fabric-integration-bundle

# Fabric Integration Bundle Agent uses this bundle for Microsoft Fabric requests that span data movement, OneLake layout, semantic modeling, analytics serving, real-time processing, governance, diagnostics, or Copilot experiences. ## Activation | Need | Route | |---|---| | Copy, orchestration, ETL, or pipeline authoring | `fabric-data-factory-patterns` or `fabric-pipeline-diagnostics` | | Lakehouse landing, open-table storage, or ingestion flow | `fabric-lakehouses` or `fabric-lakehouse-ingestion` | | Warehouse SQL serving, semantic objects, or report models | `fabric-synapse-analytics`, `fabric-semantic-model-builder`, or `semantic-layer-fabric` | | Streaming, KQL, alerting, or incident diagnosis | `fabric-realtime-analytics`, `fabric-kql-diagnostics`, or `fabric-data-activator` | | OneLake layout, shortcuts, mirroring, workspaces, or promotion | `fabric-onelake-patterns`, `fabric-shortcuts-mirroring`, or `fabric-workspaces-governance` | | Data science, notebooks, Power BI, Copilot, or IQ-style ontology alignment | `fabric-data-science`, `fabric-notebooks-spark`, `fabric-powerbi-integration`, `fabric-copilot-studio`, `semantic-layer-fabric`, or `ontology-design-patterns` | ## Coverage Matrix | Skill | Coverage | |---|---| | `fabric-powerbi-integration` | Dashboards, metrics, report consumption | | `fabric-data-factory-patterns` | Copy, ETL, orchestration, scheduled movement | | `fabric-synapse-analytics` | Warehouse SQL serving and star-schema delivery | | `fabric-copilot-studio` | Conversational experience over Fabric assets | | `fabric-lakehouses` | Open tables, parquet, and lakehouse storage | | `fabric-lakehouse-ingestion` | Landing, bronze intake, and ingestion reliability | | `fabric-notebooks-spark` | Spark notebooks and engineering workflows | | `fabric-realtime-analytics` | Event streams, KQL, and low-latency analytics | | `fabric-kql-diagnostics` | KQL issue diagnosis and investigation | | `fabric-data-activator` | Triggered actions from data signals | | `fabric-onelake-patterns` | OneLake storage topology and item layout | | `fabric-data-science` | Experiments, features, and model workflows | | `fabric-workspaces-governance` | Workspace boundaries, promotion, and access | | `fabric-shortcuts-mirroring` | Virtual access and replicated source access | | `fabric-medallion-architecture` | Bronze, Silver, Gold design | | `fabric-pipeline-diagnostics` | Pipeline failure triage and correction | | `fabric-semantic-model-builder` | Model assembly and semantic-object authoring | | `semantic-layer-fabric` | Curated semantic layer, measures, and RLS | | `ontology-design-patterns` | IQ-style ontology and governed concept alignment | ## Decision Order | Order | Agent Action | |---|---| | 1 | Agent sets workspace boundary and OneLake layout. | | 2 | Agent selects shortcut, mirroring, ingestion, or ETL ownership. | | 3 | Agent selects lakehouse, warehouse, or real-time storage shape. | | 4 | Agent adds semantic model, Power BI, or Copilot surfaces from curated data only. | | 5 | Agent adds notebooks, data science, diagnostics, or Data Activator where the workload requires them. | | 6 | Agent adds ontology alignment when IQ-style semantic meaning needs a governed concept layer. | ## Workflow | Step | Agent Action | Test | Pass | |---|---|---|---| | 1. Map the workload | Agent maps movement, storage, semantic, analytics, and operations needs. | Agent lists routed skills. | Every need maps to one primary skill. | | 2. Select the data shape | Agent records batch, open-table, warehouse, or stream as the primary shape. | Agent records one primary shape. | Shape fits latency and query style. | | 3. Define the handoff chain | Agent records source, land, refine, serve, and consume stages. | Agent reviews stage order. | Flow has no missing stage. | | 4. Add diagnostics and governance | Agent adds pipeline, KQL, and workspace diagnostics only where runtime risk exists. | Agent reviews operations scope. | Governance and diagnosis stay intentional. | | 5. Verify the fit | Agent compares routes to latency, cost, access, and consumer inputs. | Agent reviews the final route set. | Route set matches the request. | ## Verification Matrix | Track | Agent Verifies | Test | Pass | |---|---|---|---| | Storage | Lakehouse, warehouse, and OneLake roles stay distinct. | Inspect storage ownership. | Storage fits the workload. | | Movement | Shortcut, mirroring, ingestion, and ETL paths stay distinct. | Inspect source-ingest path. | Movement matches source behavior. | | Processing | Batch, Spark, science, and stream paths fit latency. | Inspect cadence and compute surface. | Processing fits the request. | | Semantics | Semantic builder, semantic layer, Power BI, and ontology cues align. | Inspect model and concept ownership. | Metrics and concepts stay governed. | | Operations | Workspace, pipeline, and KQL diagnostics are explicit. | Inspect failure and access paths. | Operators have one clear diagnosis path. | ## Common Pitfalls | Pitfall | Agent Fix | |---|---| | Using Power BI as the main transformation layer | Agent moves heavy transformation upstream. | | Using ETL when shortcuts or mirroring fit the source better | Agent selects the lighter movement pattern. | | Building Copilot or semantic models on raw ingestion outputs | Agent inserts a curated semantic layer first. | | Treating KQL diagnostics as the same concern as streaming design | Agent routes design to real-time analytics and diagnosis to KQL diagnostics. | | Treating ontology alignment as a report-label cleanup task only | Agent routes governed concept work through ontology and semantic-layer skills. |