A publisher built a RAG solution to let users interact with daily-published content and wants near-real-time updates. Which RAG pipeline steps are appropriate to perform as offline batch processes? (Choose two.)

Generating embeddings for the published content is an appropriate offline batch process because content is published daily, allowing for a scheduled, non-real-time update to the embedding store. This pre-computation reduces query-time latency. Building the search index is also suitable for offline batch processing. After new content embeddings are generated, the search index needs to be updated to include this new information, which can be done periodically without impacting real-time user queries. Generating embeddings for incoming user queries, retrieving relevant documents, and generating the final response are all real-time operations that must occur when a user submits a query to provide an immediate answer.