@worthy-ventures/metaglotta-runtime

The Metaglotta translation runtime. No framework, no DOM, no plugins.

import { Metaglotta } from '@worthy-ventures/metaglotta-runtime';

const i18n = Metaglotta({
    language: 'el',
    fallbackLanguage: 'en',
    defaultNs: 'common',
    ns: ['common', 'login'],
    fallbackNs: ['common'],
    backend: { prefix: 'https://cdn.example/langs/app' },
});

await i18n.run();
i18n.t('greeting', { name: 'Ada' });

What it does

Holds translations per language and namespace, fetches the ones it needs as {prefix}/{namespace}/{language}.json, falls back through namespaces and then languages when a key is missing, formats ICU, and emits an event when any of that changes.

What it does not do

In-context editing. That needs every rendered string traced back to its key, which means wrapping each one and watching the DOM - a different job, wanted only while authoring. The authoring build runs on @worthy-ventures/metaglotta-ngx/authoring instead.

Behaviours worth knowing

• A miss echoes the key. t('nope') returns 'nope', so a missing string is visible on the page. Pass orEmpty: true for '' instead - which is what the Angular pipe does.
• Namespace before language. The lookup exhausts each namespace across every fallback language before trying the next namespace. A key present in both login/en and common/el resolves to login/en when you asked for login.
• A missing ICU parameter renders 'invalid'. Not the raw string, and not an empty one.
• A file that fails to load costs its namespace, not the load. The record is remembered as empty so it is not re-requested by every component on the page; the failure is reported on the error event and to the console. Pass backend.fallbackOnFail: false to fail the whole load instead.