LatticeDB – Like SQLite but for graph databases

LatticeDB, an embedded single-file property-graph database developed by Jeff Hajewski, combines graph traversal, HNSW vector similarity search, and BM25 full-text search in one query layer, with performance benchmarks of 0.13 μs node lookups and 0.83 ms vector search at 1M vectors with 100% recall. The database is designed for local, relationship-heavy workloads such as Graph RAG and agent memory, and offers CLI, Python, TypeScript/Node.js, and Go bindings.

Embedded property-graph database with native vector and full-text indexing. LatticeDB is a single-file local database for connected, semantic, and textual data. It lets you traverse relationships, run vector similarity search, and do BM25 full-text search over the same dataset in one engine and one query layer. It is designed for relationship-heavy workloads on a single machine, with zero-config operation and an embedded single-writer model. LatticeDB is an embedded, single-file graph database that lets local applications query the same data by relationship, semantics, and text, then consume durable graph and application events from the same file. Workloads like Graph RAG, agent memory, and local knowledge tools are examples built on those primitives, not the definition of the engine. One file. Your entire database is a single portable file. No server, no configuration. One query layer. Graph traversal, HNSW vector similarity, and BM25 full-text — in the same query language. One event log. Durable named streams and a built-in graph changefeed share the same transaction/WAL path as graph writes. Local-first. Designed for one owning process on one machine, with WAL-backed durability. Fast. 0.13 μs node lookups. 0.83 ms vector search at 1M vectors with 100% recall. -- Find chunks similar to a query, traverse to their document, then to the author MATCH chunk:Chunk - :PART OF - doc:Document - :AUTHORED BY - author:Person WHERE chunk.embedding <= $query vector < 0.3 AND doc.content @@ "neural networks" RETURN doc.title, chunk.text, author.name ORDER BY chunk.embedding <= $query vector LIMIT 10 CLI curl -fsSL https://raw.githubusercontent.com/jeffhajewski/latticedb/main/dist/install.sh | bash Python pip install latticedb Published wheels are expected to bundle liblattice on supported platforms. Source installs can also bundle a staged native library during wheel builds with LATTICE BUNDLE LIB DIR=/path/to/lib . TypeScript / Node.js npm install @hajewski/latticedb Published package tarballs are expected to bundle liblattice on supported platforms. Source checkouts can stage the native library into the package with LATTICE BUNDLE LIB DIR=/path/to/lib npm run bundle:native . Go See bindings/go/README.md /jeffhajewski/latticedb/blob/main/bindings/go/README.md for the current cgo workflow. The default consumer path uses installed pkg-config metadata; in-repo development can use -tags repolocal against zig-out/lib . There is also a runnable graph/vector/text retrieval example in examples/go /jeffhajewski/latticedb/blob/main/examples/go . Recent binding-surface cleanups moved embedding helpers into dedicated modules and subpackages. See docs/client api migration.md /jeffhajewski/latticedb/blob/main/docs/client api migration.md for the preferred imports and current compatibility aliases. Getting Started /jeffhajewski/latticedb/blob/main/docs/getting started.md maps the shortest path for CLI, Python, TypeScript, and Go. CLI Quickstart /jeffhajewski/latticedb/blob/main/examples/cli/README.md is the smallest copy-paste example in the repo. Examples Overview /jeffhajewski/latticedb/blob/main/examples/README.md covers the larger graph/vector/text retrieval demos. A complete example: create a small knowledge graph with documents and authors, store embeddings, index text, then query across all three search modes. python from latticedb import Database from latticedb.embedding import hash embed with Database "knowledge.db", create=True, enable vectors=True, vector dimensions=128 as db: --- Build the graph --- with db.write as txn: Create authors alice = txn.create node labels= "Person" , properties={"name": "Alice", "field": "ML"} bob = txn.create node labels= "Person" , properties={"name": "Bob", "field": "Systems"} txn.create edge alice.id, bob.id, "COLLABORATES WITH" Create documents with chunks for title, text, author in "Attention Is All You Need", "The transformer architecture uses self-attention...", alice , "Scaling Laws for LLMs", "We find that model performance scales predictably...", alice , "Log-Structured Merge Trees", "LSM trees optimize write-heavy workloads...", bob , : doc = txn.create node labels= "Document" , properties={"title": title} chunk = txn.create node labels= "Chunk" , properties={"text": text} Store embedding and index text txn.set vector chunk.id, "embedding", hash embed text, dimensions=128 txn.fts index chunk.id, text txn.create edge chunk.id, doc.id, "PART OF" txn.create edge doc.id, author.id, "AUTHORED BY" txn.commit --- Query: vector search + text match + graph traversal --- results = db.query """ MATCH chunk:Chunk - :PART OF - doc:Document - :AUTHORED BY - author:Person WHERE chunk.embedding <= $query < 0.5 RETURN doc.title, chunk.text, author.name ORDER BY chunk.embedding <= $query LIMIT 5 """, parameters={"query": hash embed "transformer attention mechanism", dimensions=128 } for row in results: print f"{row 'doc.title' } by {row 'author.name' }" --- Full-text search --- for r in db.fts search "self-attention transformer" : print f"Node {r.node id}: score={r.score:.4f}" --- Aggregations --- stats = db.query """ MATCH doc:Document - :AUTHORED BY - p:Person RETURN p.name, count doc AS papers ORDER BY papers DESC """ for row in stats: print f"{row 'p.name' }: {row 'papers' } papers" js import { Database } from "@hajewski/latticedb"; import { hashEmbed } from "@hajewski/latticedb/embedding"; const db = new Database "knowledge.db", { create: true, enableVectors: true, vectorDimensions: 128, } ; await db.open ; // Build a graph await db.write async txn = { const alice = await txn.createNode { labels: "Person" , properties: { name: "Alice", field: "ML" }, } ; const doc = await txn.createNode { labels: "Document" , properties: { title: "Attention Is All You Need" }, } ; const chunk = await txn.createNode { labels: "Chunk" , properties: { text: "The transformer architecture uses self-attention..." }, } ; await txn.setVector chunk.id, "embedding", hashEmbed "transformer self-attention", 128 ; await txn.ftsIndex chunk.id, "The transformer architecture uses self-attention..." ; await txn.createEdge chunk.id, doc.id, "PART OF" ; await txn.createEdge doc.id, alice.id, "AUTHORED BY" ; } ; // Query across vector search + graph traversal const results = await db.query MATCH chunk:Chunk - :PART OF - doc:Document - :AUTHORED BY - author:Person WHERE chunk.embedding <= $query < 0.5 RETURN doc.title, chunk.text, author.name ORDER BY chunk.embedding <= $query LIMIT 5 , { query: hashEmbed "attention mechanism", 128 } ; for const row of results.rows { console.log ${row "doc.title" } by ${row "author.name" } ; } await db.close ; db, err := latticedb.Open "knowledge.db", latticedb.OpenOptions{ Create: true, EnableVectors: true, VectorDimensions: 128, } if err = nil { log.Fatal err } defer db.Close err = db.Update func tx latticedb.Tx error { node, err := tx.CreateNode latticedb.CreateNodeOptions{ Labels: string{"Chunk"}, Properties: map string latticedb.Value{"text": "The transformer architecture uses self-attention..."}, } if err = nil { return err } if err := tx.SetVector node.ID, "embedding", float32{1, 0, 0, 0} ; err = nil { return err } return tx.FTSIndex node.ID, "The transformer architecture uses self-attention..." } if err = nil { log.Fatal err } Benchmarked on Apple M1, single-threaded, with auto-scaled buffer pool. Run zig build benchmark to reproduce. For the repeated-term FTS indexing workload that previously exposed quadratic append behavior, run zig build fts-benchmark . | Operation | Latency | Throughput | Target | Status | |---|---|---|---|---| | Node lookup | 0.13 μs | 7.9M ops/sec | < 1 μs | PASS | | Node creation | 0.65 μs | 1.5M ops/sec | — | — | | Edge traversal | 9 μs | 111K ops/sec | — | — | | Full-text search 100 docs | 19 μs | 53K ops/sec | — | — | | 10-NN vector search 1M vectors | 0.83 ms | 1.2K ops/sec | < 10 ms @ 1M | PASS | 128-dimensional cosine vectors, M=16, ef construction=200, ef search=64, k=10. Run zig build vector-benchmark to reproduce. | Scale | Mean Latency | P99 Latency | Recall@10 | Memory | |---|---|---|---|---| | 1,000 | 65 μs | 70 μs | 100% | 1 MB | | 10,000 | 174 μs | 695 μs | 99% | 10 MB | | 100,000 | 438 μs | 1.2 ms | 99% | 101 MB | | 1,000,000 | 832 μs | 1.8 ms | 100% | 1,040 MB | Search latency scales sub-linearly O log N with 99–100% recall@10. Uses heuristic neighbor selection HNSW paper Algorithm 4 for diverse graph connectivity, connection page packing for ~4.5x memory reduction, and pre-normalized dot product for fast cosine distance. ef search Sensitivity 1M vectors | ef search | Mean Latency | Recall@10 | |---|---|---| | 16 | 506 μs | 57% | | 32 | 1.9 ms | 79% | | 64 | 990 μs | 100% | | 128 | 3.2 ms | 100% | | 256 | 11.6 ms | 100% | | System | Latency | Type | Source | |---|---|---|---| LatticeDB | 0.13 μs | Embedded | zig build benchmark | | RocksDB in-memory | 0.14 μs | Embedded | | Turso blog https://turso.tech/blog/microsecond-level-sql-query-latency-with-libsql-local-replicas-5e4ae19b628b marending.dev https://marending.dev/notes/sqlite-benchmarks/ Memgraph comparison https://memgraph.com/blog/memgraph-vs-neo4j-performance-benchmark-comparison LatticeDB's B+Tree achieves sub-microsecond cached lookups, matching RocksDB in-memory and outperforming SQLite on disk by 23x. | System | Latency 10-NN | Scale | Type | Source | |---|---|---|---|---| LatticeDB | 0.83 ms mean, 100% recall | 1M | Embedded | zig build vector-benchmark | | FAISS HNSW single-thread | 0.5–3 ms | 1M | Library | | Weaviate benchmarks https://docs.weaviate.io/weaviate/benchmarks/ann Qdrant benchmarks https://qdrant.tech/benchmarks/ VectorDBBench https://zilliz.com/vdbbench-leaderboard Jonathan Katz https://jkatz05.com/post/postgres/pgvector-performance-150x-speedup/ LanceDB blog https://medium.com/etoai/benchmarking-lancedb-92b01032874a Chroma docs https://docs.trychroma.com/production/administration/performance Pinecone blog https://www.pinecone.io/blog/dedicated-read-nodes/ Alex Garcia https://alexgarcia.xyz/blog/2024/sqlite-vec-stable-release/index.html LatticeDB at 1M achieves 0.83 ms mean with 100% recall@10 — faster than FAISS single-threaded HNSW and competitive with Weaviate and Qdrant server-based systems which add network overhead in practice . | System | 2-hop 100K nodes | Type | Source | |---|---|---|---| LatticeDB | 39 μs | Embedded | zig build sqlite-benchmark | | SQLite recursive CTE | 548 μs | Embedded | zig build sqlite-benchmark | | Kuzu | 19 ms | Embedded | | Neo4j blog https://neo4j.com/news/how-much-faster-is-a-graph-database-really/ LatticeDB vs SQLite — Social network graph with power-law degree distribution, adjacency cache pre-warmed: Small Scale 10K nodes, 50K edges | Workload | LatticeDB | SQLite | Speedup | |---|---|---|---| | 1-hop traversal | 560 ns | 13.0 μs | 23x | | 2-hop traversal | 3.0 μs | 37.5 μs | 13x | | 3-hop traversal | 19.1 μs | 178.5 μs | 9x | | Variable path 1..5 | 82.4 μs | 4.3 ms | 52x | Medium Scale 100K nodes, 500K edges | Workload | LatticeDB | SQLite | Speedup | |---|---|---|---| | 1-hop traversal | 8.0 μs | 290.0 μs | 36x | | 2-hop traversal | 38.7 μs | 548.3 μs | 14x | | 3-hop traversal | 197.3 μs | 1.2 ms | 6x | | Variable path 1..5 | 134.4 μs | 10.1 ms | 75x | Depth-Limited Traversal 10K nodes, 50K edges | Depth | LatticeDB | SQLite | Speedup | |---|---|---|---| | 10 | 311 μs | 121 ms | 390x | | 15 | 380 μs | 271 ms | 713x | | 25 | 318 μs | 587 ms | 1,848x | | 50 | 500 μs | 1.4 s | 2,819x | LatticeDB uses BFS with adjacency cache and bitset visited tracking. SQLite uses a recursive CTE with UNION deduplication. Both compute identical reachable node sets ~8K nodes . The gap widens at deeper depths as SQLite's CTE overhead grows with each recursion level. Run zig build graph-benchmark -- --quick to reproduce. | System | Search Latency | Type | Source | |---|---|---|---| LatticeDB | 19 μs | Embedded | zig build benchmark | | SQLite FTS5 | < 6 ms | Embedded | | LatticeDB's inverted index with BM25 scoring is ~300x faster than SQLite FTS5 and competitive with Tantivy a dedicated Rust search library . Graph - Nodes and edges with labels and arbitrary properties - Durable explicit equality indexes for scoped node and edge properties - Multi-hop traversal, variable-length paths 1..3 - ACID transactions with commit/rollback and crash recovery - MERGE, WITH, UNWIND, aggregations count , sum , avg , min , max , collect Vector Search - HNSW approximate nearest neighbor with configurable M, ef - Built-in hash embeddings or HTTP client for Ollama/OpenAI - Bulk vector node insertion for fast ingestion Full-Text Search - BM25-ranked inverted index with tokenization and stemming - Fuzzy search with configurable Levenshtein distance Cypher Query Language - MATCH, WHERE, RETURN, CREATE, DELETE, SET, REMOVE - ORDER BY, LIMIT, SKIP, DETACH DELETE - Vector distance operator: <= - Full-text search operator: @@ - Parameters: $name Operations - Single-file storage with write-ahead log for crash recovery - Durable named streams with explicit consumer offsets, manual trim, and graph changefeeds - Online freelist reuse plus lattice compact for safe physical tail reclamation - Zero configuration — open a file and start working - Embedded single-writer model for local applications - Clean C API; Python, TypeScript, and Go bindings wrap it Connected local data — Notes, documents, catalogs, citation graphs, and entity graphs Graph plus retrieval — Relationship traversal, semantic search, and lexical search over the same dataset Local knowledge tools — Embedded apps that need graph structure without running a separate server Agent memory and RAG pipelines — One example class of workload built on the graph/vector/text substrate Local development — Lightweight alternative to Neo4j or Weaviate for prototyping on one machine LatticeDB is fast, but speed is not the only thing that matters. Here are cases where a different tool is the better choice. You need multiple applications writing to the same database at the same time. LatticeDB is embedded with a single-writer model. One process opens the file and owns it. If you need many clients connecting over a network, use Neo4j, PostgreSQL, or another client-server database. Your data is fundamentally tabular. If your data fits naturally into rows and columns — sales records, user accounts, time series — a relational database like SQLite or PostgreSQL will be simpler and just as fast. Graph databases shine when relationships between records are the point, not an afterthought. You need to scale beyond a single machine. LatticeDB stores everything in one file on one machine. If you need sharding, replication, or distributed queries across billions of nodes, look at Neo4j cluster, Dgraph, or a managed service like Neptune. You need the full Cypher language. LatticeDB supports most of Cypher but not all of it. Features like OPTIONAL MATCH and CALL procedures are not yet implemented. If your queries depend on these, Neo4j is the complete implementation. You need mature tooling and ecosystem. Neo4j has visualization tools, admin dashboards, monitoring, drivers in every language, and years of community resources. PostgreSQL has decades of tooling. LatticeDB is new and lean — which is a strength for embedding, but a weakness if you need a rich operational ecosystem around your database. Written in Zig. No dependencies. git clone https://github.com/jeffhajewski/latticedb.git cd latticedb zig build build everything zig build test run tests zig build -Doptimize=ReleaseFast optimized build Getting Started /jeffhajewski/latticedb/blob/main/docs/getting started.md Durable Streams and Graph Changefeeds /jeffhajewski/latticedb/blob/main/docs/14 durable streams.md Property Indexes /jeffhajewski/latticedb/blob/main/docs/property index design.md Examples Overview /jeffhajewski/latticedb/blob/main/examples/README.md CLI Quickstart /jeffhajewski/latticedb/blob/main/examples/cli/README.md Architecture Overview /jeffhajewski/latticedb/blob/main/docs/00 introduction.md 0.10.0 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.10.0.md 0.9.6 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.9.6.md 0.9.5 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.9.5.md 0.9.0 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.9.0.md 0.8.7 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.8.7.md 0.8.6 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.8.6.md 0.8.5 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.8.5.md 0.8.4 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.8.4.md 0.8.2 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.8.2.md 0.8.0 Release Notes /jeffhajewski/latticedb/blob/main/docs/release notes 0.8.0.md Client API Migration Notes /jeffhajewski/latticedb/blob/main/docs/client api migration.md Python API Reference /jeffhajewski/latticedb/blob/main/bindings/python/README.md TypeScript API Reference /jeffhajewski/latticedb/blob/main/bindings/typescript/README.md Go API Reference /jeffhajewski/latticedb/blob/main/bindings/go/README.md C API Header /jeffhajewski/latticedb/blob/main/include/lattice.h