Vector Index

A stored embedding per memory, computed once and reused, rather than recomputed on every query.

Why it matters in a memory layer

Without one, retrieval embeds every memory on every query — measured here, 2N calls per query, 1,480 of them on a 740-memory store, for content that never changes. What makes it safe is that Memory.id is content-addressed: an edited memory is a different id and therefore a different cache entry, so a cached vector can never describe text it does not match. What makes it interesting is what the cache cannot see — a superseded belief keeps its id and its content, so its vector stays perfectly valid for something nobody should retrieve.

Connections

Taught in: Vector Stores for Data That Changes

Used in: Caching, Batching, Routing · The Write Path Dominates · Hybrid Architecture · The Underrated Default

Do not confuse with: Vector Search