feat(locale): bootstrap Latvian (lv) locale with i18n + 6 news sources

[zionappp-ui]

What this adds

Full Latvian (`lv`) locale bootstrap.

Locale changes

• `src/services/i18n.ts` — `'lv'` in `SUPPORTED_LANGUAGES` + `LANGUAGES` (`Latviešu 🇱🇻`)
• `src/utils/map-locale.ts` — `lv: 'name:lv'`
• `public/offline.html` — Latvian offline strings
• `scripts/translate-locales.mjs` + `sync-offline-translations.mjs` — `'lv'` added
• `src/locales/lv.json` — locale bundle with Latvian overlay

Sources (6 feeds, all verified live)

Source	Type	RSS	Items	Tier
LSM (public broadcaster)	mainstream	direct	50	1
Delfi LV	mainstream	direct	25	2
Latvijas Avīze	mainstream	direct	25	2
Diena	mainstream	Google News lv	44	2
TVNet	mainstream	Google News lv	100	2
Neatkarīgā	mainstream	Google News lv	100	2

LSM (Latvijas Sabiedriskie Mediji) is Latvia's public broadcaster. 3 direct RSS feeds, 3 via locale-tuned Google News (`hl=lv&gl=LV&ceid=LV:lv`).

[vercel]

@zionappp-ui is attempting to deploy a commit to the World Monitor Team on Vercel.

A member of the Team first needs to authorize it.

[greptile-apps]

Greptile Summary

Bootstraps the Latvian (lv) locale end-to-end: i18n registration, map-locale tile field, offline fallback strings, translation bundle, and 6 news feeds (3 direct RSS + 3 Google News). All domain allowlists (shared/rss-allowed-domains.json, scripts/shared/, api/_rss-allowed-domains.js, vite.config.ts) are updated consistently, source-tier registrations match feed names, and client/server feed routing is in parity.

• 6 Latvian feeds added across both src/config/feeds.ts and server/worldmonitor/news/v1/_feeds.ts with consistent direct/Google-News routing per feed, satisfying the feeds-client-server-parity test.
• lv.json locale bundle covers the same key count as en.json but several meta and widget strings appear untranslated (left in English), and three pre-existing source-tiers.json entries were re-encoded with \\uXXXX escapes as a serialiser side-effect.

Confidence Score: 4/5

Safe to merge — all domain allowlists are consistent, feed routing parity is maintained, and no data path is broken.

The wiring is solid: allowlists, source tiers, and client/server feed parity are all correct. The main rough edges are cosmetic — some Latvian locale strings were left in English (visible to users who select Latvian), and a serialiser artifact re-encoded three pre-existing source names in source-tiers.json with Unicode escapes.

src/locales/lv.json deserves a translation-completeness pass before or shortly after merge; scripts/shared/source-tiers.json and shared/source-tiers.json carry unintended Unicode-escape churn on three pre-existing entries.

Important Files Changed

Filename	Overview
src/locales/lv.json	New 3,269-line Latvian locale bundle; key offline/shell strings are translated but several meta and widget labels remain in English
server/worldmonitor/news/v1/_feeds.ts	Adds 6 Latvian feeds (3 direct RSS, 3 via Google News gnLocale helper); routing matches the client-side FULL_FEEDS entry for each feed name, satisfying the parity test
src/config/feeds.ts	Client-side FULL_FEEDS and SOURCE_TYPES updated for all 6 Latvian sources; routing (direct vs Google News) is consistent with the server _feeds.ts counterpart
api/_rss-allowed-domains.js	Adds www.lsm.lv, www.delfi.lv, www.la.lv — correctly mirrors shared/rss-allowed-domains.json and vite.config.ts; Google News proxied feeds need no new entry
shared/source-tiers.json	Adds correct tier entries for 6 Latvian sources; also re-encodes three pre-existing non-ASCII strings as \uXXXX escapes (harmless but noisy)
src/services/i18n.ts	Adds lv to SUPPORTED_LANGUAGES (now sorted alphabetically) and LANGUAGES display list; consistent with sync-offline-translations.mjs LOCALES array
vite.config.ts	Dev-proxy allowlist updated with 3 direct Latvian RSS domains, kept in sync with shared/rss-allowed-domains.json
public/offline.html	Latvian offline strings inserted in alphabetical order (between ko and nl); values match lv.json shell.offline* keys exactly
src/utils/map-locale.ts	Adds lv → name:lv tile field mapping, consistent with all other locale entries in the map

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[User selects Latvian locale] --> B[i18n.ts SUPPORTED_LANGUAGES includes 'lv']
    B --> C[Load src/locales/lv.json]
    C --> D[map-locale.ts: lv → name:lv tile field]

    E[News panel requests feeds] --> F{Feed type?}
    F -->|Direct RSS| G[RSS Proxy]
    G --> H{Domain in allowlist?}
    H -->|www.lsm.lv| I[LSM feed]
    H -->|www.delfi.lv| J[Delfi LV feed]
    H -->|www.la.lv| K[Latvijas Avize feed]

    F -->|Google News| L[news.google.com - already allowed]
    L --> M[Diena / TVNet / Neatkariga]

    N[source-tiers.json] --> O[getSourceTier lookup]
    O --> I & J & K & M

    P[public/offline.html] --> Q[lv offline strings]

Loading

_{Reviews (1): Last reviewed commit: "feat(locale): bootstrap Latvian (lv) loc..." | Re-trigger Greptile}

[greptile-apps]

Untranslated meta strings in the Latvian locale bundle

app.description ("Global Situation with AI Insights") and shell.metaDescription ("Real-time global intelligence platform…") are left in English. Comparing with de.json, both fields are fully translated there (e.g. "Globale Lage mit KI-Erkenntnissen"). These strings are used in the document <title> and <meta name="description"> tag, so Latvian users will see raw English text in their browser tab and social previews. Additionally, several widget-level labels (e.g. "chatTitle": "Widget Builder", "phaseChecking": "Checking") are also in English throughout the file. The lv.json and en.json share the same line count (3,269), suggesting the AI translation pass may have skipped some segments.

[greptile-apps]

Unrelated Unicode-escape churn on pre-existing entries

Three existing entries were re-serialised with JSON Unicode escapes as a side-effect of this PR: "El País" → "El Pa\u00eds", "Híradó" → "H\u00edrad\u00f3", "La Silla Vacía" → "La Silla Vac\u00eda". The same change appears in shared/source-tiers.json. JSON \uXXXX escapes and literal Unicode characters are parsed identically, so there is no runtime impact; but the diff noise makes it harder to review the Latvian additions and could trigger downstream churn if a future contributor normalises the file in the opposite direction. The churn is likely a serialiser artifact — worth stripping before merge.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!