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