What is LiteGraph?

LiteGraph is a property graph database with support for graph relationships, tags, labels, metadata, data, and vectors. LiteGraph is intended to be a unified database for providing persistence and retrieval for knowledge and artificial intelligence applications.

LiteGraph provides a lightweight yet powerful solution that combines the best of multiple database paradigms - graph databases for relationship modeling, vector databases for AI/ML embeddings, and document stores for flexible metadata storage.

Key Features

Multi-modal Database: Supports graph structures, relational data, vectors, and metadata in a single system
Property Graph Model: Full support for nodes, edges, labels, tags, and custom data properties
Vector Support: Native vector storage and similarity search capabilities for AI applications
Flexible Deployment: Can be run in-process (using LiteGraphClient) or as a standalone RESTful server (using LiteGraph.Server)
In-Memory Operation: Optional in-memory mode with controlled flushing to disk
Multi-tenancy: Built-in support for multiple tenants and graph isolation
Export Capabilities: Export graphs to GEXF format for visualization
No Dependencies: Leverages SQLite for a zero-configuration embedded database

Use Cases

Knowledge Management

Build knowledge graphs for organizing complex information
Create semantic networks for content relationships
Implement recommendation systems based on graph traversal

Artificial Intelligence Applications

Store and query vector embeddings from LLMs
Build RAG (Retrieval-Augmented Generation) systems
Implement similarity search for semantic content matching
Create hybrid search combining graph relationships and vector similarity

Application Development

Embed a graph database directly into applications without external dependencies
Build networks with detailed, rich relationships
Model organizational hierarchies and permissions
Track dependencies and relationships in complex systems

Research and Analysis

Network analysis and graph algorithms
Pattern detection in connected data
Path finding and route optimization
Community detection and clustering

Comparing LiteGraph

Vs Traditional Relational Databases

Native Graph Support: First-class support for nodes, edges, and graph traversal
Flexible Schema: No rigid table structures; properties can be added dynamically
Relationship-Centric: Optimized for querying connected data rather than tabular data
Multi-modal: Combines relational, graph, and vector capabilities

Vs Vector Databases

Graph Context: Vectors are enriched with graph relationships and metadata
Unified Storage: No need for separate graph and vector databases
Flexible Filtering: Use labels, tags, and expressions (data filters) when performing a vector search
Lightweight: Built on SQLite rather than requiring specialized infrastructure

Vs Metadata/Feature Stores

Relationship Modeling: Goes beyond key-value storage to model complex relationships
Type Flexibility: The Data property is an object and can be attached to any Graph, Node, or Edge. Data supports any object serializable to JSON
Built-in Search: Native support for graph traversal and vector similarity search
Version History: Timestamps for creation and updates on all entities

Vs Traditional Graph Databases

Embedded Operation: No separate server required; runs in-process with your application
SQLite Foundation: Leverages proven SQLite reliability and performance
Vector Integration: Native vector support without extensions or plugins
Simplified Deployment: Single file database with no complex configuration

Getting Started

Install via NuGet:

dotnet add package LiteGraph

Basic Usage

using LiteGraph;

LiteGraphClient graph = new LiteGraphClient(new SqliteRepository("litegraph.db"));
graph.InitializeRepository();

// Create a tenant
TenantMetadata tenant = graph.CreateTenant(new TenantMetadata { Name = "My tenant" });

// Create a graph
Graph graph = graph.CreateGraph(new Graph { TenantGUID = tenant.GUID, Name = "This is my graph!" });

// Create nodes
Node node1 = graph.CreateNode(new Node { TenantGUID = tenant.GUID, GraphGUID = graph.GUID, Name = "node1" });
Node node2 = graph.CreateNode(new Node { TenantGUID = tenant.GUID, GraphGUID = graph.GUID, Name = "node2" });
Node node3 = graph.CreateNode(new Node { TenantGUID = tenant.GUID, GraphGUID = graph.GUID, Name = "node3" });

// Create edges
Edge edge1 = graph.CreateEdge(new Edge { TenantGUID = tenant.GUID, GraphGUID = graph.GUID, From = node1.GUID, To = node2.GUID, Name = "Node 1 to node 2" });
Edge edge2 = graph.CreateEdge(new Edge { TenantGUID = tenant.GUID, GraphGUID = graph.GUID, From = node2.GUID, To = node3.GUID, Name = "Node 2 to node 3" });

// Find routes
foreach (RouteDetail route in graph.GetRoutes(
	SearchTypeEnum.DepthFirstSearch,
  tenant.GUID,
	graph.GUID,
	node1.GUID,
	node2.GUID))
{
  Console.WriteLine(...);
}

Getting Started with Docker

Using Docker Hub Image A Docker image is available in Docker Hub under jchristn/litegraph.

Quick Start with Docker Compose

Navigate to the Docker directory in the repository Use the Docker Compose start (compose-up.sh and compose-up.bat) and stop (compose-down.sh and compose-down.bat) scripts

Manual Docker Setup

docker pull jchristn/litegraph

Run with volume mounts for database and configuration

docker run -d \
-p 8701:8701 \
-v /path/to/litegraph.db:/app/litegraph.db \
-v /path/to/litegraph.json:/app/litegraph.json \
jchristn/litegraph

Ensure that you have a valid database file (e.g. litegraph.db) and configuration file (e.g. litegraph.json) exposed into your container.

REST API Server

By default, LiteGraph.Server (and by consequence, the Docker image) listens on http://localhost:8701 and is only accessible to localhost. Modify the litegraph.json file to change settings including hostname and port.