ParqDB takes a different approach: the vector index itself is stored in Parquet format. This makes the index a portable dataset that can be published, shared through object storage, and queried directly by clients (including browsers via WASM) without a database server.
More details and examples are available in our GitHub repo: https://github.com/parqdb-io/parqdb