The previous tutorial, knowledge/agentic-rag, made the case for ditching the vector database entirely and letting an agent grep the source files. That works well when your corpus is code, runbooks, or ...
The European Reference Genome Atlas has 14 repositories available. Follow their code on GitHub.