Skip to main content

Embedded · Rust · Cypher-like

The graph database for connected systems.

LoraDB is an in-process graph store with a Cypher-like query engine — small enough to embed in an agent, a robot, or a stream processor, and expressive enough to model the relationships those systems actually depend on.

QuickstartWhy LoraDBGithub7
  • Node.js · Python · WASM · Go · Ruby
  • Zero daemons · runs in your process
  • Open source · readable end-to-end
context.cypher
MATCH (a:Agent)-[:REMEMBERS]->(c:Context)
      -[:ABOUT]->(e:Entity)
WHERE c.updated_at > datetime() - duration('PT1H')
RETURN e.id, collect(c.summary) AS recent_context

The shape of the problem

Modern systems are graphs. Most databases aren’t.

Relational stores fight relational questions. Document stores fight evolving relationships. Graph platforms are often disproportionate — a service, a protocol, and a TCO that only pays off at scale, when all you wanted was a graph data structure next to the code that uses it. LoraDB is the option that was missing in the other direction: the one you reach for when the graph belongs inside your process.

Long formWhy an embedded graph at allThe argument in full — vs. SQL, vs. document stores, vs. managed graph platforms.

What LoraDB is

An engine, a store, and the surfaces that reach them.

Three places to start, depending on what you want to see first.

EngineA Cypher pipeline you can readA pragmatic subset — MATCH, WITH, WHERE, paths, aggregation — composed top to bottom in short, readable queries.StoreLabelled property graph, in processNodes, typed directed relationships, and properties — held in RAM, next to the code that uses them.SurfacesRust, Node, Python, WASM, Go, Ruby, HTTPPick the binding that fits the host process. Same Cypher, same result shape, same engine underneath.

Built for

Systems that reason over connected, evolving context.

AI agents & LLM pipelines

Tools, entities, observations and decisions as a live graph. Retrieval becomes a pattern match, not a similarity score.

Vector retrieval recipes

Context & memory systems

Model claims, evidence, citations, and contradictions as typed edges. Ask "why do we believe this?" as a traversal.

Graph patterns

Event pipelines & streams

Resolve entities, infer relationships, and enrich events in-process with Cypher rules that read top-to-bottom.

Event recipes

Embedded graph storage

A graph data structure inside your own process. No service to deploy, no protocol to speak, no daemon to babysit.

Pick a binding
Browse the full cookbook

Start in a minute

Add the package. Open a database. Write a query.

There’s no server to stand up, no protocol to speak. Opening a LoraDB is a function call — in Node.js, Python, WASM, Go, or Ruby. Same Cypher, same result shape, across every binding.

InstallCheat sheetTen-minute tour
quickstart.ts
import { createDatabase } from '@loradb/lora-node';

const db = await createDatabase();           // in-memory
// const db = await createDatabase('app', { databaseDir: './data' }); // ./data/app.loradb

await db.execute(
  "CREATE (:Person {name: 'Ada'})-[:INFLUENCED]->(:Person {name: 'Grace'})"
);

const result = await db.execute(
  "MATCH (a)-[:INFLUENCED]->(b) RETURN a.name, b.name"
);

console.log(result.rows);

Where to next

Pick a path into the docs.

InstallI just want to get startedPick a binding, install, and ship a hello-world in a minute.ConceptsI want to understand the modelNodes, relationships, properties, and how the engine sees them.ExamplesI want query examplesA copy-paste tour of the Cypher LoraDB supports.EvaluateWhat does it support — and not?The full capability surface, plus the lines we won’t pretend to cross.
Or read the docs from the topGithub7