/** * RetrievalCommand — `buff retrieval` — inspect and drive the vector retrieval layer. * * The retrieval engine (src/learning/retrieval.ts) turns large code/doc context * into token-efficient, semantically-relevant context using a local embedding * model (bge-small-en-v1.5) + the pure-JS VectorStore. It complements the * quota ledger: retrieval SAVES tokens, the ledger MANAGES quotas. * * Subcommands: * buff retrieval stats — token-savings transparency (dashboard data) * buff retrieval index