Checking each language's translation coverage at a glance

This is a report, not an editor — its whole job is to answer "how are we doing, language by language" in one table, and to send you to the actual editor the moment you want to change something.

The list is a table, not a fixed set of columns

Hebrew, Arabic, Russian, Amharic and English are what your platform happens to ship with today — they're rows in a table, not a hard-coded list five columns wide. A new customer needing Tigrinya gets a new row, not a release. This screen deliberately renders whatever rows exist, including every one your organization has switched off, because a language you can't see here is one you can't easily decide to switch back on.

The endonym column is not decoration

Each language's own name is shown written in its own script and its own reading direction. A resident choosing their language from a picker can't read the picker if it's written in a language they don't have — showing only an English name for every language is the single most common way a language switcher becomes unusable for exactly the people who need it most.

Reading the coverage number honestly

Each row's coverage percentage is a real ratio of translated keys to total keys — except when the report itself couldn't be computed, in which case the screen says "unknown" rather than quietly showing 100%. That distinction matters: a completeness check with no data behind it and a language that's genuinely fully translated would otherwise look identical, and only one of those is actually good news.