fluffychat/.github/instructions/language-list.instructions.md

applyTo
lib/pangea/languages/**,lib/pangea/learning_settings/**,lib/pangea/login/pages/language_selection_page.dart

Language List — Client

For the cross-service language list architecture, L1/L2 definitions, and CMS schema, see language-list.instructions.md.

Data Flow

1. LanguageRepo._fetch() fetches all languages from CMS REST API (/api/languages?limit=500&sort=language_name) — public, no auth
2. PLanguageStore caches the list in SharedPreferences and re-fetches roughly daily
3. Hardcoded fallback in LanguageConstants.languageList if CMS is unreachable and cache is empty

L1/L2 Filtering

PLanguageStore exposes three getters:

Getter	Filter	Used for
baseOptions	All languages (no filter)	L1 / source / native language selection
targetOptions	element.l2 (i.e., l2Support != L2SupportEnum.na)	L2 / target / learning language selection
unlocalizedTargetOptions	L2 filter + excludes regional variants (e.g., keeps "Portuguese" but not "Portuguese (Brazil)")	Course creation language filter

The l2 getter on LanguageModel returns l2Support != L2SupportEnum.na.

Key Usage Sites

• Login language selection: language_selection_page.dart — uses targetOptions for L2 chips, defaults L1 to system language
• Learning settings dialog: p_language_dialog.dart — baseOptions for L1, targetOptions for L2
• Learning settings view: settings_learning_view.dart — same pattern
• Bot chat settings: bot_chat_settings_dialog.dart — targetOptions for room target language
• Course creation: course_language_filter.dart — unlocalizedTargetOptions

Models

• LanguageModel — core model with langCode, displayName, l2Support, script, localeEmoji, voices
• L2SupportEnum — na, alpha, beta, full with localized display strings and badge rendering
• LanguageArc — L1→L2 pair, constructed from user settings

Conventions

• Display names are localized via getDisplayName(context) using l10n keys, with fallback to CMS language_name
• Regional variants show localeEmoji in place of parenthesized region: "Portuguese 🇧🇷" instead of "Portuguese (Brazil)"
• langCodeShort strips the territory: en-US → en
• RTL detection uses a hardcoded list in LanguageConstants.rtlLanguageCodes

Future Work

(No linked issues yet.)