REBUILT IN RUST · SINGLE NODE
Turn similarity
into queries.
Vectors, structured data, and search in one engine.
Rebuilt in Rust. A clear starting point for your next similarity search.
Standalone Rust implementation · Local persistence · REST API · Embedded use
01 — A SEARCH YOU CAN SEE
See it. Understand it.
One query vector, a set of candidates.
Switch the distance metric to see which points become neighbors.
Choose a query position in the controls. Distances are computed exactly in your browser, with no server connection.
Find the nearest answers
Query positionLower L2 means closer; higher COSINE and IP mean more similar. Vectors are supplied by your application.
02 — FROM CODE TO QUERY
One service.
Your first query.
Requires Rust 1.89+ and a C toolchain.
Listens on localhost:19530 by default. No external database service needed.
03 — BUILT, NOT JUST PLANNED
Know what is ready.
Vector search, structured data, and persistence together.
This release focuses on a single node. Here is what is built and what is next.
Databases, collections, schemas, partitions, aliases, automatic primary keys, dynamic fields
ImplementedFLAT / IVF_FLAT / HNSW; dense, sparse, and binary vectors
ImplementedScalar filtering, JSON paths, range search, grouping search, RRF / weighted fusion
ImplementedAtomic batch writes, WAL persistence, restart recovery, snapshot backups
ImplementedREST API, Bearer tokens, health checks, Prometheus counters
ImplementedDistributed replication and sharding, gRPC, RBAC, BM25, GPU indexes
Not implementedThis is an early single-node implementation. ANN indexes are rebuilt on demand after writes; queries currently read collection data. Large-scale performance and production validation are not yet complete. SQLite is an embedded storage dependency.
04 — SIMPLE TO FOLLOW
From request to persistence.
Call over HTTP, or embed the engine directly in Rust.
Trace the interfaces, search algorithms, and persistence boundaries in the source.
FLAT · IVF_FLAT · HNSW
BUILD SOMETHING THAT FINDS.