feat: copy management — react-i18next adopted, Settings migrated (goal 0032)

docs/SPEC.md

@@ -286,6 +286,37 @@ and [`docs/adr/0002-cicd-pipeline-phased-rollout.md`](adr/0002-cicd-pipeline-pha
   `*.module.css` files (`RunbookView.module.css` shared by `RunbookView`
   and `ActivityView`, which already reused the same card/list visual
   language). `frontend/public/style.css` deleted. `LOCKED`
+- **Copy management: `react-i18next` v17.0.11 + `i18next` v26.3.6
+  adopted, namespace-per-bounded-context JSON — `LOCKED` (partial
+  migration, goal 0032).** Owner-observed: 40 of 72 `.tsx` files carry
+  inline hardcoded copy, no i18n library present. Research first
+  framed this as a headless-CMS question (Decap/Sveltia/Keystatic/
+  Tina) — rejected on architecture fit, not quality: every git-native
+  CMS candidate needs either a hosted OAuth intermediary or a locally-
+  running backend daemon, disqualified outright by §1.1's no-hosted-
+  service/no-second-daemon constraint. The owner reframed it mid-
+  research: this is plain i18n (key → string JSON, no authoring UI, no
+  CMS product), not localization or centralized authoring — so the
+  real adopt-vs-build call is a plain i18n library, and `react-i18next`
+  is MIT, zero-server, and the standard choice for React. Resources
+  (`frontend/src/locales/en/{common,app,composition,configure,
+  views}.json`, mirroring `frontend/src`'s own bounded-context folders)
+  are imported statically and bundled at build time — no runtime
+  fetch, no network call, matching every hard constraint in §1.1.
+  Initialized once in `frontend/src/app/i18n.ts`, imported for its
+  side effect from `main.tsx`; every other folder just calls
+  `useTranslation()` from the `react-i18next` package directly, never
+  importing the init module (dependency-cruiser's existing bounded-
+  context boundaries, ADR-0012, apply unchanged). Migrated as proof:
+  `SettingsView.tsx` (`views.json`'s `settings` namespace +
+  `common.json`'s shared action verbs). The remaining ~39 files are
+  tracked debt, not silently orphaned — see `docs/goals/BACKLOG.md`'s
+  copy-management entries. `eslint-plugin-i18next` (a lint guard
+  against new hardcoded strings) was evaluated and deliberately not
+  added yet: its `no-literal-string` rule would immediately flag every
+  still-unmigrated file, failing the lint gate repo-wide rather than
+  just guarding new code — revisit once migration is far enough along
+  that the false-positive surface is small.
 - CI: GitHub Actions, all four ADR-0002 phases shipped in
   `.github/workflows/ci.yml` + `.github/workflows/release.yml`.
   `golangci-lint` v2, ESLint flat config, Vitest, `go test -race -cover`,

docs/goals/0032-copy-management.md

@@ -0,0 +1,86 @@
+# 0032 — Copy management: is Mill's UI text hardcoded, and what fixes it
+
+## Goal
+Owner-observed 2026-08-12: every string in Mill's UI is likely
+hand-typed inline JSX. Confirmed by audit: 40 of 72 `.tsx` files carry
+inline copy, zero i18n library in `frontend/package.json`, no
+locale/copy directory anywhere. Owner's explicit constraint: adopt an
+existing lightweight, **no-server** solution — "so we don't rebuild
+that locally," per `.claude/rules/architecture.md`'s adopt-over-
+hand-roll default and SPEC §1.1's no-hosted-service-dependency hard
+constraint.
+
+**Reframed by the owner mid-research (2026-08-12): this is plain i18n,
+not a CMS.** "Authoring a json for [copy] locally is fine just as long
+as we don't hardcode text and key to value etc which is i18n" —
+confirms the research instinct that "headless CMS" was the wrong frame
+for a desktop app's inline copy. The actual want: key→value JSON files
++ an i18n library, no CMS product, no authoring UI layer. Locks the
+research's localization-vs-centralization question in favor of
+straight i18n tooling (react-i18next/react-intl-class options) — the
+CMS candidates (Decap/Sveltia/Keystatic/Tina) are answering a website
+question Mill doesn't have.
+
+## Plan
+1. [x] Research DONE 2026-08-12 (all sources primary, GitHub/npm/docs
+   verified live). Verdict: "headless CMS" is the wrong frame — every
+   git-native CMS candidate (Decap, Sveltia, Keystatic, Tina) requires
+   either a hosted OAuth intermediary or a locally-running backend
+   daemon, disqualified by SPEC §1.1's no-hosted-service/no-second-
+   daemon constraint regardless of maintenance health (all four are
+   healthy — rejected on architecture fit, not quality). This is
+   copy-centralization, not localization, solved by a plain i18n
+   library. **Adopt: react-i18next v17.0.11 + i18next v26.3.6 (both
+   MIT, actively maintained, zero server)**, namespace-per-view JSON
+   (i18next's own documented convention) mirroring Mill's existing
+   frontend/src/{app,composition,configure,shared,views} structure.
+   Rejected explicitly, not to be re-proposed: Decap/Sveltia CMS (OAuth
+   intermediary or decap-server daemon required), Keystatic
+   (Next.js/Astro/Remix only, no plain-Vite path), TinaCMS (self-hosted
+   path needs its own GraphQL backend + DB adapter — "Local Mode"
+   doesn't remove this).
+2. [x] Capability map DONE 2026-08-12: today's need is English-only
+   copy centralization (no language switcher, no plan for one yet);
+   the realistic future is (a) actual localization if Mill ever ships
+   non-English UI and (b) faster live-copy iteration without touching
+   JSX. Schema decided once: namespace-per-bounded-context JSON
+   (`frontend/src/locales/en/{common,app,composition,configure,
+   views}.json`, mirroring `frontend/src`'s own folders), nested
+   per-view keys inside each namespace (e.g. `views.json`'s
+   `settings.*`) rather than one file per component — i18next's own
+   documented convention, and it keeps the file count bounded (5
+   files, not 72) while still scoping per bounded context if ever
+   needed later.
+3. [x] Migration scope estimate DONE 2026-08-12: staged, not one PR —
+   this pass ships the library + init + ONE real slice (Settings) as
+   proof; the remaining ~39 files are tracked as four Standing
+   tech-debt entries in `docs/goals/BACKLOG.md` (`app/`, `composition/`,
+   `configure/`, `views/` minus Settings), each with its own DoR/DoD
+   per `.claude/rules/delivery-discipline.md` — never a bare TODO.
+4. [x] Build DONE 2026-08-12 (partial, by design): `react-i18next`
+   v17.0.11 + `i18next` v26.3.6 installed; `frontend/src/app/i18n.ts`
+   (init, wired into `main.tsx`); `SettingsView.tsx` fully migrated
+   (`views.json`'s `settings` namespace + `common.json`'s shared action
+   verbs) as the proof-of-pattern slice, verified against the existing
+   `e2e/settings.spec.ts` (unchanged assertions still pass — translated
+   text matches the original English exactly) plus a new
+   `app/i18n.test.ts` unit test (init loads, `t()` resolves a known key,
+   interpolation works). `eslint-plugin-i18next` evaluated as a guard
+   against new hardcoded strings — deliberately NOT added yet: its
+   `no-literal-string` rule would fail the lint gate across the ~39
+   still-unmigrated files rather than just guarding new code; revisit
+   once the Standing tech-debt entries close the gap.
+
+## Acceptance
+A capability map and adopt-vs-build decision recorded in SPEC (done —
+`docs/SPEC.md`'s copy-management bullet, §1.3); the chosen pattern
+lands with a working example, not just a library install (done —
+Settings, proven live + by test); remaining hardcoded copy is tracked,
+not silently orphaned (done — four Standing tech-debt entries in
+`docs/goals/BACKLOG.md`, each independently DoR/DoD-shaped).
+
+**Status: partially delivered 2026-08-12, intentionally left OPEN, not
+archived** — per this goal's own Plan step 3, the ~39-file migration is
+staged, ongoing debt, not a silently-dropped scope. Continue by picking
+up the next Standing tech-debt entry (any of `app/`, `composition/`,
+`configure/`, `views/`), not by re-researching this goal from scratch.

docs/goals/BACKLOG.md

@@ -108,6 +108,16 @@ this pipeline and on this code)**
    2026-08-12: the guardrailed AI-node family (n8n/Make/Zapier/
    Dify taxonomy convergence), Mill's category-defining capability;
    research not started.
+10. [ ] [0032 — Copy management](0032-copy-management.md) — owner-observed
+    2026-08-12: 40 of 72 `.tsx` files carried inline hardcoded copy, no
+    i18n library. Research + adopt decision DONE and library landed
+    2026-08-12: `react-i18next` + `i18next`, namespace-per-bounded-
+    context JSON (`frontend/src/locales/en/`), init wired in
+    `app/i18n.ts`; `SettingsView.tsx` migrated as the proof-of-pattern
+    slice. Deliberately left OPEN, not archived: the remaining ~39
+    files are real, staged debt — see the four Standing tech-debt
+    entries above (`app/`, `composition/`, `configure/`, `views/`
+    minus Settings) for the rest of the migration.
 
 **Ratified 2026-08-10 (owner): three groups, A→B→C. 0001 stays standing
 live-review material, interleaved during owner reviews, not a lane.**
@@ -138,6 +148,10 @@ live-review material, interleaved during owner reviews, not a lane.**
 - [ ] [0021 — MCP dogfood gap closure](0021-mcp-dogfood-gap-closure.md) (owner-mandated 2026-08-11: orchestrator live-probes the MCP surface against the bank use cases, logs ranked gaps, fixes graduate out; phase 1 done — 4 gaps + 1 confirmed-by-design)
 - [ ] Workflow pins/favorites (tech debt, split from goal 0015's remainder 2026-08-12) — no pin/favorite concept exists anywhere in Mill today (grepped before scoping it out); needs its own small schema decision (which store owns a pin list, per-workflow or a plain ID set) before any build — deliberately not invented ad hoc under 0015's frecency-only ship. Quick Panel's workflow list sorts by frequency alone until this lands.
 - [ ] ⌘?/⌘/ multi-binding keybinding alias (tech debt, split from goal 0015's remainder 2026-08-12) — the owner's goal-0015 "bind ⌘? (and/or ⌘/) to open the palette too" ask needs a command to carry more than one `KeyCombo`; today's registry (`shared/commands.ts`) is 1:1 (`defaultBinding: KeyCombo | null`). Needs a real schema call (array vs. a small alias table) before it's buildable — real data-model infrastructure, not a self-contained UI change.
+- [ ] Copy-management migration — `app/` (tech debt, split from goal 0032 2026-08-12) — extract `app/`'s remaining hardcoded JSX copy (App.tsx's shell chrome, QuickPanel/QuickPanelApp, ApprovalPromptApp, workflowFrecency-adjacent UI, etc. — ~11 files carry inline strings) into `frontend/src/locales/en/app.json` (already scaffolded, currently `{}`) following `SettingsView.tsx`'s established pattern (`useTranslation()`/`t()`, namespace-per-bounded-context). DoR: read `docs/goals/0032-copy-management.md` for the locked i18n pattern before starting — no new library/schema decision needed, this is mechanical extraction. DoD: every `app/*.tsx` file free of inline user-facing string literals in JSX (aria-labels included), `app.json` populated, existing e2e specs touching `app/` still pass unchanged (translated text must match original English exactly).
+- [ ] Copy-management migration — `composition/` (tech debt, split from goal 0032 2026-08-12) — same extraction as above, scoped to `composition/`'s ~22 files with inline copy (canvas node cards, palette, validation messages, trigger/schedule UI) into `frontend/src/locales/en/composition.json`. Largest of the four remaining slices — consider whether it splits further once started (per-node-type vs. whole-folder) rather than treating it as one atomic PR. Same DoR/DoD shape as the `app/` entry above.
+- [ ] Copy-management migration — `configure/` (tech debt, split from goal 0032 2026-08-12) — same extraction, scoped to `configure/`'s ~14 files (connector/list/MCP-server forms, OpenAPI synth UI) into `frontend/src/locales/en/configure.json`. Same DoR/DoD shape as the `app/` entry above.
+- [ ] Copy-management migration — `views/` minus Settings (tech debt, split from goal 0032 2026-08-12) — same extraction, scoped to the remaining `views/*.tsx` files (Settings already migrated into `views.json`'s `settings` namespace) — add sibling namespaces (e.g. `views.json`'s `home`, `activity`, etc. keys) per view. Same DoR/DoD shape as the `app/` entry above. Once all four of these land, revisit `eslint-plugin-i18next`'s `no-literal-string` rule (evaluated and deliberately deferred in goal 0032 — see `docs/SPEC.md`'s copy-management bullet) as a guard against regression.
 
 **Delivered**
 - [x] [0003 — MCP authoring live dogfood](archive/0003-mcp-authoring-dogfood.md) — 2026-08-10

frontend/package.json

@@ -23,11 +23,13 @@
     "cronstrue": "^3.24.0",
     "elkjs": "^0.12.0",
     "genson-js": "^0.0.8",
+    "i18next": "^26.3.6",
     "papaparse": "^5.5.4",
     "randexp": "^0.5.3",
     "react": "^18.2.0",
     "react-dom": "^18.2.0",
     "react-dropzone": "^20.1.0",
+    "react-i18next": "^17.0.11",
     "react-is": "^19.2.8",
     "react-querybuilder": "^8.22.4",
     "recharts": "^3.10.1",

frontend/src/app/i18n.test.ts

@@ -0,0 +1,29 @@
+import { describe, expect, it } from 'vitest'
+import i18n from './i18n'
+
+// docs/goals/0032-copy-management.md's proof-of-pattern slice: confirms
+// the init actually wires resources up (a config typo or a missing
+// namespace in the `ns` list would otherwise only surface as a blank
+// string live, in whichever component happened to call t() first).
+describe('i18n init', () => {
+  it('initializes synchronously with English as the active language', () => {
+    expect(i18n.isInitialized).toBe(true)
+    expect(i18n.language).toBe('en')
+  })
+
+  it('resolves a known key from the views namespace (Settings slice)', () => {
+    expect(i18n.t('settings.title', { ns: 'views' })).toBe('Settings')
+  })
+
+  it('resolves a known key from the common namespace via the ns-prefixed form', () => {
+    expect(i18n.t('common:actions.change')).toBe('Change')
+  })
+
+  it('interpolates variables into a templated key', () => {
+    expect(i18n.t('settings.updates.updateAvailable', { ns: 'views', version: '1.2.3' })).toBe('Update available: v1.2.3')
+  })
+
+  it('falls back to the key itself for an unknown key, never throwing', () => {
+    expect(() => i18n.t('settings.doesNotExist', { ns: 'views' })).not.toThrow()
+  })
+})

frontend/src/app/i18n.ts

@@ -0,0 +1,48 @@
+import i18n from 'i18next'
+import { initReactI18next } from 'react-i18next'
+import common from '../locales/en/common.json'
+import app from '../locales/en/app.json'
+import composition from '../locales/en/composition.json'
+import configure from '../locales/en/configure.json'
+import views from '../locales/en/views.json'
+
+// docs/goals/0032-copy-management.md's locked research verdict:
+// react-i18next + i18next, namespace-per-bounded-context JSON --
+// mirroring frontend/src's own app/composition/configure/shared/views
+// folders -- chosen over every git-native headless-CMS candidate
+// (Decap/Sveltia/Keystatic/Tina), all of which need either a hosted
+// OAuth intermediary or a locally-running backend daemon, disqualified
+// by SPEC §1.1's no-hosted-service/no-second-daemon constraint. This
+// is copy-CENTRALIZATION (key -> string JSON, no authoring UI, no
+// CMS product), not localization -- there is no language switcher and
+// no plan for one yet; English is the only shipped locale. Resources
+// are imported statically and bundled at build time (Vite's own
+// resolveJsonModule support), never fetched at runtime, so this adds
+// zero network calls and zero server, matching Mill's own hard
+// constraints.
+//
+// Initialized once here and imported for its side effect from
+// app/main.tsx, before the first render -- react-i18next's
+// useTranslation() hook reads the already-initialized global i18next
+// instance from any bounded-context folder without importing this
+// module directly (app/ is the only folder allowed to import this,
+// per .claude/rules/frontend.md's dependency-cruiser boundaries; every
+// other folder just calls useTranslation() from the 'react-i18next'
+// package).
+void i18n.use(initReactI18next).init({
+  resources: {
+    en: { common, app, composition, configure, views },
+  },
+  lng: 'en',
+  fallbackLng: 'en',
+  defaultNS: 'common',
+  ns: ['common', 'app', 'composition', 'configure', 'views'],
+  interpolation: {
+    // React already escapes interpolated values when rendering JSX --
+    // i18next's own default (HTML-escaping) would double-escape them.
+    escapeValue: false,
+  },
+  returnNull: false,
+})
+
+export default i18n

frontend/src/app/main.tsx

@@ -1,5 +1,6 @@
 import React from 'react'
 import ReactDOM from 'react-dom/client'
+import './i18n'
 import './index.css'
 import '@primer/primitives/dist/css/primitives.css'
 import '@primer/primitives/dist/css/functional/themes/light.css'

frontend/src/locales/en/app.json

@@ -0,0 +1 @@
+{}

frontend/src/locales/en/common.json

@@ -0,0 +1,6 @@
+{
+  "actions": {
+    "change": "Change",
+    "clear": "Clear"
+  }
+}

frontend/src/locales/en/composition.json

@@ -0,0 +1 @@
+{}

frontend/src/locales/en/configure.json

@@ -0,0 +1 @@
+{}