Хочете будувати кар’єру в Data Science та AI? Vodafone Big Data Lab відкриває новий набір —>
Тарас КозаковWar
19 November 2026, 11:05
2026-11-19
Ukrainians created an encyclopedia of Russian occupation policy in Python and automated its filling using AI agents
The Ukrainian team has developed an online encyclopedia "Lakuzy", which systematizes information about people whose actions, according to the project authors, contribute to Russia’s imperial and colonial policy in the temporarily occupied territories of Ukraine. The database already contains over 2,200 dossiers. To work with such a volume of information, the team used Python, MkDocs and static site generation, and automated some of the filling processes using AI agents.
The Ukrainian team has developed an online encyclopedia "Lakuzy", which systematizes information about people whose actions, according to the project authors, contribute to Russia’s imperial and colonial policy in the temporarily occupied territories of Ukraine. The database already contains over 2,200 dossiers. To work with such a volume of information, the team used Python, MkDocs and static site generation, and automated some of the filling processes using AI agents.
Vasyl Symonimus, who is responsible for the knowledge base architecture, technical implementation, and automation, told dev.ua about the technical architecture and principles of the project. According to him, "Lakuzy" was created to collect and systematize information from open sources about people who have been contributing to the implementation of Russia’s occupation policy in Ukraine since 2014.
Who created "Lakuzy" and what gets into the encyclopedia
The author of the idea and the project manager is Serhiy Mokrenyuk, a veteran of the Russian-Ukrainian war, a lawyer, a representative of the National Council of Ukrainians of Crimea, and a researcher of the instruments of Russian occupation policy. Andriy Shchekun, a former political prisoner of the Kremlin and an expert in the field of preserving Ukrainian identity, is responsible for searching and verifying the sources. The technical part is led by Vasyl SYMONYMUS.
The team explains that it does not add people to the database solely based on citizenship, profession, or the fact of residence in the occupied territory. Each dossier must be based on the person’s specific actions and their consequences. The project authors conduct the assessment within the framework of the Law of Ukraine No. 3005-IX “On the condemnation and prohibition of the propaganda of Russian imperial policy in Ukraine and the decolonization of toponymy.”
Materials are collected from public sources: court decisions, state registers, media and documentary evidence. The dossiers are divided into nine areas: education, religion, culture, media, coercive apparatus, occupation authorities, militarization, economy and occupation elections. One person can belong to several categories.
The project also describes the rules for deleting and updating information. According to the team’s rules, a dossier can be deleted if the original source or all sources it relies on officially refute the published information. There is also a separate option to mark a person’s rehabilitation — in particular, in the case of a significant financial contribution to the Armed Forces of Ukraine or officially confirmed assistance to Ukrainian special services. In this case, the dossier remains in the encyclopedia, but receives an appropriate notification.
The authors emphasize that the collected materials should inform society and become the basis for future legal assessment by competent authorities. That is, the presence of a dossier in the encyclopedia is not in itself a judicial decision or an establishment of a person’s guilt.
Why did the team choose a static website instead of a CMS?
For the technical basis of "Lakuza", the team did not use a classic CMS with a database, admin panel, and dynamic backend. Instead, they built the site on the Knowledge-Base-as-Code principle — when information is stored in the form of files, and the site is generated from them automatically.
The core stack is Python, the MkDocs static site generator, and the Material theme. These are complemented by custom Python hooks, templates, and client-side scripts.
Each dossier is a separate Markdown file with YAML metadata. The team tracks its change history via Git: you can see who added or updated information, when, and based on what source.
This approach, explains the project’s technical representative, offers several advantages. The working version of the site does not have an SQL database, an authorization system, or dynamic endpoints that could become additional points of attack. At the same time, the entire database can be stored and transferred as regular text files, without being tied to a specific infrastructure provider.
The team has also separated public files and internal OSINT research materials. The repository has separate directories for published pages, primary sources, input materials, and internal archives. Service files are excluded from the final build of the site, so they should not be published as web pages or included in the search index.
For security reasons, the team does not disclose where exactly the site is hosted, what servers and providers it uses, and how the deployment process is organized.
How AI agents help populate the database
A separate challenge for an encyclopedia with thousands of dossiers is transferring information from open sources into a structured format. To reduce the amount of manual work, the team built an agent pipeline directly into the project repository.
A set of instructions and specialized skills have been developed for it. For example, ingest is responsible for parsing the source, extracting facts, and forming a dossier; naming is responsible for transliteration and naming rules; linter is responsible for checking the integrity of the database.
The team envisions the AI agent working with primary materials, but it can’t independently invent facts or alter historical logs. Changes are made in a format that can be verified via Git diff — that is, to compare what exactly was added or changed.
The developers also covered key components of the site with automated tests in Python and Node.js: hooks, templates, SEO, redirector, and search index. In particular, they tested the operation of the search engine after building the site.
Another technical feature is the search by different spellings of names. The file stores the Ukrainian spelling of the name, the Russian original, and alternative forms. During site generation, these options are added to the search index so that the user can find a person regardless of the language of the query.
An SEO description, cards for sharing on social networks, and structured Person JSON-LD metadata are also automatically generated for each dossier.
The site has been adapted for people with visual and motor impairments.
The team paid special attention to digital accessibility. According to the developer, the site was brought into compliance with DSTU EN 301 549:2022 and WCAG 2.1 level AA criteria.
Implemented solutions include full keyboard navigation, the ability to jump directly to the main content, tested contrast in light and dark themes, semantic markup, and correct language attributes for screen readers. Fonts are stored locally, and users are not forced to use unstoppable timers.
"Lakuzy" is being developed as an open knowledge base that can be accessed by searching and providing verified public sources. The team encourages reporting on materials about occupation policy figures that can supplement the encyclopedia.
“My job is to give Claude Code tasks, and he does everything himself.” An IT father created an AI service for his son over the weekend that invents new LEGO models. Now he gets organic traffic from Google and has Zelensky and Svitolina figurines in his collection
Німеччина передасть Україні 14 наземних безпілотників THeMIS. Що це за машини та як вони працюють
Німеччина передасть Україні 14 наземних безпілотників THeMIS. Імовірно, 7 таких транспортних засобів прибудуть до України до кінця цього року, вони призначені для евакуації поранених. Інші очікуються у 2023 році — їх будуть використовувати для очищення доріг. Нагадуємо, що це за зброя і що вона вміє.
Атлас зброї: Німеччина передає захисникам чергові MARS II. Які ще РСЗВ отримала Україна від західних партнерів та як вони працюють
Федеральне міністерство оборони Німеччини повідомило у Twitter, що у найближчі кілька тижнів Україні будуть передані чергові системи залпового вогню MARS II.
Раніше ми розповідали про те, які РСЗВ партнери надіслали на допомогу українським воїнам. Нагадуємо про це знову.
(Текст від 26 липня)
На український ударний дрон PUNISHER зібрали понад 1,3 млн грн. Що може гроза командних центрів ворога
Дмитро Томчук повідомив, що найлегші та найменші ударні дрони PUNISHER, які не мають аналогів у світі, тепер коштують $70 000. Виробник додав у комплекс ще одного безпілотника, і тепер їх там два. Дмитро пише, що Юлія Черешня та Павло Харузов зібрали на комплекс більше 1,3 млн грн.
Раніше він розповів журналісту dev.ua, що вміють ці «пташки», та чому саме вони можуть переламати хід війни на користь України.
(текст від 1 червня 2022 року)
США в черговому пакеті допомоги надасть Україні дрони Switchblade 600. Подробиці про ці та інші «пташки», якими користуються українські захисники
Серед списку допомоги, яку пропонує Пентагон, є дрони-камікадзе Switchblade 600, які офіційний Київ просив передати раніше.
Контракт на дослідження та розробку для 10 дронів SwitchBlade 600 виробника Aerovironment, очікується протягом наступних 30 днів, заявила прес-секретар Пентагона Джессіка Максвелл в електронному листі до News Defense.
Ми зібрали список дронів, які використовують українці проти ворога, в тому числі і подробиці про Switchblade 600.
(текст від 17 травня 2022)