How to use graphrag with other database and structured data?

[jamesmkfoo23]

I have tried the "Get Started "guide (https://microsoft.github.io/graphrag/get_started/) by installing python graphrag, indexing and querying using graphrag cli which creates the output in local lancedb.

Q1) Is it possible to create the indexes into Azure Postgres Flexible server or Neo4J using ms graphrag?

Q2) Additionally what if I have these data sets:

1. structured data which shows entity relationships (entities are in bold)
   Sample:
   {
   "relationship": "Contains::Contained by",
   "parent.class": "site",
   "parent.name": "APAC-HUB-01",
   "child.description": "Windows Server for xxxx",
   "parent.description": "Site location",
   "child.class": "win_server",
   "child.name": "server01"
   }

2. entity and attributes. Each entity attributes can be selected and expanded as required. Sample from 1 shows the name, class and description for the entity

Is it possible to use graphrag with these types of data?

[sonnt-dna]

I have tried the "Get Started "guide (https://microsoft.github.io/graphrag/get_started/) by installing python graphrag, indexing and querying using graphrag cli which creates the output in local lancedb.

Q1) Is it possible to create the indexes into Azure Postgres Flexible server or Neo4J using ms graphrag?

Q2) Additionally what if I have these data sets:

1. structured data which shows entity relationships (entities are in bold)
   Sample:
   {
   "relationship": "Contains::Contained by",
   "parent.class": "site",
   "parent.name": "APAC-HUB-01",
   "child.description": "Windows Server for xxxx",
   "parent.description": "Site location",
   "child.class": "win_server",
   "child.name": "server01"
   }
2. entity and attributes. Each entity attributes can be selected and expanded as required. Sample from 1 shows the name, class and description for the entity

Is it possible to use graphrag with these types of data?

Q1: Is it possible to create the indexes into Azure Postgres Flexible server or Neo4J using MS GraphRAG?
Currently, MS GraphRAG primarily uses LanceDB as its default vector database for indexing and querying. However, integrating with Azure Postgres Flexible Server or Neo4J would require additional customization. Below are possible approaches:

Azure Postgres Flexible Server

GraphRAG does not natively support indexing directly into Azure Postgres Flexible Server.
However, you can store extracted structured data (e.g., entity relationships, text embeddings) into PostgreSQL manually after GraphRAG processes the data.
You can modify the indexing workflow to export results from GraphRAG into PostgreSQL.
Neo4J

GraphRAG is built for graph-based retrieval, making Neo4J a natural candidate for integration.
While there is no direct Neo4J support, you could:
Use GraphRAG's extracted relationships (nodes & edges) and write them into Neo4J.
Convert extracted text embeddings into Neo4J Graph Data Science (GDS) models for graph-based vector search.
Use LanceDB for embedding storage and Neo4J for relationship indexing.

✅ Summary: While GraphRAG does not natively support Azure Postgres Flexible Server or Neo4J, custom integration is possible using manual export and processing scripts.

Q2: Can GraphRAG work with structured data showing entity relationships?
Yes, GraphRAG can handle structured data like your provided sample, but with some considerations:

GraphRAG is optimized for unstructured/semi-structured text retrieval, but it supports structured entity relationships by:
Extracting Entities & Relationships: GraphRAG can extract entities (parent.class, child.class) and map relationships (Contains::Contained by).
Indexing Entities: You can store extracted entities in LanceDB or another database.
Expanding Attributes: GraphRAG can allow attribute expansion by defining custom schema mappings.
Implementation Options:

If relationship data is static, store it separately and link to GraphRAG’s text-based retrieval.
If you want GraphRAG to handle graph-like queries, consider using Neo4J or another graph database in parallel.

✅ Summary: GraphRAG can work with structured entity relationships, but it requires additional configuration to integrate with structured graph data.

Final Recommendations
For Postgres/Neo4J Integration, you may need custom scripts to store indexed data.
For structured entity relationships, you can preprocess data into GraphRAG or use Neo4J alongside GraphRAG.