Notes from the index
Short essays on embeddable vector search, ANN index trade-offs, and the day-to-day mechanics of shipping retrieval inside a Rust binary.
2026-05-29 persistencesqliteoperations
Persistence patterns for in-process vector indexes
An in-memory vector index is easy. An in-memory vector index that survives a crash, a deploy, and a schema change is the actual work.
2026-05-21 annhnswindexing
HNSW vs flat: choosing for embeddable workloads
A flat index is honest. HNSW is fast. For embeddable workloads, the choice is rarely either-or — it's about where in your data's life each one earns its keep.
2026-05-12 architectureragembeddable
When you don't need a vector DB
Most retrieval workloads fit on one box. Here is how to know yours is one of them — and what 'no vector DB' actually buys you.