Modeling Intelligent Agents for Information Retrieval Automation

Large Language Models continue to evolve rapidly and have become a key tool in the development of intelligent systems. Their applications extend beyond traditional text generation, enabling multistep reasoning, integration with external data sources, and automation of applied tasks. This paper explores the modeling of intelligent agents based on the ReAct architecture implemented within the LangChain framework. Special attention is given to the use of new-generation open-source language models (such as Gemma, Qwen, and others) and their comparison through benchmark testing. The developed agent is capable of performing intelligent information retrieval across documents and the web, utilizing external tools, and adapting to various models. The results of the study demonstrate the practical value of the agent-based approach for organizing data search and analysis, and emphasize the promising potential of LLM-based agents in modern applied scenarios.