Neontology: Neo4j, Python and Pydantic

Easily ingest data into a openCypher / GQL (Graph Query Language) graph database like Neo4j using Python, Pydantic and pandas.

Neontology is a simple object-graph mapper which lets you use Pydantic models to define Nodes and Relationships. It imposes certain restrictions on how you model data, which aims to make life easier for most users in areas like the construction of knowledge graphs and development of graph database applications.

Neontology is inspired by projects like py2neo (which is no longer maintained), Beanie and SQLModel.

Read the documentation here.

Installation

pip install neontology

Example

from typing import ClassVar, Optional
import pandas as pd
from neontology import BaseNode, BaseRelationship, init_neontology, Neo4jConfig

# We define nodes by inheriting from BaseNode
class PersonNode(BaseNode):
    __primarylabel__: ClassVar[str] = "Person"
    __primaryproperty__: ClassVar[str] = "name"
    __secondarylabels__: ClassVar[Optional[list]] = ["Individual", "Somebody"]
    
    name: str
    age: int

# We define relationships by inheriting from BaseRelationship
class FollowsRel(BaseRelationship):
    __relationshiptype__: ClassVar[str] = "FOLLOWS"
    
    source: PersonNode
    target: PersonNode

# initialise the connection to the database
config = Neo4jConfig(
    uri="neo4j+s://mydatabaseid.databases.neo4j.io", 
    username="neo4j",
    password="<PASSWORD>"
)
init_neontology(config) 

# Define a couple of people
alice = PersonNode(name="Alice", age=40)

bob = PersonNode(name="Bob", age=40)

# Create them in the database
alice.create()
bob.create()

# Create a follows relationship between them
rel = FollowsRel(source=bob,target=alice)
rel.merge()

# We can also use pandas DataFrames to create multiple nodes
node_records = [{"name": "Freddy", "age": 42}, {"name": "Philippa", "age":42}]
node_df = pd.DataFrame.from_records(node_records)

PersonNode.merge_df(node_df)

# We can also merge relationships from a pandas DataFrame, using the primary property values of the nodes
rel_records = [
    {"source": "Freddy", "target": "Philippa"},
    {"source": "Alice", "target": "Freddy"}
]
rel_df = pd.DataFrame.from_records(rel_records)

FollowsRel.merge_df(rel_df)

Configuring your graph connection

On initialisation

You can explicitly provide access information as in the example above with a config object.

With a dotenv file or environment variables

You can use a .env file as below (or explicitly set them as environment variables), which should automatically get picked up by neontology.

# .env
NEO4J_URI=neo4j+s://myneo4j.example.com
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=<PASSWORD>

With the above environment variables defined, you can just use init_neontology() without providing any arguments.

Executing Queries

Neontology has limited, experimental support for running GQL/cypher queries.

Using a GraphConnection, you can call evaluate_query with a GQL/cypher query which returns nodes and relationships and get them back as Neontology Nodes Relationships.

Once neontology is initialized, only one connection to the database is used under the hood which can be accessed with GraphConnection.

from neontology import init_neontology, GraphConnection

# Only nodes and relationships that have already been defined can be returned as results.
class PersonNode(BaseNode):
    __primarylabel__: ClassVar[str] = "Person"
    __primaryproperty__: ClassVar[str] = "name"
    
    name: str
    age: int

class FollowsRel(BaseRelationship):
    __relationshiptype__: ClassVar[str] = "FOLLOWS"
    
    source: PersonNode
    target: PersonNode


init_neontology()

gc = GraphConnection()

cypher_query = """
MATCH (n)
WHERE n.name = $name
OPTIONAL MATCH (n)-[r:FOLLOWS]-(o:Person)
RETURN r, o
"""

results = gc.evaluate_query(cypher_query, {"name": "bob"})

results.records[0]["nodes"]["o"]    # Get the "o" result of the 1st record as a PersonNode
results.nodes                       # The nodes returned by the query as PersonNode objects
results.relationships               # The relationships as FollowsRel objects

For large or complex queries, data science or visualization/exploration, consider using a native driver or built-in interface (like Neo4j Browser/Bloom or Memgraph Lab).

Alternative Graph Engines

Neontology has experimental support for GQL/openCypher property graph databases other than Neo4j:

• Memgraph

Memgraph Engine

Memgraph is a Neo4j compatible database.

from neontology import init_neontology, MemgraphConfig
from neontology.graphengines import MemgraphEngine

config = MemgraphConfig(
                "uri": "bolt://localhost:9687",
                "username": "memgraphuser",
                "password": "<MEMGRAPH PASSWORD>"
            )

init_neontology(config)

You can also use the following environment variables and just init_neontology(MemgraphConfig()):

• MEMGRAPH_URI
• MEMGRAPH_USERNAME
• MEMGRAPH_PASSWORD