fix(ci): coordinate Kubernetes startup migrations (#13474)

.github/workflows/pr-craft-k8s-tests.yml

@@ -440,11 +440,12 @@ jobs:
           cat "${RUNNER_TEMP}/postgres-port-forward.log" || true
           exit 1
 
-      - name: Run migrations
+      - name: Wait for API startup migrations
         run: |
-          cd backend
-          alembic upgrade head
-          alembic heads --verbose
+          kubectl -n "${HELM_RELEASE_NAMESPACE}" rollout status \
+            deployment/onyx-api-server --timeout=360s
+          kubectl -n "${HELM_RELEASE_NAMESPACE}" exec deployment/onyx-api-server -- \
+            alembic current --check-heads
 
       - name: Wait for chart OpenSearch ready
         run: |

deployment/helm/charts/onyx/Chart.yaml

@@ -5,7 +5,7 @@ home: https://www.onyx.app/
 sources:
   - "https://github.com/onyx-dot-app/onyx"
 type: application
-version: 0.8.2
+version: 0.8.3
 appVersion: latest
 annotations:
   category: Productivity
@@ -17,10 +17,8 @@ annotations:
       image: docker.io/onyxdotapp/onyx-backend:latest
   artifacthub.io/changes: |
     - kind: changed
-      description: Documented that the default backend images no longer ship psql (it
-        moved to the -dev suffixed image variants); the opt-in tooling.pgInto helper
-        now needs a -dev backend image tag or a custom image providing the client
-        binary.
+      description: Made the Kubernetes CI profile coordinate API startup migrations
+        through deployment readiness before running tests.
 dependencies:
   # Helm uses the first condition path that exists: `postgresqlOperator.enabled`
   # is unset by default, so this falls through to `postgresql.enabled`. Set it

deployment/helm/charts/onyx/values-ci.yaml

@@ -123,9 +123,15 @@ webserver:
 
 api:
   replicaCount: 1
-  # The workflow runs `alembic upgrade head` once after Postgres is reachable.
-  # Keep API startup from racing that centralized migration step.
-  runStartupMigrations: false
+  runStartupMigrations: true
+  # The API port opens only after startup migrations finish and Uvicorn starts.
+  # This makes rollout status a reliable migration-completion barrier.
+  readinessProbe:
+    tcpSocket:
+      port: api-server-port
+    periodSeconds: 2
+    timeoutSeconds: 1
+    failureThreshold: 180
   resources:
     requests:
       cpu: 250m