place_document() embeds, chunks, and stores in one call — and records exactly which model produced each vector.

python

docs = db.vault("documents")
docs.place_document(
    "Long markdown article ...",
    {"kind": "design"},
)

for hit in docs.seek_text("alpha", top=3):
    print(hit.document.text[:80])
    print(hit.chunk.start, hit.chunk.end)
    print(hit.lineage.provider, hit.lineage.model, hit.lineage.revision)

Lineage is what lets you re-embed only the records produced by a stale model. It's also what powers exact citations.