langchain-arcadedb

This package contains the LangChain integration for ArcadeDB, the multi-model database with native graph, document, key-value, time-series, and vector support.

Installation

pip install langchain-arcadedb

This will also install the neo4j Python driver, which is used to connect to ArcadeDB via its native Bolt protocol.

Prerequisites

ArcadeDB must be running with the Bolt protocol plugin enabled.

Docker (quickest way)

docker run --rm -p 2480:2480 -p 7687:7687 \
    -e JAVA_OPTS="-Darcadedb.server.plugins=Bolt:com.arcadedb.bolt.BoltProtocolPlugin" \
    -e arcadedb.server.rootPassword=playwithdata \
    arcadedata/arcadedb:latest

Then create a database (e.g. via the HTTP API):

curl -X POST http://localhost:2480/api/v1/server \
    -d '{"command":"create database mydb"}' \
    -u root:playwithdata \
    -H "Content-Type: application/json"

Server / Embedded

If you run ArcadeDB from a distribution or embedded, add the Bolt plugin to your server configuration:

arcadedb.server.plugins=Bolt:com.arcadedb.bolt.BoltProtocolPlugin

See the ArcadeDB docs for details.

Quick Start

from langchain_arcadedb import ArcadeDBGraph

graph = ArcadeDBGraph(
    url="bolt://localhost:7687",
    username="root",
    password="playwithdata",
    database="mydb",
)

# Schema is auto-detected from existing data
print(graph.get_schema)

# Run Cypher queries
result = graph.query("MATCH (n:Person) RETURN n.name AS name")

Use with GraphCypherQAChain

ArcadeDBGraph implements the GraphStore protocol, so it works as a drop-in replacement wherever LangChain expects a graph store:

from langchain_arcadedb import ArcadeDBGraph
from langchain_neo4j import GraphCypherQAChain
from langchain_openai import ChatOpenAI

graph = ArcadeDBGraph(
    url="bolt://localhost:7687",
    username="root",
    password="playwithdata",
    database="mydb",
)

llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
chain = GraphCypherQAChain.from_llm(llm=llm, graph=graph, verbose=True)

answer = chain.invoke({"query": "Who does Alice know?"})

Import Graph Documents

You can import structured graph documents directly:

from langchain_arcadedb import ArcadeDBGraph, GraphDocument, Node, Relationship
from langchain_core.documents import Document

graph = ArcadeDBGraph(
    url="bolt://localhost:7687",
    username="root",
    password="playwithdata",
    database="mydb",
)

# Build a graph document
nodes = [
    Node(id="alice", type="Person", properties={"name": "Alice", "age": 30}),
    Node(id="bob", type="Person", properties={"name": "Bob", "age": 25}),
]
relationships = [
    Relationship(source=nodes[0], target=nodes[1], type="KNOWS"),
]
doc = GraphDocument(
    nodes=nodes,
    relationships=relationships,
    source=Document(page_content="Alice knows Bob"),
)

graph.add_graph_documents([doc])

Configuration

Connection parameters can be passed directly or via environment variables:

Key Features

• Bolt protocol -- connects via the standard Neo4j Python driver
• APOC-free -- schema introspection and document import use pure Cypher
• GraphStore protocol -- works with GraphCypherQAChain out of the box
• Graph document import -- batched MERGE operations grouped by type
• Lazy initialization -- connection is established on first use
• Schema auto-detection -- node labels, relationship types, and property types are inferred from sampled data

Development

# Install dependencies
uv sync --all-groups

# Run unit tests
make test

# Run integration tests (requires a running ArcadeDB instance)
make integration_tests

# Lint & format
make lint
make format

Links

• ArcadeDB Documentation
• ArcadeDB on GitHub
• LangChain Documentation