Reliable RAG systems are less about connecting an LLM to a vector database and more about the full retrieval loop: document structure, chunk quality, indexing decisions, ranking, context construction, evaluation, and latency.
This is a placeholder note for future writing.