feat: chrome translation Tier 1 — Home + Breathe + badges (ja/zh-Hans/zh-Hant)

[TrainTravel]

Summary

First tier of the chrome translation pass. Adds Japanese, Simplified Chinese, and Traditional Chinese keys to the highest-visibility t() and bilingual() call sites — Home screen, Breathe screen, and the BADGES constant.

Pairs with #39 (bilingual dedupe). Together they fix the visible "Write today / Write today" rendering for users with target=ja and primary=zh-Hant. With both PRs merged, a Japanese learner who speaks Chinese sees natural CJK chrome on Home instead of duplicated English.

Scope

~40 calls × 3 langs = ~120 new translation keys, across:

• HomeScreen.tsx — sign-out, hero subtitle, "Completed" badge, word counter (×2), badge tooltips, "More tools" toggle, secondary action bilingual() anchors (Brain Dump, Thought Garden, Free Write, One Thing at a Time, ABC List, Small Wins, Tiny Experiments, Zen Garden, Sand Timer), Emotion vocabulary card, "One or two sentences is enough." tagline
• BreatheScreen.tsx — Back, three breathing phases (inhale / hold / exhale), grounding prompt, "I'm ready" CTA
• BADGES constant + Badge type — extended with optional ja, zh-Hans, zh-Hant; all 6 badges translated (新芽, 書き手, あなたの声, 語り手, 庭師, 賢者)

Why tiered

A "full sweep" would be ~350 sites × 3 = ~1000 strings of unreviewed AI translation. That's 100× the surface area of the Japanese prompts in PR #36 that already need native review. Tier 1 covers the 80% of visual weight without committing 1000 strings; Tier 2 + Tier 3 ship after Tier 1 has had a native pass.

Test plan

• npx tsc --noEmit — clean
• npx vitest run — 188 passed / 189 (only pre-existing useJournal startFreeWrite fails)
• All existing bilingual() tests still pass (the new keys don't break the dedupe behavior shipped in feat: dedupe bilingual() identical sides + add CTA ja/zh translations #39 — Tier 1 strings differ across target/primary, so dedupe doesn't trigger for them)
• Manual on a real device: open Home with target=ja primary=en → all chrome reads naturally in Japanese where translated
• Manual: open Home with target=fr primary=zh-Hant → primary side shows Chinese, target shows French (bilingual() pair where both sides have translations)
• Manual: BreatheScreen with target=ja → phase prompts read 息を吸って... / 息を止めて... / 息を吐いて...

Known follow-ups (called out in CHANGELOG)

• "French journaling practice" hero subtitle is hardcoded. It translates literally but is semantically wrong for non-French targets (a Japanese learner sees "Japanese journaling practice → フランス語日記の練習"). Future PR makes it target-aware.
• Word counter splits on /\s+/ — produces unreliable counts for Japanese / Chinese. Counter logic itself needs a separate fix (probably a graphemes-based or character-based count for CJK targets).
• Native review of the Japanese and Chinese strings shipped here is still planned. Same path as for the Japanese prompts in PR feat: Japanese target language + edge function consolidation #36 — iTalki / r/translator. The copy-paste consensus-check template from earlier sessions still applies.

Out of scope (Tier 2 / Tier 3)

• WriteScreen, EmotionsScreen, ReflectionScreen, GratitudeScreen, FeedbackScreen
• BrainDumpScreen body, ThoughtGardenScreen interior, ProgressScreen
• LanguageSettings / Account / Font / Theme picker chrome
• Anything in src/components/ui (shadcn primitives)
• Error messages / edge-case toasts

🤖 Generated with Claude Code