Make the supporting material inspectable
Local RAG explores a document-search workflow that brings retrieval, reranking, and source-linked answers together. The question is not only whether an answer sounds useful, but whether a reader can get back to the material that supports it.
Source and scope
The linked public books-RAG repository contains Python scripts and utilities, including embedding diagnostics and streaming-related code. It is the source destination already associated with this project in the portfolio. The process diagram is a conceptual explanation, not a screenshot of a production deployment.
The answer is only one part
An answer should help a reader return to its supporting material. That makes the retrieval and ranking steps important parts of the experience: they determine which evidence can reach the response in the first place. The public repository is the starting point for examining the code and experimenting with the workflow.