Microsoft Fabric IQ
Fabric IQ (preview) is a Fabric workload for unifying data across OneLake and organizing it according to your business vocabulary. It exposes data to analytics, AI agents, and applications with consistent semantic meaning and context.
When to Use This Skill
- Creating or managing ontology items in Fabric IQ
- Defining entity types, properties, and relationship types
- Binding data from lakehouses, eventhouses, or semantic models to ontologies
- Generating ontologies from Power BI semantic models
- Configuring Fabric data agents with ontology as a source
- Working with Graph in Microsoft Fabric for traversals and graph queries
- Enabling Fabric IQ tenant settings in the admin portal
- Querying ontology graphs using the preview experience
- Building operations agents that reason across business concepts
- remediate ontology creation, data binding, or agent integration issues
- Automating Fabric IQ items via REST API or PowerShell
Prerequisites
- A Fabric workspace with a Microsoft Fabric-enabled capacity (F2+ or P1+)
- Required tenant settings enabled (see tenant-settings.md)
- Data in OneLake (lakehouse tables), an eventhouse, or Power BI semantic models
Fabric IQ Items Overview
Fabric IQ contains five items that work together:
| Item | Purpose | Shared With |
|---|---|---|
| Ontology (preview) | Enterprise vocabulary and semantic layer — entity types, relationships, properties, data bindings | IQ only |
| Graph in Microsoft Fabric (preview) | Native graph storage/compute for nodes, edges, traversals, path finding | Real-Time Intelligence |
| Fabric data agent (preview) | Conversational Q&A using generative AI, grounded in ontology | Data Science |
| Operations agent (preview) | AI agent to monitor real-time data and recommend actions | Real-Time Intelligence |
| Power BI semantic model | Curated analytics model for reporting and DAX | Power BI |
Choosing the Right Item
| Scenario | Use |
|---|---|
| Cross-domain consistency, governance, AI agent grounding | Ontology |
| Relationship-heavy questions (impact chains, shortest paths) | Graph |
| Trusted KPIs and fast visuals with dimensional modeling | Power BI semantic model |
| Operational context, stateful twins, what-if simulation | Digital twin builder (Real-Time Intelligence) |
Step-by-Step Workflows
Workflow 1: Create an Ontology from OneLake
For the complete walkthrough with all field mappings, see ontology-workflows.md.
- Navigate to your Fabric workspace and select + New item > Ontology (preview)
- Name the ontology (letters, numbers, underscores only — no spaces or dashes)
- Add entity types from the ribbon or canvas
- Bind static or time series data from OneLake sources
- Set entity type keys (unique identifier properties)
- Create relationship types between entity types and bind them to source data
- Use the preview experience to explore entity instances and the ontology graph
Workflow 2: Generate an Ontology from a Semantic Model
For the complete walkthrough, see ontology-workflows.md.
- Navigate to your Power BI semantic model in Fabric
- Select Generate Ontology from the ribbon
- Choose workspace and name the ontology
- Verify generated entity types, bindings, and relationship types
- Configure any incomplete relationship bindings manually
Workflow 3: Connect an Ontology to a Data Agent
For the complete walkthrough, see ontology-workflows.md.
- Create a Data agent item in your workspace
- Add the ontology as a knowledge source
- Add agent instructions (e.g.,
Support group by in GQL) - Test queries in the agent chat to validate semantic grounding
Workflow 4: Validate Tenant Prerequisites
Run the prereq validation script to check your environment:
./scripts/Validate-FabricIQPrereqs.ps1 -TenantId "your-tenant-id"
Key Concepts
Ontology Core Concepts
| Concept | Description |
|---|---|
| Entity type | Represents a real-world concept (e.g., Customer, Truck, Sensor) |
| Property | A fact about an entity type (e.g., name, email, temperature) |
| Entity type key | Unique identifier property for entity instances |
| Relationship type | Semantic connection between entity types (e.g., "drives", "has", "soldIn") |
| Data binding | Connects ontology definitions to concrete OneLake data sources |
| Ontology graph | Queryable instance graph built from data bindings and relationships |
Data Binding Types
| Type | Use Case | Example |
|---|---|---|
| Static | Descriptive attributes that change infrequently | Store locations, product catalog |
| Time series | Timestamped observations in columnar format | Sensor telemetry, temperature readings |
Naming Constraints
| Element | Rules |
|---|---|
| Ontology name | Letters, numbers, underscores. No spaces or dashes |
| Entity type name | 1-26 chars, alphanumeric + hyphens + underscores, start/end alphanumeric |
| Property name | 1-26 chars, alphanumeric + hyphens + underscores, unique across entity types for same type |
REST API Support
The Fabric REST API supports ontology CRUD operations:
| Operation | Supported |
|---|---|
| Create (without definition) | Yes |
| Create (with payload/definition) | Yes |
| Service principal support | Yes |
| Get | Yes |
| Update | Yes |
| Delete | Yes |
| List | Yes |
Use the Fabric CLI for command-line operations:
pip install ms-fabric-cli
fab auth login
remediate
For the full remediate guide, see remediate.md.
| Issue | Quick Fix |
|---|---|
| Unable to create ontology item | Enable all required tenant settings |
| Graph errors on new ontology | Enable User can create Graph (preview) tenant setting |
| Data agent 403 Forbidden | Enable Copilot and Azure OpenAI tenant settings |
| Generated ontology has no entity types | Ensure semantic model tables are visible (not hidden) |
| Generated ontology has no data bindings | Check semantic model mode — Import mode not supported |
| Decimal properties return null | Recreate property as Double type |
| Aggregation queries fail in data agent | Add instruction: Support group by in GQL |