rag-chunk-reorder

RAG chunk reordering, retrieval reranking, and context packing to mitigate lost-in-the-middle in LLM context windows. Drop-in reordering for LangChain, LlamaIndex, Vercel AI, and any RAG pipeline.