Restore Ontology to Phase 1 — fabric-cicd requires it before DataAgent

Author: michaeldeongreen

.github/workflows/deploy-test.yml

@@ -1,8 +1,8 @@
 # Orchestrator workflow for deploying Fabric items to the Test workspace.
 #
 # Deploys all supported items via fabric-cicd. The reusable workflow handles
-# phased deployment internally (Lakehouse first, then everything else)
-# to satisfy parameter.yml dependency resolution.
+# phased deployment internally (Lakehouse + Ontology first, then everything
+# else) to satisfy fabric-cicd dependency resolution.
 #
 # The ETL workflow (etl-test.yml) triggers automatically via workflow_run
 # once deployment completes successfully.

.github/workflows/reusable-deploy-supported.yml

@@ -91,22 +91,21 @@ jobs:
           environment = os.environ['ENVIRONMENT']
 
           # Deployment uses a 2-phase approach to satisfy item dependencies.
-          # fabric-cicd resolves \$items dynamic variables (e.g.,
-          # \$items.Lakehouse.PatternsLakehouse.\$id) by querying the live workspace,
-          # so the referenced item must already exist.
+          # fabric-cicd caches workspace state once at the start of each
+          # publish_all_items() call, so items deployed within the same call
+          # are not visible to later items' logicalId or \$items resolution.
           #
-          # Phase 1: Lakehouse only — must exist before Phase 2 so that
-          #          parameter.yml rules referencing \$items.Lakehouse resolve.
-          # Phase 2: All remaining items (Ontology, DataAgent, Notebook, etc.).
-          #          Ontology and DataAgent use logicalId references resolved by
-          #          Fabric at runtime, so they have no deploy-time ordering needs.
+          # Phase 1: Lakehouse + Ontology
+          #   - Lakehouse must exist so parameter.yml \$items.Lakehouse rules resolve.
+          #   - Ontology must exist so DataAgent's logicalId reference resolves.
+          # Phase 2: All remaining items (DataAgent, Notebook, SemanticModel, etc.).
           #
-          # On subsequent deployments, all items already exist and phases are
+          # On subsequent deployments all items already exist and phases are
           # idempotent — they simply update in place.
 
-          phase1_types = ['Lakehouse']
+          phase1_types = ['Lakehouse', 'Ontology']
 
-          print('Phase 1: Deploying Lakehouse...')
+          print('Phase 1: Deploying Lakehouse + Ontology...')
           phase1_ws = FabricWorkspace(
               repository_directory=repo_dir,
               workspace_id=workspace_id,