M.D. Gridchin1, A.D. Sitnikov2
1, 2 Bunin Yelets State University (Yelets, Russia)

1 vatis.st@mail.ru, 2 artem-sitnikov1@mail.ru

A pressing issue in information technology is the development of intelligent decision support systems (IDSS) designed to work with operating system components. The use of large language models (LLM) opens up new possibilities for creating IDSS. However, using LLMs is challenging due to the need for fine-tuning. Fully retraining such models requires significant computational resources and time. Therefore, the Low-Rank Adaptation (LoRA) algorithm is used in developing system architectures. It enables the adaptation of large neural network models to specific application problems with minimal effort.

The goal of this work is to develop an IDSS architecture that provides context-adaptive interaction with the operating system command shell.

The design of functional components has been carried out, namely, a knowledge base of commands and scripts, an execution context analysis module, a tokenizer, an encoder using the LoRA low-rank adaptation algorithm for customizing the large language model, a decoder, and a decision presentation interface. The developed architecture operates in several stages: collecting and preprocessing contextual data, vectorizing it, processing the query with a LoRA-adapted model, and generating a final recommendation with a user action prediction.

The proposed architecture can be used to implement decision support systems designed to ensure infrastructure management security, reduce human error risks (minimize administrator errors), and maintain the smooth operation of IT systems.

Gridchin M.D., Sitnikov A.D. Development of architecture for an intelligent decision support system for the command shell // Neurocomputers. 2026. V. 28. № 4. P. 97–106. DOI: https://doi.org/10.18127/j19998554-202604-10

1. Stepanova E.A., Elizarov D.A., Murzaleva E.E. Ekspluatatsiya uyazvimosti operatsionnoj sistemy Astra Linux. Prikaspijskij zhurnal: upravlenie i vysokie tekhnologii. 2025. № 1 (69). S. 86–90. (in Russian)
2. Krylov A.V., Okhtilev M.Yu., Sobolevskij V.A. i dr. Metodologicheskie i metodicheskie osnovy sozdaniya i ispol'zovaniya integrirovannykh sistem podderzhki prinyatiya reshenij. Izvestiya vysshikh uchebnykh zavedenij. Priborostroenie. 2024. T. 63. № 11. S. 963–974. (in Russian)
3. Savenkov P.A. Ispol'zovanie metodov i algoritmov mashinnogo obucheniya v sistemakh podderzhki prinyatiya upravlencheskikh reshenij. Izvestiya Tul'skogo gosudarstvennogo universiteta. Tekhnicheskie nauki. 2019. № 55. S. 23–25. (in Russian)
4. Chernomordov S.V., Druzhinina O.V., Masina O.N., Petrov A.A. Primenenie metodov mashinnogo obucheniya v zadachakh nejrosetevogo modelirovaniya upravlyaemykh tekhnicheskikh system. Nejrokomp'yutery: razrabotka, primenenie. 2022. T. 24. № 1. S. 25–35. DOI: 10.18127/j19998554-202201-03. (in Russian)
5. Samojlenko V.S., Fedotova A.Yu. Arkhitektura intellektual'noj sistemy podderzhki prinyatiya reshenij. Mezhdunarodnyj nauchno-issledo­vatel'skij zhurnal. 2024. № 8 (146). S. 14–21. (in Russian)
6. Karelin V.P. Intellektual'nye tekhnologii i sistemy iskusstvennogo intellekta dlya podderzhki prinyatiya reshenij. Vestnik Taganrogskogo instituta upravleniya i ekonomiki. 2011. № 2. S. 79–84. (in Russian)
7. Alyoubi B.A. Decision support system and knowledge-based strategic management. Procedia Computer Science. 2015. V. 65. P. 278–284. DOI: 10.1016/j.procs.2015.09.079.
8. Demidova L.A., Moroshkin N.A. Obuchenie voprosno-otvetnoj nejrosetevoj modeli na baze arkhitektury modeli LLaVA 1.5 s enkoderom Saiga Mistral 7b i algoritma nizkorangovoj adaptatsii LoRA. Sb. nauch. statej Mezhdunar. nauch.-praktich. konf. «BIG DATA i analiz vysokogo urovnya». Minsk: BGUIR. 2024. S. 401–406. (in Russian)
9. Petrov A.A., Druzhinina O.V., Masina O.N. Razvitie algoritmicheskogo obespecheniya dlya modelirovaniya nelinejnykh upravlyaemykh sistem s pereklyucheniyami. Nelinejnyj mir. 2022. T. 20. № 1. S. 5–13. DOI: 10.18127/j20700970-202201-01. (in Russian)
10. Tikhomirov M., Chernyshev D. Impact of tokenization on LLaMa Russian adaptation. arXiv.org. 2023. DOI: 10.48550/arXiv.2312.02598.
11. Chernenko V.V., Piskorskaya S.Yu. Ekspertnye sistemy. Aktual'nye problemy aviatsii i kosmonavtiki. 2012. T. 2. № 8. S. 322–323. (in Russian)
12. Bokarev D.V., Nikishov S.I. Primenenie NLP-modelej dlya izvlecheniya tsenovoj informatsii iz nestrukturirovannykh dialogov. Computational Nanotechnology. 2025. T. 12. № 1. S. 182–193. (in Russian)
13. Olad'ko A.Yu. Podsistema monitoringa i audita informatsionnoj bezopasnosti v operatsionnoj sisteme Linux. Izvestiya Yuzhnogo federal'nogo universiteta. Tekhnicheskie nauki. 2012. № 12 (137). S. 22–28. (in Russian)
14. Hu E.J., Shen Ye., Wallis Ph. et al. LoRA: Low-rank adaptation of large language models. arXiv.org. 2021. DOI: 10.48550/arXiv.2106. 09685.
15. Pichugov A.A., Namiot D.E., Zubareva E.V. Sovremennye metody obucheniya bol'shikh yazykovykh modelej s minimumom dannykh: ot odnogo primera k absolyutnomu nulyu – akademicheskij obzor. International Journal of Open Information Technologies. 2025. T. 13. № 6. S. 114–124. (in Russian)
16. Samonov A.V., Burova I.O. Metodika razrabotki avtomatizirovannykh sredstv generatsii programmnogo koda posredstvom nastrojki bol'shikh yazykovykh modelej. Voprosy kiberbezopasnosti. 2024. № 3 (61). S. 68–75. (in Russian)