🔍LatticeDB: A Single-File Database for Graphs and Vectors
A single-file database that does graphs, vectors, and text all at once?
TL;DR
LatticeDB is a new local database supporting graph traversal, vector similarity search, and full-text indexing in one file. Ideal for relationship-heavy workloads with zero-config operation.
LatticeDB just dropped: an embedded property-graph database that does vectors and text too. It's all about the single-file setup where you get graph traversal, vector similarity searches, and BM25 full-text search over a unified dataset. If you're dealing with relationship-heavy data on one machine, this is your new best friend. Zero-config operation means it just works out of the box, making it perfect for local-first designs and embedded applications. Node lookups take 0.13 μs, vector searches at 1M vectors hit 0.83 ms, and full-text search clocks in at 19 μs.
Key Points
Supports graph traversal, HNSW vector similarity, and BM25 full-text search in one layer
Node lookups take just 0.13 μs; node creation takes 0.65 μs
Vector search at 1M vectors hits 0.83 ms with 100% recall
Full-text search returns nodes with scores, aggregations return paper counts per author
Performance benchmarked on Apple M1, single-threaded, auto-scaled buffer pool
Why It Matters
If you're building a Graph RAG or an agent memory system, LatticeDB's local-first design and zero-config operation make it a game-changer. Workloads like these benefit from its ability to handle graph traversal, vector similarity searches, and full-text indexing in one file without any server setup.
Frequently Asked Questions
Why does this matter?
If you're building a Graph RAG or an agent memory system, LatticeDB's local-first design and zero-config operation make it a game-changer. Workloads like these benefit from its ability to handle graph traversal, vector similarity searches, and full-text indexing in one file without any server setup.
What happened?
LatticeDB is a new local database supporting graph traversal, vector similarity search, and full-text indexing in one file. Ideal for relationship-heavy workloads with zero-config operation.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,315 builders reading daily.