The objective of this thesis is to configure and integrate an existing Large Language Model (LLM) to interpret natural language requests and translate them into modifications within an Apache Kafka cluster.

This task involves utilizing a LLM to accurately interpret human language in the context of system configuration tasks. Additionally, it includes mapping these interpreted requests to an event hub platform that manages data stream topology and generates corresponding configuration changes.

The system will autonomously transform user requests into event stream configuration representations, apply the necessary modifications to the Kafka cluster, and submit the resulting changes as merge requests to a Git repository that maintains the current state of the deployed Kafka artifacts.

The goal of this thesis is to improve the end-to-end integration of the LLM with Apache Kafka, with a particular focus on managing active data streams. The aim is to establish a seamless and reliable workflow that automates processes and minimizes the need for manual intervention. Conducted in collaboration with SOFTEC spol. s r.o., this work aims to deliver a practical, scalable solution for event-driven system management in alignment with industry standards.