A useful application of foundation models, or large language models (LLMs), is documentation search. LLMs can expand the scope of product support, enabling users to access specific details and insights without the need to manually sift through voluminous documents. Their ability to match the underlying semantics of user queries to product documents enables contextually relevant search results. Consequently, a well defined architecture with guardrails around LLMs can improve the time needed to arrive at solutions, thus expediting problem resolution and heightening overall customer satisfaction.
This project covers solution architectures and challenges around the primary objective of developing a robust Question-Answering system tailored to product documentation. For our experiments, we use public documents about the Red Hat Openshift on AWS (ROSA) service.