feat(tui): harvest zh-Hans locale bundle

[Hmbown]

Summary

• harvest feat(i18n): extract Chinese translations into dedicated JSON file #3549 by @cy2311 as a clean current-main merge with contributor credit
• move Simplified Chinese UI strings into crates/tui/locales/zh-Hans.json without reducing shipped locale coverage
• keep the useful details-shortcut cleanup from the branch: visible help/footer/context copy now points to bare v; the old modified key path remains only as a compatibility alias

Verification

• cargo fmt --all
• cargo test -p codewhale-tui --bin codewhale-tui --locked localization::tests
• cargo test -p codewhale-tui --bin codewhale-tui --locked tool_details_help_documents_bare_v_without_alt_v
• cargo test -p codewhale-tui --bin codewhale-tui --locked activity_footer_hint
• cargo test -p codewhale-tui --bin codewhale-tui --locked activity_detail
• cargo test -p codewhale-tui --bin codewhale-tui --locked open_tool_details_pager

Harvested from PR #3549 by @cy2311.
Refs #3537.

[claude]

Claude finished @Hmbown's task in 3m 54s —— View job

---

Code Review

• Gather context and read changed files
• Review correctness and logic
• Check localization loading and coverage
• Check keybinding changes
• Summarize findings

---

Summary

Clean harvest. The JSON-loading approach for zh-Hans is architecturally sound, coverage is enforced by the existing test, and the bare-v shortcut restoration is correctly guarded. A few minor observations below.

---

Correctness

chinese_simplified lifetime — OK.
MAP.get(&format!("{id:?}")) returns Option<&'static String> (because MAP is a static LazyLock), and .map(String::as_str) correctly lifts that to Option<&'static str>. The return type matches.

.expect("valid zh-Hans.json translations") — acceptable.
This panics on first zh-Hans access if the JSON is malformed. For a compile-time-embedded asset that is gated by shipped_first_pack_has_no_missing_core_messages, this is the right failure mode (loudly wrong, not silently English). The alternative — unwrap_or_default() — would mask a corrupted bundle.

Bare v guard — tight.

KeyCode::Char('v')
    if key.modifiers == KeyModifiers::NONE
        && app.input.is_empty()
        && detail_target_cell_index(app).is_some() =>

All three conditions must hold before a keystroke is consumed, so typing v into a non-empty composer is unaffected. Alt+V at line 4911 and the macOS Option+V (√) legacy path at line 3901 are both preserved, maintaining backward compat.

Vim normal mode interaction — worth knowing.
The bare v arm appears before the vim-normal-mode arm in the match. If vim normal mode is active, input is empty, and a tool cell is selected, pressing v opens the details pager instead of entering visual mode. Empty-buffer visual selection does nothing meaningful anyway, so this is likely intentional — but documenting it would help a future reader who wonders why vim's v seems to misfire here.

---

Potential fragility

JSON key format coupling.
MAP.get(&format!("{id:?}")) keys are the Debug-derived variant names ("ComposerPlaceholder", etc.). If MessageId ever gets a custom Debug impl or per-variant #[doc(hidden)]-style renaming, the JSON keys silently stop matching. This is a known tradeoff of the translator-friendly JSON approach and is caught by shipped_first_pack_has_no_missing_core_messages on next CI run — acceptable, but worth a code-comment in chinese_simplified so future contributors don't reach for {id:?} without knowing why it works.

---

Minor / nits

CtxInspAltVHint variant name is stale.
Both the English string ("Open the matching card and press v for full details.") and the zh-Hans translation ("打开对应的卡片并按 v 查看详细信息。") correctly say v, but the MessageId variant is still named CtxInspAltVHint. Not a bug — the name is code-internal only — but it'll confuse a grep for AltV in the future. A follow-up rename is low-priority but worthwhile. Fix this →

open_tool_details_pager return value discarded — preexisting, not introduced here.
The new call site at the v arm is consistent with the existing call sites at lines 3902 and 4914. No #[must_use] annotation, and none of the callers act on success/failure. Not a regression.

ZhHant fallback spec vs. runtime behaviour.
LocaleSpec::fallback: "zh-Hans" for Traditional Chinese is informational only — the runtime fallback_translation() path always falls back to English, not zh-Hans. Pre-existing gap, not introduced here.

---

Tests

The PR's verification commands (localization::tests, tool_details_help_documents_bare_v_without_alt_v, activity_footer_hint, activity_detail, open_tool_details_pager) cover the main paths. shipped_first_pack_has_no_missing_core_messages acts as exhaustiveness enforcement for the JSON bundle.

LGTM with the note that the CtxInspAltVHint rename is a small follow-up worth filing.

[claude]

Claude finished @Hmbown's task in 2m 44s —— View job

---

Code Review

• Read changed files and diff
• Review localization.rs refactor
• Review zh-Hans.json bundle
• Review keybinding changes
• Review UI and test changes
• Summarize findings

---

Summary

Clean harvest. The JSON-backed chinese_simplified is architecturally sound, coverage is enforced by shipped_first_pack_has_no_missing_core_messages, and the bare-v restoration is correctly guarded. A few minor observations below — nothing blocking.

---

Correctness

chinese_simplified lifetime — OK.
MAP is static LazyLock<HashMap<String, String>>, so MAP.get(...) yields Option<&'static String> and .map(String::as_str) correctly produces Option<&'static str>. Return type matches the callers in tr().

.expect("valid zh-Hans.json translations") — appropriate.
This panics on first zh-Hans access if the JSON is malformed. For a compile-time-embedded asset already gated by shipped_first_pack_has_no_missing_core_messages, a loud panic is the right failure mode — unwrap_or_default() would silently fall back to English and mask a corrupted bundle.

zh-Hant fallback path — still works.
traditional_chinese at line 2761 has other => chinese_simplified(other)?. With chinese_simplified now returning MAP.get(&format!("{id:?}")).map(String::as_str), KbSelectedDetails and the updated CtxInspAltVHint both resolve via the JSON map for zh-Hant. The shipped_first_pack_has_no_missing_core_messages test covers this transitively.

Bare v guard — tight.

KeyCode::Char('v')
    if key.modifiers == KeyModifiers::NONE
        && app.input.is_empty()
        && detail_target_cell_index(app).is_some() =>

All three conditions must hold. If no tool/subagent cell is selected, the arm doesn't fire and v falls through to normal text input — better behavior than the old Alt+V arm which consumed the keystroke regardless of context.

Alt+V and macOS Option+V (√) preserved.
Both compat paths remain: macOS Option+V at line 3901 and Alt+V | Alt+Shift+V at line 4911. No backward-compat regression.

KbSelectedDetails coverage — complete.
Added to ALL_MESSAGE_IDS, english(), and all 5 non-zh-Hans locales (vietnamese, japanese, portuguese_brazil, spanish_latin_america, and in zh-Hans.json at line 211). The shipped_first_pack_has_no_missing_core_messages test enforces exhaustiveness at CI time.

---

Worth noting

Vim normal mode interaction.
The bare v arm (line 3967 in ui.rs) appears before the vim-normal-mode arm (line 4920). If vim normal mode is active, input is empty, and a tool cell is selected, pressing v opens the details pager rather than entering visual mode. Empty-buffer visual mode has no meaningful effect here, so this is likely fine — but a one-line comment in the arm would help a future reader who wonders why vim's v seems to fire early. Fix this →

CtxInspAltVHint variant name is stale.
Both the English string and the zh-Hans translation now correctly say v, but the MessageId variant is still CtxInspAltVHint. Code-internal only — not a user-facing bug — but will confuse a future grep AltV. Low-priority follow-up rename. Fix this →

JSON key format coupling.
MAP.get(&format!("{id:?}")) keys rely on Debug-derived variant names. If MessageId ever gets a custom Debug impl, JSON keys silently stop matching. Caught by shipped_first_pack_has_no_missing_core_messages on next CI run — acceptable, but a short comment in chinese_simplified would save a future contributor from wondering why {id:?} and not {id}.

---

Tests

The PR's verification commands cover the main paths. shipped_first_pack_has_no_missing_core_messages acts as exhaustiveness enforcement for the JSON bundle. Keybinding test correctly renamed from tool_details_help_documents_alt_v_without_bare_v → tool_details_help_documents_bare_v_without_alt_v with updated assertion that Alt+V no longer appears in the table.

LGTM. The two follow-up items (CtxInspAltVHint rename, vim comment) are low-priority and don't block merge.

[claude]

Claude finished @Hmbown's task in 2m 17s —— View job

---

Code Review

• Read diff and changed files
• Review correctness and logic
• Check localization loading and coverage
• Check keybinding changes
• Summarize findings

---

Summary

The follow-up commit (0bfe3f5) completes what the two earlier reviews flagged: the CtxInspAltVHint → CtxInspVHint rename is done, the vim-mode precedence comment is in place, and the cy2311 AUTHOR_MAP entry is added. One small residual item below, otherwise clean.

---

Correctness

chinese_simplified lifetime — OK.
MAP is static LazyLock<HashMap<String, String>>, so MAP.get(...) yields Option<&'static String> and .map(String::as_str) correctly produces Option<&'static str>. Return type matches all callers.

.expect("valid zh-Hans.json translations") — appropriate.
Panics on first zh-Hans access if the JSON is malformed. For a compile-time-embedded asset gated by shipped_first_pack_has_no_missing_core_messages, a loud panic is the right failure mode.

KbSelectedDetails coverage — complete.
Added to ALL_MESSAGE_IDS, english(), all 5 non-zh-Hans locale functions, and zh-Hans.json (line 211). CtxInspAltVHint stale variant is gone — grep -rn CtxInspAltVHint returns empty.

Bare v guard — tight.

KeyCode::Char('v')
    if key.modifiers == KeyModifiers::NONE
        && app.input.is_empty()
        && detail_target_cell_index(app).is_some() =>

All three conditions must hold. Alt+V | Alt+Shift+V compat arm at line 4914 and macOS Option+V (√) arm at line 3901 are both preserved.

Vim-mode precedence comment — added.
ui.rs line 3967 now documents why the bare v arm intentionally precedes vim-normal-mode. Previous reviews' request addressed.

---

Residual: KbToolDetailsPager is now orphaned

KbToolDetailsPager was replaced in the keybinding table by KbSelectedDetails, but the variant is still present in the enum, ALL_MESSAGE_IDS, english(), all five locale functions, and zh-Hans.json. It's dead code from the help-display perspective — nothing renders it to the user anymore. shipped_first_pack_has_no_missing_core_messages would need updating if you remove it, but it's not enforcing anything visible.

Not blocking, but worth a follow-up: Fix this →

---

Tests

Keybinding test correctly renamed tool_details_help_documents_bare_v_without_alt_v with a double assertion: primary chord is "v", and "Alt+V" is absent from the table. ui/tests.rs assertions updated to match the new "v raw" / "v details" string formats.

LGTM. All previous review items are resolved. KbToolDetailsPager cleanup is low-priority follow-up only.