Graphiti supports multiple graph database backends through a unified driver interface. Choose the database that best fits your deployment needs.
Supported Databases
Neo4j Production-ready graph database with full Cypher support
FalkorDB Redis-based graph database for in-memory performance
Amazon Neptune Fully managed graph database service on AWS
Kùzu Embedded graph database for local-first applications
Neo4j Driver Neo4j is the default and most widely used graph database with Graphiti.
Installation Neo4j support is included in the base Graphiti installation:
pip install graphiti-core
Connection from graphiti_core import Graphiti graphiti = Graphiti( uri = "bolt://localhost:7687" , user = "neo4j" , password = "your_password" )
Connection Parameters Parameter Type Description Default uristrNeo4j connection URI Required userstrDatabase username Required passwordstrDatabase password Required
Environment Variables NEO4J_URI = bolt://localhost:7687 NEO4J_USER = neo4j NEO4J_PASSWORD = your_password
Neo4j Desktop Setup
Create a Project
Open Neo4j Desktop and create a new project
Add a Database
Click “Add” → “Local DBMS” and set a password
Start the Database
Click the play button to start your local DBMS
Connect from Graphiti
graphiti = Graphiti( uri = "bolt://localhost:7687" , user = "neo4j" , password = "password" # Your chosen password )
Neo4j Aura (Cloud) For managed Neo4j hosting:
graphiti = Graphiti( uri = "neo4j+s://xxxxx.databases.neo4j.io" , user = "neo4j" , password = "your_password" )
FalkorDB Driver FalkorDB is a Redis module that provides graph database capabilities with in-memory performance.
Installation Install the FalkorDB driver:
pip install graphiti-core[falkordb]
Connection from graphiti_core import Graphiti from graphiti_core.driver.falkordb_driver import FalkorDriver # Create FalkorDB driver driver = FalkorDriver( host = "localhost" , port = "6379" , username = None , # Optional password = None # Optional ) # Initialize Graphiti with FalkorDB graphiti = Graphiti( graph_driver = driver)
Connection Parameters Parameter Type Description Default hoststrFalkorDB server host "localhost"portstrFalkorDB server port "6379"usernamestrUsername (optional) NonepasswordstrPassword (optional) None
Environment Variables FALKORDB_HOST = localhost FALKORDB_PORT = 6379 FALKORDB_USERNAME = # Optional FALKORDB_PASSWORD = # Optional
Running FalkorDB docker run -p 6379:6379 falkordb/falkordb:latest
FalkorDB Cloud For managed FalkorDB hosting, set your connection credentials:
driver = FalkorDriver( host = "your-instance.falkordb.cloud" , port = "6379" , username = "your_username" , password = "your_password" )
Amazon Neptune Driver Amazon Neptune is a fully managed graph database service optimized for AWS deployments.
Installation pip install graphiti-core[neptune]
Connection from graphiti_core import Graphiti from graphiti_core.driver.neptune_driver import NeptuneDriver driver = NeptuneDriver( host = "your-neptune-cluster.cluster-xxxxx.us-east-1.neptune.amazonaws.com" , port = 8182 , use_https = True ) graphiti = Graphiti( graph_driver = driver)
Connection Parameters Parameter Type Description hoststrNeptune cluster endpoint portintPort number (typically 8182) use_httpsboolUse HTTPS connection
AWS IAM Authentication Neptune supports IAM authentication for enhanced security:
driver = NeptuneDriver( host = "your-cluster.cluster-xxxxx.us-east-1.neptune.amazonaws.com" , port = 8182 , use_https = True , use_iam = True # Enable IAM authentication )
Ensure your AWS credentials are configured:
Kùzu Driver Kùzu is an embedded graph database perfect for local-first applications.
Installation pip install graphiti-core[kuzu]
Connection from graphiti_core import Graphiti from graphiti_core.driver.kuzu_driver import KuzuDriver driver = KuzuDriver( database_path = "./my_graph_db" # Local directory path ) graphiti = Graphiti( graph_driver = driver)
Connection Parameters Parameter Type Description database_pathstrPath to database directory
Use Cases Kùzu is ideal for:
Local development and testing
Desktop applications
Edge deployments
Embedded graph analytics
Driver Interface All drivers implement a unified interface, allowing you to switch databases without changing your application code:
from graphiti_core import Graphiti from graphiti_core.driver.neo4j_driver import Neo4jDriver from graphiti_core.driver.falkordb_driver import FalkorDriver # Both drivers work the same way driver = Neo4jDriver( uri = "bolt://localhost:7687" , user = "neo4j" , password = "password" ) # OR driver = FalkorDriver( host = "localhost" , port = "6379" ) graphiti = Graphiti( graph_driver = driver) # Same API regardless of driver await graphiti.add_episode( ... ) results = await graphiti.search( ... )
Passing Custom Drivers You can pass a pre-configured driver to Graphiti:
from graphiti_core import Graphiti from graphiti_core.driver.neo4j_driver import Neo4jDriver # Configure driver separately driver = Neo4jDriver( uri = "bolt://localhost:7687" , user = "neo4j" , password = "password" ) # Pass to Graphiti graphiti = Graphiti( graph_driver = driver)
This approach is useful when:
You need custom driver configuration
You want to reuse a driver across multiple Graphiti instances
You’re testing with different database backends
Database Initialization After connecting, build indices and constraints:
# Required for all drivers await graphiti.build_indices_and_constraints()
This creates:
Vector indices for semantic search
Full-text indices for keyword search
Uniqueness constraints on UUIDs
Performance-optimized indices
Always run build_indices_and_constraints() once before using a new database. This ensures optimal performance.
Choosing a Database Neo4j
FalkorDB
Neptune
Kùzu
Best for:
Production deployments
Large-scale graphs
Complex queries
Mature ecosystem
Considerations:
Requires separate database server
Commercial licensing for production
Best for:
In-memory performance
Real-time applications
Redis-based infrastructure
Simple deployment
Considerations:
Limited to available memory
Fewer features than Neo4j
Best for:
AWS-native deployments
Managed infrastructure
High availability
Compliance requirements
Considerations: Best for:
Local development
Desktop applications
Edge deployments
No server setup
Considerations:
Single-user access
Limited to single machine
Connection Pooling Drivers handle connection pooling automatically:
# Connections are pooled and reused graphiti = Graphiti( uri = "bolt://localhost:7687" , user = "neo4j" , password = "password" ) # Multiple operations share the connection pool await graphiti.add_episode( ... ) await graphiti.search( ... ) await graphiti.add_episode( ... )
Closing Connections Always close the driver when done:
try : await graphiti.add_episode( ... ) results = await graphiti.search( ... ) finally : await graphiti.close() # Close all connections
Or use as a context manager:
async with Graphiti( uri = "bolt://localhost:7687" , user = "neo4j" , password = "password" ) as graphiti: await graphiti.add_episode( ... ) results = await graphiti.search( ... ) # Automatically closed
Multi-Database Support Some drivers support multiple databases:
# Use different databases with the same driver driver = Neo4jDriver( uri = "bolt://localhost:7687" , user = "neo4j" , password = "password" ) # Clone for different database driver_production = driver.clone( database = "production" ) driver_staging = driver.clone( database = "staging" ) graphiti_prod = Graphiti( graph_driver = driver_production) graphiti_stage = Graphiti( graph_driver = driver_staging)
Next Steps
Adding Episodes Start adding content to your graph database
LLM Providers Configure your LLM provider
Embeddings Set up embedding providers