Graph index

Description

The Graph index action (previously Neo4j index) creates or drops indexes on node or relationship properties in a graph database, in the syntax of the database: Neo4j, Memgraph, FalkorDB or Apache AGE.

For similar operations on constraints, check the Graph constraint action.

Options

Graph database connection : The name of the graph database connection to update indexes on
Index updates to perform: You can specify your list of index updates here
- Update type : CREATE or DROP
- Type of object to index on: NODE or RELATIONSHIP
- Index name: optional but recommended name for your index
- Object name: the label of the node or relationship you want to update an index for
- Properties (, separated): the comma separated list of node or relationship properties to index
- Index type: RANGE for a regular property index (the default), VECTOR for a vector index
- Vector dimensions: VECTOR only, the number of numbers in each vector, for example 768 for nomic-embed-text
- Vector similarity: VECTOR only, COSINE (the default) or EUCLIDEAN
- Vector capacity: VECTOR on Memgraph only, the number of vectors to reserve room for, 1000 when empty

The index name, object name, properties, vector dimensions and vector capacity can be variables, for example ${CHUNK_INDEX} to keep a test and a production index apart. The action resolves them before it builds the statement, for create and drop, and the Cypher preview shows the resolved values. Use the same variable for the index name in Graph vector search.

Databases

Database	Range index	Vector index
Neo4j	Nodes and relationships	Nodes and relationships
Memgraph	Nodes and relationships	Nodes and relationships
FalkorDB	Nodes and relationships	Nodes and relationships
Apache AGE	Nodes, with a name	No
Amazon Neptune, Gremlin	No	No

Database

Range index

Vector index

Neo4j

Nodes and relationships

Memgraph

Nodes and relationships

FalkorDB

Nodes and relationships

Apache AGE

Nodes, with a name

Amazon Neptune, Gremlin

On Apache AGE an index is a PostgreSQL index on the table of the label, created and dropped with SQL. The action creates the label first when there are no nodes with it yet, so the index can be created before the data is loaded.

Vector indexes

A vector index is on a single property holding embeddings, see Vectors and embeddings. Neo4j, Memgraph and FalkorDB have vector indexes; the action refuses them on the other databases.

Database Notes

Database	Notes
Neo4j	Vector indexes on nodes and relationships, `CREATE VECTOR INDEX … FOR ()-[n:TYPE]-()` for relationships. Drop them by name.
Memgraph	Vector indexes on nodes and, from Memgraph 3, relationships (`CREATE VECTOR EDGE INDEX`), with a name, which you also drop them by. EUCLIDEAN uses Memgraph’s `l2sq` metric.
FalkorDB	Vector indexes on nodes and relationships, identified by label or relationship type and property like its other indexes.

Neo4j

Vector indexes on nodes and relationships, CREATE VECTOR INDEX … FOR ()-[n:TYPE]-() for relationships. Drop them by name.

Memgraph

Vector indexes on nodes and, from Memgraph 3, relationships (CREATE VECTOR EDGE INDEX), with a name, which you also drop them by. EUCLIDEAN uses Memgraph’s l2sq metric.

FalkorDB

Vector indexes on nodes and relationships, identified by label or relationship type and property like its other indexes.

For a vector index on relationships set the type of object to RELATIONSHIP and the object name to the relationship type. Search it with the Graph vector search transform, Search set to RELATIONSHIP.

Show Cypher Preview

The action dialog includes a "Show Cypher" button that displays a preview of the Cypher statements that will be executed for creating or dropping indexes. This preview shows the exact CREATE INDEX or DROP INDEX statements before execution.

The preview is read-only and helps you verify the index operations that will be performed.

Result

What this action passes on to the next action. See Action results for how results travel through a workflow.

Result	Value
True	All the index updates ran, with 0 errors. The drops run first, then the creates. An empty list is also true. An empty Index name is only fine where the database doesn’t use names: to create a RANGE index on Neo4j, which then names the index, for range indexes on Memgraph, and for every index on FalkorDB. In the other cases it stops the workflow, see the WARNING.
False	The action is never false: see the WARNING.
Result files	Passed on unchanged.
Result rows	Passed on unchanged.

Result

Value

True

All the index updates ran, with 0 errors. The drops run first, then the creates. An empty list is also true. An empty Index name is only fine where the database doesn’t use names: to create a RANGE index on Neo4j, which then names the index, for range indexes on Memgraph, and for every index on FalkorDB. In the other cases it stops the workflow, see the WARNING.

False

The action is never false: see the WARNING.

Result files

Passed on unchanged.

Result rows

Passed on unchanged.

When an index update can’t be done, the workflow stops with an error and no hop from the action is followed: the Graph database connection is empty, doesn’t exist or can’t be reached, an update has no Update type, an index is created without Object name or Properties (on Memgraph and FalkorDB, which identify an index by them, also when it is dropped, except a Memgraph vector index), a VECTOR index is created with Vector dimensions that are empty, not a number or not positive (a drop doesn’t need them), Index name is empty where the database needs one (to drop an index or create a vector index on Neo4j, for a vector index on Memgraph, for any index on Apache AGE), or the database refuses the statement. The updates before it stay done, the ones after it are not run.

On Neo4j and Apache AGE the action creates with IF NOT EXISTS and drops with IF EXISTS. Creating an index that already exists, or dropping one that doesn’t, does nothing and the action is still true. On Neo4j this is also the case when an index with the same name exists on another label or property, or an equivalent one exists with another name: nothing is created.