LlamaIndex ArcadeDB Graph Store Integration

A PropertyGraphStore implementation for LlamaIndex that integrates with ArcadeDB, enabling GraphRAG applications with multi-model database capabilities.

Features

• PropertyGraphStore Interface: Full implementation of LlamaIndex's PropertyGraphStore interface
• Multi-Model Database: Graph, Document, Key-Value, and Vector data models in one database
• Native SQL Support: Uses ArcadeDB's high-performance SQL engine for graph operations
• Vector Search: Built-in vector similarity search capabilities
• Dynamic Schema: Automatic schema creation and management
• Production Ready: Docker support, comprehensive testing, and error handling

Requirements

• Python: 3.10 or higher
• ArcadeDB Server: 24.4.1 or higher
• arcadedb-python: Python client for ArcadeDB (automatically installed as dependency)
• LlamaIndex: Core components for PropertyGraph functionality

Installation

1. Install ArcadeDB Server

Docker (Recommended):

docker run --rm -p 2480:2480 -p 2424:2424 \
  -e JAVA_OPTS="-Darcadedb.server.rootPassword=playwithdata" \
  arcadedata/arcadedb:latest

Manual Installation:

# Download and extract ArcadeDB
wget https://github.com/ArcadeData/arcadedb/releases/latest/download/arcadedb-latest.tar.gz
tar -xf arcadedb-latest.tar.gz
cd arcadedb-*

# Start server
bin/server.sh -Darcadedb.server.rootPassword=playwithdata

2. Install Python Dependencies

# Install from PyPI
pip install llama-index-graph-stores-arcadedb

# Or install from source
pip install -e ./graph_stores/llama-index-graph-stores-arcadedb

# Install LlamaIndex components
pip install llama-index-core llama-index-embeddings-openai llama-index-llms-openai

Usage

Basic Example

import os
from llama_index.core import PropertyGraphIndex, Document
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.llms.openai import OpenAI
from llama_index.graph_stores.arcadedb import ArcadeDBPropertyGraphStore

# Set your OpenAI API key
os.environ["OPENAI_API_KEY"] = "your-api-key-here"

# Initialize ArcadeDB graph store
graph_store = ArcadeDBPropertyGraphStore(
    host="localhost",
    port=2480,
    username="root",
    password="playwithdata",
    database="knowledge_graph",
    embedding_dimension=1536  # OpenAI text-embedding-ada-002
)

# Create documents
documents = [
    Document(text="Apple Inc. was founded by Steve Jobs in 1976."),
    Document(text="Steve Jobs was the CEO of Apple until 2011."),
    Document(text="Tim Cook became Apple's CEO after Steve Jobs."),
]

# Create PropertyGraphIndex
index = PropertyGraphIndex.from_documents(
    documents,
    property_graph_store=graph_store,
    embed_model=OpenAIEmbedding(model_name="text-embedding-ada-002"),
    show_progress=True,
)

# Query the knowledge graph
query_engine = index.as_query_engine()
response = query_engine.query("Who founded Apple?")
print(response)

Class Parameters

Parameter	Type	Default	Description
host	str	"localhost"	ArcadeDB server host
port	int	2480	ArcadeDB server port
username	str	"root"	Database username
password	str	"playwithdata"	Database password
database	str	"graph"	Database name
create_database_if_not_exists	bool	True	Create database if it doesn't exist
include_basic_schema	bool	True	Include basic entity types (PERSON, ORGANIZATION, LOCATION, PLACE)
embedding_dimension	int	None	Vector dimension for embeddings (optional)

Schema Configuration

The PropertyGraphStore creates schema types from multiple sources:

Always Created (Core Types):

• Entity, TextChunk (vertex types)
• MENTIONS (edge type for chunk-to-entity relationships)

Basic Schema Types (include_basic_schema=True):

• PERSON, ORGANIZATION, LOCATION, PLACE (common entity types)

Dynamic Types from LlamaIndex/LLMs:

• Additional entity and relationship types discovered during processing
• Custom types from SchemaLLMPathExtractor configurations
• Types inferred by LLMs during knowledge graph extraction

# Default configuration - includes basic entity types
graph_store = ArcadeDBPropertyGraphStore(
    host="localhost",
    port=2480,
    username="root",
    password="playwithdata",
    database="knowledge_graph",
    include_basic_schema=True  # Pre-creates common entity types
)

# Minimal configuration - only core types, let LlamaIndex handle schema
graph_store = ArcadeDBPropertyGraphStore(
    host="localhost",
    port=2480,
    username="root",
    password="playwithdata",
    database="knowledge_graph",
    include_basic_schema=False  # Only creates Entity, TextChunk, MENTIONS
)

Embedding Dimensions

Choose the correct embedding_dimension based on your embedding model:

Model	Dimension	Usage
OpenAI text-embedding-ada-002	1536	Production OpenAI
Ollama all-MiniLM-L6-v2	384	Local embeddings
OpenAI text-embedding-3-small	1536	Cost-effective OpenAI
OpenAI text-embedding-3-large	3072	High-performance OpenAI

Advanced Usage

Direct Graph Operations

# Get nodes by properties
apple_nodes = graph_store.get(properties={"name": "Apple"})

# Get relationships
triplets = graph_store.get_triplets(entity_names=["Apple"])

# Execute custom SQL queries
results = graph_store.structured_query("""
    SELECT person.name as founder, company.name as company
    FROM Person person, Company company, FOUNDED rel
    WHERE rel.out = person.@rid AND rel.in = company.@rid
""")

Docker Deployment

Docker Compose

version: '3.8'
services:
  arcadedb:
    image: arcadedata/arcadedb:latest
    ports:
      - "2480:2480"
      - "2424:2424"
    environment:
      - JAVA_OPTS=-Darcadedb.server.rootPassword=playwithdata
    volumes:
      - arcadedb_data:/home/arcadedb/databases

  app:
    build: .
    depends_on:
      - arcadedb
    environment:
      - OPENAI_API_KEY=${OPENAI_API_KEY}
      - ARCADEDB_HOST=arcadedb
      - ARCADEDB_PORT=2480

volumes:
  arcadedb_data:

Testing

Run Tests

# Install test dependencies
uv pip install pytest pytest-asyncio

# Run tests
pytest tests/ -v

Integration Tests (Requires Running ArcadeDB)

# Start ArcadeDB server first
docker run --rm -p 2480:2480 -p 2424:2424 \
  -e JAVA_OPTS="-Darcadedb.server.rootPassword=playwithdata" \
  arcadedata/arcadedb:latest

# Run integration tests
pytest tests/test_final_integration.py -v

Examples

See the examples/ directory for complete working examples:

• basic_usage.py - Simple PropertyGraphStore usage
• advanced_usage.py - Custom schema and vector search
• migration_from_neo4j.py - Migration guide from Neo4j

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

References

• ArcadeDB - Multi-model database
• LlamaIndex Framework - Framework for building LLM applications
• arcadedb-python - Python client for ArcadeDB (available on PyPI)

Support

• Documentation: ArcadeDB Docs
• Issues: GitHub Issues