Usearch assumes nodes are stored sequentially in an in memory array.
In Lantern, we get rid of the in memory storage and store usearch nodes in a postgres index.
Postgres index elements are identified by 48 bit tuple ids (BlockNumber + OffsetNumber)
We used to maintain a mapping between Usearch sequential ids and postgres tuple ids (tids).
This PR removes the need of such mapping.
This PR allows us to pass inserted tuple tid into Usearch so usearch can use that in neighbor list in place of the sequential id.
Usearch assumes nodes are stored sequentially in an in memory array. In Lantern, we get rid of the in memory storage and store usearch nodes in a postgres index. Postgres index elements are identified by 48 bit tuple ids (BlockNumber + OffsetNumber)
We used to maintain a mapping between Usearch sequential ids and postgres tuple ids (tids). This PR removes the need of such mapping. This PR allows us to pass inserted tuple tid into Usearch so usearch can use that in neighbor list in place of the sequential id.
PR making necessary changes in usearch: https://github.com/Ngalstyan4/usearch/pull/10