GeoLira is built for 30 target languages. The language selector is generated from the central locale registry and shows only languages that are fully published.
Published languages come from one 30-language architecture
The selector is generated from the central locale registry, not from a hardcoded language list. A planned locale stays hidden until it covers all logical routes and passes interface, content, knowledge, Telegram, backend and SEO checks. HTML language and text direction also come from locale metadata, including RTL support.
Publication requires full parity
The interface catalog, launch content, Source Catalog, Research Library, Scientific Glossary, Evidence Guides, Telegram catalog, spirituality content, Sacred Places and backend messages are checked for locale coverage. Adding a feature in one language must not silently remove it from another published language. Unsupported locales fall back to en_US.
Translation never changes the underlying data
Changing language does not alter source_id, station_id, DOI, physical units, record quality, canonical observation time or evidence class. Labels and formatting may change, but a translation cannot turn an authored index into a measurement, a model into an observation, or weaken privacy and limitation language.