Ck724OZ5L0EVrr3eHRNsG/b3xQJLA2hZpVCS99bLAX+hm1IHXDEnzU6by5z/5MIY794/a8NQ==}
     peerDependencies:
@@ -1189,6 +1275,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-use-previous@1.1.2':
+    resolution: {integrity: sha512-IGBQPtRFdhN6MQ8dbegVmBq1LVZluya3F1jWY+puIcQC3MHctRwTDSBWCkL/3ZcnMJLTMJ++Z+ktmvg0F89iCw==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-use-rect@1.1.1':
     resolution: {integrity: sha512-QTYuDesS0VtuHNNvMh+CjlKJ4LJickCMUAqjlE3+j8w+RlRpwyX3apEQKGFzbZGdo7XNG1tXa+bQqIE7HIXT2w==}
     peerDependencies:
@@ -1207,6 +1302,15 @@ packages:
       '@types/react':
         optional: true
 
+  '@radix-ui/react-use-size@1.1.2':
+    resolution: {integrity: sha512-giWQp+4mxjBPt4KZ0MmyuykFNWfbDxKt4x+fPkRYmgRFJSbCZFzUglvMb/Kjn38tm10YP4ufiQZDx3zna4LU6w==}
+    peerDependencies:
+      '@types/react': '*'
+      react: ^16.8 || ^17.0 || ^18.0 || ^19.0 || ^19.0.0-rc
+    peerDependenciesMeta:
+      '@types/react':
+        optional: true
+
   '@radix-ui/react-visually-hidden@1.2.3':
     resolution: {integrity: sha512-pzJq12tEaaIhqjbzpCuv/OypJY/BPavOofm+dbab+MHLajy277+1lLm6JFcGgF5eskJ6mquGirhXY2GD/8u8Ug==}
     peerDependencies:
@@ -4571,6 +4675,8 @@ snapshots:
 
   '@radix-ui/primitive@1.1.3': {}
 
+  '@radix-ui/primitive@1.1.4': {}
+
   '@radix-ui/react-accordion@1.2.12(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
     dependencies:
       '@radix-ui/primitive': 1.1.3
@@ -4647,12 +4753,24 @@ snapshots:
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-compose-refs@1.1.3(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-context@1.1.2(@types/react@18.3.28)(react@18.3.1)':
     dependencies:
       react: 18.3.1
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-context@1.1.4(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-dialog@1.1.15(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
     dependencies:
       '@radix-ui/primitive': 1.1.3
@@ -4847,6 +4965,15 @@ snapshots:
       '@types/react': 18.3.28
       '@types/react-dom': 18.3.7(@types/react@18.3.28)
 
+  '@radix-ui/react-primitive@2.1.6(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
+    dependencies:
+      '@radix-ui/react-slot': 1.3.0(@types/react@18.3.28)(react@18.3.1)
+      react: 18.3.1
+      react-dom: 18.3.1(react@18.3.1)
+    optionalDependencies:
+      '@types/react': 18.3.28
+      '@types/react-dom': 18.3.7(@types/react@18.3.28)
+
   '@radix-ui/react-roving-focus@1.1.11(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
     dependencies:
       '@radix-ui/primitive': 1.1.3
@@ -4916,6 +5043,28 @@ snapshots:
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-slot@1.3.0(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      '@radix-ui/react-compose-refs': 1.1.3(@types/react@18.3.28)(react@18.3.1)
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
+  '@radix-ui/react-switch@1.3.1(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
+    dependencies:
+      '@radix-ui/primitive': 1.1.4
+      '@radix-ui/react-compose-refs': 1.1.3(@types/react@18.3.28)(react@18.3.1)
+      '@radix-ui/react-context': 1.1.4(@types/react@18.3.28)(react@18.3.1)
+      '@radix-ui/react-primitive': 2.1.6(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)
+      '@radix-ui/react-use-controllable-state': 1.2.3(@types/react@18.3.28)(react@18.3.1)
+      '@radix-ui/react-use-previous': 1.1.2(@types/react@18.3.28)(react@18.3.1)
+      '@radix-ui/react-use-size': 1.1.2(@types/react@18.3.28)(react@18.3.1)
+      react: 18.3.1
+      react-dom: 18.3.1(react@18.3.1)
+    optionalDependencies:
+      '@types/react': 18.3.28
+      '@types/react-dom': 18.3.7(@types/react@18.3.28)
+
   '@radix-ui/react-tooltip@1.2.8(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
     dependencies:
       '@radix-ui/primitive': 1.1.3
@@ -4950,6 +5099,14 @@ snapshots:
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-use-controllable-state@1.2.3(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      '@radix-ui/react-use-effect-event': 0.0.3(@types/react@18.3.28)(react@18.3.1)
+      '@radix-ui/react-use-layout-effect': 1.1.2(@types/react@18.3.28)(react@18.3.1)
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-use-effect-event@0.0.2(@types/react@18.3.28)(react@18.3.1)':
     dependencies:
       '@radix-ui/react-use-layout-effect': 1.1.1(@types/react@18.3.28)(react@18.3.1)
@@ -4957,6 +5114,13 @@ snapshots:
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-use-effect-event@0.0.3(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      '@radix-ui/react-use-layout-effect': 1.1.2(@types/react@18.3.28)(react@18.3.1)
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-use-escape-keydown@1.1.1(@types/react@18.3.28)(react@18.3.1)':
     dependencies:
       '@radix-ui/react-use-callback-ref': 1.1.1(@types/react@18.3.28)(react@18.3.1)
@@ -4970,12 +5134,24 @@ snapshots:
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-use-layout-effect@1.1.2(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-use-previous@1.1.1(@types/react@18.3.28)(react@18.3.1)':
     dependencies:
       react: 18.3.1
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-use-previous@1.1.2(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-use-rect@1.1.1(@types/react@18.3.28)(react@18.3.1)':
     dependencies:
       '@radix-ui/rect': 1.1.1
@@ -4990,6 +5166,13 @@ snapshots:
     optionalDependencies:
       '@types/react': 18.3.28
 
+  '@radix-ui/react-use-size@1.1.2(@types/react@18.3.28)(react@18.3.1)':
+    dependencies:
+      '@radix-ui/react-use-layout-effect': 1.1.2(@types/react@18.3.28)(react@18.3.1)
+      react: 18.3.1
+    optionalDependencies:
+      '@types/react': 18.3.28
+
   '@radix-ui/react-visually-hidden@1.2.3(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)':
     dependencies:
       '@radix-ui/react-primitive': 2.1.3(@types/react-dom@18.3.7(@types/react@18.3.28))(@types/react@18.3.28)(react-dom@18.3.1(react@18.3.1))(react@18.3.1)
diff --git a/src/components/ui/switch.tsx b/src/components/ui/switch.tsx
new file mode 100644
index 00000000..86da2355
--- /dev/null
+++ b/src/components/ui/switch.tsx
@@ -0,0 +1,27 @@
+import * as React from 'react';
+
+import * as SwitchPrimitive from '@radix-ui/react-switch';
+
+import { cn } from '@/lib/ui/utils';
+
+function Switch({ className, ...props }: React.ComponentProps<typeof SwitchPrimitive.Root>) {
+  return (
+    <SwitchPrimitive.Root
+      className={cn(
+        'peer data-[state=checked]:bg-primary data-[state=unchecked]:bg-input focus-visible:border-ring focus-visible:ring-ring/50 dark:data-[state=unchecked]:bg-input/80 inline-flex h-[1.15rem] w-8 shrink-0 items-center rounded-full border border-transparent shadow-xs transition-all outline-none focus-visible:ring-[3px] disabled:cursor-not-allowed disabled:opacity-50',
+        className
+      )}
+      data-slot='switch'
+      {...props}
+    >
+      <SwitchPrimitive.Thumb
+        className={cn(
+          'bg-background dark:data-[state=unchecked]:bg-foreground dark:data-[state=checked]:bg-primary-foreground pointer-events-none block size-4 rounded-full ring-0 transition-transform data-[state=checked]:translate-x-[calc(100%-2px)] data-[state=unchecked]:translate-x-0'
+        )}
+        data-slot='switch-thumb'
+      />
+    </SwitchPrimitive.Root>
+  );
+}
+
+export { Switch };
diff --git a/src/features/checks/components/ChecksPanel.test.tsx b/src/features/checks/components/ChecksPanel.test.tsx
index 66d82326..979e85ae 100644
--- a/src/features/checks/components/ChecksPanel.test.tsx
+++ b/src/features/checks/components/ChecksPanel.test.tsx
@@ -111,14 +111,37 @@ describe('ChecksPanel — Show Ignored toggle', () => {
     // Default OFF: the inactive snippet is not shown.
     expect(screen.queryByText('and and (in JDG 4:1)')).not.toBeInTheDocument();
 
-    await user.click(screen.getByRole('checkbox', { name: 'Show Ignored' }));
+    await user.click(screen.getByRole('switch', { name: 'Show Ignored' }));
     expect(screen.getByText('and and (in JDG 4:1)')).toBeInTheDocument();
     expect(screen.getByTestId('inactive-label')).toHaveTextContent('Ignore Here');
   });
 
   it('does not render the toggle when there are no inactive findings', () => {
     renderPanel({ active: [active('JDG 4:1', 'the the')], inactive: [] });
-    expect(screen.queryByRole('checkbox', { name: 'Show Ignored' })).not.toBeInTheDocument();
+    expect(screen.queryByRole('switch', { name: 'Show Ignored' })).not.toBeInTheDocument();
+  });
+
+  it('renders revealed ignored findings in a dedicated section BELOW the toggle (revised #278 mock)', async () => {
+    const resolved: ResolvedFindings = {
+      active: [active('JDG 4:1', 'the the')],
+      inactive: [inactive('JDG 4:2', 'and and', 'occurrence')],
+    };
+    const { user } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+
+    await user.click(screen.getByRole('switch', { name: 'Show Ignored' }));
+
+    const ignoredSection = screen.getByTestId('ignored-section');
+    // The ignored finding lives inside the dedicated below-toggle section...
+    expect(ignoredSection).toHaveTextContent('and and (in JDG 4:2)');
+    expect(ignoredSection).toHaveTextContent('Verse 2');
+    // ...and NOT interleaved into the active group above the toggle.
+    expect(ignoredSection).not.toHaveTextContent('the the (in JDG 4:1)');
+
+    // The toggle (switch) appears in the DOM before the ignored section.
+    const toggle = screen.getByRole('switch', { name: 'Show Ignored' });
+    expect(toggle.compareDocumentPosition(ignoredSection)).toBe(Node.DOCUMENT_POSITION_FOLLOWING);
   });
 
   it('resets the toggle (not persisted) across remounts', async () => {
@@ -129,7 +152,7 @@ describe('ChecksPanel — Show Ignored toggle', () => {
     const { user, unmount } = renderWithProviders(
       <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
     );
-    await user.click(screen.getByRole('checkbox', { name: 'Show Ignored' }));
+    await user.click(screen.getByRole('switch', { name: 'Show Ignored' }));
     expect(screen.getByText('and and (in JDG 4:1)')).toBeInTheDocument();
     unmount();
 
diff --git a/src/features/checks/components/ChecksPanel.tsx b/src/features/checks/components/ChecksPanel.tsx
index 7f780245..73fab32e 100644
--- a/src/features/checks/components/ChecksPanel.tsx
+++ b/src/features/checks/components/ChecksPanel.tsx
@@ -6,9 +6,9 @@ import {
   AccordionItem,
   AccordionTrigger,
 } from '@/components/ui/accordion';
-import { Checkbox } from '@/components/ui/checkbox';
 import { Label } from '@/components/ui/label';
 import { Separator } from '@/components/ui/separator';
+import { Switch } from '@/components/ui/switch';
 
 import { type ResolvedFinding, type ResolvedFindings } from '../checks.types';
 
@@ -49,55 +49,74 @@ interface VerseGroup {
 }
 
 /**
- * Group resolved findings by verse (`snt_id`), preserving first-seen verse order
- * and, within each verse, active findings before inactive ones. Only verses that
- * actually have rows to show produce a group (§5.1, §6.4).
+ * Group one bucket of resolved findings by verse (`snt_id`), preserving
+ * first-seen verse order. Only verses that actually have rows produce a group
+ * (§5.1, §6.4). Active and inactive findings are grouped **separately** now
+ * (revised #278 mock): active findings render above the "Show Ignored" toggle,
+ * ignored findings in their own dedicated section below it — so each call groups
+ * a single, already-filtered bucket rather than merging the two.
  */
-const groupByVerse = (
-  active: ResolvedFinding[],
-  inactive: ResolvedFinding[],
-  showIgnored: boolean
-): VerseGroup[] => {
+const groupByVerse = (findings: ResolvedFinding[]): VerseGroup[] => {
   const order: string[] = [];
-  const byVerse = new Map<string, { active: ResolvedFinding[]; inactive: ResolvedFinding[] }>();
-
-  const bucket = (sntId: string) => {
-    let entry = byVerse.get(sntId);
-    if (!entry) {
-      entry = { active: [], inactive: [] };
-      byVerse.set(sntId, entry);
+  const byVerse = new Map<string, ResolvedFinding[]>();
+
+  for (const f of findings) {
+    const sntId = f.finding.snt_id;
+    let rows = byVerse.get(sntId);
+    if (!rows) {
+      rows = [];
+      byVerse.set(sntId, rows);
       order.push(sntId);
     }
-    return entry;
-  };
-
-  for (const f of active) bucket(f.finding.snt_id).active.push(f);
-  if (showIgnored) {
-    for (const f of inactive) bucket(f.finding.snt_id).inactive.push(f);
+    rows.push(f);
   }
 
-  return order
-    .map(sntId => {
-      const entry = byVerse.get(sntId) ?? { active: [], inactive: [] };
-      return {
-        sntId,
-        heading: verseHeading(sntId),
-        rows: [...entry.active, ...entry.inactive],
-      };
-    })
-    .filter(group => group.rows.length > 0);
+  return order.map(sntId => ({
+    sntId,
+    heading: verseHeading(sntId),
+    rows: byVerse.get(sntId) ?? [],
+  }));
 };
 
+/** Render a list of verse groups with separators between (one fewer than groups). */
+const renderGroups = (
+  groups: VerseGroup[],
+  globalIgnoresAvailable: boolean,
+  handlers: Pick<
+    ChecksPanelProps,
+    'onIgnoreHere' | 'onIgnoreEverywhere' | 'onUndo' | 'onStopIgnoringEverywhere'
+  >
+) =>
+  groups.map((group, index) => (
+    <div key={group.sntId}>
+      {index > 0 && <Separator className='my-2' />}
+      <h4 className='text-sm font-bold'>{group.heading}</h4>
+      {group.rows.map(row => (
+        <FindingRow
+          key={row.occurrenceKey}
+          globalIgnoresAvailable={globalIgnoresAvailable}
+          resolved={row}
+          onIgnoreEverywhere={handlers.onIgnoreEverywhere}
+          onIgnoreHere={handlers.onIgnoreHere}
+          onStopIgnoringEverywhere={handlers.onStopIgnoringEverywhere}
+          onUndo={handlers.onUndo}
+        />
+      ))}
+    </div>
+  ));
+
 /**
  * The Checks-tab body (§5.1, §6.4–§6.6).
  *
  * A per-check accordion (only "Repeated Words" is wired; the structure
- * anticipates sibling checks) containing verse-grouped findings, a bottom
- * "Show Ignored" toggle, a zero state, and an inline error line. Purely
- * presentational — it receives the cascade-resolved buckets and the suppression
- * callbacks; it owns neither the check query nor the suppression hook (Phase 4
- * composes them), and only the session-local "Show Ignored" toggle is local
- * state (deliberately **not** persisted, per revised #278).
+ * anticipates sibling checks) containing the active verse-grouped findings, a
+ * "Show Ignored" toggle, and — when toggled on — a dedicated **ignored** section
+ * *below* the toggle (revised #278 mock layout: ignored occurrences appear below
+ * the active flags, visually distinct/dimmed). Also a zero state and an inline
+ * error line. Purely presentational — it receives the cascade-resolved buckets
+ * and the suppression callbacks; it owns neither the check query nor the
+ * suppression hook (Phase 4 composes them), and only the session-local "Show
+ * Ignored" toggle is local state (deliberately **not** persisted, per #278).
  */
 export const ChecksPanel: React.FC<ChecksPanelProps> = ({
   resolved,
@@ -111,12 +130,15 @@ export const ChecksPanel: React.FC<ChecksPanelProps> = ({
   const [showIgnored, setShowIgnored] = useState(false);
 
   const { active, inactive } = resolved;
-  const groups = groupByVerse(active, inactive, showIgnored);
+  const handlers = { onIgnoreHere, onIgnoreEverywhere, onUndo, onStopIgnoringEverywhere };
+  const activeGroups = groupByVerse(active);
+  const ignoredGroups = groupByVerse(inactive);
+  const hasIgnored = inactive.length > 0;
+  const showingIgnored = showIgnored && hasIgnored;
   // The zero state must never show on error (§9.2 — an error line over an empty
   // section instead) and only when there is genuinely nothing to display: no
   // active findings AND nothing currently revealed by the "Show Ignored" toggle.
-  const showZeroState = !isError && groups.length === 0;
-  const hasIgnored = inactive.length > 0;
+  const showZeroState = !isError && activeGroups.length === 0 && !showingIgnored;
 
   return (
     <div className='px-1 py-3'>
@@ -134,39 +156,27 @@ export const ChecksPanel: React.FC<ChecksPanelProps> = ({
               {showZeroState ? (
                 <p className='py-4 text-center text-sm font-bold'>No issues found</p>
               ) : (
-                <div>
-                  {groups.map((group, index) => (
-                    <div key={group.sntId}>
-                      {index > 0 && <Separator className='my-2' />}
-                      <h4 className='text-sm font-bold'>{group.heading}</h4>
-                      {group.rows.map(row => (
-                        <FindingRow
-                          key={row.occurrenceKey}
-                          globalIgnoresAvailable={globalIgnoresAvailable}
-                          resolved={row}
-                          onIgnoreEverywhere={onIgnoreEverywhere}
-                          onIgnoreHere={onIgnoreHere}
-                          onStopIgnoringEverywhere={onStopIgnoringEverywhere}
-                          onUndo={onUndo}
-                        />
-                      ))}
-                    </div>
-                  ))}
-                </div>
+                <div>{renderGroups(activeGroups, globalIgnoresAvailable, handlers)}</div>
               )}
 
               {hasIgnored && (
                 <div className='mt-3 flex items-center gap-2 border-t pt-3'>
-                  <Checkbox
+                  <Switch
                     checked={showIgnored}
                     id='checks-show-ignored'
-                    onCheckedChange={value => setShowIgnored(value === true)}
+                    onCheckedChange={setShowIgnored}
                   />
                   <Label className='text-sm font-normal' htmlFor='checks-show-ignored'>
                     Show Ignored
                   </Label>
                 </div>
               )}
+
+              {showingIgnored && (
+                <div className='mt-3' data-testid='ignored-section'>
+                  {renderGroups(ignoredGroups, globalIgnoresAvailable, handlers)}
+                </div>
+              )}
             </AccordionContent>
           </AccordionItem>
         </Accordion>
diff --git a/src/features/checks/components/FindingRow.tsx b/src/features/checks/components/FindingRow.tsx
index fd1fe80d..883f6da7 100644
--- a/src/features/checks/components/FindingRow.tsx
+++ b/src/features/checks/components/FindingRow.tsx
@@ -131,7 +131,7 @@ export const FindingRow: React.FC<FindingRowProps> = ({
   return (
     <div className='py-2 opacity-50' data-testid='finding-row'>
       <p className='text-sm'>{snippet}</p>
-      <div className='mt-2 flex flex-wrap items-center gap-2'>
+      <div className='mt-2 flex flex-wrap items-center justify-between gap-2'>
         <span className='text-muted-foreground text-xs font-medium' data-testid='inactive-label'>
           {label}
         </span>

From e8aa9643c08ffe1fc1416327df1b3b9676f91003 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 10:48:08 -0400
Subject: [PATCH 17/33] fix(checks): pass target language code through the PM
 open-chapter path

The repeated-words check sends greek-room a `lang_code` keyed on the ISO
639-3 target language code; its legitimate-duplicate whitelist (e.g.
"truly truly" for "eng") only matches when the code is correct.

The PM "open chapter" path builds its ProjectItem from the project
progress endpoint and silently omitted targetLanguageCode, so the check
sent "<unknown>" and legitimate duplicates were flagged (the translator
path was unaffected because its assignment response already carried the
code).

- Add the now-server-provided targetLanguageCode to ChapterAssignmentProgress
  and pass it into the ProjectItem built in ProjectDetailPage.
- Make ProjectItem.targetLanguageCode and ChapterAssignmentProgress
  .targetLanguageCode required (was optional) so the compiler forces every
  builder to supply it; the runtime "<unknown>" fallback is retained for
  empty values so the check degrades instead of crashing.
- Update the unit test's missing-code case to use an empty string (still
  exercises the runtime fallback now that the field is typed required).

Requires the matching fluent-api change that surfaces targetLanguageCode on
GET /projects/{id}/chapter-assignments/progress.
---
 .../checks/hooks/useRepeatedWordsCheck.test.ts    | 11 +++++++----
 .../projects/components/ProjectDetailPage.tsx     |  4 ++++
 src/lib/types.ts                                  | 15 ++++++++++++---
 3 files changed, 23 insertions(+), 7 deletions(-)

diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
index 2a974e37..34d5c697 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -160,13 +160,16 @@ describe('buildRepeatedWordsRequest', () => {
     expect(request.lang_code).not.toBe('English');
   });
 
-  it('falls back to "<unknown>" when targetLanguageCode is missing (no crash)', () => {
+  it('falls back to "<unknown>" when targetLanguageCode is empty (no crash)', () => {
+    // `targetLanguageCode` is a required `string` (BUG #3 made it non-optional so
+    // the compiler catches any ProjectItem constructor that omits it). The runtime
+    // fallback still degrades gracefully: an empty/whitespace code becomes
+    // "<unknown>" on the wire rather than producing undefined/empty.
     const request = buildRepeatedWordsRequest(
-      makeProjectItem({ targetLanguage: 'English', targetLanguageCode: undefined }),
+      makeProjectItem({ targetLanguage: 'English', targetLanguageCode: '' }),
       [{ verseNumber: 3, content: 'truly truly' }]
     );
-    // A missing code must NOT produce undefined/empty on the wire; greek-room
-    // just won't match any legitimate whitelist for an unknown code.
+    // greek-room just won't match any legitimate whitelist for an unknown code.
     expect(request.lang_code).toBe('<unknown>');
     expect(request.lang_name).toBe('English');
   });
diff --git a/src/features/projects/components/ProjectDetailPage.tsx b/src/features/projects/components/ProjectDetailPage.tsx
index 41e477ee..4518c08b 100644
--- a/src/features/projects/components/ProjectDetailPage.tsx
+++ b/src/features/projects/components/ProjectDetailPage.tsx
@@ -328,6 +328,10 @@ export const ProjectDetailPage: React.FC<ProjectDetailPageProps> = ({
         bibleId: assignment.bibleId,
         bibleName: projectSource,
         targetLanguage: projectTargetLanguageName,
+        // ISO 639-3 code for the repeated-words check's `lang_code` (BUG #3):
+        // the progress endpoint now surfaces this so the PM path no longer
+        // sends "<unknown>".
+        targetLanguageCode: assignment.targetLanguageCode,
         bookId: assignment.bookId,
         book: assignment.bookNameEng,
         chapterStatus: assignment.status,
diff --git a/src/lib/types.ts b/src/lib/types.ts
index 76aa3f8b..4486e833 100644
--- a/src/lib/types.ts
+++ b/src/lib/types.ts
@@ -86,6 +86,13 @@ export interface ChapterAssignmentProgress {
   bookId: number;
   bookCode: string;
   sourceLangCode: string;
+  /**
+   * ISO 639-3 target language CODE, e.g. "eng" (sent as the repeated-words
+   * check's `lang_code`). Required so the PM "open chapter" path can populate
+   * `ProjectItem.targetLanguageCode`; if it were optional the field could be
+   * silently dropped and the check would send "<unknown>" (BUG #3).
+   */
+  targetLanguageCode: string;
   bookNameEng: string;
   chapterNumber: number;
   assignedUser: AssignmentUser | null;
@@ -120,10 +127,12 @@ export interface ProjectItem {
    * ISO 639-3 target language CODE, e.g. "eng". Sent as the check's `lang_code`
    * — greek-room keys its legitimate-duplicate whitelist on this code, so the
    * display name must NOT be used here. See phase-04 manual smoke (BUG #2).
-   * Optional: older API responses may omit it; the check falls back to
-   * "<unknown>" rather than crashing.
+   * Required so the compiler forces every `ProjectItem` builder to supply it
+   * (the PM "open chapter" path silently omitted it — BUG #3). The check still
+   * degrades to "<unknown>" at runtime if the value is somehow empty, rather
+   * than crashing.
    */
-  targetLanguageCode?: string;
+  targetLanguageCode: string;
   bookId: number;
   book: string;
   chapterStatus: string;

From 2790e7151d5df8eed4d9bb010aab13f6285434b6 Mon Sep 17 00:00:00 2001
From: "coderabbitai[bot]"
 <136622811+coderabbitai[bot]@users.noreply.github.com>
Date: Thu, 25 Jun 2026 18:31:54 +0000
Subject: [PATCH 18/33] =?UTF-8?q?=F0=9F=93=9D=20CodeRabbit=20Chat:=20Add?=
 =?UTF-8?q?=20unit=20tests=20for=20PR=20changes?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 src/components/ui/switch.test.tsx             |  92 ++++
 .../hooks/useResourceStatePersistence.test.ts | 398 ++++++++++++++++++
 .../checks/components/ChecksPanel.test.tsx    | 112 +++++
 .../hooks/useRepeatedWordsCheck.test.ts       |  62 +++
 .../checks/hooks/useResolvedFindings.test.ts  |  84 ++++
 .../checks/hooks/useSuppressions.test.ts      |  91 ++++
 6 files changed, 839 insertions(+)
 create mode 100644 src/components/ui/switch.test.tsx
 create mode 100644 src/features/bible/hooks/useResourceStatePersistence.test.ts

diff --git a/src/components/ui/switch.test.tsx b/src/components/ui/switch.test.tsx
new file mode 100644
index 00000000..e90a2ea8
--- /dev/null
+++ b/src/components/ui/switch.test.tsx
@@ -0,0 +1,92 @@
+import { fireEvent, render, screen } from '@testing-library/react';
+import { describe, expect, it, vi } from 'vitest';
+
+import { Switch } from './switch';
+
+describe('Switch', () => {
+  it('renders without crashing', () => {
+    render(<Switch />);
+    expect(document.querySelector('[data-slot="switch"]')).toBeInTheDocument();
+  });
+
+  it('renders the thumb element', () => {
+    render(<Switch />);
+    expect(document.querySelector('[data-slot="switch-thumb"]')).toBeInTheDocument();
+  });
+
+  it('is unchecked by default', () => {
+    render(<Switch />);
+    const switchEl = document.querySelector('[data-slot="switch"]');
+    expect(switchEl).toHaveAttribute('data-state', 'unchecked');
+  });
+
+  it('renders checked when the checked prop is true', () => {
+    render(<Switch checked onCheckedChange={() => {}} />);
+    const switchEl = document.querySelector('[data-slot="switch"]');
+    expect(switchEl).toHaveAttribute('data-state', 'checked');
+  });
+
+  it('renders unchecked when the checked prop is false', () => {
+    render(<Switch checked={false} onCheckedChange={() => {}} />);
+    const switchEl = document.querySelector('[data-slot="switch"]');
+    expect(switchEl).toHaveAttribute('data-state', 'unchecked');
+  });
+
+  it('calls onCheckedChange when clicked', async () => {
+    const onCheckedChange = vi.fn();
+    render(<Switch checked={false} onCheckedChange={onCheckedChange} />);
+    const switchEl = document.querySelector('[data-slot="switch"]') as HTMLElement;
+    fireEvent.click(switchEl);
+    expect(onCheckedChange).toHaveBeenCalledTimes(1);
+    expect(onCheckedChange).toHaveBeenCalledWith(true);
+  });
+
+  it('calls onCheckedChange with false when toggled off', () => {
+    const onCheckedChange = vi.fn();
+    render(<Switch checked onCheckedChange={onCheckedChange} />);
+    const switchEl = document.querySelector('[data-slot="switch"]') as HTMLElement;
+    fireEvent.click(switchEl);
+    expect(onCheckedChange).toHaveBeenCalledWith(false);
+  });
+
+  it('is accessible as a switch role', () => {
+    render(<Switch aria-label='Toggle feature' />);
+    expect(screen.getByRole('switch', { name: 'Toggle feature' })).toBeInTheDocument();
+  });
+
+  it('merges extra className into the root element', () => {
+    render(<Switch className='custom-class' />);
+    const switchEl = document.querySelector('[data-slot="switch"]');
+    expect(switchEl).toHaveClass('custom-class');
+  });
+
+  it('is disabled when the disabled prop is set', () => {
+    render(<Switch disabled />);
+    const switchEl = document.querySelector('[data-slot="switch"]') as HTMLButtonElement;
+    expect(switchEl).toBeDisabled();
+  });
+
+  it('does not call onCheckedChange when disabled', () => {
+    const onCheckedChange = vi.fn();
+    render(<Switch disabled checked={false} onCheckedChange={onCheckedChange} />);
+    const switchEl = document.querySelector('[data-slot="switch"]') as HTMLElement;
+    fireEvent.click(switchEl);
+    expect(onCheckedChange).not.toHaveBeenCalled();
+  });
+
+  it('forwards a ref or any other native prop (aria-label) to the underlying element', () => {
+    render(<Switch aria-label='Dark mode' />);
+    const switchEl = document.querySelector('[data-slot="switch"]');
+    expect(switchEl).toHaveAttribute('aria-label', 'Dark mode');
+  });
+
+  it('can be identified by an id linked from a label', () => {
+    render(
+      <>
+        <label htmlFor='my-switch'>My Switch</label>
+        <Switch id='my-switch' />
+      </>
+    );
+    expect(screen.getByRole('switch', { name: 'My Switch' })).toBeInTheDocument();
+  });
+});
\ No newline at end of file
diff --git a/src/features/bible/hooks/useResourceStatePersistence.test.ts b/src/features/bible/hooks/useResourceStatePersistence.test.ts
new file mode 100644
index 00000000..8b03b9a6
--- /dev/null
+++ b/src/features/bible/hooks/useResourceStatePersistence.test.ts
@@ -0,0 +1,398 @@
+import { createElement, type ReactNode } from 'react';
+
+import { QueryClientProvider } from '@tanstack/react-query';
+import { http, HttpResponse } from 'msw';
+import { describe, expect, it, vi } from 'vitest';
+
+import { config } from '@/lib/config';
+import { server } from '@/test/msw/server';
+import { act, createTestQueryClient, renderHook, waitFor } from '@/test/render';
+
+import {
+  type FetchResourceState,
+  useResourceState,
+  useSaveResourceState,
+} from './useResourceStatePersistence';
+
+const API_BASE = config.api.url;
+const editorStateUrl = (id: number) =>
+  `${API_BASE}/chapter-assignments/${id}/editor-state`;
+
+const makeWrapper = () => {
+  const queryClient = createTestQueryClient();
+  return ({ children }: { children: ReactNode }) =>
+    createElement(QueryClientProvider, { client: queryClient }, children);
+};
+
+// ---------------------------------------------------------------------------
+// Test data factories
+// ---------------------------------------------------------------------------
+
+/** A minimal valid FetchResourceState (all required fields plus defaults for new optional ones). */
+const makeState = (over: Partial<FetchResourceState> = {}): FetchResourceState => ({
+  activeResource: 'ult',
+  languageCode: 'eng',
+  tabStatus: false,
+  ...over,
+});
+
+// ---------------------------------------------------------------------------
+// useResourceState — fetching the new optional fields
+// ---------------------------------------------------------------------------
+
+describe('useResourceState — new field: activeLeftTab', () => {
+  it('returns activeLeftTab "checks" when the server includes it', async () => {
+    const state = makeState({ activeLeftTab: 'checks' });
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(state))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.activeLeftTab).toBe('checks');
+  });
+
+  it('returns activeLeftTab "resources" when the server includes it', async () => {
+    const state = makeState({ activeLeftTab: 'resources' });
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(state))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.activeLeftTab).toBe('resources');
+  });
+
+  it('returns undefined activeLeftTab when the server omits it (old row, backward-compatible)', async () => {
+    // Old rows do not include activeLeftTab; the key must simply be absent.
+    const state = makeState(); // no activeLeftTab
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(state))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.activeLeftTab).toBeUndefined();
+  });
+});
+
+describe('useResourceState — new field: checkOccurrenceRules', () => {
+  it('returns checkOccurrenceRules when the server includes them', async () => {
+    const rules = {
+      'JDG 4:3|the the|0': 'suppress' as const,
+      'JDG 4:5|and and|0': 'surface' as const,
+    };
+    const state = makeState({ checkOccurrenceRules: rules });
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(state))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.checkOccurrenceRules).toEqual(rules);
+  });
+
+  it('returns empty checkOccurrenceRules when the server sends an empty map', async () => {
+    const state = makeState({ checkOccurrenceRules: {} });
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(state))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.checkOccurrenceRules).toEqual({});
+  });
+
+  it('returns undefined checkOccurrenceRules when the server omits it (old row, backward-compatible)', async () => {
+    const state = makeState(); // no checkOccurrenceRules
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(state))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.checkOccurrenceRules).toBeUndefined();
+  });
+});
+
+describe('useResourceState — edge cases', () => {
+  it('returns null when the chapter has no saved state (404)', async () => {
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json(null, { status: 404 }))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data).toBeNull();
+  });
+
+  it('errors when the server returns a non-404 error', async () => {
+    server.use(
+      http.get(editorStateUrl(99), () => HttpResponse.json({}, { status: 500 }))
+    );
+
+    const { result } = renderHook(() => useResourceState(99), {
+      wrapper: makeWrapper(),
+    });
+
+    await waitFor(() => expect(result.current.isError).toBe(true));
+  });
+
+  it('does not fire when chapterAssignmentId is 0 (falsy guard)', async () => {
+    const calls = { count: 0 };
+    server.use(
+      http.get(`${API_BASE}/chapter-assignments/0/editor-state`, () => {
+        calls.count += 1;
+        return HttpResponse.json({});
+      })
+    );
+
+    renderHook(() => useResourceState(0), { wrapper: makeWrapper() });
+
+    await new Promise(r => setTimeout(r, 50));
+    expect(calls.count).toBe(0);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// useSaveResourceState — persisting the new optional fields
+// ---------------------------------------------------------------------------
+
+describe('useSaveResourceState — new fields in payload', () => {
+  it('sends activeLeftTab "checks" in the resources blob', async () => {
+    const captured: { body: unknown } = { body: undefined };
+    server.use(
+      http.put(editorStateUrl(10), async ({ request }) => {
+        captured.body = await request.json();
+        return HttpResponse.json({});
+      })
+    );
+
+    const wrapper = makeWrapper();
+    const { result } = renderHook(() => useSaveResourceState(), { wrapper });
+
+    await act(async () => {
+      result.current.mutate({
+        chapterAssignmentId: 10,
+        resourceState: {
+          resources: {
+            bookCode: 'GEN',
+            chapterNumber: 1,
+            verseNumber: 1,
+            activeResource: 'ult',
+            languageCode: 'eng',
+            tabStatus: false,
+            activeLeftTab: 'checks',
+          },
+        },
+      });
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(captured.body).toMatchObject({
+      resources: { activeLeftTab: 'checks' },
+    });
+  });
+
+  it('sends checkOccurrenceRules in the resources blob', async () => {
+    const rules = { 'GEN 1:1|the the|0': 'suppress' as const };
+    const captured: { body: unknown } = { body: undefined };
+    server.use(
+      http.put(editorStateUrl(10), async ({ request }) => {
+        captured.body = await request.json();
+        return HttpResponse.json({});
+      })
+    );
+
+    const wrapper = makeWrapper();
+    const { result } = renderHook(() => useSaveResourceState(), { wrapper });
+
+    await act(async () => {
+      result.current.mutate({
+        chapterAssignmentId: 10,
+        resourceState: {
+          resources: {
+            bookCode: 'GEN',
+            chapterNumber: 1,
+            verseNumber: 1,
+            activeResource: 'ult',
+            languageCode: 'eng',
+            tabStatus: false,
+            checkOccurrenceRules: rules,
+          },
+        },
+      });
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(captured.body).toMatchObject({
+      resources: { checkOccurrenceRules: rules },
+    });
+  });
+
+  it('sends both activeLeftTab and checkOccurrenceRules together', async () => {
+    const rules = {
+      'JDG 4:3|the the|0': 'suppress' as const,
+      'JDG 4:5|and and|0': 'surface' as const,
+    };
+    const captured: { body: unknown } = { body: undefined };
+    server.use(
+      http.put(editorStateUrl(42), async ({ request }) => {
+        captured.body = await request.json();
+        return HttpResponse.json({});
+      })
+    );
+
+    const wrapper = makeWrapper();
+    const { result } = renderHook(() => useSaveResourceState(), { wrapper });
+
+    await act(async () => {
+      result.current.mutate({
+        chapterAssignmentId: 42,
+        resourceState: {
+          resources: {
+            bookCode: 'JDG',
+            chapterNumber: 4,
+            verseNumber: 3,
+            activeResource: 'ult',
+            languageCode: 'eng',
+            tabStatus: true,
+            activeLeftTab: 'checks',
+            checkOccurrenceRules: rules,
+          },
+        },
+      });
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(captured.body).toMatchObject({
+      resources: {
+        activeLeftTab: 'checks',
+        checkOccurrenceRules: rules,
+      },
+    });
+  });
+
+  it('can omit activeLeftTab and checkOccurrenceRules (backward-compatible with old callers)', async () => {
+    const captured: { body: unknown } = { body: undefined };
+    server.use(
+      http.put(editorStateUrl(5), async ({ request }) => {
+        captured.body = await request.json();
+        return HttpResponse.json({});
+      })
+    );
+
+    const wrapper = makeWrapper();
+    const { result } = renderHook(() => useSaveResourceState(), { wrapper });
+
+    await act(async () => {
+      result.current.mutate({
+        chapterAssignmentId: 5,
+        resourceState: {
+          resources: {
+            bookCode: 'GEN',
+            chapterNumber: 1,
+            verseNumber: 1,
+            activeResource: 'ult',
+            languageCode: 'eng',
+            tabStatus: false,
+            // no activeLeftTab, no checkOccurrenceRules
+          },
+        },
+      });
+    });
+
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect((captured.body as { resources: FetchResourceState }).resources.activeLeftTab).toBeUndefined();
+    expect((captured.body as { resources: FetchResourceState }).resources.checkOccurrenceRules).toBeUndefined();
+  });
+
+  it('errors and does not mutate cache on server failure', async () => {
+    server.use(
+      http.put(editorStateUrl(10), () => HttpResponse.json({}, { status: 500 }))
+    );
+
+    // Silence the expected Logger.logException call by spying without throwing
+    vi.spyOn(console, 'error').mockImplementation(() => {});
+
+    const wrapper = makeWrapper();
+    const { result } = renderHook(() => useSaveResourceState(), { wrapper });
+
+    await act(async () => {
+      result.current.mutate({
+        chapterAssignmentId: 10,
+        resourceState: {
+          resources: {
+            bookCode: 'GEN',
+            chapterNumber: 1,
+            verseNumber: 1,
+            activeResource: 'ult',
+            languageCode: 'eng',
+            tabStatus: false,
+          },
+        },
+      });
+    });
+
+    await waitFor(() => expect(result.current.isError).toBe(true));
+  });
+
+  it('updates the react-query cache on success so a subsequent useResourceState read is fresh', async () => {
+    const savedState = makeState({
+      activeLeftTab: 'checks',
+      checkOccurrenceRules: { 'JDG 4:3|the the|0': 'suppress' },
+    });
+
+    server.use(
+      http.put(editorStateUrl(10), () => HttpResponse.json({})),
+      http.get(editorStateUrl(10), () => HttpResponse.json(savedState))
+    );
+
+    // Share a single QueryClient so the mutation cache update is visible to the query.
+    const queryClient = createTestQueryClient();
+    const wrapper = ({ children }: { children: ReactNode }) =>
+      createElement(QueryClientProvider, { client: queryClient }, children);
+
+    const saveHook = renderHook(() => useSaveResourceState(), { wrapper });
+    const stateHook = renderHook(() => useResourceState(10), { wrapper });
+
+    // First, fetch the initial state.
+    await waitFor(() => expect(stateHook.result.current.isSuccess).toBe(true));
+
+    // Perform a save with the new fields.
+    await act(async () => {
+      saveHook.result.current.mutate({
+        chapterAssignmentId: 10,
+        resourceState: { resources: { ...savedState, bookCode: 'JDG', chapterNumber: 4, verseNumber: 3 } },
+      });
+    });
+
+    await waitFor(() => expect(saveHook.result.current.isSuccess).toBe(true));
+
+    // The query cache should now reflect the new fields via onSuccess -> setQueryData.
+    const cached = queryClient.getQueryData<FetchResourceState | null>(['resource-state', 10]);
+    expect(cached?.activeLeftTab).toBe('checks');
+    expect(cached?.checkOccurrenceRules).toEqual({ 'JDG 4:3|the the|0': 'suppress' });
+  });
+});
\ No newline at end of file
diff --git a/src/features/checks/components/ChecksPanel.test.tsx b/src/features/checks/components/ChecksPanel.test.tsx
index 979e85ae..8e5a4045 100644
--- a/src/features/checks/components/ChecksPanel.test.tsx
+++ b/src/features/checks/components/ChecksPanel.test.tsx
@@ -163,3 +163,115 @@ describe('ChecksPanel — Show Ignored toggle', () => {
     expect(screen.queryByText('and and (in JDG 4:1)')).not.toBeInTheDocument();
   });
 });
+
+describe('ChecksPanel — verseHeading helper (via grouping)', () => {
+  it('handles a snt_id with no colon by rendering the raw id as the heading', () => {
+    renderPanel({ active: [active('NOCOLON', 'the the')], inactive: [] });
+    expect(screen.getByRole('heading', { name: 'NOCOLON' })).toBeInTheDocument();
+  });
+
+  it('handles a snt_id whose verse segment after the colon is not a number', () => {
+    renderPanel({ active: [active('JDG 4:abc', 'the the')], inactive: [] });
+    expect(screen.getByRole('heading', { name: 'JDG 4:abc' })).toBeInTheDocument();
+  });
+
+  it('renders multi-digit verse numbers correctly', () => {
+    renderPanel({ active: [active('REV 22:21', 'the the')], inactive: [] });
+    expect(screen.getByRole('heading', { name: 'Verse 21' })).toBeInTheDocument();
+  });
+});
+
+describe('ChecksPanel — "Show Ignored" toggle visibility', () => {
+  it('does not render the ignored section before the toggle is clicked even with inactive findings', () => {
+    const resolved: ResolvedFindings = {
+      active: [],
+      inactive: [inactive('JDG 4:1', 'the the', 'legitimate')],
+    };
+    renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+    expect(screen.queryByTestId('ignored-section')).not.toBeInTheDocument();
+  });
+
+  it('shows the zero state while inactive findings exist but toggle is off', () => {
+    // No active findings + inactive findings present + toggle off = zero state visible
+    // because showingIgnored is false and activeGroups is empty.
+    const resolved: ResolvedFindings = {
+      active: [],
+      inactive: [inactive('JDG 4:1', 'the the', 'occurrence')],
+    };
+    renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+    expect(screen.getByText('No issues found')).toBeInTheDocument();
+    expect(screen.queryByTestId('ignored-section')).not.toBeInTheDocument();
+  });
+
+  it('hides the zero state once the toggle is turned on (ignored items now showing)', async () => {
+    const resolved: ResolvedFindings = {
+      active: [],
+      inactive: [inactive('JDG 4:1', 'the the', 'occurrence')],
+    };
+    const { user } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+    // Toggle is OFF: zero state is visible.
+    expect(screen.getByText('No issues found')).toBeInTheDocument();
+
+    await user.click(screen.getByRole('switch', { name: 'Show Ignored' }));
+
+    // Toggle is ON: zero state must be hidden; ignored section appears.
+    expect(screen.queryByText('No issues found')).not.toBeInTheDocument();
+    expect(screen.getByTestId('ignored-section')).toBeInTheDocument();
+  });
+
+  it('supports all three inactive reasons (occurrence / global / legitimate) in the ignored section', async () => {
+    const resolved: ResolvedFindings = {
+      active: [],
+      inactive: [
+        inactive('JDG 4:1', 'the the', 'occurrence', 0),
+        inactive('JDG 4:2', 'and and', 'global', 0),
+        inactive('JDG 4:3', 'for for', 'legitimate', 0),
+      ],
+    };
+    const { user } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...handlers()} />
+    );
+    await user.click(screen.getByRole('switch', { name: 'Show Ignored' }));
+
+    const section = screen.getByTestId('ignored-section');
+    expect(section).toHaveTextContent('Verse 1');
+    expect(section).toHaveTextContent('Verse 2');
+    expect(section).toHaveTextContent('Verse 3');
+  });
+});
+
+describe('ChecksPanel — callback wiring', () => {
+  it('passes onIgnoreHere through to the active FindingRow', async () => {
+    const h = handlers();
+    const resolved: ResolvedFindings = {
+      active: [active('JDG 4:1', 'the the')],
+      inactive: [],
+    };
+    const { user } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...h} />
+    );
+    await user.click(screen.getByRole('button', { name: 'Ignore Here' }));
+    expect(h.onIgnoreHere).toHaveBeenCalledWith('JDG 4:1|the the|0');
+  });
+
+  it('passes onUndo through to the inactive FindingRow (after toggle)', async () => {
+    const h = handlers();
+    const inactiveFinding = inactive('JDG 4:1', 'the the', 'occurrence');
+    const resolved: ResolvedFindings = {
+      active: [active('JDG 4:2', 'and and')],
+      inactive: [inactiveFinding],
+    };
+    const { user } = renderWithProviders(
+      <ChecksPanel globalIgnoresAvailable isError={false} resolved={resolved} {...h} />
+    );
+    await user.click(screen.getByRole('switch', { name: 'Show Ignored' }));
+    await user.click(screen.getByRole('button', { name: 'Undo Ignore' }));
+    expect(h.onUndo).toHaveBeenCalledWith(inactiveFinding);
+  });
+});
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
index 34d5c697..f05ca5ed 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -275,4 +275,66 @@ describe('useRepeatedWordsCheck — error handling', () => {
     );
     await waitFor(() => expect(result.current.isError).toBe(true));
   });
+
+  it('retains prior successful data on refetch failure (TanStack placeholderData behaviour)', async () => {
+    // First call succeeds with 2 findings.
+    const calls = mockPost(makeResponse(2));
+    const { result, rerender } = renderHook(
+      (saveCounter: number) =>
+        useRepeatedWordsCheck({
+          projectItem: makeProjectItem(),
+          verses: [{ verseNumber: 3, content: 'the the' }],
+          saveCounter,
+          enabled: true,
+        }),
+      { initialProps: 0, wrapper: makeWrapper() }
+    );
+    await waitFor(() => expect(result.current.isSuccess).toBe(true));
+    expect(result.current.data?.result?.findings).toHaveLength(2);
+    expect(calls.count).toBe(1);
+
+    // Second call returns a 502 — the prior data should still be on `data`.
+    mockPost(makeResponse(0), 502);
+    rerender(1);
+    await waitFor(() => expect(result.current.isError).toBe(true));
+    // TanStack keeps stale data after a refetch failure.
+    expect(result.current.data?.result?.findings).toHaveLength(2);
+  });
+});
+
+describe('buildRepeatedWordsRequest — additional edge cases', () => {
+  it('sends only verses with non-empty content (mixed empty and non-empty)', () => {
+    const request = buildRepeatedWordsRequest(makeProjectItem(), [
+      { verseNumber: 1, content: 'real text' },
+      { verseNumber: 2, content: '' },
+      { verseNumber: 3, content: '   ' },
+      { verseNumber: 4, content: 'more text' },
+    ]);
+    expect(request.verses).toHaveLength(2);
+    expect(request.verses.map(v => v.snt_id)).toEqual(['JDG 4:1', 'JDG 4:4']);
+  });
+
+  it('sends an empty verses array when all content is blank', () => {
+    const request = buildRepeatedWordsRequest(makeProjectItem(), [
+      { verseNumber: 1, content: '' },
+      { verseNumber: 2, content: '   ' },
+    ]);
+    expect(request.verses).toHaveLength(0);
+  });
+
+  it('builds snt_id from chapterNumber on projectItem (not a hardcoded chapter)', () => {
+    const request = buildRepeatedWordsRequest(
+      makeProjectItem({ chapterNumber: 12 }),
+      [{ verseNumber: 5, content: 'text' }]
+    );
+    expect(request.verses[0].snt_id).toBe('JDG 12:5');
+  });
+
+  it('uses whitespace-trimmed targetLanguageCode — empty after trim falls back to <unknown>', () => {
+    const request = buildRepeatedWordsRequest(
+      makeProjectItem({ targetLanguageCode: '   ' }),
+      [{ verseNumber: 1, content: 'text' }]
+    );
+    expect(request.lang_code).toBe('<unknown>');
+  });
 });
diff --git a/src/features/checks/hooks/useResolvedFindings.test.ts b/src/features/checks/hooks/useResolvedFindings.test.ts
index 511e2240..2f01efc1 100644
--- a/src/features/checks/hooks/useResolvedFindings.test.ts
+++ b/src/features/checks/hooks/useResolvedFindings.test.ts
@@ -229,4 +229,88 @@ describe('resolveFindings — output shape', () => {
     expect(result.active).toEqual([]);
     expect(result.inactive).toEqual([]);
   });
+
+  it('puts inactiveReason on inactive findings', () => {
+    const result = resolveFindings([makeFinding({ legitimate: true })], NO_OCCURRENCE, NO_GLOBAL);
+    expect(result.inactive[0]).toHaveProperty('inactiveReason', 'legitimate');
+  });
+
+  it('includes occurrenceKey on inactive findings', () => {
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|0': 'suppress' };
+    const result = resolveFindings([makeFinding()], occ, NO_GLOBAL);
+    expect(result.inactive[0].occurrenceKey).toBe('JDG 4:3|the the|0');
+  });
+});
+
+describe('resolveFindings — multi-verse, multi-pair scenarios', () => {
+  it('handles findings from multiple verses independently', () => {
+    const occ: OccurrenceRules = { 'JDG 4:3|the the|0': 'suppress' };
+    const findings = [
+      makeFinding({ snt_id: 'JDG 4:3', repeated_word: 'the the' }),
+      makeFinding({ snt_id: 'JDG 4:4', repeated_word: 'the the' }),
+    ];
+    const result = resolveFindings(findings, occ, NO_GLOBAL);
+    // The JDG 4:3 occurrence is suppressed; JDG 4:4 one is not (it's a different verse).
+    expect(result.active).toHaveLength(1);
+    expect(result.active[0].finding.snt_id).toBe('JDG 4:4');
+    expect(result.inactive).toHaveLength(1);
+    expect(result.inactive[0].finding.snt_id).toBe('JDG 4:3');
+  });
+
+  it('handles multiple different pairs in the same verse', () => {
+    const global: GlobalRules = { 'and and': 'suppress' };
+    const findings = [
+      makeFinding({ repeated_word: 'the the' }),
+      makeFinding({ repeated_word: 'and and', start_position: 10 }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, global);
+    expect(result.active).toHaveLength(1);
+    expect(result.active[0].finding.repeated_word).toBe('the the');
+    expect(result.inactive).toHaveLength(1);
+    expect(result.inactive[0].finding.repeated_word).toBe('and and');
+    expect(result.inactive[0].inactiveReason).toBe('global');
+  });
+
+  it('preserves input order of findings in the output buckets', () => {
+    // Two findings, same pair and verse, out of order by start_position.
+    // Ordinals are assigned by start_position, but the output ordering
+    // reflects the *input* order (the caller's order is preserved).
+    const findings = [
+      makeFinding({ start_position: 40, surf: 'B the the' }),
+      makeFinding({ start_position: 10, surf: 'A the the' }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
+    // Both active; input order preserved (start_position:40 first, then 10).
+    expect(result.active[0].finding.surf).toBe('B the the');
+    expect(result.active[1].finding.surf).toBe('A the the');
+    // But ordinals follow start_position: start_position:10 is ordinal 0, 40 is ordinal 1.
+    expect(result.active[0].ordinal).toBe(1); // was start_position:40
+    expect(result.active[1].ordinal).toBe(0); // was start_position:10
+  });
+});
+
+describe('resolveFindings — occurrence key uniqueness', () => {
+  it('builds unique occurrence keys for same pair across different verses', () => {
+    const findings = [
+      makeFinding({ snt_id: 'JDG 4:3' }),
+      makeFinding({ snt_id: 'JDG 4:4' }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
+    const keys = result.active.map(r => r.occurrenceKey);
+    expect(new Set(keys).size).toBe(2); // all unique
+    expect(keys).toContain('JDG 4:3|the the|0');
+    expect(keys).toContain('JDG 4:4|the the|0');
+  });
+
+  it('builds unique occurrence keys for different pairs in the same verse', () => {
+    const findings = [
+      makeFinding({ repeated_word: 'the the', start_position: 0 }),
+      makeFinding({ repeated_word: 'and and', start_position: 10 }),
+    ];
+    const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
+    const keys = result.active.map(r => r.occurrenceKey);
+    expect(new Set(keys).size).toBe(2);
+    expect(keys).toContain('JDG 4:3|the the|0');
+    expect(keys).toContain('JDG 4:3|and and|0');
+  });
 });
diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 76d11fd7..65d71d4a 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -311,4 +311,95 @@ describe('useSuppressions — global actions', () => {
       })
     );
   });
+
+  it('stopIgnoringEverywhere rolls back on PUT failure', async () => {
+    mockGet({
+      settings: { checkIgnoredWordPairs: { 'the the': 'suppress' } },
+      updatedAt: null,
+    });
+    mockPut(500);
+    const { result } = setup({}, vi.fn());
+    await waitFor(() =>
+      expect(result.current.globalRules).toEqual({ 'the the': 'suppress' })
+    );
+
+    act(() => result.current.stopIgnoringEverywhere('the the'));
+    // Optimistically removed, then rolls back on 500.
+    await waitFor(() => expect(result.current.globalRules).toEqual({ 'the the': 'suppress' }));
+  });
+
+  it('purge-local runs for NFC-composed vs decomposed pair in occurrence key', async () => {
+    // The occurrence key stores the NFC-normalized pair; global write purges by
+    // normalizing the repeated_word. Composed "caf\u00e9" must purge a key that
+    // stores the decomposed form "cafe\u0301" (or vice versa).
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    mockPut();
+    const save = vi.fn();
+    const decomposedKey = 'GEN 1:1|cafe\u0301 cafe\u0301|0';
+    const { result } = setup({ [decomposedKey]: 'suppress' }, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    // ignoreEverywhere with the composed form — should purge the decomposed occurrence rule.
+    act(() => result.current.ignoreEverywhere('caf\u00e9 caf\u00e9'));
+    // The save callback should have been called with an empty map (decomposed key purged).
+    expect(save).toHaveBeenCalledWith({});
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Occurrence rules propagation
+// ---------------------------------------------------------------------------
+
+describe('useSuppressions — occurrence rules propagation', () => {
+  it('passes through the injected occurrenceRules unchanged in the returned value', async () => {
+    const rules: OccurrenceRules = {
+      'JDG 4:3|the the|0': 'suppress',
+      'JDG 4:5|and and|0': 'surface',
+    };
+    const { result } = setup(rules, vi.fn());
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.occurrenceRules).toEqual(rules);
+  });
+
+  it('updates the occurrenceRules reference when the prop changes (via rerender)', async () => {
+    const save = vi.fn();
+    const { result, rerender } = setup({}, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    const newRules: OccurrenceRules = { 'JDG 4:1|the the|0': 'suppress' };
+    rerender({ occurrenceRules: newRules, saveOccurrenceRules: save });
+    expect(result.current.occurrenceRules).toEqual(newRules);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// settingsProbeResolved timing
+// ---------------------------------------------------------------------------
+
+describe('useSuppressions — settingsProbeResolved gates the check query', () => {
+  it('starts false before the probe returns', () => {
+    // The beforeEach sets up a default GET handler; the probe resolves
+    // asynchronously, so at the very start it must be false.
+    const { result } = setup();
+    expect(result.current.settingsProbeResolved).toBe(false);
+  });
+
+  it('becomes true after a 200 response', async () => {
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+  });
+
+  it('becomes true even after a 404 (unavailable but resolved)', async () => {
+    mockGet({ message: 'Not Found' }, 404);
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.globalIgnoresAvailable).toBe(false);
+  });
+
+  it('becomes true even after a non-404 server error (conservatively unavailable)', async () => {
+    mockGet({}, 503);
+    const { result } = setup();
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+    expect(result.current.globalIgnoresAvailable).toBe(false);
+  });
 });

From 63b1d4619f7c0da156bf970b3804df9323f4dd5c Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 14:52:27 -0400
Subject: [PATCH 19/33] test(checks): clean up CodeRabbit-generated unit tests
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Remove the one generated test that asserted stale findings survive a failed
  re-check; the query key includes saveCounter, so a post-save failure starts a
  new key with no cached data — the asserted retention never happens.
- Make the useRepeatedWordsCheck stale-data comments honest (the panel renders
  from DraftingPage-held state on error, not from TanStack keeping data).
- Run prettier over the generated test files (missing EOL newlines + wrapping)
  so lint/format:check pass.

Also documents the intentional defensive optional-chain guard on
targetLanguageCode at the API trust boundary.
---
 src/components/ui/switch.test.tsx             |  2 +-
 .../hooks/useResourceStatePersistence.test.ts | 53 +++++++------------
 .../hooks/useRepeatedWordsCheck.test.ts       | 39 +++-----------
 .../checks/hooks/useRepeatedWordsCheck.ts     | 21 ++++++--
 .../checks/hooks/useResolvedFindings.test.ts  |  5 +-
 .../checks/hooks/useSuppressions.test.ts      |  4 +-
 6 files changed, 47 insertions(+), 77 deletions(-)

diff --git a/src/components/ui/switch.test.tsx b/src/components/ui/switch.test.tsx
index e90a2ea8..dbe7f580 100644
--- a/src/components/ui/switch.test.tsx
+++ b/src/components/ui/switch.test.tsx
@@ -89,4 +89,4 @@ describe('Switch', () => {
     );
     expect(screen.getByRole('switch', { name: 'My Switch' })).toBeInTheDocument();
   });
-});
\ No newline at end of file
+});
diff --git a/src/features/bible/hooks/useResourceStatePersistence.test.ts b/src/features/bible/hooks/useResourceStatePersistence.test.ts
index 8b03b9a6..5619c84b 100644
--- a/src/features/bible/hooks/useResourceStatePersistence.test.ts
+++ b/src/features/bible/hooks/useResourceStatePersistence.test.ts
@@ -15,8 +15,7 @@ import {
 } from './useResourceStatePersistence';
 
 const API_BASE = config.api.url;
-const editorStateUrl = (id: number) =>
-  `${API_BASE}/chapter-assignments/${id}/editor-state`;
+const editorStateUrl = (id: number) => `${API_BASE}/chapter-assignments/${id}/editor-state`;
 
 const makeWrapper = () => {
   const queryClient = createTestQueryClient();
@@ -43,9 +42,7 @@ const makeState = (over: Partial<FetchResourceState> = {}): FetchResourceState =
 describe('useResourceState — new field: activeLeftTab', () => {
   it('returns activeLeftTab "checks" when the server includes it', async () => {
     const state = makeState({ activeLeftTab: 'checks' });
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(state))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(state)));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -57,9 +54,7 @@ describe('useResourceState — new field: activeLeftTab', () => {
 
   it('returns activeLeftTab "resources" when the server includes it', async () => {
     const state = makeState({ activeLeftTab: 'resources' });
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(state))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(state)));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -72,9 +67,7 @@ describe('useResourceState — new field: activeLeftTab', () => {
   it('returns undefined activeLeftTab when the server omits it (old row, backward-compatible)', async () => {
     // Old rows do not include activeLeftTab; the key must simply be absent.
     const state = makeState(); // no activeLeftTab
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(state))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(state)));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -92,9 +85,7 @@ describe('useResourceState — new field: checkOccurrenceRules', () => {
       'JDG 4:5|and and|0': 'surface' as const,
     };
     const state = makeState({ checkOccurrenceRules: rules });
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(state))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(state)));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -106,9 +97,7 @@ describe('useResourceState — new field: checkOccurrenceRules', () => {
 
   it('returns empty checkOccurrenceRules when the server sends an empty map', async () => {
     const state = makeState({ checkOccurrenceRules: {} });
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(state))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(state)));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -120,9 +109,7 @@ describe('useResourceState — new field: checkOccurrenceRules', () => {
 
   it('returns undefined checkOccurrenceRules when the server omits it (old row, backward-compatible)', async () => {
     const state = makeState(); // no checkOccurrenceRules
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(state))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(state)));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -135,9 +122,7 @@ describe('useResourceState — new field: checkOccurrenceRules', () => {
 
 describe('useResourceState — edge cases', () => {
   it('returns null when the chapter has no saved state (404)', async () => {
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json(null, { status: 404 }))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json(null, { status: 404 })));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -148,9 +133,7 @@ describe('useResourceState — edge cases', () => {
   });
 
   it('errors when the server returns a non-404 error', async () => {
-    server.use(
-      http.get(editorStateUrl(99), () => HttpResponse.json({}, { status: 500 }))
-    );
+    server.use(http.get(editorStateUrl(99), () => HttpResponse.json({}, { status: 500 })));
 
     const { result } = renderHook(() => useResourceState(99), {
       wrapper: makeWrapper(),
@@ -324,14 +307,16 @@ describe('useSaveResourceState — new fields in payload', () => {
     });
 
     await waitFor(() => expect(result.current.isSuccess).toBe(true));
-    expect((captured.body as { resources: FetchResourceState }).resources.activeLeftTab).toBeUndefined();
-    expect((captured.body as { resources: FetchResourceState }).resources.checkOccurrenceRules).toBeUndefined();
+    expect(
+      (captured.body as { resources: FetchResourceState }).resources.activeLeftTab
+    ).toBeUndefined();
+    expect(
+      (captured.body as { resources: FetchResourceState }).resources.checkOccurrenceRules
+    ).toBeUndefined();
   });
 
   it('errors and does not mutate cache on server failure', async () => {
-    server.use(
-      http.put(editorStateUrl(10), () => HttpResponse.json({}, { status: 500 }))
-    );
+    server.use(http.put(editorStateUrl(10), () => HttpResponse.json({}, { status: 500 })));
 
     // Silence the expected Logger.logException call by spying without throwing
     vi.spyOn(console, 'error').mockImplementation(() => {});
@@ -384,7 +369,9 @@ describe('useSaveResourceState — new fields in payload', () => {
     await act(async () => {
       saveHook.result.current.mutate({
         chapterAssignmentId: 10,
-        resourceState: { resources: { ...savedState, bookCode: 'JDG', chapterNumber: 4, verseNumber: 3 } },
+        resourceState: {
+          resources: { ...savedState, bookCode: 'JDG', chapterNumber: 4, verseNumber: 3 },
+        },
       });
     });
 
@@ -395,4 +382,4 @@ describe('useSaveResourceState — new fields in payload', () => {
     expect(cached?.activeLeftTab).toBe('checks');
     expect(cached?.checkOccurrenceRules).toEqual({ 'JDG 4:3|the the|0': 'suppress' });
   });
-});
\ No newline at end of file
+});
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
index f05ca5ed..a121f711 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -275,31 +275,6 @@ describe('useRepeatedWordsCheck — error handling', () => {
     );
     await waitFor(() => expect(result.current.isError).toBe(true));
   });
-
-  it('retains prior successful data on refetch failure (TanStack placeholderData behaviour)', async () => {
-    // First call succeeds with 2 findings.
-    const calls = mockPost(makeResponse(2));
-    const { result, rerender } = renderHook(
-      (saveCounter: number) =>
-        useRepeatedWordsCheck({
-          projectItem: makeProjectItem(),
-          verses: [{ verseNumber: 3, content: 'the the' }],
-          saveCounter,
-          enabled: true,
-        }),
-      { initialProps: 0, wrapper: makeWrapper() }
-    );
-    await waitFor(() => expect(result.current.isSuccess).toBe(true));
-    expect(result.current.data?.result?.findings).toHaveLength(2);
-    expect(calls.count).toBe(1);
-
-    // Second call returns a 502 — the prior data should still be on `data`.
-    mockPost(makeResponse(0), 502);
-    rerender(1);
-    await waitFor(() => expect(result.current.isError).toBe(true));
-    // TanStack keeps stale data after a refetch failure.
-    expect(result.current.data?.result?.findings).toHaveLength(2);
-  });
 });
 
 describe('buildRepeatedWordsRequest — additional edge cases', () => {
@@ -323,18 +298,16 @@ describe('buildRepeatedWordsRequest — additional edge cases', () => {
   });
 
   it('builds snt_id from chapterNumber on projectItem (not a hardcoded chapter)', () => {
-    const request = buildRepeatedWordsRequest(
-      makeProjectItem({ chapterNumber: 12 }),
-      [{ verseNumber: 5, content: 'text' }]
-    );
+    const request = buildRepeatedWordsRequest(makeProjectItem({ chapterNumber: 12 }), [
+      { verseNumber: 5, content: 'text' },
+    ]);
     expect(request.verses[0].snt_id).toBe('JDG 12:5');
   });
 
   it('uses whitespace-trimmed targetLanguageCode — empty after trim falls back to <unknown>', () => {
-    const request = buildRepeatedWordsRequest(
-      makeProjectItem({ targetLanguageCode: '   ' }),
-      [{ verseNumber: 1, content: 'text' }]
-    );
+    const request = buildRepeatedWordsRequest(makeProjectItem({ targetLanguageCode: '   ' }), [
+      { verseNumber: 1, content: 'text' },
+    ]);
     expect(request.lang_code).toBe('<unknown>');
   });
 });
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.ts b/src/features/checks/hooks/useRepeatedWordsCheck.ts
index 6ee4f655..24c41134 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.ts
@@ -59,6 +59,12 @@ export const buildRepeatedWordsRequest = (
   // an empty/undefined value. greek-room simply won't match any legitimate
   // whitelist for an unknown code (duplicates are still flagged) — a safe
   // degradation that keeps the check running instead of crashing.
+  // The `?.` here is intentionally defensive against a RUNTIME value the TYPE
+  // says can't happen: `targetLanguageCode` is typed non-optional, but the API
+  // can still omit it. We keep the type non-optional (so genuinely missing
+  // usages elsewhere stay caught) and guard only at this trust boundary, hence
+  // the targeted disable of the "unnecessary optional chain" rule.
+  // eslint-disable-next-line @typescript-eslint/no-unnecessary-condition -- defensive guard against an undefined runtime value the type does not model
   const langCode = projectItem.targetLanguageCode?.trim() ? projectItem.targetLanguageCode : null;
 
   return {
@@ -122,7 +128,13 @@ export interface UseRepeatedWordsCheckParams {
  *    `refetchInterval` can be added later without reshaping callers (D3/D9).
  *  - `failed` / `cancelled` ⇒ fluent-api surfaces these as HTTP 502 errors, so
  *    they arrive here as `query.isError` (handled by the panel's inline error,
- *    W9). TanStack retains the last successful `data` on refetch failure.
+ *    W9).
+ *
+ * Note on stale data: TanStack only retains the previous `data` on a refetch of
+ * the *same* query key. Because the key includes `saveCounter`, each auto-save
+ * is effectively a NEW key — so a failed re-check after a save leaves `data`
+ * `undefined`, not the prior findings. The panel handles this purely via
+ * `isError` (it does not rely on stale `data` surviving the error).
  *
  * Follows the fetch pattern in `features/bible/hooks/useBibleTarget.ts`.
  */
@@ -146,8 +158,11 @@ export const useRepeatedWordsCheck = ({
       }
     },
     enabled: enabled && hasContent,
-    // Keep the last successful findings visible on refetch failure (W9); the
-    // panel renders an inline error line above them.
+    // No automatic retry: a failed check surfaces immediately as `isError` so
+    // the panel can render its inline error line (W9). (We do NOT rely on
+    // TanStack keeping stale `data` here — the `saveCounter` query key changes
+    // on each save, so the prior findings are not retained across a failed
+    // re-check; see the stale-data note in the docblock above.)
     retry: false,
   });
 };
diff --git a/src/features/checks/hooks/useResolvedFindings.test.ts b/src/features/checks/hooks/useResolvedFindings.test.ts
index 2f01efc1..c6a31a04 100644
--- a/src/features/checks/hooks/useResolvedFindings.test.ts
+++ b/src/features/checks/hooks/useResolvedFindings.test.ts
@@ -291,10 +291,7 @@ describe('resolveFindings — multi-verse, multi-pair scenarios', () => {
 
 describe('resolveFindings — occurrence key uniqueness', () => {
   it('builds unique occurrence keys for same pair across different verses', () => {
-    const findings = [
-      makeFinding({ snt_id: 'JDG 4:3' }),
-      makeFinding({ snt_id: 'JDG 4:4' }),
-    ];
+    const findings = [makeFinding({ snt_id: 'JDG 4:3' }), makeFinding({ snt_id: 'JDG 4:4' })];
     const result = resolveFindings(findings, NO_OCCURRENCE, NO_GLOBAL);
     const keys = result.active.map(r => r.occurrenceKey);
     expect(new Set(keys).size).toBe(2); // all unique
diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 65d71d4a..1aa8d431 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -319,9 +319,7 @@ describe('useSuppressions — global actions', () => {
     });
     mockPut(500);
     const { result } = setup({}, vi.fn());
-    await waitFor(() =>
-      expect(result.current.globalRules).toEqual({ 'the the': 'suppress' })
-    );
+    await waitFor(() => expect(result.current.globalRules).toEqual({ 'the the': 'suppress' }));
 
     act(() => result.current.stopIgnoringEverywhere('the the'));
     // Optimistically removed, then rolls back on 500.

From 9507d09fa4e45ad068a6c7d16abccb1cdcc840af Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 15:47:44 -0400
Subject: [PATCH 20/33] fix(checks): recover occurrence-key pair without
 split('|')

The occurrence key is "{snt_id}|{repeated_word}|{ordinal}". purgeLocalForPair
previously recovered the pair via split('|')[1], which assumes repeated_word
contains no '|'. The two outer fields (snt_id, ordinal) are provably pipe-free,
so slice the middle field between the first and last '|' instead, making the
purge correct even if a verse-derived pair contains a pipe. Adds regression
tests for pairs containing '|', '"', and '\'.
---
 .../checks/hooks/useResolvedFindings.ts       | 12 ++++--
 .../checks/hooks/useSuppressions.test.ts      | 41 +++++++++++++++++++
 src/features/checks/hooks/useSuppressions.ts  | 32 ++++++++++++---
 3 files changed, 77 insertions(+), 8 deletions(-)

diff --git a/src/features/checks/hooks/useResolvedFindings.ts b/src/features/checks/hooks/useResolvedFindings.ts
index c4edffb6..7108d5d3 100644
--- a/src/features/checks/hooks/useResolvedFindings.ts
+++ b/src/features/checks/hooks/useResolvedFindings.ts
@@ -45,9 +45,15 @@ export const normalizePair = (repeatedWord: string): string => repeatedWord.norm
 /**
  * Build the stable occurrence-identity key `"{snt_id}|{repeated_word}|{ordinal}"`.
  * The `repeated_word` segment is NFC-normalized so a stored rule matches a fresh
- * finding regardless of composed/decomposed accent representation. The `|`
- * separator cannot appear in a `snt_id`, and `repeated_word` is the final
- * segment-pair, so keys are unambiguous (§7.1).
+ * finding regardless of composed/decomposed accent representation (§7.1).
+ *
+ * The `|` separator is chosen because the two *outer* fields are provably
+ * pipe-free — `snt_id` is `"{bookCode} {chapter}:{verse}"` and `ordinal` is a
+ * number — so the key always splits into exactly three logical parts even if the
+ * verse-derived `repeated_word` itself were to contain a `|`. Consumers that need
+ * to recover the pair therefore slice the *middle* field (between the first and
+ * last `|`) rather than `split('|')[1]`; see `purgeLocalForPair` in
+ * `useSuppressions.ts`.
  */
 export const buildOccurrenceKey = (sntId: string, repeatedWord: string, ordinal: number): string =>
   `${sntId}|${normalizePair(repeatedWord)}|${ordinal}`;
diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 1aa8d431..2a9a8efa 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -113,6 +113,28 @@ describe('purgeLocalForPair', () => {
     expect(out.changed).toBe(false);
     expect(out.rules).toBe(rules);
   });
+
+  it('CR-1 regression: purges a pair that itself contains the "|" delimiter', () => {
+    // The pair is the *middle* field; the outer fields (snt_id, ordinal) are
+    // pipe-free, so a "|" inside the repeated_word must not break matching.
+    // A naive split('|')[1] would read "a " and fail to purge this rule.
+    const rules: OccurrenceRules = {
+      'JDG 4:3|a | a|0': 'suppress',
+      'JDG 4:3|and and|0': 'suppress',
+    };
+    const { rules: next, changed } = purgeLocalForPair(rules, 'a | a');
+    expect(changed).toBe(true);
+    expect(next).toEqual({ 'JDG 4:3|and and|0': 'suppress' });
+  });
+
+  it('CR-1 regression: handles pairs with other awkward characters (quote, backslash)', () => {
+    const rules: OccurrenceRules = {
+      'GEN 1:1|a" a"|0': 'suppress',
+      'GEN 1:2|b\\ b\\|0': 'surface',
+    };
+    expect(purgeLocalForPair(rules, 'a" a"').rules).toEqual({ 'GEN 1:2|b\\ b\\|0': 'surface' });
+    expect(purgeLocalForPair(rules, 'b\\ b\\').rules).toEqual({ 'GEN 1:1|a" a"|0': 'suppress' });
+  });
 });
 
 // ---------------------------------------------------------------------------
@@ -326,6 +348,25 @@ describe('useSuppressions — global actions', () => {
     await waitFor(() => expect(result.current.globalRules).toEqual({ 'the the': 'suppress' }));
   });
 
+  it('CR-1 regression: ignoreEverywhere purges a current-chapter rule whose pair contains "|"', async () => {
+    // End-to-end through the real purge path (purgeCurrentChapter → purgeLocalForPair).
+    // The occurrence key for a pipe-containing pair must still be purged on a
+    // global "Ignore Everywhere"; otherwise the just-clicked panel keeps a stale
+    // occurrence rule (occurrence beats global) for that pair.
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    mockPut();
+    const save = vi.fn();
+    const { result } = setup(
+      { 'JDG 4:3|a | a|0': 'suppress', 'JDG 4:3|and and|0': 'suppress' },
+      save
+    );
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreEverywhere('a | a'));
+    // Only the pipe-containing rule is purged; the unrelated one stands.
+    expect(save).toHaveBeenCalledWith({ 'JDG 4:3|and and|0': 'suppress' });
+  });
+
   it('purge-local runs for NFC-composed vs decomposed pair in occurrence key', async () => {
     // The occurrence key stores the NFC-normalized pair; global write purges by
     // normalizing the repeated_word. Composed "caf\u00e9" must purge a key that
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index 5816838c..08edd0fc 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -141,12 +141,36 @@ export const deleteRule = <T extends Record<string, RuleVerdict>>(rules: T, key:
   return next;
 };
 
+/**
+ * Extract the `repeated_word` segment from an occurrence key
+ * (`"{snt_id}|{repeated_word}|{ordinal}"`).
+ *
+ * We deliberately do **not** `split('|')`: the `repeated_word` is the *middle*
+ * field, and while `snt_id` (`"{bookCode} {chapter}:{verse}"`) and `ordinal` (a
+ * number) provably contain no `|`, the pair itself is verse-derived text and is
+ * not guaranteed to be pipe-free. A naive `split('|')[1]` would mis-slice such a
+ * pair (e.g. a key `"JDG 4:3|a | a|0"` would yield `"a "` instead of `"a | a"`)
+ * and silently fail to purge that rule. Because the two *outer* fields are
+ * pipe-free, the pair is exactly the substring between the first and last `|`,
+ * which we recover unambiguously regardless of any `|` inside it. Returns
+ * `undefined` for a malformed key (fewer than two delimiters).
+ */
+const extractPairFromKey = (key: string): string | undefined => {
+  const first = key.indexOf('|');
+  const last = key.lastIndexOf('|');
+  if (first === -1 || last === first) {
+    return undefined;
+  }
+  return key.slice(first + 1, last);
+};
+
 /**
  * Remove every occurrence rule that targets `repeatedWord` (NFC-normalized) in
  * the current chapter's occurrence map — used by "purge-local on global write"
  * (§6.5). The occurrence key is `"{snt_id}|{repeated_word}|{ordinal}"`, so we
- * match the normalized middle segment. Returns a new map (or the same instance
- * if nothing matched, to avoid needless writes).
+ * match the normalized middle segment (see {@link extractPairFromKey} for why we
+ * slice between the outer delimiters rather than `split('|')`). Returns a new map
+ * (or the same instance if nothing matched, to avoid needless writes).
  */
 export const purgeLocalForPair = (
   occurrenceRules: OccurrenceRules,
@@ -156,9 +180,7 @@ export const purgeLocalForPair = (
   let changed = false;
   const next: OccurrenceRules = {};
   for (const [key, verdict] of Object.entries(occurrenceRules)) {
-    const segments = key.split('|');
-    // key = snt_id | repeated_word | ordinal  (snt_id has no '|')
-    const pair = segments.length >= 3 ? segments[1] : undefined;
+    const pair = extractPairFromKey(key);
     if (pair !== undefined && normalizePair(pair) === target) {
       changed = true;
       continue; // drop it

From b1f4978e724eb806ff7eca869cae22e5c725859d Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 16:08:21 -0400
Subject: [PATCH 21/33] refactor(checks): dedupe LeftTab via single source in
 useResourceStatePersistence

LeftPanel re-declared the identical 'resources' | 'checks' union already
exported from useResourceStatePersistence (the editor-state activeLeftTab
field, which DraftingPage imports). Import the type from there and
re-export it so LeftPanel's public type surface is unchanged. Type-only;
no runtime or persisted-shape change.
---
 src/features/checks/components/LeftPanel.tsx | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/src/features/checks/components/LeftPanel.tsx b/src/features/checks/components/LeftPanel.tsx
index f3e79a20..ae54337f 100644
--- a/src/features/checks/components/LeftPanel.tsx
+++ b/src/features/checks/components/LeftPanel.tsx
@@ -1,9 +1,15 @@
 import { type ReactNode } from 'react';
 
+import { type LeftTab } from '@/features/bible/hooks/useResourceStatePersistence';
 import { cn } from '@/lib/utils';
 
-/** Which tab of the drafting-page left panel is showing (persisted, W11). */
-export type LeftTab = 'resources' | 'checks';
+/**
+ * Which tab of the drafting-page left panel is showing (persisted, W11).
+ * Single source of truth lives in `useResourceStatePersistence`, where it is
+ * the editor-state `activeLeftTab` field; re-exported here so consumers of the
+ * panel's type surface can keep importing it from `LeftPanel`.
+ */
+export type { LeftTab };
 
 export interface LeftPanelProps {
   /** Controlled active tab; lifted to `DraftingPage` so it can be persisted

From 3c5119dbdf3bd7b540ac2f0267d3a239f691953d Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 16:18:14 -0400
Subject: [PATCH 22/33] a11y(checks): complete LeftPanel tab/tabpanel ARIA
 wiring

Wire the drafting-page left-panel tabs to their shared body container:
each tab gets a stable id + aria-controls, and the panel div now has
role=tabpanel, an id, and aria-labelledby tracking the active tab. Tidies
our own unmerged tab markup so assistive tech announces the controlled
panel. Markup/attribute-only; no behavior, state, or wire-contract change.
Adds two unit tests covering the aria-controls and aria-labelledby wiring.
---
 .../checks/components/LeftPanel.test.tsx      | 24 ++++++++++++++++
 src/features/checks/components/LeftPanel.tsx  | 28 +++++++++++++++++--
 2 files changed, 50 insertions(+), 2 deletions(-)

diff --git a/src/features/checks/components/LeftPanel.test.tsx b/src/features/checks/components/LeftPanel.test.tsx
index 07fd26ca..fb19cc99 100644
--- a/src/features/checks/components/LeftPanel.test.tsx
+++ b/src/features/checks/components/LeftPanel.test.tsx
@@ -37,6 +37,30 @@ describe('LeftPanel', () => {
     );
   });
 
+  it('wires both tabs to the shared tabpanel via aria-controls', () => {
+    render(<LeftPanel {...baseProps} activeTab='resources' />);
+    const panel = screen.getByRole('tabpanel');
+    const panelId = panel.getAttribute('id');
+    expect(panelId).toBeTruthy();
+    expect(screen.getByRole('tab', { name: /Resources/ })).toHaveAttribute(
+      'aria-controls',
+      panelId
+    );
+    expect(screen.getByRole('tab', { name: /Checks/ })).toHaveAttribute('aria-controls', panelId);
+  });
+
+  it('labels the tabpanel by the active tab (aria-labelledby tracks the active tab)', () => {
+    const { rerender } = render(<LeftPanel {...baseProps} activeTab='resources' />);
+    const resourcesTabId = screen.getByRole('tab', { name: /Resources/ }).getAttribute('id');
+    expect(resourcesTabId).toBeTruthy();
+    expect(screen.getByRole('tabpanel')).toHaveAttribute('aria-labelledby', resourcesTabId);
+
+    rerender(<LeftPanel {...baseProps} activeTab='checks' />);
+    const checksTabId = screen.getByRole('tab', { name: /Checks/ }).getAttribute('id');
+    expect(checksTabId).toBeTruthy();
+    expect(screen.getByRole('tabpanel')).toHaveAttribute('aria-labelledby', checksTabId);
+  });
+
   it('calls onTabChange when a tab is clicked', () => {
     const onTabChange = vi.fn();
     render(<LeftPanel {...baseProps} activeTab='resources' onTabChange={onTabChange} />);
diff --git a/src/features/checks/components/LeftPanel.tsx b/src/features/checks/components/LeftPanel.tsx
index ae54337f..22cd43cd 100644
--- a/src/features/checks/components/LeftPanel.tsx
+++ b/src/features/checks/components/LeftPanel.tsx
@@ -26,15 +26,29 @@ export interface LeftPanelProps {
   checksContent: ReactNode;
 }
 
+/**
+ * Stable element ids wiring the tabs to the shared panel (W11, §6.6). The two
+ * tabs share a single body container; `aria-labelledby` on the panel points at
+ * whichever tab is currently active, so screen readers announce the panel's
+ * controlling tab. (CR-5: complete the ARIA tab/tabpanel relationship.)
+ */
+const TAB_IDS: Record<LeftTab, string> = {
+  resources: 'left-panel-tab-resources',
+  checks: 'left-panel-tab-checks',
+};
+const PANEL_ID = 'left-panel-tabpanel';
+
 interface TabButtonProps {
+  id: string;
   label: string;
   isActive: boolean;
   onClick: () => void;
   showDot?: boolean;
 }
 
-const TabButton: React.FC<TabButtonProps> = ({ label, isActive, onClick, showDot = false }) => (
+const TabButton: React.FC<TabButtonProps> = ({ id, label, isActive, onClick, showDot = false }) => (
   <button
+    aria-controls={PANEL_ID}
     aria-selected={isActive}
     className={cn(
       'flex items-center gap-1.5 border-b-2 px-1 pb-2 text-base font-semibold transition-colors',
@@ -42,6 +56,7 @@ const TabButton: React.FC<TabButtonProps> = ({ label, isActive, onClick, showDot
         ? 'border-blue-600 text-blue-600'
         : 'text-muted-foreground hover:text-foreground border-transparent'
     )}
+    id={id}
     role='tab'
     type='button'
     onClick={onClick}
@@ -85,6 +100,7 @@ export const LeftPanel: React.FC<LeftPanelProps> = ({
     <div className='flex h-full flex-col'>
       <div aria-label='Left panel' className='flex items-center gap-6 px-1 pt-4' role='tablist'>
         <TabButton
+          id={TAB_IDS.resources}
           isActive={activeTab === 'resources'}
           label='Resources'
           onClick={() => onTabChange('resources')}
@@ -93,6 +109,7 @@ export const LeftPanel: React.FC<LeftPanelProps> = ({
             and is visible whatever tab is active (mock shows it while the
             Resources tab is active too) — §5.1, W11. */}
         <TabButton
+          id={TAB_IDS.checks}
           isActive={activeTab === 'checks'}
           label='Checks'
           showDot={hasActiveFindings}
@@ -100,7 +117,14 @@ export const LeftPanel: React.FC<LeftPanelProps> = ({
         />
       </div>
 
-      <div className='min-h-0 flex-1'>
+      {/* Single shared panel: `aria-labelledby` tracks the active tab so the
+          tab/tabpanel relationship is complete for assistive tech (CR-5). */}
+      <div
+        aria-labelledby={TAB_IDS[activeTab]}
+        className='min-h-0 flex-1'
+        id={PANEL_ID}
+        role='tabpanel'
+      >
         {activeTab === 'resources' ? resourcesContent : checksContent}
       </div>
     </div>

From f5b1283733ba7884de55a2e4c99f6ddeca9a110f Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 16:59:33 -0400
Subject: [PATCH 23/33] fix(checks): forward the trimmed lang_code, not the
 padded value

buildRepeatedWordsRequest trimmed targetLanguageCode only to test
truthiness but forwarded the untrimmed value, so a padded code (" spa ")
would reach the wire verbatim. greek-room keys its legitimate-duplicate
whitelist on the exact code, so a padded value silently disables
suppression. Compute the trimmed code once and forward that; the
<unknown> fallback for missing/empty codes is unchanged. Add a test for
the padded-code case.
---
 .../checks/hooks/useRepeatedWordsCheck.test.ts        | 11 +++++++++++
 src/features/checks/hooks/useRepeatedWordsCheck.ts    |  8 +++++++-
 2 files changed, 18 insertions(+), 1 deletion(-)

diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
index a121f711..a4917b09 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -310,4 +310,15 @@ describe('buildRepeatedWordsRequest — additional edge cases', () => {
     ]);
     expect(request.lang_code).toBe('<unknown>');
   });
+
+  it('sends the TRIMMED code, not the padded value (a padded " spa " becomes "spa")', () => {
+    // greek-room keys its legitimate-duplicate whitelist on the EXACT code, so a
+    // padded code (" spa ") would match nothing and silently disable suppression
+    // (the BUG #2 failure mode). The guard must forward the trimmed value, not
+    // just use the trim to test truthiness. See phase-04 manual smoke (BUG #2).
+    const request = buildRepeatedWordsRequest(makeProjectItem({ targetLanguageCode: ' spa ' }), [
+      { verseNumber: 1, content: 'text' },
+    ]);
+    expect(request.lang_code).toBe('spa');
+  });
 });
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.ts b/src/features/checks/hooks/useRepeatedWordsCheck.ts
index 24c41134..137b2268 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.ts
@@ -64,8 +64,14 @@ export const buildRepeatedWordsRequest = (
   // can still omit it. We keep the type non-optional (so genuinely missing
   // usages elsewhere stay caught) and guard only at this trust boundary, hence
   // the targeted disable of the "unnecessary optional chain" rule.
+  // Trim ONCE and forward the trimmed value: greek-room keys its
+  // legitimate-duplicate whitelist on the exact code, so a padded code
+  // (" spa ") must be normalized to "spa" before it goes on the wire — testing
+  // truthiness on a trimmed copy while forwarding the untrimmed original would
+  // silently disable suppression (the BUG #2 failure mode).
   // eslint-disable-next-line @typescript-eslint/no-unnecessary-condition -- defensive guard against an undefined runtime value the type does not model
-  const langCode = projectItem.targetLanguageCode?.trim() ? projectItem.targetLanguageCode : null;
+  const trimmedCode = projectItem.targetLanguageCode?.trim();
+  const langCode = trimmedCode ? trimmedCode : null;
 
   return {
     lang_code: langCode ?? '<unknown>',

From 78d42f3d9d939fcfe55d6f991555748dc3fed74e Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 17:26:07 -0400
Subject: [PATCH 24/33] fix(checks): restore occurrence rules when global
 ignore PUT fails

writeGlobal optimistically purges the current chapter's occurrence rules
(a persisted write via saveOccurrenceRules) before the /self/settings PUT,
but the failure catch only rolled back the global map. Snapshot the
occurrence map before the purge and restore it in the catch so a failed
global write no longer drops the user's per-occurrence ignores for that pair.
Adds a regression test for the rollback.
---
 .../checks/hooks/useSuppressions.test.ts      | 26 +++++++++++++++++++
 src/features/checks/hooks/useSuppressions.ts  | 12 ++++++++-
 2 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 2a9a8efa..8948201d 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -275,6 +275,32 @@ describe('useSuppressions — global actions', () => {
     await waitFor(() => expect(result.current.globalRules).toEqual({}));
   });
 
+  it('CR-7 regression: a failed global PUT restores the optimistically-purged occurrence rules', async () => {
+    // The optimistic global write purges the current-chapter occurrence rules for
+    // the pair (a *persisted* write through saveOccurrenceRules). If the PUT then
+    // fails, the global flag rolls back — and so must the occurrence map, or the
+    // user's per-occurrence ignores for that pair would be permanently lost.
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    mockPut(500);
+    const save = vi.fn();
+    const original: OccurrenceRules = {
+      'JDG 4:3|the the|0': 'suppress',
+      'JDG 4:3|and and|0': 'suppress',
+    };
+    const { result } = setup(original, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreEverywhere('the the'));
+
+    // First call is the optimistic purge (the "the the" occurrence rule dropped).
+    expect(save).toHaveBeenNthCalledWith(1, { 'JDG 4:3|and and|0': 'suppress' });
+    // After the PUT 500, the catch restores the full pre-purge occurrence map.
+    await waitFor(() => expect(save).toHaveBeenCalledTimes(2));
+    expect(save).toHaveBeenNthCalledWith(2, original);
+    // Global map also rolled back to its prior (empty) state.
+    expect(result.current.globalRules).toEqual({});
+  });
+
   it('stopIgnoringEverywhere deletes the global entry and PUTs the reduced blob', async () => {
     mockGet({
       settings: { checkIgnoredWordPairs: { 'the the': 'suppress', 'and and': 'suppress' } },
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index 08edd0fc..d489d138 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -340,6 +340,10 @@ export const useSuppressions = ({
     (repeatedWord: string, mutate: (rules: GlobalRules) => GlobalRules) => {
       const pair = normalizePair(repeatedWord);
       const previous = globalRules;
+      // Snapshot the occurrence map *before* the optimistic purge so we can
+      // fully roll back if the PUT fails (the purge below is a persisted write
+      // through the single editor-state writer, not just local state).
+      const previousOccurrence = occurrenceRef.current;
       const next = mutate(previous);
       // Optimistic: the finding greys/un-greys immediately.
       setGlobalRules(next);
@@ -354,10 +358,16 @@ export const useSuppressions = ({
           // Rollback: the flag returns to its prior state; re-click allowed.
           // (No queued retry for v1 — the visible flip is the failure notice.)
           setGlobalRules(previous);
+          // Also restore the occurrence rules the optimistic purge removed —
+          // otherwise the user's per-occurrence ignores for this pair would be
+          // permanently lost while the global write rolled back. Re-saving an
+          // identical map (when nothing was purged) is harmless: the editor-state
+          // save is debounced/idempotent.
+          saveOccurrenceRules(previousOccurrence);
         }
       });
     },
-    [globalRules, purgeCurrentChapter]
+    [globalRules, purgeCurrentChapter, saveOccurrenceRules]
   );
 
   const ignoreEverywhere = useCallback(

From 7cb14d897b8aac8fd40e1455e8ecd1b5fac4c61e Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 17:55:38 -0400
Subject: [PATCH 25/33] docs(checks): align reason label to "Ignore Everywhere"
 in review summary

---
 .../repeated-word-check/checks-ui-integration-summary.md        | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-summary.md b/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
index 12af7694..e51a1fe1 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-summary.md
@@ -18,7 +18,7 @@ Give translators the user-facing half of the Repeated Word Check: a **Checks tab
 
 4. **Chapter-wide check on every successful verse auto-save** (W3/W4): a TanStack `useQuery` keyed on `(chapterAssignmentId, saveCounter)` sending all drafted verses; `snt_id` = `"{bookCode} {chapter}:{verse}"` (USFM, the smoke-test convention); no extra debounce beyond the existing 2 s save debounce. Drafting mode only (W10).
 
-5. **Three-layer active/inactive cascade, most-specific-non-silent-wins** (W5/W6): Greek Room's `legitimate` verdict → user-global word-pair rule → occurrence rule `(snt_id, repeated_word, ordinal)`; rules are tri-state maps (`absent/'suppress'/'surface'`), which is what makes per-occurrence **undo** of a global rule or machine verdict possible. The dot counts active findings only; ignored items are hidden by default and revealed dimmed (with "Default Ignore"/"Ignore Here"/"Ignore Always" reason labels and "Undo Ignore") under a **"Show Ignored" toggle** (default off, not persisted), matching the revised #278.
+5. **Three-layer active/inactive cascade, most-specific-non-silent-wins** (W5/W6): Greek Room's `legitimate` verdict → user-global word-pair rule → occurrence rule `(snt_id, repeated_word, ordinal)`; rules are tri-state maps (`absent/'suppress'/'surface'`), which is what makes per-occurrence **undo** of a global rule or machine verdict possible. The dot counts active findings only; ignored items are hidden by default and revealed dimmed (with "Default Ignore"/"Ignore Here"/"Ignore Everywhere" reason labels and "Undo Ignore") under a **"Show Ignored" toggle** (default off, not persisted), matching the revised #278.
 
 6. **Graceful degradation, either-PR-lands-first** (W8/W9): the UI feature-detects `GET /self/settings` (404 ⇒ "Ignore Everywhere" simply not rendered — capability hidden, never a dead control); check failures surface as one inline red line in the panel while the last successful findings stay rendered (TanStack default). No toasts/banners — none exist in this codebase. **"Ignore Everywhere" shows a confirm dialog** (per revised #278) while still being reversible via undo.
 

From 57a041fc679f0a6eb98e3ee936c9aa07338f989b Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Thu, 25 Jun 2026 17:58:11 -0400
Subject: [PATCH 26/33] test(switch): drop redundant misnamed aria-label test

---
 src/components/ui/switch.test.tsx | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/src/components/ui/switch.test.tsx b/src/components/ui/switch.test.tsx
index dbe7f580..a7433124 100644
--- a/src/components/ui/switch.test.tsx
+++ b/src/components/ui/switch.test.tsx
@@ -74,12 +74,6 @@ describe('Switch', () => {
     expect(onCheckedChange).not.toHaveBeenCalled();
   });
 
-  it('forwards a ref or any other native prop (aria-label) to the underlying element', () => {
-    render(<Switch aria-label='Dark mode' />);
-    const switchEl = document.querySelector('[data-slot="switch"]');
-    expect(switchEl).toHaveAttribute('aria-label', 'Dark mode');
-  });
-
   it('can be identified by an id linked from a label', () => {
     render(
       <>

From a54e249ecf8e4cf09b11ce3119b8344a516dae5d Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 28 Jun 2026 11:14:57 -0400
Subject: [PATCH 27/33] fix(checks): a11y label on resource toggle + sort verse
 groups (CR-10/CR-12)

CR-10 (DraftingPage): give the resource-toggle icon Button an accessible
name via aria-label (Show/Hide Resources, mirroring the tooltip), and make
the active-checks notification dot decorative (aria-hidden) so a screen
reader announces the button action rather than the dot's label. The dot
keeps its data-testid.

CR-12 (ChecksPanel): groupByVerse() now sorts emitted groups by parsed verse
number (segment after the last ':'; leading digits of a range like '4-5';
malformed -> POSITIVE_INFINITY so it sorts last and the never-blank-heading
guarantee holds) instead of relying on first-seen input order. Doc comment
updated accordingly.

Behavior-neutral: no test changes required (no test queried the dot by
label; no test asserted first-seen ordering).
---
 .../bible/components/DraftingHeader.tsx       | 11 +++++-
 .../bible/components/DraftingUI.test.tsx      |  1 +
 .../checks/components/ChecksPanel.tsx         | 37 +++++++++++++------
 3 files changed, 36 insertions(+), 13 deletions(-)

diff --git a/src/features/bible/components/DraftingHeader.tsx b/src/features/bible/components/DraftingHeader.tsx
index a452f4c0..15c3d613 100644
--- a/src/features/bible/components/DraftingHeader.tsx
+++ b/src/features/bible/components/DraftingHeader.tsx
@@ -104,6 +104,11 @@ export const DraftingHeader: React.FC<DraftingHeaderProps> = ({
               <Tooltip>
                 <TooltipTrigger asChild>
                   <Button
+                    aria-label={
+                      showResources
+                        ? t('hideResources', 'Hide Resources')
+                        : t('showResources', 'Show Resources')
+                    }
                     aria-pressed={showResources}
                     className='bg-primary relative flex cursor-pointer items-center gap-2'
                     type='button'
@@ -112,10 +117,12 @@ export const DraftingHeader: React.FC<DraftingHeaderProps> = ({
                     <BookText color='#ffffff' />
                     {/* S5: mirror the active-checks notification dot on the
                         panel-toggle button while the panel is closed, so the
-                        translator is still notified (#277). */}
+                        translator is still notified (#277). The dot is
+                        decorative (aria-hidden) so a screen reader announces the
+                        button action, not the dot (CR-10). */}
                     {!showResources && activeFindingsCount > 0 && (
                       <span
-                        aria-label='Active checks present'
+                        aria-hidden='true'
                         className='absolute -top-1 -right-1 inline-block h-2.5 w-2.5 rounded-full bg-blue-600 ring-2 ring-white'
                         data-testid='checks-toggle-notification-dot'
                       />
diff --git a/src/features/bible/components/DraftingUI.test.tsx b/src/features/bible/components/DraftingUI.test.tsx
index 1da97403..624bd1e7 100644
--- a/src/features/bible/components/DraftingUI.test.tsx
+++ b/src/features/bible/components/DraftingUI.test.tsx
@@ -131,6 +131,7 @@ const mockProjectItem: ProjectItem = {
   bibleId: 1,
   bibleName: 'WEB',
   targetLanguage: 'Spanish',
+  targetLanguageCode: 'spa',
   bookId: 1,
   book: 'Genesis',
   chapterStatus: ChapterAssignmentStatus.DRAFT,
diff --git a/src/features/checks/components/ChecksPanel.tsx b/src/features/checks/components/ChecksPanel.tsx
index 73fab32e..66affa35 100644
--- a/src/features/checks/components/ChecksPanel.tsx
+++ b/src/features/checks/components/ChecksPanel.tsx
@@ -49,12 +49,25 @@ interface VerseGroup {
 }
 
 /**
- * Group one bucket of resolved findings by verse (`snt_id`), preserving
- * first-seen verse order. Only verses that actually have rows produce a group
- * (§5.1, §6.4). Active and inactive findings are grouped **separately** now
- * (revised #278 mock): active findings render above the "Show Ignored" toggle,
- * ignored findings in their own dedicated section below it — so each call groups
- * a single, already-filtered bucket rather than merging the two.
+ * Parse the verse number out of a `snt_id` (segment after the last `:`) for
+ * sort ordering. For a verse range (e.g. `"JDG 4:4-5"`) the leading number
+ * before the dash is used. A malformed `snt_id` (no colon / no leading digits)
+ * yields `Number.POSITIVE_INFINITY` so it sorts to the end — preserving the
+ * never-blank-heading guarantee (`verseHeading` still renders the raw id).
+ */
+const verseNum = (sntId: string): number => {
+  const afterColon = sntId.slice(sntId.lastIndexOf(':') + 1).trim();
+  const leadingDigits = /^\d+/.exec(afterColon)?.[0];
+  return leadingDigits ? Number(leadingDigits) : Number.POSITIVE_INFINITY;
+};
+
+/**
+ * Group one bucket of resolved findings by verse (`snt_id`), sorted by verse
+ * number. Only verses that actually have rows produce a group (§5.1, §6.4).
+ * Active and inactive findings are grouped **separately** now (revised #278
+ * mock): active findings render above the "Show Ignored" toggle, ignored
+ * findings in their own dedicated section below it — so each call groups a
+ * single, already-filtered bucket rather than merging the two.
  */
 const groupByVerse = (findings: ResolvedFinding[]): VerseGroup[] => {
   const order: string[] = [];
@@ -71,11 +84,13 @@ const groupByVerse = (findings: ResolvedFinding[]): VerseGroup[] => {
     rows.push(f);
   }
 
-  return order.map(sntId => ({
-    sntId,
-    heading: verseHeading(sntId),
-    rows: byVerse.get(sntId) ?? [],
-  }));
+  return order
+    .map(sntId => ({
+      sntId,
+      heading: verseHeading(sntId),
+      rows: byVerse.get(sntId) ?? [],
+    }))
+    .sort((a, b) => verseNum(a.sntId) - verseNum(b.sntId));
 };
 
 /** Render a list of verse groups with separators between (one fewer than groups). */

From 897d0f8b209a88389ed76757999e68b148637287 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 28 Jun 2026 11:33:14 -0400
Subject: [PATCH 28/33] fix(checks): preserve sibling self/settings + guard
 stale global-write rollback

Two CodeRabbit findings in useSuppressions.ts (PR #320), both in the
global-write path, fixed together:

CR-14 (data-loss forward-risk): PUT /self/settings is a full-replace of the
whole JSONB blob (the API upsert overwrites `settings` wholesale, no
server-side merge), but we were sending only `{ checkIgnoredWordPairs }`, so
any sibling settings key would be dropped on every global-ignore write. Carry
the full GET `settings` snapshot through ProbeResult, hold it in a ref, and
send `{ settings: { ...snapshot, checkIgnoredWordPairs: rules } }` so siblings
ride through untouched. SelfSettings gains an index signature for unknown
sibling keys. Stays on the camelCase /self/settings boundary. Regression test
added: GET returns checkIgnoredWordPairs + a sibling -> PUT body still carries
the sibling unchanged.

CR-15 (concurrency, successor to CR-7): writeGlobal's catch restored call-time
snapshots unconditionally, so a failed PUT could clobber a newer in-flight
edit. Add a writeSeqRef sequence guard: each call captures mySeq, and the catch
only rolls back when mySeq is still the latest issued write.
---
 .../checks/hooks/useSuppressions.test.ts      | 27 +++++++
 src/features/checks/hooks/useSuppressions.ts  | 74 +++++++++++++++----
 2 files changed, 88 insertions(+), 13 deletions(-)

diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 8948201d..e1186ec0 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -301,6 +301,33 @@ describe('useSuppressions — global actions', () => {
     expect(result.current.globalRules).toEqual({});
   });
 
+  it('CR-14 regression: a global write preserves sibling settings keys read on GET (full-replace merge)', async () => {
+    // The PUT /self/settings is a full-replace of the whole JSONB blob (the API
+    // upsert overwrites `settings` wholesale, no server-side merge). So a global
+    // write must echo back every sibling key it read on GET, or those keys would
+    // be silently dropped. Here GET returns checkIgnoredWordPairs + an unrelated
+    // sibling; after ignore-everywhere the PUT body must still carry the sibling
+    // unchanged, with only checkIgnoredWordPairs updated.
+    mockGet({
+      settings: { checkIgnoredWordPairs: {}, someOtherFeatureSetting: { enabled: true } },
+      updatedAt: null,
+    });
+    const put = mockPut();
+    const { result } = setup({}, vi.fn());
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    act(() => result.current.ignoreEverywhere('the the'));
+
+    await waitFor(() =>
+      expect(put.body).toEqual({
+        settings: {
+          someOtherFeatureSetting: { enabled: true },
+          checkIgnoredWordPairs: { 'the the': 'suppress' },
+        },
+      })
+    );
+  });
+
   it('stopIgnoringEverywhere deletes the global entry and PUTs the reduced blob', async () => {
     mockGet({
       settings: { checkIgnoredWordPairs: { 'the the': 'suppress', 'and and': 'suppress' } },
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index d489d138..04996c97 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -43,10 +43,22 @@ import { normalizePair } from './useResolvedFindings';
 // /self/settings wire shape (Fluent-domain → camelCase, conventions §B).
 // ---------------------------------------------------------------------------
 
-/** The user-settings blob behind `/self/settings`. Tolerant of absent keys. */
+/**
+ * The user-settings blob behind `/self/settings`. Tolerant of absent keys.
+ *
+ * The index signature lets us carry any *sibling* settings keys (owned by other
+ * features) through the read-modify-write opaquely: the PUT is a full-replace of
+ * the whole JSONB blob (§8.1, and the API `upsert` replaces `settings` wholesale
+ * with no server-side merge), so we must echo back every key we read on GET or it
+ * would be dropped (CR-14). Spreading siblings does not rename anything — this
+ * stays on the camelCase `/self/settings` boundary, NOT the snake_case greek-room
+ * pass-through (see checks.types.ts:1-19).
+ */
 interface SelfSettings {
   /** Global word-pair rules, keyed by NFC-normalized `repeated_word`. */
   checkIgnoredWordPairs?: GlobalRules;
+  /** Unknown sibling settings owned by other features — carried through untouched. */
+  [key: string]: unknown;
 }
 
 /** GET response: the server returns `{ settings }`, possibly `null` (W8). */
@@ -65,13 +77,19 @@ const SELF_SETTINGS_URL = `${config.api.url}/self/settings`;
 interface ProbeResult {
   available: boolean;
   rules: GlobalRules;
+  /**
+   * The full `settings` blob we read on GET (`{}` when absent/unavailable). Held
+   * so a later PUT can full-replace the blob while preserving sibling keys we
+   * don't model (CR-14).
+   */
+  snapshot: SelfSettings;
 }
 
 /** Pull the global rules out of a settings blob, tolerating any odd shape. */
 const extractGlobalRules = (settings: SelfSettings | null): GlobalRules => {
   const pairs = settings?.checkIgnoredWordPairs;
   if (pairs && typeof pairs === 'object') {
-    return pairs;
+    return pairs as GlobalRules;
   }
   return {};
 };
@@ -89,31 +107,40 @@ const probeSelfSettings = async (): Promise<ProbeResult> => {
       headers: { 'Content-Type': 'application/json' },
     });
     if (res.status === 404) {
-      return { available: false, rules: {} };
+      return { available: false, rules: {}, snapshot: {} };
     }
     if (!res.ok) {
       // Other non-2xx (401/500/…): treat as unavailable for the session.
-      return { available: false, rules: {} };
+      return { available: false, rules: {}, snapshot: {} };
     }
     const body = (await res.json()) as SelfSettingsResponse | null;
-    return { available: true, rules: extractGlobalRules(body?.settings ?? null) };
+    const settings = body?.settings ?? null;
+    return {
+      available: true,
+      rules: extractGlobalRules(settings),
+      // Keep the whole blob so a later PUT can preserve sibling keys (CR-14).
+      snapshot: settings ?? {},
+    };
   } catch (error) {
     Logger.logException(error, { context: 'Repeated-words self/settings probe failed' });
-    return { available: false, rules: {} };
+    return { available: false, rules: {}, snapshot: {} };
   }
 };
 
 /**
  * PUT the full settings blob (full-replace, §8.1). The caller supplies the
- * complete `checkIgnoredWordPairs` map it wants persisted; this hook never
- * sends a partial patch.
+ * complete `checkIgnoredWordPairs` map it wants persisted plus the full
+ * `snapshot` of the blob we last read on GET; we spread the snapshot first so
+ * any sibling keys owned by other features ride through untouched and are not
+ * dropped by the server-side full-replace (CR-14). This hook never sends a
+ * partial patch.
  */
-const putSelfSettings = async (rules: GlobalRules): Promise<void> => {
+const putSelfSettings = async (rules: GlobalRules, snapshot: SelfSettings): Promise<void> => {
   const res = await fetch(SELF_SETTINGS_URL, {
     method: 'PUT',
     credentials: 'include',
     headers: { 'Content-Type': 'application/json' },
-    body: JSON.stringify({ settings: { checkIgnoredWordPairs: rules } }),
+    body: JSON.stringify({ settings: { ...snapshot, checkIgnoredWordPairs: rules } }),
   });
   if (!res.ok) {
     throw new Error('Failed to save global ignore rules');
@@ -270,17 +297,24 @@ export const useSuppressions = ({
     };
   }, []);
 
+  // Full `settings` blob from the GET probe, held so a later PUT can full-replace
+  // the blob while preserving sibling keys other features own (CR-14). Defaults
+  // to `{}` until the probe resolves (and stays `{}` when unavailable).
+  const snapshotRef = useRef<SelfSettings>({});
+
   useEffect(() => {
     if (probeStarted.current) {
       return;
     }
     probeStarted.current = true;
-    void probeSelfSettings().then(({ available, rules }) => {
+    void probeSelfSettings().then(({ available, rules, snapshot }) => {
       if (!mounted.current) {
         return;
       }
       setGlobalIgnoresAvailable(available);
       setGlobalRules(available ? rules : {});
+      // Carry the full GET blob through to the next PUT (sibling-preserving).
+      snapshotRef.current = available ? snapshot : {};
       setSettingsProbeResolved(true);
     });
   }, []);
@@ -290,6 +324,12 @@ export const useSuppressions = ({
   const occurrenceRef = useRef(occurrenceRules);
   occurrenceRef.current = occurrenceRules;
 
+  // Monotonic counter identifying the most-recently-issued global write. Each
+  // `writeGlobal` captures its own `mySeq`; a failed PUT only rolls back when its
+  // `mySeq` is still the latest issued write, so an old failing PUT can't clobber
+  // a newer in-flight edit's optimistic state (CR-15; successor to CR-7).
+  const writeSeqRef = useRef(0);
+
   // --- Occurrence actions (write through the single editor-state writer) ---
 
   const ignoreHere = useCallback(
@@ -344,17 +384,25 @@ export const useSuppressions = ({
       // fully roll back if the PUT fails (the purge below is a persisted write
       // through the single editor-state writer, not just local state).
       const previousOccurrence = occurrenceRef.current;
+      // Mark this as the latest issued write. On failure we only roll back if no
+      // newer write was issued in the meantime (CR-15 sequence guard).
+      const mySeq = ++writeSeqRef.current;
       const next = mutate(previous);
       // Optimistic: the finding greys/un-greys immediately.
       setGlobalRules(next);
       // Purge current-chapter occurrence rules for the pair so the just-clicked
       // panel doesn't appear to ignore the action (occurrence beats global).
       purgeCurrentChapter(pair);
-      void putSelfSettings(next).catch(error => {
+      // Send the full blob, preserving any sibling settings keys (CR-14).
+      void putSelfSettings(next, snapshotRef.current).catch(error => {
         Logger.logException(error, {
           context: 'Repeated-words global ignore write failed; rolling back',
         });
-        if (mounted.current) {
+        // Only the latest issued write may revert its own optimistic changes. If
+        // a newer edit was issued while this PUT was in flight, the call-time
+        // snapshots below are stale and would clobber that newer edit (CR-15), so
+        // we skip the rollback and let the newer write own the state.
+        if (mounted.current && mySeq === writeSeqRef.current) {
           // Rollback: the flag returns to its prior state; re-click allowed.
           // (No queued retry for v1 — the visible flip is the failure notice.)
           setGlobalRules(previous);

From fac40322ff5d3c55807615d58ea2e97a6a58cb95 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 28 Jun 2026 12:56:06 -0400
Subject: [PATCH 29/33] refactor(checks): send only checkIgnoredWordPairs on
 PUT; drop client snapshot cache

---
 .../checks/hooks/useSuppressions.test.ts      | 19 +++---
 src/features/checks/hooks/useSuppressions.ts  | 62 +++++++------------
 2 files changed, 32 insertions(+), 49 deletions(-)

diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index e1186ec0..12085333 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -301,13 +301,13 @@ describe('useSuppressions — global actions', () => {
     expect(result.current.globalRules).toEqual({});
   });
 
-  it('CR-14 regression: a global write preserves sibling settings keys read on GET (full-replace merge)', async () => {
-    // The PUT /self/settings is a full-replace of the whole JSONB blob (the API
-    // upsert overwrites `settings` wholesale, no server-side merge). So a global
-    // write must echo back every sibling key it read on GET, or those keys would
-    // be silently dropped. Here GET returns checkIgnoredWordPairs + an unrelated
-    // sibling; after ignore-everywhere the PUT body must still carry the sibling
-    // unchanged, with only checkIgnoredWordPairs updated.
+  it('single-key full-replace: the PUT body carries only checkIgnoredWordPairs (no client-side sibling echo)', async () => {
+    // The blob is single-key today; the client sends only the key it owns and the
+    // server full-replaces (the API write schema strips unknown keys, so a second
+    // setting cannot exist without a server-schema change that must add server-side
+    // merge — see the API implementation note). Even if GET happens to return an
+    // unknown sibling, the client must NOT echo it back: it sends only
+    // checkIgnoredWordPairs.
     mockGet({
       settings: { checkIgnoredWordPairs: {}, someOtherFeatureSetting: { enabled: true } },
       updatedAt: null,
@@ -320,10 +320,7 @@ describe('useSuppressions — global actions', () => {
 
     await waitFor(() =>
       expect(put.body).toEqual({
-        settings: {
-          someOtherFeatureSetting: { enabled: true },
-          checkIgnoredWordPairs: { 'the the': 'suppress' },
-        },
+        settings: { checkIgnoredWordPairs: { 'the the': 'suppress' } },
       })
     );
   });
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index 04996c97..456734a8 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -46,19 +46,20 @@ import { normalizePair } from './useResolvedFindings';
 /**
  * The user-settings blob behind `/self/settings`. Tolerant of absent keys.
  *
- * The index signature lets us carry any *sibling* settings keys (owned by other
- * features) through the read-modify-write opaquely: the PUT is a full-replace of
- * the whole JSONB blob (§8.1, and the API `upsert` replaces `settings` wholesale
- * with no server-side merge), so we must echo back every key we read on GET or it
- * would be dropped (CR-14). Spreading siblings does not rename anything — this
- * stays on the camelCase `/self/settings` boundary, NOT the snake_case greek-room
- * pass-through (see checks.types.ts:1-19).
+ * Single-key today: `checkIgnoredWordPairs` is the only setting in the blob, so
+ * the PUT is a full-replace of the whole JSONB blob (§8.1; the API `upsert`
+ * replaces `settings` wholesale with no server-side merge) and the client simply
+ * sends the one key it owns. The API write schema strips unknown keys, so a
+ * second setting cannot be introduced without editing the server schema — at
+ * which point the server is responsible for read-merging keys (the API carries an
+ * implementation note gating that). We therefore deliberately do NOT cache/echo a
+ * client-side snapshot of sibling keys here. This stays on the camelCase
+ * `/self/settings` boundary, NOT the snake_case greek-room pass-through (see
+ * checks.types.ts:1-19).
  */
 interface SelfSettings {
   /** Global word-pair rules, keyed by NFC-normalized `repeated_word`. */
   checkIgnoredWordPairs?: GlobalRules;
-  /** Unknown sibling settings owned by other features — carried through untouched. */
-  [key: string]: unknown;
 }
 
 /** GET response: the server returns `{ settings }`, possibly `null` (W8). */
@@ -77,12 +78,6 @@ const SELF_SETTINGS_URL = `${config.api.url}/self/settings`;
 interface ProbeResult {
   available: boolean;
   rules: GlobalRules;
-  /**
-   * The full `settings` blob we read on GET (`{}` when absent/unavailable). Held
-   * so a later PUT can full-replace the blob while preserving sibling keys we
-   * don't model (CR-14).
-   */
-  snapshot: SelfSettings;
 }
 
 /** Pull the global rules out of a settings blob, tolerating any odd shape. */
@@ -107,40 +102,38 @@ const probeSelfSettings = async (): Promise<ProbeResult> => {
       headers: { 'Content-Type': 'application/json' },
     });
     if (res.status === 404) {
-      return { available: false, rules: {}, snapshot: {} };
+      return { available: false, rules: {} };
     }
     if (!res.ok) {
       // Other non-2xx (401/500/…): treat as unavailable for the session.
-      return { available: false, rules: {}, snapshot: {} };
+      return { available: false, rules: {} };
     }
     const body = (await res.json()) as SelfSettingsResponse | null;
     const settings = body?.settings ?? null;
     return {
       available: true,
       rules: extractGlobalRules(settings),
-      // Keep the whole blob so a later PUT can preserve sibling keys (CR-14).
-      snapshot: settings ?? {},
     };
   } catch (error) {
     Logger.logException(error, { context: 'Repeated-words self/settings probe failed' });
-    return { available: false, rules: {}, snapshot: {} };
+    return { available: false, rules: {} };
   }
 };
 
 /**
- * PUT the full settings blob (full-replace, §8.1). The caller supplies the
- * complete `checkIgnoredWordPairs` map it wants persisted plus the full
- * `snapshot` of the blob we last read on GET; we spread the snapshot first so
- * any sibling keys owned by other features ride through untouched and are not
- * dropped by the server-side full-replace (CR-14). This hook never sends a
- * partial patch.
+ * PUT the settings blob (full-replace, §8.1). The blob is single-key today
+ * (`checkIgnoredWordPairs`), so the caller supplies the complete map it wants
+ * persisted and we send the whole blob. A second setting cannot exist without a
+ * server-schema change, at which point the server merges keys (it carries an
+ * implementation note gating that) — so we deliberately send only the key we own
+ * and do NOT echo a client-cached snapshot.
  */
-const putSelfSettings = async (rules: GlobalRules, snapshot: SelfSettings): Promise<void> => {
+const putSelfSettings = async (rules: GlobalRules): Promise<void> => {
   const res = await fetch(SELF_SETTINGS_URL, {
     method: 'PUT',
     credentials: 'include',
     headers: { 'Content-Type': 'application/json' },
-    body: JSON.stringify({ settings: { ...snapshot, checkIgnoredWordPairs: rules } }),
+    body: JSON.stringify({ settings: { checkIgnoredWordPairs: rules } }),
   });
   if (!res.ok) {
     throw new Error('Failed to save global ignore rules');
@@ -297,24 +290,17 @@ export const useSuppressions = ({
     };
   }, []);
 
-  // Full `settings` blob from the GET probe, held so a later PUT can full-replace
-  // the blob while preserving sibling keys other features own (CR-14). Defaults
-  // to `{}` until the probe resolves (and stays `{}` when unavailable).
-  const snapshotRef = useRef<SelfSettings>({});
-
   useEffect(() => {
     if (probeStarted.current) {
       return;
     }
     probeStarted.current = true;
-    void probeSelfSettings().then(({ available, rules, snapshot }) => {
+    void probeSelfSettings().then(({ available, rules }) => {
       if (!mounted.current) {
         return;
       }
       setGlobalIgnoresAvailable(available);
       setGlobalRules(available ? rules : {});
-      // Carry the full GET blob through to the next PUT (sibling-preserving).
-      snapshotRef.current = available ? snapshot : {};
       setSettingsProbeResolved(true);
     });
   }, []);
@@ -393,8 +379,8 @@ export const useSuppressions = ({
       // Purge current-chapter occurrence rules for the pair so the just-clicked
       // panel doesn't appear to ignore the action (occurrence beats global).
       purgeCurrentChapter(pair);
-      // Send the full blob, preserving any sibling settings keys (CR-14).
-      void putSelfSettings(next, snapshotRef.current).catch(error => {
+      // Send the full blob (single-key full-replace, §8.1).
+      void putSelfSettings(next).catch(error => {
         Logger.logException(error, {
           context: 'Repeated-words global ignore write failed; rolling back',
         });

From b0cd51f31687fa8e3b2e4d6dc9a891a030c617a6 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 28 Jun 2026 16:40:12 -0400
Subject: [PATCH 30/33] fix(checks): serialize global settings writes so PUTs
 commit in UI-action order

The sequence guard added earlier only protects the client rollback path (a
stale failed write can't clobber a newer in-flight edit's optimistic state). It
does not prevent two overlapping *successful* full-replace PUTs to
/self/settings from committing out of order on the server (last-writer-wins, no
server merge), which could persist an older blob even though the UI shows the
newer edit and have it read back on the next session.

Chain each global write onto the previous one via a putChainRef promise so a
PUT does not start until the prior one settles. A leading no-op .catch keeps a
failed write from breaking the chain for later writes; each write still reports
its own failure in its trailing .catch (the rollback guard there is unchanged).
Net ~6 lines, no API change.

Adds a regression test: with the first PUT held open, a second global toggle
does not issue its PUT until the first completes, and the two bodies arrive in
UI-action order.
---
 .../checks/hooks/useSuppressions.test.ts      | 55 ++++++++++++++++++
 src/features/checks/hooks/useSuppressions.ts  | 57 ++++++++++++-------
 2 files changed, 92 insertions(+), 20 deletions(-)

diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 12085333..164a4493 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -417,6 +417,61 @@ describe('useSuppressions — global actions', () => {
     expect(save).toHaveBeenCalledWith({ 'JDG 4:3|and and|0': 'suppress' });
   });
 
+  it('CR-15 follow-up: serializes overlapping global writes so PUTs reach the server in UI-action order', async () => {
+    // Two quick global toggles must not race on the server. Each full-replace PUT
+    // is chained after the previous one settles, so the second PUT does not even
+    // start until the first has finished — guaranteeing the persisted blob ends in
+    // the same order as the user actions (an older PUT can't land after a newer
+    // one). We prove this by making the FIRST PUT hang until we release it: while
+    // it is in flight the second PUT must NOT have started; after release, both
+    // arrive in order.
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+
+    const order: Array<Record<string, unknown>> = [];
+    let releaseFirst!: () => void;
+    const firstReleased = new Promise<void>(resolve => {
+      releaseFirst = resolve;
+    });
+    let putCount = 0;
+    server.use(
+      http.put(SETTINGS_URL, async ({ request }) => {
+        const body = (await request.json()) as Record<string, unknown>;
+        putCount += 1;
+        if (putCount === 1) {
+          // Hold the first write open until the test releases it.
+          await firstReleased;
+        }
+        order.push(body);
+        return HttpResponse.json({ settings: {}, updatedAt: null }, { status: 200 });
+      })
+    );
+
+    const { result } = setup({}, vi.fn());
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    // First global toggle — its PUT starts and hangs (held open above).
+    act(() => result.current.ignoreEverywhere('aaa'));
+    await waitFor(() => expect(putCount).toBe(1));
+
+    // Second global toggle while the first PUT is still in flight. Its PUT is
+    // chained behind the first and must NOT have started yet (serialized — not
+    // fired concurrently). The optimistic state already reflects both.
+    act(() => result.current.ignoreEverywhere('bbb'));
+    expect(result.current.globalRules).toEqual({ aaa: 'suppress', bbb: 'suppress' });
+    // Still only one PUT in flight; the second is queued behind it.
+    expect(putCount).toBe(1);
+    expect(order).toHaveLength(0); // first hasn't completed, second hasn't started
+
+    // Release the first; now the second proceeds. Both complete in order A→B.
+    releaseFirst();
+    await waitFor(() => expect(order).toHaveLength(2));
+    expect(order[0]).toEqual({ settings: { checkIgnoredWordPairs: { aaa: 'suppress' } } });
+    // Second carries BOTH toggles (state accumulated), proving B persisted last.
+    expect(order[1]).toEqual({
+      settings: { checkIgnoredWordPairs: { aaa: 'suppress', bbb: 'suppress' } },
+    });
+  });
+
   it('purge-local runs for NFC-composed vs decomposed pair in occurrence key', async () => {
     // The occurrence key stores the NFC-normalized pair; global write purges by
     // normalizing the repeated_word. Composed "caf\u00e9" must purge a key that
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index 456734a8..76dccd5d 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -316,6 +316,16 @@ export const useSuppressions = ({
   // a newer in-flight edit's optimistic state (CR-15; successor to CR-7).
   const writeSeqRef = useRef(0);
 
+  // Tail of the global-write chain. Each PUT is appended after the previous one
+  // settles, so the full-replace PUTs reach the server strictly in UI-action
+  // order — an older successful write can no longer commit after a newer one.
+  // (The `/self/settings` blob is last-writer-wins with no server merge, §8.1, so
+  // out-of-order delivery of two overlapping PUTs would otherwise persist the
+  // stale blob even though the UI shows the newer edit.) The seq-guard above
+  // protects the client rollback; this chain protects the server write order.
+  // CR-15 follow-up.
+  const putChainRef = useRef<Promise<unknown>>(Promise.resolve());
+
   // --- Occurrence actions (write through the single editor-state writer) ---
 
   const ignoreHere = useCallback(
@@ -379,27 +389,34 @@ export const useSuppressions = ({
       // Purge current-chapter occurrence rules for the pair so the just-clicked
       // panel doesn't appear to ignore the action (occurrence beats global).
       purgeCurrentChapter(pair);
-      // Send the full blob (single-key full-replace, §8.1).
-      void putSelfSettings(next).catch(error => {
-        Logger.logException(error, {
-          context: 'Repeated-words global ignore write failed; rolling back',
+      // Send the full blob (single-key full-replace, §8.1), chained after the
+      // previous global write so the PUTs commit on the server in UI-action
+      // order. The leading `.catch(() => {})` swallows the *previous* link's
+      // rejection so one failed write doesn't break the chain for later ones;
+      // each write still reports its own failure in the trailing `.catch`.
+      putChainRef.current = putChainRef.current
+        .catch(() => {})
+        .then(() => putSelfSettings(next))
+        .catch(error => {
+          Logger.logException(error, {
+            context: 'Repeated-words global ignore write failed; rolling back',
+          });
+          // Only the latest issued write may revert its own optimistic changes.
+          // If a newer edit was issued while this PUT was in flight, the
+          // call-time snapshots below are stale and would clobber that newer edit
+          // (CR-15), so we skip the rollback and let the newer write own the state.
+          if (mounted.current && mySeq === writeSeqRef.current) {
+            // Rollback: the flag returns to its prior state; re-click allowed.
+            // (No queued retry for v1 — the visible flip is the failure notice.)
+            setGlobalRules(previous);
+            // Also restore the occurrence rules the optimistic purge removed —
+            // otherwise the user's per-occurrence ignores for this pair would be
+            // permanently lost while the global write rolled back. Re-saving an
+            // identical map (when nothing was purged) is harmless: the editor-state
+            // save is debounced/idempotent.
+            saveOccurrenceRules(previousOccurrence);
+          }
         });
-        // Only the latest issued write may revert its own optimistic changes. If
-        // a newer edit was issued while this PUT was in flight, the call-time
-        // snapshots below are stale and would clobber that newer edit (CR-15), so
-        // we skip the rollback and let the newer write own the state.
-        if (mounted.current && mySeq === writeSeqRef.current) {
-          // Rollback: the flag returns to its prior state; re-click allowed.
-          // (No queued retry for v1 — the visible flip is the failure notice.)
-          setGlobalRules(previous);
-          // Also restore the occurrence rules the optimistic purge removed —
-          // otherwise the user's per-occurrence ignores for this pair would be
-          // permanently lost while the global write rolled back. Re-saving an
-          // identical map (when nothing was purged) is harmless: the editor-state
-          // save is debounced/idempotent.
-          saveOccurrenceRules(previousOccurrence);
-        }
-      });
     },
     [globalRules, purgeCurrentChapter, saveOccurrenceRules]
   );

From e786c83e046f68f646eddeada8cf8d6d6d29c315 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 28 Jun 2026 17:22:58 -0400
Subject: [PATCH 31/33] fix(checks): roll back global write to last committed
 baseline; merge-restore only purged occurrence keys
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

CodeRabbit thread 3488525698 (NEW-C, follow-up to the CR-15 serialization).

With global PUTs now serialized, a queued write B can capture a call-time
snapshot that still holds an earlier write A's optimistic-but-unconfirmed
state. On a double-failure (A fails, B fails) the seq-guard lets B roll back,
which previously restored that phantom state. B now rolls back to a
committedGlobalRef updated only on PUT success, so an unconfirmed change can't
survive a rollback.

Also restore occurrence rules surgically: capture exactly the keys the
optimistic purge removed and merge only those back into the live occurrence
map, instead of overwriting the whole map with a stale call-time snapshot — so
an unrelated occurrence edit made during the in-flight window is preserved.

committedGlobalRef is written only in the success path and read only in the
catch, keeping the happy path unchanged. +2 regression tests (186 total).
---
 .../checks/hooks/useSuppressions.test.ts      | 80 +++++++++++++++++++
 src/features/checks/hooks/useSuppressions.ts  | 79 ++++++++++++------
 2 files changed, 135 insertions(+), 24 deletions(-)

diff --git a/src/features/checks/hooks/useSuppressions.test.ts b/src/features/checks/hooks/useSuppressions.test.ts
index 164a4493..d15bf546 100644
--- a/src/features/checks/hooks/useSuppressions.test.ts
+++ b/src/features/checks/hooks/useSuppressions.test.ts
@@ -301,6 +301,40 @@ describe('useSuppressions — global actions', () => {
     expect(result.current.globalRules).toEqual({});
   });
 
+  it('NEW-C: rollback merges back only the purged keys, preserving a later unrelated occurrence edit', async () => {
+    // Regression for CR thread 3488525698. The old rollback restored the WHOLE
+    // call-time occurrence snapshot, which would wipe any occurrence edit the user
+    // made while the global PUT was in flight. We now merge back only the keys the
+    // optimistic purge removed, into the *current* map — so an unrelated edit that
+    // landed in the meantime survives the rollback.
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+    mockPut(500);
+    const save = vi.fn();
+    const original: OccurrenceRules = { 'JDG 4:3|the the|0': 'suppress' };
+    const { result, rerender } = setup(original, save);
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    // Global "Ignore Everywhere" for "the the": optimistically purges that key.
+    act(() => result.current.ignoreEverywhere('the the'));
+    expect(save).toHaveBeenNthCalledWith(1, {});
+
+    // While the PUT is in flight, an unrelated occurrence edit lands (e.g. the
+    // user clicks "Ignore Here" on a different finding). Simulate the new map
+    // flowing back in through the prop (the single editor-state writer).
+    const afterEdit: OccurrenceRules = { 'JDG 4:5|and and|0': 'suppress' };
+    rerender({ occurrenceRules: afterEdit, saveOccurrenceRules: save });
+
+    // PUT 500 → rollback merges ONLY the purged "the the" key back into the
+    // current map, keeping the unrelated "and and" edit intact (not overwritten
+    // by the stale pre-purge snapshot).
+    await waitFor(() => expect(save).toHaveBeenCalledTimes(2));
+    expect(save).toHaveBeenNthCalledWith(2, {
+      'JDG 4:5|and and|0': 'suppress',
+      'JDG 4:3|the the|0': 'suppress',
+    });
+    expect(result.current.globalRules).toEqual({});
+  });
+
   it('single-key full-replace: the PUT body carries only checkIgnoredWordPairs (no client-side sibling echo)', async () => {
     // The blob is single-key today; the client sends only the key it owns and the
     // server full-replaces (the API write schema strips unknown keys, so a second
@@ -472,6 +506,52 @@ describe('useSuppressions — global actions', () => {
     });
   });
 
+  it('NEW-C: a queued write rolls back to the last COMMITTED baseline, not a call-time snapshot holding an earlier unconfirmed write', async () => {
+    // Regression for CR thread 3488525698 (the half our serialization surfaced).
+    // Writes are serialized: A then B. B is issued while A is still optimistic but
+    // unconfirmed, so B's call-time `previous` already contains A's change. If A
+    // FAILS and B also FAILS, the seq-guard lets B (the latest) roll back. Rolling
+    // back to B's call-time snapshot would resurrect A's never-persisted state;
+    // we instead roll back to the last *server-confirmed* baseline (here: empty).
+    mockGet({ settings: { checkIgnoredWordPairs: {} }, updatedAt: null });
+
+    let releaseFirst!: () => void;
+    const firstReleased = new Promise<void>(resolve => {
+      releaseFirst = resolve;
+    });
+    let putCount = 0;
+    server.use(
+      http.put(SETTINGS_URL, async () => {
+        putCount += 1;
+        if (putCount === 1) {
+          await firstReleased;
+        }
+        // Both writes fail — neither ever commits, so the confirmed baseline stays empty.
+        return HttpResponse.json({ message: 'boom' }, { status: 500 });
+      })
+    );
+
+    const { result } = setup({}, vi.fn());
+    await waitFor(() => expect(result.current.settingsProbeResolved).toBe(true));
+
+    // Write A: optimistic {aaa}; its PUT starts and hangs (held open).
+    act(() => result.current.ignoreEverywhere('aaa'));
+    await waitFor(() => expect(putCount).toBe(1));
+
+    // Write B while A is unconfirmed: B's call-time `previous` is {aaa} (A's
+    // optimistic state). Optimistic now shows both.
+    act(() => result.current.ignoreEverywhere('bbb'));
+    expect(result.current.globalRules).toEqual({ aaa: 'suppress', bbb: 'suppress' });
+
+    // Release A (fails), then B runs (also fails). Only B rolls back (latest seq).
+    releaseFirst();
+    await waitFor(() => expect(putCount).toBe(2));
+
+    // B rolls back to the last COMMITTED baseline (empty), NOT to {aaa} — A's
+    // change was never persisted, so it must not survive as a phantom.
+    await waitFor(() => expect(result.current.globalRules).toEqual({}));
+  });
+
   it('purge-local runs for NFC-composed vs decomposed pair in occurrence key', async () => {
     // The occurrence key stores the NFC-normalized pair; global write purges by
     // normalizing the repeated_word. Composed "caf\u00e9" must purge a key that
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index 76dccd5d..1c9ca9f7 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -290,6 +290,15 @@ export const useSuppressions = ({
     };
   }, []);
 
+  // Last global-rules map the server has *confirmed* (probe result, then each
+  // successful PUT). Rollback restores from THIS, never from a call-time
+  // snapshot: with writes serialized (see `putChainRef`), a later write B can be
+  // issued while an earlier write A is still optimistic-but-unconfirmed, so B's
+  // call-time `previous` may contain A's never-persisted change. Reverting to the
+  // last *committed* baseline can't resurrect that phantom state (CR `3488525698`).
+  // Only ever written in the success path; only ever read in the catch.
+  const committedGlobalRef = useRef<GlobalRules>({});
+
   useEffect(() => {
     if (probeStarted.current) {
       return;
@@ -301,6 +310,7 @@ export const useSuppressions = ({
       }
       setGlobalIgnoresAvailable(available);
       setGlobalRules(available ? rules : {});
+      committedGlobalRef.current = available ? rules : {};
       setSettingsProbeResolved(true);
     });
   }, []);
@@ -361,13 +371,27 @@ export const useSuppressions = ({
 
   // --- Global actions (optimistic + rollback; purge current-chapter local) ---
 
-  /** Apply a current-chapter purge for `pair` through the single writer. */
+  /**
+   * Apply a current-chapter purge for `pair` through the single writer. Returns
+   * the occurrence entries it removed (key → verdict) so a failed global write
+   * can restore *just those* keys — see the rollback in `writeGlobal`. Empty
+   * when nothing matched (no write issued).
+   */
   const purgeCurrentChapter = useCallback(
-    (repeatedWord: string) => {
-      const { rules, changed } = purgeLocalForPair(occurrenceRef.current, repeatedWord);
-      if (changed) {
-        saveOccurrenceRules(rules);
+    (repeatedWord: string): OccurrenceRules => {
+      const before = occurrenceRef.current;
+      const { rules, changed } = purgeLocalForPair(before, repeatedWord);
+      if (!changed) {
+        return {};
+      }
+      const purged: OccurrenceRules = {};
+      for (const [key, verdict] of Object.entries(before)) {
+        if (!(key in rules)) {
+          purged[key] = verdict;
+        }
       }
+      saveOccurrenceRules(rules);
+      return purged;
     },
     [saveOccurrenceRules]
   );
@@ -376,10 +400,6 @@ export const useSuppressions = ({
     (repeatedWord: string, mutate: (rules: GlobalRules) => GlobalRules) => {
       const pair = normalizePair(repeatedWord);
       const previous = globalRules;
-      // Snapshot the occurrence map *before* the optimistic purge so we can
-      // fully roll back if the PUT fails (the purge below is a persisted write
-      // through the single editor-state writer, not just local state).
-      const previousOccurrence = occurrenceRef.current;
       // Mark this as the latest issued write. On failure we only roll back if no
       // newer write was issued in the meantime (CR-15 sequence guard).
       const mySeq = ++writeSeqRef.current;
@@ -387,8 +407,9 @@ export const useSuppressions = ({
       // Optimistic: the finding greys/un-greys immediately.
       setGlobalRules(next);
       // Purge current-chapter occurrence rules for the pair so the just-clicked
-      // panel doesn't appear to ignore the action (occurrence beats global).
-      purgeCurrentChapter(pair);
+      // panel doesn't appear to ignore the action (occurrence beats global). Keep
+      // exactly which entries were removed so a failed PUT restores only those.
+      const purgedOccurrence = purgeCurrentChapter(pair);
       // Send the full blob (single-key full-replace, §8.1), chained after the
       // previous global write so the PUTs commit on the server in UI-action
       // order. The leading `.catch(() => {})` swallows the *previous* link's
@@ -396,25 +417,35 @@ export const useSuppressions = ({
       // each write still reports its own failure in the trailing `.catch`.
       putChainRef.current = putChainRef.current
         .catch(() => {})
-        .then(() => putSelfSettings(next))
+        .then(async () => {
+          await putSelfSettings(next);
+          // Persisted: this map is now the confirmed baseline future rollbacks
+          // restore from. (Only the success path writes this ref.)
+          committedGlobalRef.current = next;
+        })
         .catch(error => {
           Logger.logException(error, {
             context: 'Repeated-words global ignore write failed; rolling back',
           });
           // Only the latest issued write may revert its own optimistic changes.
-          // If a newer edit was issued while this PUT was in flight, the
-          // call-time snapshots below are stale and would clobber that newer edit
-          // (CR-15), so we skip the rollback and let the newer write own the state.
+          // If a newer edit was issued while this PUT was in flight, rolling back
+          // would clobber that newer edit (CR-15), so we skip and let the newer
+          // write own the state.
           if (mounted.current && mySeq === writeSeqRef.current) {
-            // Rollback: the flag returns to its prior state; re-click allowed.
-            // (No queued retry for v1 — the visible flip is the failure notice.)
-            setGlobalRules(previous);
-            // Also restore the occurrence rules the optimistic purge removed —
-            // otherwise the user's per-occurrence ignores for this pair would be
-            // permanently lost while the global write rolled back. Re-saving an
-            // identical map (when nothing was purged) is harmless: the editor-state
-            // save is debounced/idempotent.
-            saveOccurrenceRules(previousOccurrence);
+            // Roll the global flag back to the last *server-confirmed* state —
+            // NOT the call-time `previous`, which (with serialized writes) could
+            // still hold an earlier write's never-persisted optimistic change
+            // (CR `3488525698`). Re-click allowed; the visible flip is the only
+            // failure notice (no queued retry for v1).
+            setGlobalRules(committedGlobalRef.current);
+            // Restore only the occurrence entries this write's purge removed,
+            // merging them back into the *current* map so unrelated occurrence
+            // edits made meanwhile survive (CR `3488525698`). Skip the write
+            // entirely when nothing was purged.
+            const purgedKeys = Object.keys(purgedOccurrence);
+            if (purgedKeys.length > 0) {
+              saveOccurrenceRules({ ...occurrenceRef.current, ...purgedOccurrence });
+            }
           }
         });
     },

From 594efbd96b4f7ddf27e8022b8047a2084f822e38 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Fri, 3 Jul 2026 16:33:38 -0400
Subject: [PATCH 32/33] feat(flags): consume feature flags to hide the Repeated
 Word Check when AI is unwired

Add a fail-closed feature-flag consumer (src/features/flags/) over the
API's GET /config/features endpoint, and use it to hide the Repeated
Word Check end-to-end when its flag is off:
feat(flags): consume feature flags to hide the Repeated Word Check when AI is unwired

Add a fail-closed feature-flag consumer (src/features/flags/) over the
API's GET /config/features endpoint, and use it to hide the Repeated
Word Check end-to-end when its flag is off:

- useFeatureFlags / useFeatureFlag / FeatureGate + types, fail-closed
  (every flag reads false while loading or on error), retry disabled.
- DraftingUI consumes the flag once: gates the useRepeatedWordsCheck
  query's `enabled` (no known-failing POST when off), forces the
  active-findings count to 0, and forces the left panel to Resources.
- DraftingResourceSidebar / LeftPanel gain showChecksTab (default true);
  LeftPanel hides the Checks tab + notification dot when off.
- Add an unlinked, login-gated, read-only feature-flags diagnostics page
  at /_authenticated/debug (proposal D8).

All gates green (lint/format/typecheck/test/build).

- useFeatureFlags / useFeatureFlag / FeatureGate + types, fail-closed
  (every flag reads false while loading or on error), retry disabled.
- DraftingUI consumes the flag once: gates the useRepeatedWordsCheck
  query's  (no known-failing POST when off), forces the
  active-findings count to 0, and forces the left panel to Resources.
- DraftingResourceSidebar / LeftPanel gain showChecksTab (default true);
  LeftPanel hides the Checks tab + notification dot when off.
- Add an unlinked, login-gated, read-only feature-flags diagnostics page
  at /_authenticated/debug (proposal D8).

All gates green (lint/format/typecheck/test/build).
---
 .../components/DraftingResourceSidebar.tsx    |   6 +
 .../bible/components/DraftingUI.test.tsx      |  75 ++++++++++
 src/features/bible/components/DraftingUI.tsx  |  32 +++-
 .../checks/components/LeftPanel.test.tsx      |  32 ++++
 src/features/checks/components/LeftPanel.tsx  |  40 +++--
 .../flags/FeatureFlagsDiagnostics.test.tsx    |  91 ++++++++++++
 .../flags/FeatureFlagsDiagnostics.tsx         | 108 ++++++++++++++
 src/features/flags/FeatureGate.tsx            |  34 +++++
 src/features/flags/flags.types.ts             |  40 +++++
 src/features/flags/index.ts                   |  26 ++++
 src/features/flags/useFeatureFlags.test.tsx   | 139 ++++++++++++++++++
 src/features/flags/useFeatureFlags.ts         | 115 +++++++++++++++
 src/routeTree.gen.ts                          |  21 +++
 src/routes/_authenticated/debug.tsx           |  15 ++
 14 files changed, 759 insertions(+), 15 deletions(-)
 create mode 100644 src/features/flags/FeatureFlagsDiagnostics.test.tsx
 create mode 100644 src/features/flags/FeatureFlagsDiagnostics.tsx
 create mode 100644 src/features/flags/FeatureGate.tsx
 create mode 100644 src/features/flags/flags.types.ts
 create mode 100644 src/features/flags/index.ts
 create mode 100644 src/features/flags/useFeatureFlags.test.tsx
 create mode 100644 src/features/flags/useFeatureFlags.ts
 create mode 100644 src/routes/_authenticated/debug.tsx

diff --git a/src/features/bible/components/DraftingResourceSidebar.tsx b/src/features/bible/components/DraftingResourceSidebar.tsx
index 1e958799..12c5510f 100644
--- a/src/features/bible/components/DraftingResourceSidebar.tsx
+++ b/src/features/bible/components/DraftingResourceSidebar.tsx
@@ -32,6 +32,10 @@ interface DraftingResourceSidebarProps {
   activeFindingsCount: number;
   /** Body for the Checks tab (the `ChecksPanel`, composed by the parent). */
   checksContent: ReactNode;
+  /** Whether the Checks tab is shown at all. When the Repeated Word Check
+   *  feature is disabled (feature-flags proposal D6/D7) the tab is hidden so it
+   *  looks unimplemented; forwarded straight to `LeftPanel`. Defaults to true. */
+  showChecksTab?: boolean;
 }
 
 export const DraftingResourceSidebar: React.FC<DraftingResourceSidebarProps> = ({
@@ -53,6 +57,7 @@ export const DraftingResourceSidebar: React.FC<DraftingResourceSidebarProps> = (
   onTabChange,
   activeFindingsCount,
   checksContent,
+  showChecksTab = true,
 }) => {
   const [resourcePanelWidth, setResourcePanelWidth] = useState(25); // percentage
   const isDraggingRef = useRef(false);
@@ -139,6 +144,7 @@ export const DraftingResourceSidebar: React.FC<DraftingResourceSidebarProps> = (
               onResourceChange={setCurrentResource}
             />
           }
+          showChecksTab={showChecksTab}
           onTabChange={onTabChange}
         />
       </div>
diff --git a/src/features/bible/components/DraftingUI.test.tsx b/src/features/bible/components/DraftingUI.test.tsx
index 624bd1e7..36589816 100644
--- a/src/features/bible/components/DraftingUI.test.tsx
+++ b/src/features/bible/components/DraftingUI.test.tsx
@@ -90,6 +90,18 @@ vi.mock('@/features/checks/hooks/useResolvedFindings', () => ({
   useResolvedFindings: () => ({ active: [], inactive: [] }),
 }));
 
+// Mock the feature-flags selector. The Repeated Word Check tab/query are gated
+// on `useFeatureFlag('repeatedWordCheck')` (feature-flags proposal D5–D7); this
+// hook wraps TanStack Query / fetch, so mocking it keeps DraftingUI renderable
+// with no network I/O. Default: feature ON, so the existing tests exercise the
+// full Checks surface; individual tests flip `mockFeatureFlag` to false to
+// assert the hidden state.
+const mockFeatureFlag = vi.fn<(name: string) => boolean>(() => true);
+
+vi.mock('@/features/flags', () => ({
+  useFeatureFlag: (name: string) => mockFeatureFlag(name) as unknown,
+}));
+
 // Mock ResourcePanel
 vi.mock('@/features/resources/components/ResourcePanel', () => ({
   ResourcePanel: ({
@@ -203,6 +215,9 @@ describe('DraftingUI', () => {
   beforeEach(() => {
     vi.clearAllMocks();
 
+    // Default: Repeated Word Check feature ON (see mock definition above).
+    mockFeatureFlag.mockReturnValue(true);
+
     mockUseAddTranslatedVerse.mockReturnValue({
       mutateAsync: vi.fn(),
       isPending: false,
@@ -419,4 +434,64 @@ describe('DraftingUI', () => {
 
     expect(handleNextClickMock).toHaveBeenCalled();
   });
+
+  // --- Repeated Word Check feature flag (feature-flags proposal D5–D7) -------
+  describe('Repeated Word Check feature flag', () => {
+    const openResourcePanel = async (user: ReturnType<typeof userEvent.setup>) => {
+      const toggleButton = screen.getByRole('button', { pressed: false });
+      await user.click(toggleButton);
+    };
+
+    it('shows the Checks tab when the feature is enabled', async () => {
+      const user = userEvent.setup();
+      mockFeatureFlag.mockReturnValue(true);
+
+      render(
+        <DraftingUI
+          projectItem={mockProjectItem}
+          sourceVerses={mockSourceVerses}
+          targetVerses={mockTargetVerses}
+          userdetail={{ id: 1 } as unknown as User}
+        />
+      );
+
+      await openResourcePanel(user);
+
+      expect(screen.getByRole('tab', { name: 'Resources' })).toBeInTheDocument();
+      expect(screen.getByRole('tab', { name: 'Checks' })).toBeInTheDocument();
+    });
+
+    it('hides the Checks tab (feature looks unimplemented) when the feature is disabled', async () => {
+      const user = userEvent.setup();
+      mockFeatureFlag.mockReturnValue(false);
+
+      render(
+        <DraftingUI
+          projectItem={mockProjectItem}
+          sourceVerses={mockSourceVerses}
+          targetVerses={mockTargetVerses}
+          userdetail={{ id: 1 } as unknown as User}
+        />
+      );
+
+      await openResourcePanel(user);
+
+      // Resources tab remains; the Checks tab is gone entirely.
+      expect(screen.getByRole('tab', { name: 'Resources' })).toBeInTheDocument();
+      expect(screen.queryByRole('tab', { name: 'Checks' })).not.toBeInTheDocument();
+    });
+
+    it('only queries the flag for the repeatedWordCheck feature', () => {
+      render(
+        <DraftingUI
+          projectItem={mockProjectItem}
+          sourceVerses={mockSourceVerses}
+          targetVerses={mockTargetVerses}
+          userdetail={{ id: 1 } as unknown as User}
+        />
+      );
+
+      expect(mockFeatureFlag).toHaveBeenCalledWith('repeatedWordCheck');
+    });
+  });
 });
diff --git a/src/features/bible/components/DraftingUI.tsx b/src/features/bible/components/DraftingUI.tsx
index ebdea4ca..31c423a6 100644
--- a/src/features/bible/components/DraftingUI.tsx
+++ b/src/features/bible/components/DraftingUI.tsx
@@ -20,6 +20,7 @@ import { ChecksPanel } from '@/features/checks/components/ChecksPanel';
 import { useRepeatedWordsCheck } from '@/features/checks/hooks/useRepeatedWordsCheck';
 import { useResolvedFindings } from '@/features/checks/hooks/useResolvedFindings';
 import { useSuppressions } from '@/features/checks/hooks/useSuppressions';
+import { useFeatureFlag } from '@/features/flags';
 import { type BibleVerse } from '@/features/resources/hooks/hooks';
 import { Logger } from '@/lib/services/logger';
 import {
@@ -196,6 +197,15 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
 
   // --- Repeated Word Check wiring (Phase 4, §6.2/§6.6, W3/W10/W11) ----------
 
+  // Feature flag: is the Repeated Word Check enabled in this environment? The
+  // whole Checks feature depends on fluent-ai, which isn't hosted everywhere,
+  // so the flag lets this code ship hidden until AI is wired (feature-flags
+  // proposal D6/D7). Fail-closed: `useFeatureFlag` returns false while loading
+  // or on endpoint error, so the tab and the AI query stay off by default. This
+  // is the single consumption point — the flag is threaded down as booleans so
+  // the leaf components stay flag-agnostic.
+  const checksEnabled = useFeatureFlag('repeatedWordCheck');
+
   // The single writer for the occurrence-rule map: `useSuppressions` does the
   // read-modify-write and hands the next full map back here; updating state
   // makes it ride the existing debounced editor-state save (one writer, §7.1).
@@ -225,7 +235,11 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
     saveCounter,
     // Wait for the settings probe so the first render is cascade-correct, and
     // never run in read-only `/view` or when the chapter is empty (W10/§9.1).
-    enabled: !readOnly && hasContent && settingsProbeResolved,
+    // Also gate on the feature flag: when the Repeated Word Check is off (or
+    // still loading / errored — fail-closed) we suppress the query entirely so
+    // no known-failing POST to /ai/tools/... fires in environments where
+    // fluent-ai isn't wired (feature-flags proposal D5/D7).
+    enabled: checksEnabled && !readOnly && hasContent && settingsProbeResolved,
   });
 
   const resolved = useResolvedFindings({
@@ -235,8 +249,17 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
   });
 
   // Active-finding count drives both notification dots; computed once here and
-  // threaded to the tab and the closed-panel toggle button (S5, §6.4).
-  const activeFindingsCount = resolved.active.length;
+  // threaded to the tab and the closed-panel toggle button (S5, §6.4). When the
+  // Checks feature is disabled it is forced to 0 so neither dot can flash while
+  // the flag is resolving (the query is already suppressed above; this guards
+  // the count explicitly).
+  const activeFindingsCount = checksEnabled ? resolved.active.length : 0;
+
+  // When the Checks tab is hidden (feature off), a persisted `activeLeftTab ===
+  // 'checks'` must not strand the panel on an empty/absent tab — fall back to
+  // Resources for rendering. The persisted value itself is left untouched, so
+  // if the feature is later enabled the translator's last tab is restored.
+  const effectiveActiveLeftTab: LeftTab = checksEnabled ? activeLeftTab : 'resources';
 
   const handleTabChange = useCallback((tab: LeftTab) => {
     setActiveLeftTab(tab);
@@ -517,7 +540,7 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
         {showResources && isInitializedRef.current && (
           <DraftingResourceSidebar
             activeFindingsCount={activeFindingsCount}
-            activeLeftTab={activeLeftTab}
+            activeLeftTab={effectiveActiveLeftTab}
             checksContent={
               <ChecksPanel
                 globalIgnoresAvailable={globalIgnoresAvailable}
@@ -543,6 +566,7 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
             setCurrentResource={setCurrentResource}
             setOpenResourcePanel={setOpenResourcePanel}
             setSelectedPanel={setSelectedPanel}
+            showChecksTab={checksEnabled}
             onTabChange={handleTabChange}
           />
         )}
diff --git a/src/features/checks/components/LeftPanel.test.tsx b/src/features/checks/components/LeftPanel.test.tsx
index fb19cc99..a54818d0 100644
--- a/src/features/checks/components/LeftPanel.test.tsx
+++ b/src/features/checks/components/LeftPanel.test.tsx
@@ -82,4 +82,36 @@ describe('LeftPanel', () => {
     render(<LeftPanel {...baseProps} activeFindingsCount={1} activeTab='checks' />);
     expect(screen.getByTestId('checks-notification-dot')).toBeInTheDocument();
   });
+
+  // --- showChecksTab (feature-flags proposal D6/D7) --------------------------
+  describe('showChecksTab', () => {
+    it('renders the Checks tab by default (prop omitted)', () => {
+      render(<LeftPanel {...baseProps} />);
+      expect(screen.getByRole('tab', { name: /Checks/ })).toBeInTheDocument();
+    });
+
+    it('hides the Checks tab entirely when showChecksTab is false', () => {
+      render(<LeftPanel {...baseProps} showChecksTab={false} />);
+      expect(screen.getByRole('tab', { name: /Resources/ })).toBeInTheDocument();
+      expect(screen.queryByRole('tab', { name: /Checks/ })).not.toBeInTheDocument();
+    });
+
+    it('suppresses the notification dot when the Checks tab is hidden', () => {
+      render(<LeftPanel {...baseProps} activeFindingsCount={5} showChecksTab={false} />);
+      expect(screen.queryByTestId('checks-notification-dot')).not.toBeInTheDocument();
+    });
+
+    it('falls back to the Resources body even if activeTab is checks while hidden', () => {
+      render(<LeftPanel {...baseProps} activeTab='checks' showChecksTab={false} />);
+      expect(screen.getByText('resources-body')).toBeInTheDocument();
+      expect(screen.queryByText('checks-body')).not.toBeInTheDocument();
+    });
+
+    it('labels the tabpanel by the Resources tab when the Checks tab is hidden', () => {
+      render(<LeftPanel {...baseProps} activeTab='checks' showChecksTab={false} />);
+      const resourcesTabId = screen.getByRole('tab', { name: /Resources/ }).getAttribute('id');
+      expect(resourcesTabId).toBeTruthy();
+      expect(screen.getByRole('tabpanel')).toHaveAttribute('aria-labelledby', resourcesTabId);
+    });
+  });
 });
diff --git a/src/features/checks/components/LeftPanel.tsx b/src/features/checks/components/LeftPanel.tsx
index 22cd43cd..55e61d57 100644
--- a/src/features/checks/components/LeftPanel.tsx
+++ b/src/features/checks/components/LeftPanel.tsx
@@ -24,6 +24,15 @@ export interface LeftPanelProps {
   resourcesContent: ReactNode;
   /** Body for the Checks tab (the `ChecksPanel`, composed by the parent). */
   checksContent: ReactNode;
+  /**
+   * Whether the "Checks" tab is shown at all. When the Repeated Word Check
+   * feature is disabled (feature-flags proposal D6/D7), the tab — and its
+   * notification dot — are hidden entirely so the panel looks as if Checks were
+   * never implemented. The parent (`DraftingUI`) is responsible for also forcing
+   * `activeTab` back to `'resources'` when this is false, so the shared panel
+   * never tries to render an absent tab's body. Defaults to true.
+   */
+  showChecksTab?: boolean;
 }
 
 /**
@@ -93,39 +102,48 @@ export const LeftPanel: React.FC<LeftPanelProps> = ({
   activeFindingsCount,
   resourcesContent,
   checksContent,
+  showChecksTab = true,
 }) => {
   const hasActiveFindings = activeFindingsCount > 0;
+  // When the Checks tab is hidden, the shared panel must never try to render the
+  // (absent) checks body. The parent forces `activeTab` back to 'resources', but
+  // we also guard here so this component is correct in isolation.
+  const effectiveTab: LeftTab = showChecksTab ? activeTab : 'resources';
 
   return (
     <div className='flex h-full flex-col'>
       <div aria-label='Left panel' className='flex items-center gap-6 px-1 pt-4' role='tablist'>
         <TabButton
           id={TAB_IDS.resources}
-          isActive={activeTab === 'resources'}
+          isActive={effectiveTab === 'resources'}
           label='Resources'
           onClick={() => onTabChange('resources')}
         />
         {/* The notification dot sits to the right of the "Checks" label only,
             and is visible whatever tab is active (mock shows it while the
-            Resources tab is active too) — §5.1, W11. */}
-        <TabButton
-          id={TAB_IDS.checks}
-          isActive={activeTab === 'checks'}
-          label='Checks'
-          showDot={hasActiveFindings}
-          onClick={() => onTabChange('checks')}
-        />
+            Resources tab is active too) — §5.1, W11. The whole tab (and thus the
+            dot) is hidden when the Checks feature is disabled (feature-flags
+            proposal D6/D7), so it looks as if Checks were never implemented. */}
+        {showChecksTab && (
+          <TabButton
+            id={TAB_IDS.checks}
+            isActive={effectiveTab === 'checks'}
+            label='Checks'
+            showDot={hasActiveFindings}
+            onClick={() => onTabChange('checks')}
+          />
+        )}
       </div>
 
       {/* Single shared panel: `aria-labelledby` tracks the active tab so the
           tab/tabpanel relationship is complete for assistive tech (CR-5). */}
       <div
-        aria-labelledby={TAB_IDS[activeTab]}
+        aria-labelledby={TAB_IDS[effectiveTab]}
         className='min-h-0 flex-1'
         id={PANEL_ID}
         role='tabpanel'
       >
-        {activeTab === 'resources' ? resourcesContent : checksContent}
+        {effectiveTab === 'resources' ? resourcesContent : checksContent}
       </div>
     </div>
   );
diff --git a/src/features/flags/FeatureFlagsDiagnostics.test.tsx b/src/features/flags/FeatureFlagsDiagnostics.test.tsx
new file mode 100644
index 00000000..a12058e8
--- /dev/null
+++ b/src/features/flags/FeatureFlagsDiagnostics.test.tsx
@@ -0,0 +1,91 @@
+import { type ReactNode } from 'react';
+
+import { QueryClientProvider } from '@tanstack/react-query';
+import userEvent from '@testing-library/user-event';
+import { http, HttpResponse } from 'msw';
+import { afterEach, describe, expect, it, vi } from 'vitest';
+
+import { config } from '@/lib/config';
+import { server } from '@/test/msw/server';
+import { createTestQueryClient, render, screen, waitFor } from '@/test/render';
+
+import { FeatureFlagsDiagnostics } from './FeatureFlagsDiagnostics';
+
+const FEATURES_URL = `${config.api.url}/config/features`;
+
+const makeWrapper = () => {
+  const queryClient = createTestQueryClient();
+  const Wrapper = ({ children }: { children: ReactNode }) => (
+    <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+  );
+  return Wrapper;
+};
+
+const mockFeatures = (features: Record<string, boolean>, status = 200) =>
+  server.use(http.get(FEATURES_URL, () => HttpResponse.json({ features }, { status })));
+
+afterEach(() => {
+  vi.restoreAllMocks();
+});
+
+describe('FeatureFlagsDiagnostics', () => {
+  it('renders each flag with its published On/Off state once loaded', async () => {
+    mockFeatures({ repeatedWordCheck: true });
+    render(<FeatureFlagsDiagnostics />, { wrapper: makeWrapper() });
+
+    await waitFor(() =>
+      expect(screen.getByTestId('flag-state-repeatedWordCheck')).toHaveTextContent('On')
+    );
+    expect(screen.getByTestId('flag-row-repeatedWordCheck')).toBeInTheDocument();
+    expect(screen.queryByTestId('flags-error')).not.toBeInTheDocument();
+  });
+
+  it('shows the flag as Off when the API reports it off', async () => {
+    mockFeatures({ repeatedWordCheck: false });
+    render(<FeatureFlagsDiagnostics />, { wrapper: makeWrapper() });
+
+    await waitFor(() =>
+      expect(screen.getByTestId('flag-state-repeatedWordCheck')).toHaveTextContent('Off')
+    );
+  });
+
+  it('surfaces an error banner and fail-closed (Off) values when the endpoint errors', async () => {
+    mockFeatures({ repeatedWordCheck: true }, 500);
+    render(<FeatureFlagsDiagnostics />, { wrapper: makeWrapper() });
+
+    await waitFor(() => expect(screen.getByTestId('flags-error')).toBeInTheDocument());
+    // Even though the (failing) server would have said `true`, the page shows Off.
+    expect(screen.getByTestId('flag-state-repeatedWordCheck')).toHaveTextContent('Off');
+  });
+
+  it('reflects the flag map even for an unknown/new flag key (enumerates the map)', async () => {
+    // The API is the source of truth; a flag added there before the FeatureName
+    // union is updated should still render, proving the page is data-driven.
+    mockFeatures({ repeatedWordCheck: false, someNewFlag: true });
+    render(<FeatureFlagsDiagnostics />, { wrapper: makeWrapper() });
+
+    await waitFor(() =>
+      expect(screen.getByTestId('flag-state-someNewFlag')).toHaveTextContent('On')
+    );
+  });
+
+  it('refetches when Refresh is clicked', async () => {
+    let calls = 0;
+    server.use(
+      http.get(FEATURES_URL, () => {
+        calls += 1;
+        return HttpResponse.json({ features: { repeatedWordCheck: false } });
+      })
+    );
+    const user = userEvent.setup();
+    render(<FeatureFlagsDiagnostics />, { wrapper: makeWrapper() });
+
+    await waitFor(() =>
+      expect(screen.getByTestId('flag-state-repeatedWordCheck')).toHaveTextContent('Off')
+    );
+    const callsAfterLoad = calls;
+
+    await user.click(screen.getByRole('button', { name: 'Refresh' }));
+    await waitFor(() => expect(calls).toBeGreaterThan(callsAfterLoad));
+  });
+});
diff --git a/src/features/flags/FeatureFlagsDiagnostics.tsx b/src/features/flags/FeatureFlagsDiagnostics.tsx
new file mode 100644
index 00000000..51e0b8be
--- /dev/null
+++ b/src/features/flags/FeatureFlagsDiagnostics.tsx
@@ -0,0 +1,108 @@
+import React from 'react';
+
+import { config } from '@/lib/config';
+
+import { useFeatureFlags } from './useFeatureFlags';
+
+/**
+ * Read-only diagnostics page for the published feature flags (proposal D8).
+ *
+ * Purpose: a human (or a support engineer) can hit one URL and see exactly what
+ * `GET /config/features` reports for the environment they're in — the single
+ * source of truth for "is the Repeated Word Check meant to be visible here?".
+ * This is deliberately unlinked (no nav entry): you reach it only if you know
+ * the path. It is login-gated by the `_authenticated` layout but has no role
+ * gate, so any signed-in user can self-serve this diagnostic.
+ *
+ * It is strictly read-only: it renders whatever the API returned and never
+ * writes or overrides a flag. The flags themselves are owned by fluent-api
+ * (env-sourced), so there is nothing to toggle here — mirroring the proposal's
+ * "the API owns the truth" stance (D3/D4).
+ *
+ * Note: because {@link useFeatureFlags} fails closed, a network error shows every
+ * flag as "Off" *and* surfaces the error banner — that combination is itself the
+ * signal that the values shown are the safe defaults, not confirmed state.
+ */
+export const FeatureFlagsDiagnostics: React.FC = () => {
+  const { features, isLoading, isError, query } = useFeatureFlags();
+
+  // Enumerate whatever the effective map contains rather than a hard-coded list,
+  // so a flag added later (API + `FeatureName`) appears here automatically with
+  // no change to this page.
+  const entries = Object.entries(features) as Array<[string, boolean]>;
+
+  return (
+    <div className='mx-auto max-w-2xl px-8 py-10'>
+      <header className='mb-6'>
+        <h1 className='text-2xl font-bold'>Feature flags</h1>
+        <p className='text-muted-foreground mt-1 text-sm'>
+          Read-only view of what <code className='font-mono'>GET /config/features</code> reports for
+          this environment. Values are owned by the API and cannot be changed here.
+        </p>
+        <p className='text-muted-foreground mt-1 text-xs'>
+          Source: <span className='font-mono'>{config.api.url}/config/features</span> · Environment:{' '}
+          <span className='font-mono'>{config.environment.current}</span>
+        </p>
+      </header>
+
+      {isLoading && (
+        <p className='text-muted-foreground text-sm' data-testid='flags-loading'>
+          Loading flags…
+        </p>
+      )}
+
+      {isError && (
+        <div
+          className='mb-4 rounded-md border border-red-300 bg-red-50 px-4 py-3 text-sm text-red-700'
+          data-testid='flags-error'
+          role='alert'
+        >
+          Could not reach the feature-flags endpoint. Showing fail-closed defaults (every flag
+          treated as <strong>Off</strong>) — these are safe defaults, not confirmed values.
+        </div>
+      )}
+
+      <table className='w-full border-collapse text-sm' data-testid='flags-table'>
+        <thead>
+          <tr className='border-b text-left'>
+            <th className='py-2 pr-4 font-semibold'>Flag</th>
+            <th className='py-2 font-semibold'>State</th>
+          </tr>
+        </thead>
+        <tbody>
+          {entries.map(([name, enabled]) => (
+            <tr key={name} className='border-b last:border-b-0' data-testid={`flag-row-${name}`}>
+              <td className='py-2 pr-4 font-mono'>{name}</td>
+              <td className='py-2'>
+                <span
+                  className={
+                    enabled
+                      ? 'inline-flex items-center rounded-full bg-green-100 px-2.5 py-0.5 text-xs font-semibold text-green-800'
+                      : 'inline-flex items-center rounded-full bg-gray-100 px-2.5 py-0.5 text-xs font-semibold text-gray-600'
+                  }
+                  data-testid={`flag-state-${name}`}
+                >
+                  {enabled ? 'On' : 'Off'}
+                </span>
+              </td>
+            </tr>
+          ))}
+        </tbody>
+      </table>
+
+      {entries.length === 0 && !isLoading && (
+        <p className='text-muted-foreground mt-4 text-sm'>No feature flags are defined.</p>
+      )}
+
+      <button
+        className='text-primary mt-6 cursor-pointer text-sm font-semibold underline-offset-2 hover:underline'
+        type='button'
+        onClick={() => void query.refetch()}
+      >
+        Refresh
+      </button>
+    </div>
+  );
+};
+
+export default FeatureFlagsDiagnostics;
diff --git a/src/features/flags/FeatureGate.tsx b/src/features/flags/FeatureGate.tsx
new file mode 100644
index 00000000..6ae7f827
--- /dev/null
+++ b/src/features/flags/FeatureGate.tsx
@@ -0,0 +1,34 @@
+import { type ReactNode } from 'react';
+
+import { type FeatureName } from './flags.types';
+import { useFeatureFlag } from './useFeatureFlags';
+
+export interface FeatureGateProps {
+  /** The flag that must be on for the children to render. */
+  feature: FeatureName;
+  /** Rendered only when the named flag is on. */
+  children: ReactNode;
+  /**
+   * Optional fallback rendered when the flag is off (including while loading or
+   * on error — the gate is fail-closed, D7). Defaults to nothing, i.e. the
+   * gated UI simply disappears as though it were not implemented.
+   */
+  fallback?: ReactNode;
+}
+
+/**
+ * Render children only when a feature flag is on (proposal D6/D7).
+ *
+ * **Fails closed:** while the flag map is loading, or if the endpoint errors,
+ * the flag reads `false` (see {@link useFeatureFlag}), so the children are
+ * hidden and the optional `fallback` (default: nothing) is shown instead. This
+ * is a pure render-wrapper — it does not gate side effects such as network
+ * queries; a caller that also needs to suppress a query should read the flag
+ * directly and fold it into that query's `enabled`.
+ */
+export const FeatureGate: React.FC<FeatureGateProps> = ({ feature, children, fallback = null }) => {
+  const enabled = useFeatureFlag(feature);
+  return <>{enabled ? children : fallback}</>;
+};
+
+export default FeatureGate;
diff --git a/src/features/flags/flags.types.ts b/src/features/flags/flags.types.ts
new file mode 100644
index 00000000..dba0d105
--- /dev/null
+++ b/src/features/flags/flags.types.ts
@@ -0,0 +1,40 @@
+/**
+ * Feature-flag types — the fluent-web mirror of the fluent-api contract.
+ *
+ * The API owns the truth (env-sourced) and publishes a read-only projection at
+ * `GET /config/features` (proposal D3/D4). fluent-web never decides policy — it
+ * only reflects what the API reports. See
+ * `fluent-api/src/lib/features.ts` for the source of these keys.
+ *
+ * The wire shape is a **named map** so new flags are purely additive: adding a
+ * feature later is a new key here + a new `<FeatureGate>`/`useFeatureFlag`
+ * usage, with no change to the hook or the gate primitive (D6).
+ */
+
+/**
+ * Known feature-flag wire keys (camelCase), mirroring the API's `FLAGS`
+ * registry. Kept as a string-literal union (not an open `string`) so a typo in
+ * `useFeatureFlag('...')` / `<FeatureGate feature="...">` is a compile error and
+ * so the diagnostics page can enumerate the known flags.
+ */
+export type FeatureName = 'repeatedWordCheck';
+
+/** The published feature map: every known flag, always present as a boolean. */
+export type Features = Record<FeatureName, boolean>;
+
+/** The `GET /config/features` response envelope. */
+export interface FeaturesResponse {
+  features: Features;
+}
+
+/**
+ * The safe, fail-closed default map (D7): every known flag is **off**. Used
+ * while flags are loading or when the endpoint errors/is unreachable, so gated
+ * AI UI is never surfaced before the API has confirmed the feature is on.
+ *
+ * Declared as a function (not a shared const) so callers always get a fresh
+ * object and can't accidentally mutate a shared default.
+ */
+export const failClosedFeatures = (): Features => ({
+  repeatedWordCheck: false,
+});
diff --git a/src/features/flags/index.ts b/src/features/flags/index.ts
new file mode 100644
index 00000000..7234b618
--- /dev/null
+++ b/src/features/flags/index.ts
@@ -0,0 +1,26 @@
+/**
+ * Feature flags — the fluent-web consumer of fluent-api's published feature map
+ * (`GET /config/features`). See the proposal at
+ * `fluent-api/docs/proposals/feature-flags/feature-flags-suggestion.md`.
+ *
+ * Public surface:
+ *   - `useFeatureFlags()` — the full, fail-closed feature map + query status.
+ *   - `useFeatureFlag(name)` — a single boolean (fail-closed).
+ *   - `<FeatureGate feature=...>` — render children only when a flag is on.
+ *   - types: `FeatureName`, `Features`, `FeaturesResponse`.
+ */
+export { FeatureGate, type FeatureGateProps } from './FeatureGate';
+export { FeatureFlagsDiagnostics } from './FeatureFlagsDiagnostics';
+export {
+  useFeatureFlags,
+  useFeatureFlag,
+  fetchFeatureFlags,
+  FEATURE_FLAGS_QUERY_KEY,
+  type UseFeatureFlagsResult,
+} from './useFeatureFlags';
+export {
+  type FeatureName,
+  type Features,
+  type FeaturesResponse,
+  failClosedFeatures,
+} from './flags.types';
diff --git a/src/features/flags/useFeatureFlags.test.tsx b/src/features/flags/useFeatureFlags.test.tsx
new file mode 100644
index 00000000..2ee56e0e
--- /dev/null
+++ b/src/features/flags/useFeatureFlags.test.tsx
@@ -0,0 +1,139 @@
+import { type ReactNode } from 'react';
+
+import { QueryClientProvider } from '@tanstack/react-query';
+import { http, HttpResponse } from 'msw';
+import { afterEach, describe, expect, it, vi } from 'vitest';
+
+import { config } from '@/lib/config';
+import { server } from '@/test/msw/server';
+import { createTestQueryClient, render, renderHook, screen, waitFor } from '@/test/render';
+
+import { FeatureGate } from './FeatureGate';
+import { failClosedFeatures } from './flags.types';
+import { useFeatureFlag, useFeatureFlags } from './useFeatureFlags';
+
+const FEATURES_URL = `${config.api.url}/config/features`;
+
+/**
+ * A QueryClientProvider wrapper pinned to a single test client (retries off, no
+ * cache) so a `rerender` doesn't reset the cache. Mirrors the pattern in
+ * `useRepeatedWordsCheck.test.ts`. (The hook itself also sets `retry: false`, so
+ * the error case resolves immediately.)
+ */
+const makeWrapper = () => {
+  const queryClient = createTestQueryClient();
+  const Wrapper = ({ children }: { children: ReactNode }) => (
+    <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+  );
+  return Wrapper;
+};
+
+/** Register a GET handler for /config/features returning the given map. */
+const mockFeatures = (features: Record<string, boolean>, status = 200) => {
+  const calls = { count: 0 };
+  server.use(
+    http.get(FEATURES_URL, () => {
+      calls.count += 1;
+      return HttpResponse.json({ features }, { status });
+    })
+  );
+  return calls;
+};
+
+afterEach(() => {
+  vi.restoreAllMocks();
+});
+
+describe('failClosedFeatures', () => {
+  it('returns every known flag off, and a fresh object each call', () => {
+    const a = failClosedFeatures();
+    const b = failClosedFeatures();
+    expect(a).toEqual({ repeatedWordCheck: false });
+    expect(a).not.toBe(b); // not a shared mutable singleton
+  });
+});
+
+describe('useFeatureFlags — fail-closed semantics', () => {
+  it('reports every flag off while loading (before the response lands)', () => {
+    mockFeatures({ repeatedWordCheck: true });
+    const { result } = renderHook(() => useFeatureFlags(), { wrapper: makeWrapper() });
+    // First synchronous render: data hasn't arrived → fail-closed.
+    expect(result.current.isLoading).toBe(true);
+    expect(result.current.features.repeatedWordCheck).toBe(false);
+  });
+
+  it('reflects the published map once loaded', async () => {
+    mockFeatures({ repeatedWordCheck: true });
+    const { result } = renderHook(() => useFeatureFlags(), { wrapper: makeWrapper() });
+    await waitFor(() => expect(result.current.isLoading).toBe(false));
+    expect(result.current.features.repeatedWordCheck).toBe(true);
+    expect(result.current.isError).toBe(false);
+  });
+
+  it('publishes a flag as off when the API says off', async () => {
+    mockFeatures({ repeatedWordCheck: false });
+    const { result } = renderHook(() => useFeatureFlags(), { wrapper: makeWrapper() });
+    await waitFor(() => expect(result.current.isLoading).toBe(false));
+    expect(result.current.features.repeatedWordCheck).toBe(false);
+  });
+
+  it('fails closed (all flags off) when the endpoint errors', async () => {
+    mockFeatures({ repeatedWordCheck: true }, 500);
+    const { result } = renderHook(() => useFeatureFlags(), { wrapper: makeWrapper() });
+    await waitFor(() => expect(result.current.isError).toBe(true));
+    // Even though the (unreachable) server would have said `true`, we hide.
+    expect(result.current.features.repeatedWordCheck).toBe(false);
+  });
+});
+
+describe('useFeatureFlag — single-flag selector', () => {
+  it('returns the boolean for a named flag once loaded', async () => {
+    mockFeatures({ repeatedWordCheck: true });
+    const { result } = renderHook(() => useFeatureFlag('repeatedWordCheck'), {
+      wrapper: makeWrapper(),
+    });
+    expect(result.current).toBe(false); // fail-closed during load
+    await waitFor(() => expect(result.current).toBe(true));
+  });
+});
+
+describe('FeatureGate', () => {
+  it('renders children only after the flag resolves on (hidden while loading)', async () => {
+    mockFeatures({ repeatedWordCheck: true });
+    render(
+      <FeatureGate feature='repeatedWordCheck'>
+        <div>gated-content</div>
+      </FeatureGate>,
+      { wrapper: makeWrapper() }
+    );
+    // Fail-closed: not present on first render.
+    expect(screen.queryByText('gated-content')).not.toBeInTheDocument();
+    await waitFor(() => expect(screen.getByText('gated-content')).toBeInTheDocument());
+  });
+
+  it('renders the fallback (and never the children) when the flag is off', async () => {
+    mockFeatures({ repeatedWordCheck: false });
+    render(
+      <FeatureGate fallback={<div>fallback-content</div>} feature='repeatedWordCheck'>
+        <div>gated-content</div>
+      </FeatureGate>,
+      { wrapper: makeWrapper() }
+    );
+    await waitFor(() => expect(screen.getByText('fallback-content')).toBeInTheDocument());
+    expect(screen.queryByText('gated-content')).not.toBeInTheDocument();
+  });
+
+  it('stays hidden (children never appear) when the endpoint errors', async () => {
+    const calls = mockFeatures({ repeatedWordCheck: true }, 500);
+    render(
+      <FeatureGate feature='repeatedWordCheck'>
+        <div>gated-content</div>
+      </FeatureGate>,
+      { wrapper: makeWrapper() }
+    );
+    await waitFor(() => expect(calls.count).toBeGreaterThan(0));
+    // Give React Query a tick to settle into the error state.
+    await new Promise(r => setTimeout(r, 20));
+    expect(screen.queryByText('gated-content')).not.toBeInTheDocument();
+  });
+});
diff --git a/src/features/flags/useFeatureFlags.ts b/src/features/flags/useFeatureFlags.ts
new file mode 100644
index 00000000..5c2f97fb
--- /dev/null
+++ b/src/features/flags/useFeatureFlags.ts
@@ -0,0 +1,115 @@
+import { useQuery, type UseQueryResult } from '@tanstack/react-query';
+
+import { config } from '@/lib/config';
+import { Logger } from '@/lib/services/logger';
+
+import {
+  type FeatureName,
+  type Features,
+  type FeaturesResponse,
+  failClosedFeatures,
+} from './flags.types';
+
+/**
+ * Query key for the published feature map. Exported so tests and the
+ * diagnostics page can reference/invalidate it without restating the string.
+ */
+export const FEATURE_FLAGS_QUERY_KEY = ['feature-flags'] as const;
+
+/**
+ * How long the flag map is considered fresh. Flags change only on deploy/config
+ * (never at runtime — proposal D1/D8), so there is no value in refetching them
+ * often; a long staleTime avoids a request storm without hiding a real change
+ * for longer than a page reload.
+ */
+const FLAGS_STALE_TIME_MS = 5 * 60 * 1000; // 5 minutes
+
+/** GET the published feature map from fluent-api. Follows the fetch pattern in
+ *  `features/bible/hooks/useBibleTarget.ts` (plain fetch + credentials). The
+ *  endpoint is unauthenticated (proposal Q1), but we still send credentials for
+ *  consistency and so it keeps working if the reviewer decides to gate it. */
+export const fetchFeatureFlags = async (): Promise<Features> => {
+  const res = await fetch(`${config.api.url}/config/features`, {
+    method: 'GET',
+    credentials: 'include',
+    headers: {
+      'Content-Type': 'application/json',
+    },
+  });
+
+  if (!res.ok) {
+    throw new Error(`Failed to fetch feature flags (HTTP ${res.status})`);
+  }
+
+  const body = (await res.json()) as FeaturesResponse;
+  return body.features;
+};
+
+export interface UseFeatureFlagsResult {
+  /**
+   * The effective feature map. **Never undefined** — while loading or on error
+   * this is the fail-closed map (every flag `false`), so callers can read
+   * `features.repeatedWordCheck` directly without null-checks and get the safe
+   * (hidden) answer by default (D7).
+   */
+  features: Features;
+  /** True while the flags are being fetched for the first time. */
+  isLoading: boolean;
+  /** True if the fetch errored/was unreachable (flags are fail-closed). */
+  isError: boolean;
+  /** Raw query handle, for a diagnostics page that wants status/refetch. */
+  query: UseQueryResult<Features>;
+}
+
+/**
+ * Fetch and expose the published feature flags (proposal D6/D7).
+ *
+ * **Fails closed.** The returned `features` map is the fail-closed default
+ * (everything off) until the API confirms a flag is on — so gated AI UI stays
+ * hidden while the request is in flight or if it fails, which is the whole point
+ * (don't surface features whose backend isn't there). The map is always the
+ * full `Features` shape, so consumers never branch on undefined.
+ */
+export const useFeatureFlags = (): UseFeatureFlagsResult => {
+  const query = useQuery<Features>({
+    queryKey: FEATURE_FLAGS_QUERY_KEY,
+    queryFn: async () => {
+      try {
+        return await fetchFeatureFlags();
+      } catch (error) {
+        Logger.logException(error instanceof Error ? error : new Error(String(error)), {
+          context: 'Feature-flags fetch failed',
+        });
+        throw error;
+      }
+    },
+    staleTime: FLAGS_STALE_TIME_MS,
+    // No retry. The query already fails *closed* — a blip just keeps gated UI
+    // hidden (the safe state) and self-corrects on the next natural refetch, so
+    // retrying would only delay when `isError` (and the diagnostics banner) can
+    // be observed without changing the user-visible outcome. Keeping it off also
+    // means the fail-closed answer is available immediately.
+    retry: false,
+  });
+
+  return {
+    // Fail-closed: use live data only once it has actually arrived; otherwise
+    // (loading OR error) every flag reads false.
+    features: query.data ?? failClosedFeatures(),
+    isLoading: query.isLoading,
+    isError: query.isError,
+    query,
+  };
+};
+
+/**
+ * Thin selector for a single flag (D6). Returns a plain boolean, fail-closed by
+ * construction (it reads the fail-closed map from {@link useFeatureFlags}).
+ *
+ * @example
+ *   const checksEnabled = useFeatureFlag('repeatedWordCheck');
+ */
+export const useFeatureFlag = (name: FeatureName): boolean => {
+  const { features } = useFeatureFlags();
+  return features[name];
+};
diff --git a/src/routeTree.gen.ts b/src/routeTree.gen.ts
index 02b0706e..c1f42595 100644
--- a/src/routeTree.gen.ts
+++ b/src/routeTree.gen.ts
@@ -17,6 +17,7 @@ import { Route as AuthenticatedRouteImport } from './routes/_authenticated'
 import { Route as AuthenticatedIndexRouteImport } from './routes/_authenticated/index'
 import { Route as LegalTermsRouteImport } from './routes/legal/terms'
 import { Route as LegalPrivacyRouteImport } from './routes/legal/privacy'
+import { Route as AuthenticatedDebugRouteImport } from './routes/_authenticated/debug'
 import { Route as AuthenticatedUsersIndexRouteImport } from './routes/_authenticated/users/index'
 import { Route as AuthenticatedProjectsIndexRouteImport } from './routes/_authenticated/projects/index'
 import { Route as AuthenticatedProjectsProjectIdIndexRouteImport } from './routes/_authenticated/projects/$projectId/index'
@@ -62,6 +63,11 @@ const LegalPrivacyRoute = LegalPrivacyRouteImport.update({
   path: '/privacy',
   getParentRoute: () => LegalRoute,
 } as any)
+const AuthenticatedDebugRoute = AuthenticatedDebugRouteImport.update({
+  id: '/debug',
+  path: '/debug',
+  getParentRoute: () => AuthenticatedRoute,
+} as any)
 const AuthenticatedUsersIndexRoute = AuthenticatedUsersIndexRouteImport.update({
   id: '/users/',
   path: '/users/',
@@ -98,6 +104,7 @@ export interface FileRoutesByFullPath {
   '/legal': typeof LegalRouteWithChildren
   '/login': typeof LoginRoute
   '/reset-password': typeof ResetPasswordRoute
+  '/debug': typeof AuthenticatedDebugRoute
   '/legal/privacy': typeof LegalPrivacyRoute
   '/legal/terms': typeof LegalTermsRoute
   '/projects/': typeof AuthenticatedProjectsIndexRoute
@@ -111,6 +118,7 @@ export interface FileRoutesByTo {
   '/legal': typeof LegalRouteWithChildren
   '/login': typeof LoginRoute
   '/reset-password': typeof ResetPasswordRoute
+  '/debug': typeof AuthenticatedDebugRoute
   '/legal/privacy': typeof LegalPrivacyRoute
   '/legal/terms': typeof LegalTermsRoute
   '/': typeof AuthenticatedIndexRoute
@@ -127,6 +135,7 @@ export interface FileRoutesById {
   '/legal': typeof LegalRouteWithChildren
   '/login': typeof LoginRoute
   '/reset-password': typeof ResetPasswordRoute
+  '/_authenticated/debug': typeof AuthenticatedDebugRoute
   '/legal/privacy': typeof LegalPrivacyRoute
   '/legal/terms': typeof LegalTermsRoute
   '/_authenticated/': typeof AuthenticatedIndexRoute
@@ -144,6 +153,7 @@ export interface FileRouteTypes {
     | '/legal'
     | '/login'
     | '/reset-password'
+    | '/debug'
     | '/legal/privacy'
     | '/legal/terms'
     | '/projects/'
@@ -157,6 +167,7 @@ export interface FileRouteTypes {
     | '/legal'
     | '/login'
     | '/reset-password'
+    | '/debug'
     | '/legal/privacy'
     | '/legal/terms'
     | '/'
@@ -172,6 +183,7 @@ export interface FileRouteTypes {
     | '/legal'
     | '/login'
     | '/reset-password'
+    | '/_authenticated/debug'
     | '/legal/privacy'
     | '/legal/terms'
     | '/_authenticated/'
@@ -248,6 +260,13 @@ declare module '@tanstack/react-router' {
       preLoaderRoute: typeof LegalPrivacyRouteImport
       parentRoute: typeof LegalRoute
     }
+    '/_authenticated/debug': {
+      id: '/_authenticated/debug'
+      path: '/debug'
+      fullPath: '/debug'
+      preLoaderRoute: typeof AuthenticatedDebugRouteImport
+      parentRoute: typeof AuthenticatedRoute
+    }
     '/_authenticated/users/': {
       id: '/_authenticated/users/'
       path: '/users'
@@ -287,6 +306,7 @@ declare module '@tanstack/react-router' {
 }
 
 interface AuthenticatedRouteChildren {
+  AuthenticatedDebugRoute: typeof AuthenticatedDebugRoute
   AuthenticatedIndexRoute: typeof AuthenticatedIndexRoute
   AuthenticatedProjectsIndexRoute: typeof AuthenticatedProjectsIndexRoute
   AuthenticatedUsersIndexRoute: typeof AuthenticatedUsersIndexRoute
@@ -296,6 +316,7 @@ interface AuthenticatedRouteChildren {
 }
 
 const AuthenticatedRouteChildren: AuthenticatedRouteChildren = {
+  AuthenticatedDebugRoute: AuthenticatedDebugRoute,
   AuthenticatedIndexRoute: AuthenticatedIndexRoute,
   AuthenticatedProjectsIndexRoute: AuthenticatedProjectsIndexRoute,
   AuthenticatedUsersIndexRoute: AuthenticatedUsersIndexRoute,
diff --git a/src/routes/_authenticated/debug.tsx b/src/routes/_authenticated/debug.tsx
new file mode 100644
index 00000000..eea7180d
--- /dev/null
+++ b/src/routes/_authenticated/debug.tsx
@@ -0,0 +1,15 @@
+import { createFileRoute } from '@tanstack/react-router';
+
+import { FeatureFlagsDiagnostics } from '@/features/flags';
+
+/**
+ * Unlinked, read-only feature-flags diagnostics page (feature-flags proposal
+ * D8). Login-gated by the `_authenticated` layout but intentionally has no role
+ * gate and no nav entry — you reach it only by knowing the path (`/debug`) — so
+ * any signed-in user can self-serve "what does `GET /config/features` report
+ * here?" without exposing a control surface (there is nothing to toggle; the API
+ * owns the values).
+ */
+export const Route = createFileRoute('/_authenticated/debug')({
+  component: FeatureFlagsDiagnostics,
+});

From 0e6dae90efaa3f516313270955f78d7df43215b9 Mon Sep 17 00:00:00 2001
From: Joshua Lansford <Joshua@Lansfords.com>
Date: Sun, 5 Jul 2026 14:25:24 -0400
Subject: [PATCH 33/33] fix(flags): address CodeRabbit review on #337

- rename wire field targetLanguageCode -> targetLangCode (match sourceLangCode / new api key)
- checks: narrow RepeatedWordsRequest.project_id to string (fluent-ai strict contract)
- suppressions: skip /self/settings probe when the check feature is gated off
- flags: fail-closed merge of /config/features body so partial/malformed 200s stay hidden
- checks: include HTTP status in repeated-words error message
- diagnostics: drop redundant entries cast
- model /self/settings updatedAt on the client type; reconcile doc
- tests: assert request body; replace flaky setTimeout negatives with waitFor
---
 .../checks-ui-integration-suggestion.md       |  2 +-
 .../bible/components/DraftingUI.test.tsx      |  2 +-
 src/features/bible/components/DraftingUI.tsx  |  6 ++-
 src/features/checks/checks.types.ts           |  7 ++-
 .../hooks/useRepeatedWordsCheck.test.ts       | 49 +++++++++++++------
 .../checks/hooks/useRepeatedWordsCheck.ts     |  9 ++--
 src/features/checks/hooks/useSuppressions.ts  | 35 +++++++++++--
 .../dashboard/user/UserHomePage.test.tsx      |  2 +-
 .../flags/FeatureFlagsDiagnostics.tsx         |  5 +-
 src/features/flags/useFeatureFlags.ts         |  9 +++-
 .../projects/components/ProjectDetailPage.tsx |  2 +-
 src/lib/types.ts                              |  8 +--
 12 files changed, 102 insertions(+), 34 deletions(-)

diff --git a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
index af497233..dac70a6d 100644
--- a/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
+++ b/docs/proposals/repeated-word-check/checks-ui-integration-suggestion.md
@@ -379,7 +379,7 @@ Otherwise it keeps the editor-state idiom: **full-replace upsert** on PUT, one Z
 | Method | Path             | Auth                    | Body                         | Response                                                                                                                                                            |
 | ------ | ---------------- | ----------------------- | ---------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
 | `GET`  | `/self/settings` | `authenticateUser` only | —                            | `200` `{ settings: UserSettings \| null, updatedAt }` — `settings: null` when the user has no row yet (mirrors editor-state's no-state-yet response; **not** a 404) |
-| `PUT`  | `/self/settings` | `authenticateUser` only | `{ settings: UserSettings }` | `200` saved state; `400` on schema violation                                                                                                                        |
+| `PUT`  | `/self/settings` | `authenticateUser` only | `{ settings: UserSettings }` | `200` saved state (`{ settings, updatedAt }`); `422` on schema violation (a well-formed request whose body fails validation)                                        |
 
 - **No permission middleware beyond `authenticateUser`.** Every caller owns exactly their own settings row, and the handler scopes all reads/writes to `currentUser.id` from context. (Editor-state additionally guards with `CONTENT_UPDATE` + chapter-assignment participation because it is tied to a chapter resource; `self/settings` has no resource to scope, so session identity alone is the right guard — and the `self` prefix makes that explicit.)
 - **PUT is full-replace.** The client GETs, merges in memory, PUTs the whole blob — the editor-state write pattern. Concurrent tabs are last-writer-wins, identical to editor-state's accepted behavior; for an ignore-list this worst case is "one ignore from another tab is lost until re-clicked."
diff --git a/src/features/bible/components/DraftingUI.test.tsx b/src/features/bible/components/DraftingUI.test.tsx
index 36589816..faf16435 100644
--- a/src/features/bible/components/DraftingUI.test.tsx
+++ b/src/features/bible/components/DraftingUI.test.tsx
@@ -143,7 +143,7 @@ const mockProjectItem: ProjectItem = {
   bibleId: 1,
   bibleName: 'WEB',
   targetLanguage: 'Spanish',
-  targetLanguageCode: 'spa',
+  targetLangCode: 'spa',
   bookId: 1,
   book: 'Genesis',
   chapterStatus: ChapterAssignmentStatus.DRAFT,
diff --git a/src/features/bible/components/DraftingUI.tsx b/src/features/bible/components/DraftingUI.tsx
index 31c423a6..97e45325 100644
--- a/src/features/bible/components/DraftingUI.tsx
+++ b/src/features/bible/components/DraftingUI.tsx
@@ -222,7 +222,11 @@ export const DraftingUI: React.FC<DraftingUIProps> = ({
     ignoreEverywhere,
     undoOccurrence,
     stopIgnoringEverywhere,
-  } = useSuppressions({ occurrenceRules, saveOccurrenceRules });
+    // Gate the once-per-session `GET /self/settings` probe on the same feature
+    // flag as the check query: when the Repeated Word Check is off (or still
+    // loading / errored — fail-closed) the Checks UI is hidden, so there is no
+    // reason to fetch user settings for it (W2, feature-flags proposal D5/D7).
+  } = useSuppressions({ occurrenceRules, saveOccurrenceRules, enabled: checksEnabled });
 
   // What the translator currently sees is what gets checked (§6.2) — feed the
   // live drafting verses, not a refetch.
diff --git a/src/features/checks/checks.types.ts b/src/features/checks/checks.types.ts
index f969eae9..cdbb1977 100644
--- a/src/features/checks/checks.types.ts
+++ b/src/features/checks/checks.types.ts
@@ -40,7 +40,12 @@ export interface VerseInput {
 export interface RepeatedWordsRequest {
   lang_code: string;
   lang_name: string;
-  project_id: string | number;
+  // fluent-ai declares `project_id: str` (Pydantic v2, strict — it does NOT
+  // coerce an int to a str), so this MUST be a string on the wire. The builder
+  // already sends `String(projectItem.projectUnitId)`; typing it `string` (not
+  // `string | number`) matches that reality and keeps a stray numeric id from
+  // ever compiling through. See W1 and phase-04 manual smoke (2026-06-23).
+  project_id: string;
   project_name: string;
   verses: VerseInput[];
 }
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
index a4917b09..0aecc818 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.test.ts
@@ -44,8 +44,8 @@ const makeProjectItem = (over: Partial<ProjectItem> = {}): ProjectItem => ({
   // `lang_name`, NOT as `lang_code` (greek-room keys its legitimate-duplicate
   // whitelist on the ISO code). See phase-04 manual smoke (BUG #2, 2026-06-23).
   targetLanguage: 'Spanish',
-  // `targetLanguageCode` is the ISO 639-3 code the check must send as `lang_code`.
-  targetLanguageCode: 'spa',
+  // `targetLangCode` is the ISO 639-3 code the check must send as `lang_code`.
+  targetLangCode: 'spa',
   bookId: 7,
   // `book` is the human display name; the check must NOT use it for snt_id.
   book: 'Judges',
@@ -151,7 +151,7 @@ describe('buildRepeatedWordsRequest', () => {
     // silently disables legitimate-duplicate suppression. lang_code MUST be the
     // code, lang_name the name. See phase-04 manual smoke (BUG #2, 2026-06-23).
     const request = buildRepeatedWordsRequest(
-      makeProjectItem({ targetLanguage: 'English', targetLanguageCode: 'eng' }),
+      makeProjectItem({ targetLanguage: 'English', targetLangCode: 'eng' }),
       [{ verseNumber: 3, content: 'truly truly' }]
     );
     expect(request.lang_code).toBe('eng');
@@ -160,13 +160,13 @@ describe('buildRepeatedWordsRequest', () => {
     expect(request.lang_code).not.toBe('English');
   });
 
-  it('falls back to "<unknown>" when targetLanguageCode is empty (no crash)', () => {
-    // `targetLanguageCode` is a required `string` (BUG #3 made it non-optional so
+  it('falls back to "<unknown>" when targetLangCode is empty (no crash)', () => {
+    // `targetLangCode` is a required `string` (BUG #3 made it non-optional so
     // the compiler catches any ProjectItem constructor that omits it). The runtime
     // fallback still degrades gracefully: an empty/whitespace code becomes
     // "<unknown>" on the wire rather than producing undefined/empty.
     const request = buildRepeatedWordsRequest(
-      makeProjectItem({ targetLanguage: 'English', targetLanguageCode: '' }),
+      makeProjectItem({ targetLanguage: 'English', targetLangCode: '' }),
       [{ verseNumber: 3, content: 'truly truly' }]
     );
     // greek-room just won't match any legitimate whitelist for an unknown code.
@@ -190,7 +190,7 @@ describe('buildRepeatedWordsRequest', () => {
 describe('useRepeatedWordsCheck — enabled gating', () => {
   it('does not fire when disabled', async () => {
     const calls = mockPost();
-    renderHook(
+    const { result } = renderHook(
       () =>
         useRepeatedWordsCheck({
           projectItem: makeProjectItem(),
@@ -200,14 +200,17 @@ describe('useRepeatedWordsCheck — enabled gating', () => {
         }),
       { wrapper: makeWrapper() }
     );
-    // Give any rogue request a tick to land.
-    await new Promise(r => setTimeout(r, 50));
+    // A disabled query is `fetchStatus: 'idle'` — assert it stays there instead
+    // of sleeping on a fixed timer to "prove a negative" (W7). If the query
+    // ever fired, `fetchStatus` would flip to 'fetching' and this would fail.
+    await waitFor(() => expect(result.current.fetchStatus).toBe('idle'));
+    expect(result.current.isFetching).toBe(false);
     expect(calls.count).toBe(0);
   });
 
   it('does not fire when no verse has content (even if enabled)', async () => {
     const calls = mockPost();
-    renderHook(
+    const { result } = renderHook(
       () =>
         useRepeatedWordsCheck({
           projectItem: makeProjectItem(),
@@ -217,12 +220,15 @@ describe('useRepeatedWordsCheck — enabled gating', () => {
         }),
       { wrapper: makeWrapper() }
     );
-    await new Promise(r => setTimeout(r, 50));
+    // With no verse content the query is disabled (`enabled && hasContent`), so
+    // it stays `fetchStatus: 'idle'` — assert that rather than a fixed sleep (W7).
+    await waitFor(() => expect(result.current.fetchStatus).toBe('idle'));
+    expect(result.current.isFetching).toBe(false);
     expect(calls.count).toBe(0);
   });
 
   it('fires and returns the completed envelope when enabled with content', async () => {
-    mockPost(makeResponse(2));
+    const calls = mockPost(makeResponse(2));
     const { result } = renderHook(
       () =>
         useRepeatedWordsCheck({
@@ -236,6 +242,19 @@ describe('useRepeatedWordsCheck — enabled gating', () => {
     await waitFor(() => expect(result.current.isSuccess).toBe(true));
     expect(result.current.data?.status).toBe('completed');
     expect(result.current.data?.result?.findings).toHaveLength(2);
+
+    // Assert the request body that actually went on the wire (W6): the hook
+    // must POST the snake_case D8 shape assembled by buildRepeatedWordsRequest
+    // — the ISO code as lang_code, the display name as lang_name, and the
+    // project id as a STRING (fluent-ai's strict `str`).
+    expect(calls.count).toBe(1);
+    expect(calls.lastBody).toMatchObject({
+      lang_code: 'spa',
+      lang_name: 'Spanish',
+      project_id: '7',
+      project_name: 'Test Project',
+      verses: [{ snt_id: 'JDG 4:3', text: 'the the' }],
+    });
   });
 });
 
@@ -304,8 +323,8 @@ describe('buildRepeatedWordsRequest — additional edge cases', () => {
     expect(request.verses[0].snt_id).toBe('JDG 12:5');
   });
 
-  it('uses whitespace-trimmed targetLanguageCode — empty after trim falls back to <unknown>', () => {
-    const request = buildRepeatedWordsRequest(makeProjectItem({ targetLanguageCode: '   ' }), [
+  it('uses whitespace-trimmed targetLangCode — empty after trim falls back to <unknown>', () => {
+    const request = buildRepeatedWordsRequest(makeProjectItem({ targetLangCode: '   ' }), [
       { verseNumber: 1, content: 'text' },
     ]);
     expect(request.lang_code).toBe('<unknown>');
@@ -316,7 +335,7 @@ describe('buildRepeatedWordsRequest — additional edge cases', () => {
     // padded code (" spa ") would match nothing and silently disable suppression
     // (the BUG #2 failure mode). The guard must forward the trimmed value, not
     // just use the trim to test truthiness. See phase-04 manual smoke (BUG #2).
-    const request = buildRepeatedWordsRequest(makeProjectItem({ targetLanguageCode: ' spa ' }), [
+    const request = buildRepeatedWordsRequest(makeProjectItem({ targetLangCode: ' spa ' }), [
       { verseNumber: 1, content: 'text' },
     ]);
     expect(request.lang_code).toBe('spa');
diff --git a/src/features/checks/hooks/useRepeatedWordsCheck.ts b/src/features/checks/hooks/useRepeatedWordsCheck.ts
index 137b2268..9c96881a 100644
--- a/src/features/checks/hooks/useRepeatedWordsCheck.ts
+++ b/src/features/checks/hooks/useRepeatedWordsCheck.ts
@@ -60,7 +60,7 @@ export const buildRepeatedWordsRequest = (
   // whitelist for an unknown code (duplicates are still flagged) — a safe
   // degradation that keeps the check running instead of crashing.
   // The `?.` here is intentionally defensive against a RUNTIME value the TYPE
-  // says can't happen: `targetLanguageCode` is typed non-optional, but the API
+  // says can't happen: `targetLangCode` is typed non-optional, but the API
   // can still omit it. We keep the type non-optional (so genuinely missing
   // usages elsewhere stay caught) and guard only at this trust boundary, hence
   // the targeted disable of the "unnecessary optional chain" rule.
@@ -70,7 +70,7 @@ export const buildRepeatedWordsRequest = (
   // truthiness on a trimmed copy while forwarding the untrimmed original would
   // silently disable suppression (the BUG #2 failure mode).
   // eslint-disable-next-line @typescript-eslint/no-unnecessary-condition -- defensive guard against an undefined runtime value the type does not model
-  const trimmedCode = projectItem.targetLanguageCode?.trim();
+  const trimmedCode = projectItem.targetLangCode?.trim();
   const langCode = trimmedCode ? trimmedCode : null;
 
   return {
@@ -101,7 +101,10 @@ export const postRepeatedWordsCheck = async (
   });
 
   if (!res.ok) {
-    throw new Error('Failed to run repeated-words check');
+    // Include the HTTP status so a failed check is actionable in logs/telemetry
+    // (e.g. fluent-api maps a fluent-ai failure/unreachable to 502) rather than
+    // an opaque generic message (W5).
+    throw new Error(`Failed to run repeated-words check (HTTP ${res.status})`);
   }
 
   return (await res.json()) as RepeatedWordsResponse;
diff --git a/src/features/checks/hooks/useSuppressions.ts b/src/features/checks/hooks/useSuppressions.ts
index 1c9ca9f7..37594cd1 100644
--- a/src/features/checks/hooks/useSuppressions.ts
+++ b/src/features/checks/hooks/useSuppressions.ts
@@ -62,9 +62,20 @@ interface SelfSettings {
   checkIgnoredWordPairs?: GlobalRules;
 }
 
-/** GET response: the server returns `{ settings }`, possibly `null` (W8). */
+/**
+ * GET response shape. The API returns `{ settings, updatedAt }` (see
+ * fluent-api `self-settings.service.ts` `toResponse`), where `settings` is
+ * `null` when the user has no row yet (W8). The client only reads `settings`
+ * — `updatedAt` is modeled here for contract fidelity (the doc §8.1 lists it)
+ * but is deliberately ignored: there is no optimistic-concurrency / ETag path
+ * in v1 (§8.3), so the timestamp has no consumer. Kept optional so an older
+ * deployment that omits it still parses.
+ */
 interface SelfSettingsResponse {
   settings: SelfSettings | null;
+  /** Server-side last-write timestamp (ISO string) or `null`. Unused by the
+   * client today; present only to match the wire contract. */
+  updatedAt?: string | null;
 }
 
 const SELF_SETTINGS_URL = `${config.api.url}/self/settings`;
@@ -228,6 +239,18 @@ export interface UseSuppressionsParams {
    * never opens its own editor-state PUT (the "one writer" rule, §7.1).
    */
   saveOccurrenceRules: (next: OccurrenceRules) => void;
+  /**
+   * Whether the Repeated Word Check feature is enabled in this environment
+   * (the `repeatedWordCheck` flag; feature-flags proposal D5/D7). Defaults to
+   * `true` so existing callers/tests are unaffected. When `false`, the
+   * once-per-session `GET /self/settings` probe is **skipped entirely** — the
+   * whole Checks UI is hidden behind the flag, so there is no reason to fetch
+   * user settings for a feature that can't be seen (W2). The global stores stay
+   * at their fail-closed defaults (empty rules, `globalIgnoresAvailable=false`)
+   * and `settingsProbeResolved` stays `false`, which is consistent with the
+   * check query being suppressed by the same flag upstream.
+   */
+  enabled?: boolean;
 }
 
 export interface UseSuppressionsResult {
@@ -272,6 +295,7 @@ export interface UseSuppressionsResult {
 export const useSuppressions = ({
   occurrenceRules,
   saveOccurrenceRules,
+  enabled = true,
 }: UseSuppressionsParams): UseSuppressionsResult => {
   // --- Global half: probe once per session, then hold the map locally so we
   // can apply optimistic updates and roll them back without a query refetch. ---
@@ -300,7 +324,12 @@ export const useSuppressions = ({
   const committedGlobalRef = useRef<GlobalRules>({});
 
   useEffect(() => {
-    if (probeStarted.current) {
+    // Skip the probe entirely while the feature is off (W2): the Checks UI is
+    // hidden behind the flag, so fetching `/self/settings` for it would be a
+    // wasted request in every environment where fluent-ai isn't wired. The
+    // probe fires once if/when `enabled` becomes true (e.g. the flag resolves
+    // after mount), guarded by `probeStarted` so it still runs at most once.
+    if (!enabled || probeStarted.current) {
       return;
     }
     probeStarted.current = true;
@@ -313,7 +342,7 @@ export const useSuppressions = ({
       committedGlobalRef.current = available ? rules : {};
       setSettingsProbeResolved(true);
     });
-  }, []);
+  }, [enabled]);
 
   // Latest occurrence map for read-modify-write inside callbacks without
   // making the callbacks change identity on every keystroke.
diff --git a/src/features/dashboard/user/UserHomePage.test.tsx b/src/features/dashboard/user/UserHomePage.test.tsx
index 62d6d2f8..4a473ded 100644
--- a/src/features/dashboard/user/UserHomePage.test.tsx
+++ b/src/features/dashboard/user/UserHomePage.test.tsx
@@ -37,7 +37,7 @@ const assignment = (overrides: Partial<UserChapterAssignment>): UserChapterAssig
   bibleName: 'Test Bible',
   chapterStatus: 'draft',
   targetLanguage: 'Spanish',
-  targetLanguageCode: 'spa',
+  targetLangCode: 'spa',
   sourceLangCode: 'eng',
   bookCode: 'GEN',
   bookId: 1,
diff --git a/src/features/flags/FeatureFlagsDiagnostics.tsx b/src/features/flags/FeatureFlagsDiagnostics.tsx
index 51e0b8be..9a344d6c 100644
--- a/src/features/flags/FeatureFlagsDiagnostics.tsx
+++ b/src/features/flags/FeatureFlagsDiagnostics.tsx
@@ -28,8 +28,9 @@ export const FeatureFlagsDiagnostics: React.FC = () => {
 
   // Enumerate whatever the effective map contains rather than a hard-coded list,
   // so a flag added later (API + `FeatureName`) appears here automatically with
-  // no change to this page.
-  const entries = Object.entries(features) as Array<[string, boolean]>;
+  // no change to this page. `features` is `Record<FeatureName, boolean>`, so
+  // `Object.entries` is already `[string, boolean][]` — no cast needed (W4).
+  const entries = Object.entries(features);
 
   return (
     <div className='mx-auto max-w-2xl px-8 py-10'>
diff --git a/src/features/flags/useFeatureFlags.ts b/src/features/flags/useFeatureFlags.ts
index 5c2f97fb..abc34144 100644
--- a/src/features/flags/useFeatureFlags.ts
+++ b/src/features/flags/useFeatureFlags.ts
@@ -41,8 +41,15 @@ export const fetchFeatureFlags = async (): Promise<Features> => {
     throw new Error(`Failed to fetch feature flags (HTTP ${res.status})`);
   }
 
+  // The body is cast, not schema-validated, so a 200 with a malformed/partial
+  // `features` object (missing a known key, or `features` absent entirely) must
+  // NOT be trusted verbatim — that would let an unknown flag read `undefined`
+  // (falsy, but off-contract) and bypass the fail-closed guarantee. Spread the
+  // known-off defaults FIRST, then overlay whatever the API sent: every known
+  // flag is present as a boolean, and only keys the API explicitly reports as
+  // `true` turn on (W3, D7).
   const body = (await res.json()) as FeaturesResponse;
-  return body.features;
+  return { ...failClosedFeatures(), ...body.features };
 };
 
 export interface UseFeatureFlagsResult {
diff --git a/src/features/projects/components/ProjectDetailPage.tsx b/src/features/projects/components/ProjectDetailPage.tsx
index 4518c08b..8ecc3052 100644
--- a/src/features/projects/components/ProjectDetailPage.tsx
+++ b/src/features/projects/components/ProjectDetailPage.tsx
@@ -331,7 +331,7 @@ export const ProjectDetailPage: React.FC<ProjectDetailPageProps> = ({
         // ISO 639-3 code for the repeated-words check's `lang_code` (BUG #3):
         // the progress endpoint now surfaces this so the PM path no longer
         // sends "<unknown>".
-        targetLanguageCode: assignment.targetLanguageCode,
+        targetLangCode: assignment.targetLangCode,
         bookId: assignment.bookId,
         book: assignment.bookNameEng,
         chapterStatus: assignment.status,
diff --git a/src/lib/types.ts b/src/lib/types.ts
index 4486e833..fc21397a 100644
--- a/src/lib/types.ts
+++ b/src/lib/types.ts
@@ -89,10 +89,10 @@ export interface ChapterAssignmentProgress {
   /**
    * ISO 639-3 target language CODE, e.g. "eng" (sent as the repeated-words
    * check's `lang_code`). Required so the PM "open chapter" path can populate
-   * `ProjectItem.targetLanguageCode`; if it were optional the field could be
+   * `ProjectItem.targetLangCode`; if it were optional the field could be
    * silently dropped and the check would send "<unknown>" (BUG #3).
    */
-  targetLanguageCode: string;
+  targetLangCode: string;
   bookNameEng: string;
   chapterNumber: number;
   assignedUser: AssignmentUser | null;
@@ -132,7 +132,7 @@ export interface ProjectItem {
    * degrades to "<unknown>" at runtime if the value is somehow empty, rather
    * than crashing.
    */
-  targetLanguageCode: string;
+  targetLangCode: string;
   bookId: number;
   book: string;
   chapterStatus: string;
@@ -268,7 +268,7 @@ export interface UserChapterAssignment {
   /** Human-readable target language display NAME, e.g. "English". */
   targetLanguage: string;
   /** ISO 639-3 target language CODE, e.g. "eng" (sent as the check's lang_code). */
-  targetLanguageCode: string;
+  targetLangCode: string;
   sourceLangCode: string;
   bookCode: string;
   bookId: number;