v0.3.15

[0.3.15] - 2026-09-04

Added

• skills/sqldw-cli -- added a read-only Capacity Metrics workflow that discovers the installed metrics model, adapts to timestamped or fixed-window schema variants, identifies costly Warehouse and Lakehouse SQL endpoint items, and analyzes every Query Insights request overlapping the Capacity Metrics spike range. Results keep fixed-window capacity health separate from broader item history, disclose timeframe and 30-day Query Insights limits, and treat SQL statement candidates as best-effort correlation because Capacity Metrics Operation Id and Query Insights distributed_statement_id are different identifiers.
• onelake-catalog-govern-cli -- audits and safely remediates Microsoft Fabric OneLake catalog governance across domains, workspaces, capacities, protection, and curation, with separate permission-aware modes for tenant admins and data owners.

Changed

• skills/sqldw-cli -- expanded composite diagnostics for failed and canceled requests, SQL pool pressure, resource concentration, Lakehouse table health, performance regressions, optimization targets, and user/application activity. Custom SQL pool guidance now uses recurring historical contention and stable application classifiers rather than converting Capacity Metrics CU seconds or Query Insights CPU milliseconds into pool percentages.
• skills/sqldw-cli -- made operations follow-ups user-facing: results now turn evidence into concrete actions on the investigated workload, SQL item, capacity, or correlation report, while retaining timezone, retention, lag, and confidence caveats as limitations rather than skill-development suggestions.

Fixed

• skills/sqldw-cli -- corrected pressure intervals to use the complete pool-state event stream and exact pool matching, included canceled requests in non-success analysis, limited Lakehouse health checks to Lakehouse SQL analytics endpoints, and stopped recommending result-set caching while the feature is unavailable.
• skills/sqldw-cli -- retained command-less and legitimate Query Insights-referencing requests in historical custom-pool profiles, excluding only agent-labeled diagnostics, and widened duration and CPU aggregation to bigint.
• semantic-model-authoring -- preserves existing Prep data for AI configuration during unrelated semantic model edits and uses the Power BI modeling MCP for read-only metadata discovery when available.
• synapse-migration now handles Dedicated SQL Pool DACPAC and zipped SQL-project schema and code migrations more reliably, validates generated Spark SQL, safely resumes interrupted operations, and isolates concurrent migrations across multiple datamarts.