PR：i18n settings 拆分与翻译质量门禁（含 messages no-emoji）

CONTRIBUTING.md

@@ -112,6 +112,12 @@ bun run typecheck
 # 如果更改影响运行逻辑，执行端到端验证或 bun run test
 ```
 
+### i18n 变更（翻译质量/抽查）
+
+如果 PR 涉及 i18n 文案（尤其是 `settings` / `dashboard` / `myUsage`），请遵循：
+- 规则说明：`docs/i18n-translation-quality.md`
+- PR Checklist：`docs/i18n-pr-checklist.md`
+
 CI 会在 PR 上运行 `Docker Build Test`（见 `.github/CI_CD_SETUP.md`）。如需验证容器构建，可本地执行：
 
 ```bash

biome.json

@@ -1,5 +1,5 @@
 {
-  "$schema": "https://biomejs.dev/schemas/2.3.10/schema.json",
+  "$schema": "https://biomejs.dev/schemas/2.3.11/schema.json",
   "vcs": {
     "enabled": true,
     "clientKind": "git",

docs/i18n-pr-checklist.md

@@ -0,0 +1,36 @@
+# i18n PR checklist
+
+> 中文对照版: [i18n-pr-checklist.zh-CN.md](i18n-pr-checklist.zh-CN.md)
+
+This checklist is for changes that affect i18n messages, especially `settings`, `dashboard`, and `myUsage`.
+
+## Required (automation)
+
+- [ ] Run placeholder audit (scoped):
+  - `bun run i18n:audit-placeholders`
+- [ ] If the PR is meant to eliminate placeholders, ensure fail mode is clean:
+  - `bun run i18n:audit-placeholders:fail`
+- [ ] Run messages no-emoji audit (fail mode):
+  - `bun run i18n:audit-messages-no-emoji:fail`
+- [ ] Run unit tests relevant to i18n/settings split:
+  - `bunx vitest run tests/unit/i18n/settings-split-guards.test.ts`
+  - `bunx vitest run tests/unit/i18n/settings-index-modules-load.test.ts`
+
+## Required (review evidence)
+
+- [ ] Include in PR description:
+  - the audit output diff (before/after) or a short summary (counts by locale + key modules)
+  - any allowlist changes with reasons (keep allowlist minimal and auditable)
+
+## Required (manual spotcheck)
+
+For at least `ja` and `zh-TW`:
+- [ ] Settings pages (key areas): provider list, provider form, request filters, notifications
+- [ ] Dashboard key widgets
+- [ ] My Usage page
+
+Attach screenshots (or provide a local path) for the key pages above.
+
+## Rules to follow
+
+See `docs/i18n-translation-quality.md` for R1/R2/R3 rules and allowlist conventions.

docs/i18n-pr-checklist.zh-CN.md

@@ -0,0 +1,36 @@
+# i18n PR 检查清单
+
+> English: [i18n-pr-checklist.md](i18n-pr-checklist.md)
+
+此清单适用于会影响 i18n messages 的变更，尤其是 `settings`、`dashboard`、`myUsage`。
+
+## 必做（自动化）
+
+- [ ] 运行 placeholder 审计（scoped）：
+  - `bun run i18n:audit-placeholders`
+- [ ] 若该 PR 目标是清零 placeholders，确保 fail 模式无命中：
+  - `bun run i18n:audit-placeholders:fail`
+- [ ] 运行 messages no-emoji 审计（fail 模式）：
+  - `bun run i18n:audit-messages-no-emoji:fail`
+- [ ] 运行与 i18n/settings split 相关的单元测试：
+  - `bunx vitest run tests/unit/i18n/settings-split-guards.test.ts`
+  - `bunx vitest run tests/unit/i18n/settings-index-modules-load.test.ts`
+
+## 必做（Review 证据）
+
+- [ ] 在 PR 描述中包含：
+  - 审计输出 diff（before/after）或简短摘要（按 locale 统计 + 关键模块）
+  - 如有 allowlist 变更，说明原因（保持 allowlist 最小且可审计）
+
+## 必做（人工抽查）
+
+至少覆盖 `ja` 与 `zh-TW`：
+- [ ] Settings 页面（关键区域）：provider list、provider form、request filters、notifications
+- [ ] Dashboard 关键组件
+- [ ] My Usage 页面
+
+为上述关键页面附上截图（或提供本地路径）。
+
+## 需要遵守的规则
+
+参见 `docs/i18n-translation-quality.md` 获取 R1/R2/R3 规则与 allowlist 约定。

docs/i18n-pr-evidence-2026-01-11.md

@@ -0,0 +1,68 @@
+# i18n no-emoji / docs zh evidence (2026-01-11)
+
+This document is intended to be copy-pasted into a PR description.
+
+## Summary
+
+- Emoji cleanup (messages JSON):
+  - Before: 20 strings contained emoji (top files: `provider-chain.json`, `settings/data.json` across locales)
+  - After: 0 hits from `rg -n --pcre2 "\\p{Extended_Pictographic}" messages`
+- Placeholder audit (zh-CN equality placeholders): OK
+- Quality gates: `lint` / `typecheck` / `test` / `build` all pass
+
+## Emoji cleanup details (messages JSON)
+
+Key locations that were cleaned (keys unchanged, only values updated):
+- `messages/<locale>/provider-chain.json`
+  - `provider-chain.timeline.circuitTriggered`
+  - `provider-chain.timeline.systemErrorNote`
+- `messages/<locale>/settings/data.json`
+  - `settings.data.import.warningMerge`
+  - `settings.data.import.warningOverwrite`
+
+Commands:
+- Before/after scan:
+  - `rg -n --pcre2 "\\p{Extended_Pictographic}" messages`
+- Optional structured listing (prints masked preview, no emoji characters):
+  - `node scripts/audit-messages-emoji.js --format=tsv`
+
+## Placeholder audit (R1)
+
+- `bun run i18n:audit-placeholders:fail`
+- Expected output:
+  - `OK: no zh-CN placeholder candidates found in split settings.`
+
+## No-emoji gate (R4)
+
+- Script (codepoints only, no emoji printing):
+  - `bun run i18n:audit-messages-no-emoji:fail`
+- Regression test (part of `bun run test`):
+  - `tests/unit/i18n/audit-messages-no-emoji-script.test.ts`
+
+## Full regression commands
+
+- `bun run lint`
+- `bun run typecheck`
+- `bun run test`
+- `bun run build`
+
+## Related commits (local)
+
+- `564ab845` chore: add messages emoji audit script (I18NE-010)
+- `aaa9fc7d` fix: remove emoji from messages warnings (I18NE-040)
+- `80d20686` test: add messages no-emoji audit gate (I18NE-050)
+- `2ee38f59` docs: add zh-CN i18n docs (I18NE-020)
+- `44eeb5e9` docs: add messages no-emoji rule (I18NE-060)
+- `92ebaf0e` chore: run full regression checks (I18NE-070)
+
+## Manual spotcheck (ja / zh-TW)
+
+Due to environment limitations (no GUI/browser automation in this run), screenshots are not attached here.
+
+Recommended steps:
+1. Start dev server: `bun run dev` (port 13500)
+2. Open pages for both `ja` and `zh-TW` locales:
+   - Settings: `/settings` (providers list/form, request filters, notifications)
+   - Dashboard: `/dashboard` (key widgets)
+   - My Usage: `/my-usage`
+3. Attach screenshots to the PR (or provide local file paths) and label each with locale + route.

docs/i18n-settings-split.md

@@ -0,0 +1,37 @@
+# i18n settings split
+
+> 中文对照版: [i18n-settings-split.zh-CN.md](i18n-settings-split.zh-CN.md)
+
+This repository splits `messages/<locale>/settings.json` into smaller JSON chunks under `messages/<locale>/settings/`.
+
+## Layout
+- `messages/<locale>/settings/*.json`: settings top-level object parts
+- `messages/<locale>/settings/strings.json`: top-level string keys that belong directly under `settings`
+- `messages/<locale>/settings/providers/*.json`: `settings.providers` object parts
+- `messages/<locale>/settings/providers/strings.json`: provider-level string keys
+- `messages/<locale>/settings/providers/form/*.json`: `settings.providers.form` object parts
+- `messages/<locale>/settings/providers/form/strings.json`: provider form string keys
+
+Runtime composition happens in `messages/<locale>/settings/index.ts` and is imported by `messages/<locale>/index.ts`.
+
+## Verification
+- Translation quality rules and audit commands:
+  - `docs/i18n-translation-quality.md`
+
+- Sync keys across locales (canonical: zh-CN):
+  - `node scripts/sync-settings-keys.js`
+
+- Unit tests:
+  - `bun run test`
+
+- Scoped coverage for split-related modules:
+  - `bunx vitest run --coverage --coverage.include=scripts/sync-settings-keys.js --coverage.include=messages/**/settings/index.ts`
+
+- Typecheck:
+  - `bun run typecheck`
+
+- Lint:
+  - `bun run lint`
+
+- Production build:
+  - `bun run build`

docs/i18n-settings-split.zh-CN.md

@@ -0,0 +1,39 @@
+# i18n settings split（拆分说明）
+
+> English: [i18n-settings-split.md](i18n-settings-split.md)
+
+本仓库将 `messages/<locale>/settings.json` 拆分为更小的 JSON 文件，存放在 `messages/<locale>/settings/` 目录下。
+
+## 目录结构（Layout）
+
+- `messages/<locale>/settings/*.json`: settings 顶层对象的各个子模块
+- `messages/<locale>/settings/strings.json`: 直接属于 `settings` 顶层的字符串 key
+- `messages/<locale>/settings/providers/*.json`: `settings.providers` 对象拆分后的子模块
+- `messages/<locale>/settings/providers/strings.json`: provider 级别的字符串 key
+- `messages/<locale>/settings/providers/form/*.json`: `settings.providers.form` 对象拆分后的子模块
+- `messages/<locale>/settings/providers/form/strings.json`: provider form 的字符串 key
+
+运行时拼装发生在 `messages/<locale>/settings/index.ts`，并由 `messages/<locale>/index.ts` 引入。
+
+## 验证（Verification）
+
+- 翻译质量规则与审计命令：
+  - `docs/i18n-translation-quality.md`
+
+- 跨 locale 同步 key（canonical: zh-CN）：
+  - `node scripts/sync-settings-keys.js`
+
+- 单元测试：
+  - `bun run test`
+
+- 针对 split 相关模块的 scoped coverage：
+  - `bunx vitest run --coverage --coverage.include=scripts/sync-settings-keys.js --coverage.include=messages/**/settings/index.ts`
+
+- Typecheck：
+  - `bun run typecheck`
+
+- Lint：
+  - `bun run lint`
+
+- Production build：
+  - `bun run build`

docs/i18n-translation-quality.md

@@ -0,0 +1,64 @@
+# i18n translation quality rules (R1/R2/R3)
+
+> 中文对照版: [i18n-translation-quality.zh-CN.md](i18n-translation-quality.zh-CN.md)
+
+This document defines the **scope** and **executable rules** for i18n translation quality in this repo.
+Downstream scripts and review checklists should follow this document as the source of truth.
+
+## Scope
+
+Must-translate scope (at least):
+- `settings` (split settings messages under `messages/<locale>/settings/`)
+- `dashboard` (`messages/<locale>/dashboard.json`)
+- `myUsage` (`messages/<locale>/myUsage.json`)
+
+Locales: `zh-CN` is canonical. Other supported locales: `en`, `ja`, `ru`, `zh-TW`.
+
+## Rule R1: No zh-CN placeholders in non-canonical locales
+
+For any non-canonical locale:
+- If a leaf string **equals** the `zh-CN` leaf string at the same key path, and the `zh-CN` value contains Han characters, it is treated as a **placeholder candidate**.
+- Placeholder candidates should be **fixed** (translated), or **explicitly allowlisted** with a documented reason.
+
+Executable check:
+- `bun run i18n:audit-placeholders`
+- To fail the command on any findings: add `--fail`.
+  - `bun run i18n:audit-placeholders:fail`
+
+Allowlist (auditable, minimal):
+- `scripts/audit-settings-placeholders.allowlist.json`
+- Supported filters: `key`, `keyPrefix`, `keyRegex`, `valueRegex`, plus `glossary` terms.
+
+## Rule R2: Placeholders/tokens must be preserved
+
+When updating translations, **do not change**:
+- keys / JSON structure
+- placeholder tokens (e.g. `{name}`, `{count}`, `{resetTime}`)
+- URL / command snippets unless intentionally translated and verified safe
+
+Recommended verification:
+- unit tests under `tests/unit/i18n/`
+- spot-check affected UI pages for the locale (see the PR checklist)
+
+## Rule R3: Glossary and consistent terminology
+
+Maintain a short glossary for terms that should be consistent across locales (brand, model names, product terms).
+
+Initial glossary (expand as needed, but keep it minimal and reviewed):
+- Provider / Model / API / HTTP/2
+- Claude / OpenAI / Codex (names should not be translated)
+
+## Rule R4: No emoji in messages JSON
+
+`messages/**/*.json` must not contain emoji characters.
+
+Executable check:
+- `bun run i18n:audit-messages-no-emoji:fail`
+
+Notes:
+- The audit output prints file path + key path + Unicode codepoints (without printing emoji characters).
+
+## Notes
+
+- Prefer fixing translations over expanding allowlists.
+- Every allowlist entry must have a clear reason in the allowlist file (and ideally referenced in the PR description).

docs/i18n-translation-quality.zh-CN.md

@@ -0,0 +1,64 @@
+# i18n 翻译质量规则（R1/R2/R3）
+
+> English: [i18n-translation-quality.md](i18n-translation-quality.md)
+
+本文档定义本仓库 i18n 翻译质量的 **scope** 与 **可执行规则**。
+下游脚本与 review checklist 应以本文档为真相源。
+
+## Scope
+
+必须翻译的范围（至少包含）：
+- `settings`（拆分后的 settings messages，位于 `messages/<locale>/settings/`）
+- `dashboard`（`messages/<locale>/dashboard.json`）
+- `myUsage`（`messages/<locale>/myUsage.json`）
+
+Locales：`zh-CN` 为 canonical。其他支持的 locales：`en`、`ja`、`ru`、`zh-TW`。
+
+## Rule R1：非 canonical locale 禁止出现 zh-CN placeholder
+
+对于任意非 canonical locale：
+- 若某个 leaf string 在相同 key path 下 **等于** `zh-CN` 的 leaf string，且 `zh-CN` 的值包含汉字，则视为 **placeholder candidate**。
+- placeholder candidates 应被 **修复**（翻译），或以明确理由 **加入 allowlist**。
+
+可执行检查：
+- `bun run i18n:audit-placeholders`
+- 如需在任意命中时让命令失败：添加 `--fail`。
+  - `bun run i18n:audit-placeholders:fail`
+
+Allowlist（可审计、保持最小）：
+- `scripts/audit-settings-placeholders.allowlist.json`
+- 支持的过滤器：`key`、`keyPrefix`、`keyRegex`、`valueRegex`，以及 `glossary` terms。
+
+## Rule R2：必须保留 placeholders/tokens
+
+更新翻译时，**不要改动**：
+- keys / JSON structure
+- placeholder tokens（例如 `{name}`、`{count}`、`{resetTime}`）
+- URL / command snippets（除非明确要翻译且已验证安全）
+
+建议的验证方式：
+- `tests/unit/i18n/` 下的单元测试
+- 对受影响 locale 的 UI 页面做 spot-check（见 PR checklist）
+
+## Rule R3：Glossary 与术语一致性
+
+维护一份简短 glossary，用于跨 locale 保持一致的术语（品牌、模型名、产品术语等）。
+
+初始 glossary（按需扩展，但保持最小并经过 review）：
+- Provider / Model / API / HTTP/2
+- Claude / OpenAI / Codex（名称不翻译）
+
+## Rule R4：messages JSON 禁止 Emoji
+
+`messages/**/*.json` 中不允许出现 Emoji 字符。
+
+可执行检查：
+- `bun run i18n:audit-messages-no-emoji:fail`
+
+说明：
+- 审计输出包含文件路径 + key path + Unicode codepoints（不会直接打印 Emoji 字符本身）。
+
+## Notes
+
+- 优先修复翻译，而不是扩展 allowlists。
+- 每条 allowlist 记录都必须在 allowlist 文件中写清原因（最好在 PR 描述中也有引用）。

messages/en/index.ts

@@ -11,7 +11,7 @@ import notifications from "./notifications.json";
 import providerChain from "./provider-chain.json";
 import providers from "./providers.json";
 import quota from "./quota.json";
-import settings from "./settings.json";
+import settings from "./settings";
 import ui from "./ui.json";
 import usage from "./usage.json";
 import users from "./users.json";

messages/en/provider-chain.json

@@ -86,7 +86,7 @@
     "remaining": "{count} attempts remaining",
     "status": "Status",
     "alreadyBroken": "Circuit already broken",
-    "circuitTriggered": "⚠️ Circuit breaker triggered",
+    "circuitTriggered": "Warning: Circuit breaker triggered",
     "errorDetails": "Error Details",
     "systemError": "Network/System Error",
     "systemErrorFailed": "Network/System Error (Attempt {attempt})",
@@ -97,7 +97,7 @@
     "errorMeaning": "Meaning: {meaning}",
     "meaning": "Meaning",
     "notCountedInCircuit": "This error is not counted in provider circuit breaker",
-    "systemErrorNote": "ℹ️ This error is not counted in provider circuit breaker",
+    "systemErrorNote": "Note: This error is not counted in provider circuit breaker",
     "reselection": "Reselecting Provider",
     "reselect": "Reselecting Provider",
     "excluded": "Excluded: {providers}",