From Proof of Concept to Product: Challenges of RAG Systems
The previous chapters introduced Retrieval-Augmented Generation (RAG) and the first practical implementations of it using text splitters, embeddings, and vector databases. These simple techniques may work in some cases and are often good enough for proof of concepts, but real-world projects and data vary a lot, and therefore, more advanced strategies may be required. Effective implementation of RAG applications poses specific challenges, like chunking, dealing with multimodality, updating documents, compliance, and evaluating the RAG performance.
The level of granularity in chunking is crucial for RAG systems to achieve precise retrieval results. Excessively large chunks may result in ...
Get Building LLMs for Production now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
