Summary
In this chapter, we initially discussed what the problems of naïve RAG are. This allowed us to see a number of add-ons that can be used to solve the sore points of naïve RAG. Using these add-ons is the basis of what is now called the advanced RAG paradigm. Over time, the community then moved toward a more flexible and modular structure that is now called modular RAG.
We then saw how to scale this structure in the presence of big data. Like any LLM-based application, there are computational and cost challenges when you have to take the system from a development environment to a production environment. In addition, both LLMs and RAGs can have security and privacy risks. These are important points, especially when these products are open to the public. Today, there is an increasing focus on compliance and more and more regulations are being considered.
Finally, we saw that some issues remain open, such as the relationship with long-context LLMs or the multimodal...
