Generative AI for Low-Level NETCONF Configuration in Network Management Based on YANG Models
Gergely Hollósi, Dániel Ficzere, Pál Varga
Abstract
The NETCONF protocol, standardized by the IETF, is a cutting-edge solution for configuring network entities and offers an alternative to SNMP in modern network devices. Due to the complexity of configuration protocols and the challenges in creating valid configurations, generative AI solutions are promising for converting textual prompts into configuration descriptors. However, the potential of LLMs to generate NETCONF configurations has not been explored in the literature. This paper addresses this gap by evaluating the performance of five different LLMs – including Llama3, an open-source, on-premises capable model – in creating NETCONF configurations using the widespread YANG data models. In order to create valid network configurations using generative AI, this paper proposes a pipeline for integrating domain knowledge into LLMs without additional training and highlights common shortcomings and errors that prevent the generation of valid configurations. The findings indicate that the use of LLMs is promising for this task, but the current state-of-the-art is not yet mature enough for immediate industrial application in complex cases.