Built from zero since 9 August 2026 by a team of three, and already working end to end. We assembled our own corpus of Uzbek law without an API, because lex.uz has none: 227,369 fragments from 2,017 acts — 1,172 Russian and 845 Uzbek documents, 25 codes and 103 Supreme Court Plenum rulings. Nine case-category matrices produce a deterministic analysis. Citation verification runs at every start-up and stops the engine if a citation differs from the act by a single word
it caught a fabricated citation in our own first matrix. The system detects where the Russian translation lags behind the authoritative Uzbek text, and returns an article as it stood on the date of the events. Our parser found and fixed a defect that silently dropped half of the Administrative Offences Code: articles inserted by amendments carry superscript numbers. 22 screens in production, 18 test suites, search in 10–120 ms, full case analysis in about 3 seconds. The code is public on GitHub.