lattice

CLI for the LatticeDB embedded graph, vector, and full-text database

TLDR

Create a new database file

$ lattice create [knowledge.lattice]

Import graph data from JSON

$ lattice import [knowledge.lattice] --file [graph.json]

Print node/edge counts

$ lattice count [knowledge.lattice]

Run a Cypher file

$ lattice exec [knowledge.lattice] --file [query.cypher]

Check page checksums

$ lattice check [knowledge.lattice]

Compact free pages at the end of the file

$ lattice compact [knowledge.lattice]

Take a hot backup without closing the database

$ lattice backup [knowledge.lattice]

SYNOPSIS

lattice command database [options]

lattice is the command-line client for LatticeDB, an embedded, single-file property-graph engine with HNSW vector search and BM25 full-text search. There is no server: one process opens one file. The query language is a Cypher subset with `<=>` for vector distance and `@@` for full-text.Install via the project's `dist/install.sh` or build with Zig (`zig build` produces `zig-out/bin/lattice`). Language bindings (Python, TypeScript, Go, C) share the same file format.

COMMANDS

create file

Create an empty single-file database.

import file --file json

Load graph data from JSON.

exec file --file query.cypher

Run a Cypher query from a file.

count file

Print node and edge counts.

check file

Validate main-file page checksums. A sibling `-wal` file is reported if present but not fully validated.

compact file

Checkpoint and truncate a contiguous free tail. Does not relocate live pages; zero bytes reclaimed is normal.

backup file

Hot backup of an open database.

CAVEATS

Single-writer: only one process should open a given file for writes. Cypher coverage is incomplete (no `OPTIONAL MATCH` or `CALL` yet). `hash_embed` in examples is a placeholder, not a semantic model. Compact only trims a free tail.

HISTORY

Written in Zig by Jeff Hajewski as a local-first graph+vector store for RAG and agent memory.

RESOURCES

Source code · Homepage · Documentation