[Question]: I already have a graph knowledge stored in neo4j. is there a way to directly use it for Property Graph index?

[justinzyw]

Question Validation

• [X] I have searched both the documentation and discord for an answer.

Question

I already have a graph knowledge stored in neo4j. is there a way to directly use it for Property Graph index?

[dosubot[bot]]

Yes, you can directly use your existing graph knowledge stored in Neo4j for a Property Graph index. Here is how you can connect to and use an existing graph:

from llama_index.graph_stores.neo4j import Neo4jPropertyGraphStore
from llama_index.core import PropertyGraphIndex
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.llms.openai import OpenAI

graph_store = Neo4jPropertyGraphStore(
    username="neo4j",
    password="neo4j",
    url="bolt://localhost:7687",
    database="neo4j"
)

index = PropertyGraphIndex.from_existing(
    property_graph_store=graph_store,
    llm=OpenAI(model="gpt-3.5-turbo", temperature=0.3),
    embed_model=OpenAIEmbedding(model_name="text-embedding-3-small"),
)

From here, you can still insert more documents:

from llama_index.core import Document

document = Document(text="LlamaIndex is great!")

index.insert(document)

And retrieve nodes:

nodes = index.as_retriever(include_text=False).retrieve("LlamaIndex")

print(nodes[0].text)

For full details on construction, retrieval, and querying of a property graph, see the full docs page ^[1][2][3][4].

To continue talking to Dosu, mention @dosu.

[logan-markewich]

Yes*

But you are restricted to

• text to cypher retriever
• cypher template retriever
• custom retrievers

[justinzyw]

many thanks!

I am not familiar with these retrievers and will look into them.

it would be great that llama index could read an existing external graph data as document and index it as property graph

[logan-markewich]

yea thats pretty ambitious/difficult to do actually

[justinzyw]

by reading the documents, I found that there is a way to manually insert entities and relations as below. I can live with it too as I can somehow transform my knowledge graph into insert statements.. The only problem is that, as I already have the id of each entity, would it be possible to directly set the id of entity? @dosu

from llama_index.core.graph_stores.types import EntityNode, ChunkNode, Relation

Create a two entity nodes

entity1 = EntityNode(label="PERSON", name="Logan", properties={"age": 28}) entity2 = EntityNode(label="ORGANIZATION", name="LlamaIndex")

Create a relation

relation = Relation( label="WORKS_FOR", source_id=entity1.id, target_id=entity2.id, properties={"since": 2023}, )

[dosubot[bot]]

Yes, it is possible to directly set the ID of an entity when creating EntityNode instances. You can do this by providing the id parameter during the creation of the EntityNode. Here is an example:

from llama_index.core.graph_stores.types import EntityNode, Relation

# Create two entity nodes with custom IDs
entity1 = EntityNode(id="custom_id_1", label="PERSON", name="Logan", properties={"age": 28})
entity2 = EntityNode(id="custom_id_2", label="ORGANIZATION", name="LlamaIndex")

# Create a relation
relation = Relation(
    label="WORKS_FOR",
    source_id=entity1.id,
    target_id=entity2.id,
    properties={"since": 2023},
)

This allows you to use your existing IDs directly when creating the entities ^[1][2].

To continue talking to Dosu, mention @dosu.