Cayley — Open-Source Graph Database Inspired by Google's Knowledge Graph

Introduction

Cayley is a graph database written in Go, originally created at Google and open-sourced for community use. It provides a flexible query layer supporting Gizmo (JavaScript-based) and GraphQL, backed by pluggable storage engines ranging from in-memory to PostgreSQL and MongoDB.

What Cayley Does

• Stores and queries graph-structured data using subject-predicate-object triples (quads)
• Supports multiple query languages: Gizmo (JS), GraphQL, and MQL
• Pluggable backend storage: BoltDB, PostgreSQL, MongoDB, in-memory, and more
• Provides a built-in web UI for visual graph exploration and query execution
• Imports and exports standard RDF formats (N-Quads, JSON-LD, Turtle)

Architecture Overview

Cayley uses a quad store model where each fact is stored as a quad (subject, predicate, object, label). The query layer compiles traversal expressions into an iterator-based execution plan that operates against a pluggable backend. Storage engines implement a common interface, allowing swaps between BoltDB for embedded use, PostgreSQL for durability, or in-memory stores for testing.

Self-Hosting & Configuration

• Install via Go toolchain or download prebuilt binaries from GitHub releases
• Configure storage backend with --dbpath flag or YAML config file
• Default HTTP port is 64210; override with --host and --port flags
• Load initial data with cayley load --quads=<file> supporting N-Quads and JSON-LD
• Deploy with Docker using the official cayleygraph/cayley image

Key Features

• Multiple query languages let teams pick what fits their mental model
• Pluggable backends mean you can start embedded and scale to PostgreSQL later
• Built-in web visualizer for interactive graph exploration
• Native Go library for embedding directly into Go applications
• RDF-compatible data model aligns with linked data standards

Comparison with Similar Tools

• Neo4j — More mature with Cypher, but requires JVM and commercial license for clustering
• Dgraph — Distributed GraphQL-native DB with higher operational complexity
• ArangoDB — Multi-model but heavier; Cayley is lighter for pure graph workloads
• Apache Jena — Full semantic web stack in Java; Cayley is simpler and Go-native

FAQ

Q: What query language should I start with? A: Gizmo (JavaScript-based) is the most documented and flexible for traversals. GraphQL works well if you already know your schema.

Q: Can Cayley handle large datasets? A: Yes, with PostgreSQL or MongoDB backends it handles millions of quads. For billions, consider a distributed graph DB instead.

Q: Is Cayley still maintained? A: The project has slower release cadence but remains functional and accepts contributions.

Q: Can I use Cayley as an embedded library? A: Yes, import the Go packages directly and use the graph API without running a server.

Sources

• https://github.com/cayleygraph/cayley
• https://cayley.gitbook.io/cayley