DOI: 10.11118/ecos2026_037

WEB APPLICATION FOR GENERATING TECHNICAL DOCUMENTS USING LARGE LANGUAGE MODELS

Martin Šulc¹, Ivo Pisařovic¹

¹ Department of Informatics, Faculty of Business and Economics, Mendel University in Brno, Zemědělská 1, 613 00 Brno, Czech Republic

This paper investigates the viability of leveraging large language models (LLMs) to generate formal corporate documents, moving beyond rigid, traditional template-based approaches. The study aims to verify whether combining reference samples, natural language instructions, and parallel multi-model processing can yield a practical enterprise tool. The proposed microservices architecture utilizes a Node.js/LangChain backend to abstract calls across four LLM providers (OpenAI, Anthropic, Gemini, and Ollama). This is supported by Python microservices for input parsing and multi-format document generation (Word, PDF, Excel), a Vue.js frontend wizard, and PostgreSQL/Redis persistence. System validation involved end-to-end scenarios across five document types, benchmarking seven models, and user testing with ten participants. Results demonstrate the solution’s efficacy. Furthermore, comparative analysis reveals that local open-source models achieve generation quality comparable to mid-range cloud services for Czech-language tasks, albeit with trade-offs in increased latency and initial hardware investments.

Klíčová slova: large language models, document generation, web application, microservices

stránky: 338-347, online: 2026