feat(i18n): offer every ISO 639-1 language

The table was thirty languages typed by hand. It is now generated:
utils/generate_languages.py takes the 184 alpha-2 codes from pycountry,
the display names from CLDR through babel, and the writing direction from
CLDR character order. 159 languages carry their endonym; the remaining 25
have no CLDR entry and carry their English ISO name.

That corrects the right-to-left set, which had four entries and needs ten
— dv, ks, ps, sd, ug and yi were simply missed.

Only 29 languages ship an interface catalogue, so the other 155 render in
English until one is filled. make i18n-ui fills app/i18n/ui/ for them, and
make i18n now covers the interface strings as well; neither asks for a
string a shipped catalogue already answers, so hand-written entries stay.

184 entries do not fit on a screen, so the language menu scrolls inside
itself. overscroll-behavior keeps the page behind it from moving once the
list reaches its end.

babel and pycountry are dev dependencies: the generator needs them, the
application does not.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Makefile

@@ -144,6 +144,15 @@ i18n: env config
 			--url "$$LIBRETRANSLATE_URL" \
 			--api-key "$$LIBRETRANSLATE_API_KEY"
 
+.PHONY: i18n-ui
+i18n-ui: env
+	# Fill missing interface translations in app/i18n/ui/ via LibreTranslate.
+	@$(call _require_env,LIBRETRANSLATE_URL); \
+		$(PYTHON) utils/i18n_sync.py \
+			--catalog ui \
+			--url "$$LIBRETRANSLATE_URL" \
+			--api-key "$$LIBRETRANSLATE_API_KEY"
+
 .PHONY: lint-actions
 lint-actions:
 	# Lint GitHub Actions workflows.

README.md

@@ -22,8 +22,8 @@ A lightweight, Docker-powered portfolio/landing-page generator—fully customiza
   Auto-cache assets for lightning-fast loading.  
 - **Responsive Design**  
   Built on Bootstrap; looks great on desktop, tablet & mobile.  
-- **30 Languages**  
+- **184 Languages**  
-  Browser-negotiated, RTL-aware, with machine translation for your own content.  
+  Every ISO 639-1 code, browser-negotiated, RTL-aware, with machine translation for your own content.  
 - **YAML-Driven**  
   All content & structure defined in a simple `config.yaml`.  
 - **CLI Control**  
@@ -143,16 +143,19 @@ company:
 
 ## 🌍 Languages
 
-The interface ships in 30 languages. `/` serves the best match for the visitor's
+Every ISO 639-1 language — all 184 two-letter codes — has a URL, a display
+name in its own script and a writing direction. The interface ships translated
+for 29 of them; the rest fall back to English string by string until a
+catalogue is filled. `/` serves the best match for the visitor's
 `Accept-Language` header, `/<code>/` forces one, and a switcher in the navbar
-lists them all. Right-to-left languages (`ar`, `fa`, `he`, `ur`) get
+lists them all. The ten right-to-left languages get `dir="rtl"` and Bootstrap's RTL
-`dir="rtl"` and Bootstrap's RTL stylesheet automatically.
+stylesheet automatically.
 
 Translations live in two catalogues, both keyed by the English source string:
 
 | Path | Tracked | Holds |
 | --- | --- | --- |
-| `app/i18n/ui/<code>.yaml` | yes | Interface strings, shipped complete for all 29 non-English languages. English is the source and has no file. |
+| `app/i18n/ui/<code>.yaml` | yes | Interface strings. Shipped for 29 languages; English is the source and has no file. |
 | `app/i18n/content/<code>.yaml` | no | Your `config.yaml` prose, generated per deployment. |
 
 A string with no catalogue entry falls back to English, so a half-filled
@@ -165,8 +168,10 @@ instance — set `LIBRETRANSLATE_URL` in `.env`, then:
 make i18n
 ```
 
-Existing entries are never overwritten, so corrections you make by hand survive
+This fills the interface strings of the languages that ship no catalogue as
-later runs. Only prose (`description`, `text`, `warning`, `info`, `subtitel`) is
+well. Existing entries are never overwritten, and a string the shipped
+catalogue already covers is never requested, so corrections you make by hand
+survive later runs. Only prose (`description`, `text`, `warning`, `info`, `subtitel`) is
 filled automatically; `name` and `title` are left to you, because a machine
 cannot tell the menu label "Pictures" from the brand "Mastodon". Write those
 into the content catalogue yourself when you want them translated.
@@ -179,7 +184,7 @@ into the content catalogue yourself when you want them translated.
 * Secure with SSL/TLS.
 * Swap to a production database if needed.
 
-Because every page carries a canonical URL and 30 `hreflang` alternates, two
+Because every page carries a canonical URL and 184 `hreflang` alternates, two
 details of the proxy setup now matter:
 
 * **Set `TRUSTED_HOSTS`** in `.env` to your public hostname(s), comma-separated.

app/cypress/e2e/i18n.spec.js

@@ -53,7 +53,7 @@ describe('Language switcher', () => {
     cy.visit('/en/');
   });
 
-  it('names the active language and offers all thirty', () => {
+  it('names the active language and offers every ISO 639-1 code', () => {
     cy.get('#navbarDropdownLanguage')
       .should('have.attr', 'data-bs-toggle', 'dropdown')
       .and('contain.text', 'English');
@@ -61,7 +61,7 @@ describe('Language switcher', () => {
     cy.get('#navbarDropdownLanguage')
       .parent('.nav-item')
       .find('> .dropdown-menu a.dropdown-item')
-      .should('have.length', 30);
+      .should('have.length', 184);
   });
 
   it('marks the active language', () => {
@@ -200,6 +200,21 @@ describe('Translated interface details', () => {
     );
   });
 
+  it('scrolls inside the language menu instead of past the page', () => {
+    cy.viewport(1280, 720);
+    cy.visit('/en/');
+    cy.get('#navbarDropdownLanguage').click();
+
+    cy.get('.dropdown-menu.language-menu').should($menu => {
+      const menu = $menu[0];
+      expect(menu.scrollHeight, 'taller than it shows').to.be.greaterThan(
+        menu.clientHeight,
+      );
+      expect(menu.getBoundingClientRect().height).to.be.lessThan(720);
+      expect(getComputedStyle(menu).overflowY).to.eq('auto');
+    });
+  });
+
   it('offers the switcher in the header only', () => {
     cy.viewport(1280, 720);
     cy.visit('/en/');
@@ -215,7 +230,7 @@ describe('Search engine metadata', () => {
   });
 
   it('declares an alternate for every language plus a default', () => {
-    cy.get('link[rel="alternate"][hreflang]').should('have.length', 31);
+    cy.get('link[rel="alternate"][hreflang]').should('have.length', 185);
     cy.get('link[rel="alternate"][hreflang="x-default"]').should('exist');
     cy.get('link[rel="alternate"][hreflang="ja"]')
       .should('have.attr', 'href')

app/static/css/default.css

@@ -120,6 +120,12 @@ div#navbarNavfooter .navbar-nav {
   flex-wrap: nowrap;
 }
 
+.dropdown-menu.language-menu {
+  max-height: 60vh;
+  overflow-y: auto;
+  overscroll-behavior: contain;
+}
+
 main, footer, header, nav {
   position: relative;
   box-shadow: 

app/templates/moduls/navigation.html.j2

@@ -103,7 +103,7 @@
                 <a class="nav-link dropdown-toggle btn btn-light" id="navbarDropdownLanguage" role="button" data-bs-toggle="dropdown" data-bs-display="dynamic" aria-expanded="false" title="{{ t('Language') }}">
                     <i class="fa-solid fa-language"></i> {{ languages[lang] }}
                 </a>
-                <ul class="dropdown-menu">
+                <ul class="dropdown-menu language-menu">
                     {% for code, endonym in languages.items() %}
                     <li>
                         <a class="dropdown-item {% if code == lang %}active{% endif %}"

app/utils/i18n.py

@@ -11,46 +11,17 @@ from pathlib import Path
 
 import yaml
 
+try:
+    from app.utils.languages import LANGUAGES, RTL_LANGUAGES
+except ImportError:  # pragma: no cover - supports running from the app/ directory.
+    from utils.languages import LANGUAGES, RTL_LANGUAGES
+
 I18N_DIR = Path(__file__).resolve().parent.parent / "i18n"
 UI_DIR = I18N_DIR / "ui"
 CONTENT_DIR = I18N_DIR / "content"
 
 SOURCE_LANGUAGE = "en"
 
-LANGUAGES = {
-    "en": "English",
-    "zh": "中文",
-    "hi": "हिन्दी",
-    "es": "Español",
-    "fr": "Français",
-    "ar": "العربية",
-    "bn": "বাংলা",
-    "pt": "Português",
-    "ru": "Русский",
-    "ur": "اردو",
-    "id": "Bahasa Indonesia",
-    "de": "Deutsch",
-    "ja": "日本語",
-    "tr": "Türkçe",
-    "ko": "한국어",
-    "vi": "Tiếng Việt",
-    "it": "Italiano",
-    "th": "ไทย",
-    "pl": "Polski",
-    "nl": "Nederlands",
-    "uk": "Українська",
-    "fa": "فارسی",
-    "ro": "Română",
-    "el": "Ελληνικά",
-    "cs": "Čeština",
-    "sv": "Svenska",
-    "hu": "Magyar",
-    "he": "עברית",
-    "da": "Dansk",
-    "fi": "Suomi",
-}
-
-RTL_LANGUAGES = frozenset({"ar", "fa", "he", "ur"})
 
 AUTOFILL_KEYS = frozenset({"description", "text", "warning", "info", "subtitel"})
 

app/utils/languages.py

@@ -0,0 +1,196 @@
+"""ISO 639-1 languages, their display names and writing direction.
+
+Generated by utils/generate_languages.py — edit that script, not this file.
+
+Display names are CLDR endonyms where CLDR covers the language and the English
+ISO 639-1 name for the 25 codes it does not.
+"""
+
+LANGUAGES = {
+    "en": "English",
+    "aa": "Qafar",
+    "ab": "Аԥсшәа",
+    "ae": "Avestan",
+    "af": "Afrikaans",
+    "ak": "Akan",
+    "am": "አማርኛ",
+    "an": "aragonés",
+    "ar": "العربية",
+    "as": "অসমীয়া",
+    "av": "Avaric",
+    "ay": "Aymara",
+    "az": "azərbaycan",
+    "ba": "башҡорт теле",
+    "be": "беларуская",
+    "bg": "български",
+    "bi": "Bislama",
+    "bm": "bamanakan",
+    "bn": "বাংলা",
+    "bo": "བོད་སྐད་",
+    "br": "brezhoneg",
+    "bs": "bosanski",
+    "ca": "català",
+    "ce": "нохчийн",
+    "ch": "Chamorro",
+    "co": "corsu",
+    "cr": "Cree",
+    "cs": "čeština",
+    "cu": "церковнослове́нскїй",
+    "cv": "чӑваш",
+    "cy": "Cymraeg",
+    "da": "dansk",
+    "de": "Deutsch",
+    "dv": "ދިވެހިބަސް",
+    "dz": "རྫོང་ཁ",
+    "ee": "eʋegbe",
+    "el": "Ελληνικά",
+    "eo": "Esperanto",
+    "es": "español",
+    "et": "eesti",
+    "eu": "euskara",
+    "fa": "فارسی",
+    "ff": "Pulaar",
+    "fi": "suomi",
+    "fj": "Fijian",
+    "fo": "føroyskt",
+    "fr": "français",
+    "fy": "Frysk",
+    "ga": "Gaeilge",
+    "gd": "Gàidhlig",
+    "gl": "galego",
+    "gn": "avañe’ẽ",
+    "gu": "ગુજરાતી",
+    "gv": "Gaelg",
+    "ha": "Hausa",
+    "he": "עברית",
+    "hi": "हिन्दी",
+    "ho": "Hiri Motu",
+    "hr": "hrvatski",
+    "ht": "Kreyòl Ayisyen",
+    "hu": "magyar",
+    "hy": "հայերեն",
+    "hz": "Herero",
+    "ia": "interlingua",
+    "id": "Indonesia",
+    "ie": "Interlingue",
+    "ig": "Igbo",
+    "ii": "ꆈꌠꉙ",
+    "ik": "Inupiaq",
+    "io": "Ido",
+    "is": "íslenska",
+    "it": "italiano",
+    "iu": "ᐃᓄᒃᑎᑐᑦ",
+    "ja": "日本語",
+    "jv": "Jawa",
+    "ka": "ქართული",
+    "kg": "Kongo",
+    "ki": "Gikuyu",
+    "kj": "Kuanyama",
+    "kk": "қазақ тілі",
+    "kl": "kalaallisut",
+    "km": "ខ្មែរ",
+    "kn": "ಕನ್ನಡ",
+    "ko": "한국어",
+    "kr": "Kanuri",
+    "ks": "کٲشُر",
+    "ku": "kurdî (kurmancî)",
+    "kv": "Komi",
+    "kw": "kernewek",
+    "ky": "кыргызча",
+    "la": "Latina",
+    "lb": "Lëtzebuergesch",
+    "lg": "Luganda",
+    "li": "Limburgan",
+    "ln": "lingála",
+    "lo": "ລາວ",
+    "lt": "lietuvių",
+    "lu": "Tshiluba",
+    "lv": "latviešu",
+    "mg": "Malagasy",
+    "mh": "Marshallese",
+    "mi": "Māori",
+    "mk": "македонски",
+    "ml": "മലയാളം",
+    "mn": "монгол",
+    "mr": "मराठी",
+    "ms": "Melayu",
+    "mt": "Malti",
+    "my": "မြန်မာ",
+    "na": "Nauru",
+    "nb": "norsk bokmål",
+    "nd": "isiNdebele",
+    "ne": "नेपाली",
+    "ng": "Ndonga",
+    "nl": "Nederlands",
+    "nn": "norsk nynorsk",
+    "no": "norsk",
+    "nr": "isiNdebele",
+    "nv": "Diné Bizaad",
+    "ny": "Nyanja",
+    "oc": "occitan",
+    "oj": "Ojibwa",
+    "om": "Oromoo",
+    "or": "ଓଡ଼ିଆ",
+    "os": "ирон",
+    "pa": "ਪੰਜਾਬੀ",
+    "pi": "Pali",
+    "pl": "polski",
+    "ps": "پښتو",
+    "pt": "português",
+    "qu": "Runasimi",
+    "rm": "rumantsch",
+    "rn": "Ikirundi",
+    "ro": "română",
+    "ru": "русский",
+    "rw": "Ikinyarwanda",
+    "sa": "संस्कृत भाषा",
+    "sc": "sardu",
+    "sd": "سنڌي",
+    "se": "davvisámegiella",
+    "sg": "Sängö",
+    "sh": "Serbo-Croatian",
+    "si": "සිංහල",
+    "sk": "slovenčina",
+    "sl": "slovenščina",
+    "sm": "Samoan",
+    "sn": "chiShona",
+    "so": "Soomaali",
+    "sq": "shqip",
+    "sr": "српски",
+    "ss": "siSwati",
+    "st": "Sesotho",
+    "su": "Basa Sunda",
+    "sv": "svenska",
+    "sw": "Kiswahili",
+    "ta": "தமிழ்",
+    "te": "తెలుగు",
+    "tg": "тоҷикӣ",
+    "th": "ไทย",
+    "ti": "ትግርኛ",
+    "tk": "türkmen dili",
+    "tl": "Tagalog",
+    "tn": "Setswana",
+    "to": "lea fakatonga",
+    "tr": "Türkçe",
+    "ts": "Xitsonga",
+    "tt": "татар",
+    "tw": "Twi",
+    "ty": "Tahitian",
+    "ug": "ئۇيغۇرچە",
+    "uk": "українська",
+    "ur": "اردو",
+    "uz": "o‘zbek",
+    "ve": "Tshivenḓa",
+    "vi": "Tiếng Việt",
+    "vo": "Volapük",
+    "wa": "walon",
+    "wo": "Wolof",
+    "xh": "IsiXhosa",
+    "yi": "ייִדיש",
+    "yo": "Èdè Yorùbá",
+    "za": "Vahcuengh",
+    "zh": "中文",
+    "zu": "isiZulu",
+}
+
+RTL_LANGUAGES = frozenset({"ar", "dv", "fa", "he", "ks", "ps", "sd", "ug", "ur", "yi"})

pyproject.toml

@@ -18,8 +18,10 @@ dependencies = [
 
 [project.optional-dependencies]
 dev = [
+    "babel",
     "bandit",
     "pip-audit",
+    "pycountry",
     "ruff",
     "yamllint",
 ]

tests/unit/test_i18n.py

@@ -35,12 +35,12 @@ class TestNegotiate(unittest.TestCase):
 
 class TestDirection(unittest.TestCase):
     def test_every_right_to_left_language_is_marked(self):
-        for code in ("ar", "fa", "he", "ur"):
+        for code in ("ar", "dv", "fa", "he", "ks", "ps", "sd", "ug", "ur", "yi"):
             with self.subTest(code=code):
                 self.assertEqual(i18n.direction(code), "rtl")
 
     def test_other_languages_are_left_to_right(self):
-        for code in set(i18n.LANGUAGES) - {"ar", "fa", "he", "ur"}:
+        for code in set(i18n.LANGUAGES) - set(i18n.RTL_LANGUAGES):
             with self.subTest(code=code):
                 self.assertEqual(i18n.direction(code), "ltr")
 
@@ -157,9 +157,17 @@ class TestCatalogMerge(unittest.TestCase):
 
 
 class TestShippedCatalogs(unittest.TestCase):
-    def test_thirty_languages_are_offered(self):
+    def test_every_iso_639_1_language_is_offered(self):
-        self.assertEqual(len(i18n.LANGUAGES), 30)
+        self.assertEqual(len(i18n.LANGUAGES), 184)
-        self.assertIn(i18n.SOURCE_LANGUAGE, i18n.LANGUAGES)
+        self.assertEqual(i18n.SOURCE_LANGUAGE, next(iter(i18n.LANGUAGES)))
+        self.assertGreaterEqual(
+            set(i18n.LANGUAGES), {"en", "de", "zh", "ar", "he", "vo", "za", "cu"}
+        )
+
+    def test_every_display_name_is_filled_in(self):
+        blank = [code for code, name in i18n.LANGUAGES.items() if not name.strip()]
+
+        self.assertFalse(blank, f"No display name for: {blank}")
 
     def test_every_code_is_usable_in_the_route_converter(self):
         offenders = [
@@ -168,24 +176,21 @@ class TestShippedCatalogs(unittest.TestCase):
 
         self.assertFalse(offenders, f"Not usable in the route converter: {offenders}")
 
-    def test_every_language_but_the_source_ships_a_ui_catalog(self):
+    def test_every_shipped_catalog_names_a_known_language(self):
-        missing = [
+        stray = [
-            code
+            path.stem
-            for code in i18n.LANGUAGES
+            for path in i18n.UI_DIR.glob("*.yaml")
-            if code != i18n.SOURCE_LANGUAGE
+            if path.stem not in i18n.LANGUAGES or path.stem == i18n.SOURCE_LANGUAGE
-            and not (i18n.UI_DIR / f"{code}.yaml").is_file()
         ]
 
-        self.assertFalse(missing, f"No UI catalogue for: {missing}")
+        self.assertFalse(stray, f"Catalogue for an unknown language: {stray}")
 
     def test_ui_catalogs_cover_exactly_the_interface_strings(self):
         expected = set(i18n.UI_STRINGS)
         mismatched = {}
 
-        for code in i18n.LANGUAGES:
+        for path in sorted(i18n.UI_DIR.glob("*.yaml")):
-            if code == i18n.SOURCE_LANGUAGE:
+            code = path.stem
-                continue
-            path = i18n.UI_DIR / f"{code}.yaml"
             entries = yaml.safe_load(path.read_text(encoding="utf-8"))
             if set(entries) != expected:
                 mismatched[code] = {

tests/unit/test_i18n_sync.py

@@ -224,7 +224,7 @@ class TestSync(unittest.TestCase):
         session = Mock(get=Mock(return_value=listing), post=Mock(return_value=answer))
 
         with patch.object(i18n_sync.requests, "Session", return_value=session):
-            i18n_sync.sync("http://lt", "", set(sources), ["de"])
+            i18n_sync.sync("http://lt", "", set(sources), ["de"], self.directory)
         return session
 
     def test_missing_entries_are_written(self):
@@ -267,13 +267,26 @@ class TestSync(unittest.TestCase):
 
         self.assertEqual(self.path.read_text(encoding="utf-8"), original)
 
-    def test_the_content_directory_is_created(self):
+    def test_a_string_the_shipped_catalog_covers_is_not_requested(self):
+        ui = Path(tempfile.mkdtemp())
+        self.addCleanup(shutil.rmtree, ui, True)
+        self.addCleanup(setattr, i18n, "UI_DIR", i18n.UI_DIR)
+        i18n.UI_DIR = ui
+        (ui / "de.yaml").write_text("Close: Schliessen\n", encoding="utf-8")
+
+        session = self._run(["Close", "Hello"])
+
+        self.assertEqual(session.post.call_count, 1)
+        self.assertNotIn("Close", yaml.safe_load(self.path.read_text(encoding="utf-8")))
+
+    def test_the_catalog_directory_is_created(self):
         nested = self.directory / "content"
-        i18n.CONTENT_DIR = nested
+        self.directory = nested
+        self.path = nested / "de.yaml"
 
         self._run(["Hello"])
 
-        self.assertTrue((nested / "de.yaml").is_file())
+        self.assertTrue(self.path.is_file())
 
 
 if __name__ == "__main__":

utils/generate_languages.py

@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""Regenerate app/utils/languages.py from ISO 639-1 and CLDR.
+
+Every ISO 639-1 alpha-2 code becomes an entry. The display name is the CLDR
+endonym where CLDR knows the language, and the English ISO name for the
+remainder. Writing direction comes from CLDR's character order.
+
+Needs the dev extra: pip install -e ".[dev]"
+"""
+
+import subprocess
+import sys
+from pathlib import Path
+
+import pycountry
+from babel import Locale
+from babel.localedata import locale_identifiers
+
+TARGET = Path(__file__).resolve().parents[1] / "app" / "utils" / "languages.py"
+
+HEADER = '''"""ISO 639-1 languages, their display names and writing direction.
+
+Generated by utils/generate_languages.py — edit that script, not this file.
+
+Display names are CLDR endonyms where CLDR covers the language and the English
+ISO 639-1 name for the {fallback} codes it does not.
+"""
+
+LANGUAGES = {{
+'''
+
+FOOTER = """}}
+
+RTL_LANGUAGES = frozenset({rtl})
+"""
+
+
+def collect():
+    """Return (code -> display name) and the set of right-to-left codes."""
+    cldr = {code for code in locale_identifiers() if len(code) == 2 and code.isalpha()}
+    names, rtl, fallbacks = {}, set(), 0
+
+    for language in sorted(
+        pycountry.languages, key=lambda item: getattr(item, "alpha_2", "")
+    ):
+        code = getattr(language, "alpha_2", None)
+        if not code:
+            continue
+        if code in cldr:
+            locale = Locale.parse(code)
+            names[code] = locale.get_display_name(code)
+            if locale.character_order == "right-to-left":
+                rtl.add(code)
+        else:
+            names[code] = language.name
+            fallbacks += 1
+
+    return names, rtl, fallbacks
+
+
+def render(names, rtl, fallbacks):
+    """Return the module source, English first and the rest by code."""
+    ordered = ["en"] + [code for code in sorted(names) if code != "en"]
+    body = "".join(f'    "{code}": "{names[code]}",\n' for code in ordered)
+    listed = ", ".join(f'"{code}"' for code in sorted(rtl))
+    return (
+        HEADER.format(fallback=fallbacks) + body + FOOTER.format(rtl="{" + listed + "}")
+    )
+
+
+def main():
+    names, rtl, fallbacks = collect()
+    TARGET.write_text(render(names, rtl, fallbacks), encoding="utf-8")
+    subprocess.run([sys.executable, "-m", "ruff", "format", str(TARGET)], check=False)
+    print(
+        f"{len(names)} languages, {len(rtl)} right-to-left, "
+        f"{fallbacks} without a CLDR endonym"
+    )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

utils/i18n_sync.py

@@ -144,9 +144,17 @@ def translate(session, url, api_key, text, target):
     return translated
 
 
-def sync(url, api_key, sources, targets):
+def sync(url, api_key, sources, targets, directory):
-    """Fill and write the content catalogue of every language in ``targets``."""
+    """Fill and write the catalogue of every language in ``targets``.
-    i18n.CONTENT_DIR.mkdir(parents=True, exist_ok=True)
+
+    Args:
+        url: base URL of the LibreTranslate instance.
+        api_key: API key, or an empty string.
+        sources: English strings to translate.
+        targets: language codes to fill.
+        directory: catalogue directory to write into.
+    """
+    directory.mkdir(parents=True, exist_ok=True)
     session = requests.Session()
 
     available = supported_languages(url, session)
@@ -155,12 +163,21 @@ def sync(url, api_key, sources, targets):
             print(f"- {target}: not offered by {url}, skipped")
             continue
 
-        path = i18n.CONTENT_DIR / f"{target}.yaml"
+        path = directory / f"{target}.yaml"
         catalog = load_existing(path)
         if catalog is None:
             continue
 
-        missing = sorted(source for source in sources if source not in catalog)
+        shipped = (
+            {}
+            if directory == i18n.UI_DIR
+            else i18n.read_catalog(i18n.UI_DIR / f"{target}.yaml")
+        )
+        missing = sorted(
+            source
+            for source in sources
+            if source not in catalog and source not in shipped
+        )
         if not missing:
             print(f"- {target}: complete")
             continue
@@ -201,6 +218,15 @@ def main(argv=None):
         default=DEFAULT_CONFIG_PATH,
         help=f"Configuration to read the prose from (default: {DEFAULT_CONFIG_PATH}).",
     )
+    parser.add_argument(
+        "--catalog",
+        choices=("content", "ui"),
+        default="content",
+        help=(
+            "content: your configuration's prose, generated per deployment. "
+            "ui: the interface strings that ship with the project."
+        ),
+    )
     parser.add_argument(
         "--languages",
         nargs="+",
@@ -209,12 +235,18 @@ def main(argv=None):
     )
     args = parser.parse_args(argv)
 
-    config = yaml.safe_load(args.config.read_text(encoding="utf-8"))
+    if args.catalog == "ui":
-    sources = collect_sources(config)
+        directory = i18n.UI_DIR
-    print(f"{len(sources)} translatable string(s) in {args.config}")
+        sources = set(i18n.UI_STRINGS)
+        print(f"{len(sources)} interface string(s)")
+    else:
+        directory = i18n.CONTENT_DIR
+        config = yaml.safe_load(args.config.read_text(encoding="utf-8"))
+        sources = collect_sources(config) | set(i18n.UI_STRINGS)
+        print(f"{len(sources)} string(s) in {args.config} and the interface")
 
     try:
-        sync(args.url.rstrip("/"), args.api_key, sources, args.languages)
+        sync(args.url.rstrip("/"), args.api_key, sources, args.languages, directory)
     except BackendError as error:
         print(f"ERROR: {error}", file=sys.stderr)
         return 1