slAIces: an LLM Chatbot for Simplifying Experiments with the SLICES-RI

SLICES-RI is the outcome of several years of evolution of the concept of a networking test platform, that has recently transformed into a scientific instrument for the assessment of new research challenges in the Digital Infrastructures domain. Making large-scale scientific instruments easily accessible has always been challenging, often hindered by the complexity of the underlying platforms, the size and diversity of resources and the low-level settings that experimenters need to configure. Emerging technologies such as artificial intelligence(AI) have ignited a wave of innovation, inspiring companies, professionals, and researchers to incorporate chatbot assistants into their projects. These assistants revolutionize user experience by simplifying access to information and resources, significantly increasing productivity and presenting a familiar interface. In this work, we introduce slAIces, an LLM-based chatbot specifically designed to ease access to the SLICES-RI. slAIces utilizes the Generative Pre-trained Transformer(GPT) GPT-4 model to create a sophisticated Retrieval Augmented Generation(RAG) system. This system provides external knowledge by leveraging a properly preprocessed dataset, which includes documentation from SLICES-RI and all integrated testbeds. The contribution of this study lies in its methodology and recommendations for enriching the information on the test platform to enhance the quality of the service provided. We demonstrate that this chatbot significantly reduces the learning curve for new experimenters to become acquainted with the infrastructure. It exposes the appropriate level of abstraction, enabling experimenters to conduct complex experiments mobilizing the extensive and diverse resources available within a large-scale infrastructure like SLICES-RI.