What Real-Time Intelligence is for#
Real-Time Intelligence (RTI) is Fabric's end-to-end answer to event-driven and streaming scenarios: ingesting data as it arrives, transforming it, storing it, modeling it, analyzing it, and firing automated actions off it — all in one purpose-built solution [S1].
Two items do most of the work. Eventstream is a no-code pipeline for collecting, transforming, and routing high-volume real-time data to multiple destinations at once [S1]. Activator watches data flowing through Real-Time hub, KQL queries, or Power BI reports and fires actions — notifications, pipeline runs, Spark jobs, Dataflow runs, User Data Function calls, or Power Automate flows — when a threshold, pattern, or KQL-defined condition is met [S1]. Everything surfaces in Real-Time hub, the tenant-wide catalog of data in motion — the moving-data counterpart to OneLake, cataloging eventstreams, KQL tables, Microsoft-managed streams, and Fabric/Azure item events for every workload [S1] [S2].
Getting data in: Eventstream#
Eventstream's connector catalog spans Azure-native sources (Event Hubs, Event Grid, IoT Hub), database change-data-capture connectors (Azure SQL, PostgreSQL, MongoDB, Oracle, Cosmos DB), generic protocols (Kafka, MQTT, HTTP), competing cloud services (Kinesis, Pub/Sub), and Fabric-native events [S3]. A drag-and-drop processor filters, aggregates, groups, and joins the stream before routing it — to Eventhouse, a Lakehouse, a Spark notebook, or straight into Activator — with multiple destinations running independently off one stream [S3].
A worked example#
Say you want to watch a location's weather. The Real-Time Weather source needs only a place name or coordinates — no Azure subscription, no Event Hub, no credentials [S4]. Point it at Eventhouse and weather data lands continuously. Add an Activator rule: "if temperature crosses a threshold, notify me." Once enabled, the rule runs continuously and fires only when the condition is actually met, not on a fixed schedule [S5]. The same pattern covers operational data: an Activator rule can watch a file landing or a OneLake table update and invoke a Copy job directly, no Data Factory pipeline required, the moment the condition is true [S6].
What goes wrong#
- Assuming Eventstream routes anywhere by default. Within a workspace it freely routes to Real-Time Intelligence and other Fabric items, but cross-workspace and external destinations are blocked by default under Workspace Outbound Access Protection [S7].
- Treating Activator as just alerting. It also directly triggers pipelines, notebooks, Spark jobs, dataflows, and Copy jobs — build it into execution logic, not only monitoring [S1] [S5].
- Not previewing a rule first. Activator can estimate how often a candidate rule would have fired against historical data, the way to tune thresholds and avoid alert spam before enabling it [S5].