Chunking strategy for governments and internal org docs

A developer seeking advice on building a retrieval-augmented generation (RAG) system for government and internal organizational documents asks about optimal chunk sizes for BM25 and semantic search (cosine similarity) retrieval, and what data should be stored in a relational database (PostgreSQL) versus a vector database (OpenSearch or Milvus) for efficient retrieval. The query also requests recommendations for research papers or codebases to reference.

while creating rag for chunking strategy using BM25+Semantic search-cosine similarity for retrieval - the data it has governments documents like rules and laws and internal org docs, what would chunk size can be preferred, what are data best to store in relation db postgres and vector db Opensearch,/milvus separately, what each db should store to retrieve easily, fast and efficient Suggest any good research paper or any codebase i can look