Vector databases

Building and operating an app's vector databases from the command line, where a corpus is normally loaded and tuned. add loads documents and the platform extracts, chunks, and embeds them; list, status, rm, and search cover the rest of day-to-day use: see what corpora exist, check per-document ingest state, remove a document, or query a corpus to sanity-check it before wiring it into the app. Configuration lives on the corpus, not in code: config shows and changes how documents are chunked and searched, split into settings that take effect on the next search and settings that require reprocessing every document. Rebuilds never touch what is live. revectorize builds a new version alongside the current one, search --candidate compares the two, promote cuts over with no downtime, and drop discards a version you do not want. This is the maintainer's side of the same corpus the app searches through the SDK.