Shared translation config implemented: identical texts on different pages of the site are translated once and reused. The cache works at the language pair and URL pattern level — requests to external services (delivery, geolocation) also fall into the shared cache. The load on language models has decreased, and the processing speed of repeated requests has increased.