Microsoft Fabric connector
Microsoft Fabric appears in Kanva's Azure source and dataset pickers. Warehouse and Lakehouse imports are available when your administrator has installed the qualified reader and approved the target. Each manual import reads the selected data under Fabric's current permissions.
Connection and import workflow
When your administrator has a supported, approved target, create a connection under New data source > Azure > Microsoft Fabric. Name it, choose the approved item, continue, save the write-only Entra client secret, test the connection and select Use connection. Your administrator controls application identity, item access and the applicable SQL or OneLake permissions. Kanva does not accept custom SQL hosts or connection strings in this form.
Under New dataset > Azure > Microsoft Fabric, select that saved connection and an approved SQL table or view. Choose columns, optionally preview them, then select Import data. Required ordering columns and decimal precision come from approved metadata. Timestamp columns need an administrator-approved UTC interpretation.
Successful imports become ordinary Kanva datasets. A manual refresh reads a new full snapshot. Empty or failed refreshes preserve your existing data. The interface reports cancellation as pending until the worker has stopped. Previously imported data can remain usable after a source is disconnected, subject to dataset permissions. Later source data or permission changes do not change the saved snapshot. Kanva does not monitor the source to update or reauthorize that saved data.
Import limits
The qualified Warehouse and Lakehouse runtimes support Linux ARM64 with the pinned ODBC 18 reader. Each capture query has an 8 MiB SQL-traffic limit; preview queries have a 1 MiB limit. These can be stricter than the shared row and dataset limits. Exceeding a limit fails the operation and preserves the current dataset. Masked columns and views with masked dependencies are unavailable.
Lakehouse data limits
A Lakehouse combines file-based storage with an SQL endpoint. This connector reads the values visible through that SQL endpoint. It may lag recent OneLake changes, omit unsupported Delta columns or expose strings shortened by Fabric. Kanva displays and retains these limits. It does not provide a direct, lossless OneLake export.
Warehouse and Lakehouse require separate validation. A working Warehouse connection does not establish Lakehouse support. A Lakehouse import reads the selected data allowed by Fabric for the configured identity; it does not require ongoing access-mode monitoring.
API integration
Fabric uses the existing source, discovery, preview, dataset and refresh operations. There is no separate Fabric REST or SignalR endpoint. Requests use typed result contract version 2 and the closed microsoftFabric/fabricSql/entraClientCredentials configuration. Availability and approved binding responses determine whether a client may offer connection setup. An installed adapter alone does not authorize execution.
The normal import workflow requests automatic publication. Existing authorized candidate review APIs remain available under their usual revision and digest checks. Connector scheduling and automatic retraining are outside this delivery.