Why every RAG project I've built ends up fighting the pipeline — and what I'm doing about it

The author describes a recurring pain point in RAG projects: iterating on chunking, embedding, and retrieval strategies requires rewiring the entire pipeline, so most teams ship suboptimal solutions. To address this, they are building klay+, a composable RAG infrastructure layer where each component (ingestion, chunking, embedding, retrieval) is independently swappable without changing application code. The standout feature is parallel projections, which let teams test new strategies side by side with production indexes before committing.