Direct integration of graphs into Fabric IQ ontology (low-code / reuse approach)

[arabalca]

Hi team,

I would like to propose an enhancement regarding the use of graphs within Microsoft Fabric IQ.

Currently, graphs are available as a standalone element (in preview) and are part of the Fabric IQ ecosystem, which opens up strong capabilities for modeling complex relationships, agents, and advanced intelligence scenarios.

Proposal

It would be very valuable to enable the ability to directly incorporate a graph into the ontology, in a similar way to how other data sources (such as a Lakehouse) can be added today.

The idea would be to:

• Reference an existing graph within the ontology, without needing to replicate its definition
• Reuse already defined graph structures as part of the semantic model
• Maintain a single source of truth for relationships

Motivation

In many real scenarios:

• Graphs may be used initially as standalone components (for example, when an ontology is not yet defined)
• Or they may serve specific use cases such as relationship exploration, navigation, or recommendations

In these cases, it would be highly beneficial to later integrate those graphs into the ontology seamlessly, without having to rebuild or duplicate the logic.

Suggested approach

• Allow adding a graph as a resource within the ontology, similar to other sources
• Provide a declarative configuration (for example, JSON-based) to simplify integration
• Enable a low-code experience to map graph nodes and relationships to ontology entities
• Support direct reuse of existing graph relationships

Expected benefits

• Avoid duplication of relationship definitions
• Improve reuse of existing assets
• Enable a more natural integration between graphs and business semantics
• Enhance scenarios involving AI, agents, and advanced analytics
• Increase flexibility in solution evolution (graph → ontology or vice versa)

Closing

I believe this improvement would significantly enhance the usability and evolution of Fabric IQ, allowing teams to work more flexibly across graphs and ontologies without friction.

Happy to provide more detailed use cases if needed.

Thanks for your continued innovation