fix(vertex): switch CLOUD_ML_REGION default from us-east5 to global

.github/scripts/model-discovery.py

@@ -12,7 +12,7 @@
 ``anthropic`` publisher.
 
 Required env vars:
-    GCP_REGION                 - GCP region (e.g. us-east5)
+    GCP_REGION                 - GCP region (e.g. global)
     GCP_PROJECT                - GCP project ID
 
 Optional env vars:

.github/workflows/pull-reviews.yml

@@ -15,7 +15,7 @@ jobs:
     with:
       s3_bucket: pull-reviews
       gcp_project_id: ambient-code-platform
-      gcp_region: us-east5
+      gcp_region: global
       gcp_workload_identity_provider: projects/888214980327/locations/global/workloadIdentityPools/github/providers/ambient-code-github-actions
     secrets:
       S3_ENDPOINT: ${{ secrets.S3_ENDPOINT }}

components/ambient-control-plane/internal/config/config.go

@@ -67,7 +67,7 @@ func Load() (*ControlPlaneConfig, error) {
 		AnthropicAPIKey:       os.Getenv("ANTHROPIC_API_KEY"),
 		VertexEnabled:         os.Getenv("USE_VERTEX") == "1" || os.Getenv("USE_VERTEX") == "true",
 		VertexProjectID:       os.Getenv("ANTHROPIC_VERTEX_PROJECT_ID"),
-		VertexRegion:          envOrDefault("CLOUD_ML_REGION", "us-east5"),
+		VertexRegion:          envOrDefault("CLOUD_ML_REGION", "global"),
 		VertexCredentialsPath: envOrDefault("GOOGLE_APPLICATION_CREDENTIALS", "/app/vertex/ambient-code-key.json"),
 		VertexSecretName:      envOrDefault("VERTEX_SECRET_NAME", "ambient-vertex"),
 		VertexSecretNamespace: envOrDefault("VERTEX_SECRET_NAMESPACE", "ambient-code"),

components/backend/handlers/display_name.go

@@ -141,14 +141,11 @@ func getAnthropicClient(ctx context.Context, projectName string) (anthropic.Clie
 	// Check if Vertex AI is enabled (cluster-wide setting)
 	if isVertexEnabled() {
 		// For Vertex AI, use the vertex package with Google Application Default Credentials
-		// Required env vars: GOOGLE_APPLICATION_CREDENTIALS, ANTHROPIC_VERTEX_PROJECT_ID, CLOUD_ML_REGION
 		region := os.Getenv("CLOUD_ML_REGION")
 		gcpProjectID := os.Getenv("ANTHROPIC_VERTEX_PROJECT_ID")
 
-		// Default to us-east5 - claude-haiku-4-5 is not available in global region
-		// See: https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude
-		if region == "" || region == "global" {
-			region = "us-east5"
+		if region == "" {
+			region = "global"
 		}
 		if gcpProjectID == "" {
 			return anthropic.Client{}, false, fmt.Errorf("ANTHROPIC_VERTEX_PROJECT_ID is required when USE_VERTEX is enabled (check backend deployment env vars)")

components/backend/handlers/display_name_test.go

@@ -347,14 +347,14 @@ var _ = Describe("Display Name Handler", Label(test_constants.LabelUnit, test_co
 			It("Should detect Vertex AI enabled configuration", func() {
 				os.Setenv("USE_VERTEX", "1")
 				os.Setenv("ANTHROPIC_VERTEX_PROJECT_ID", "test-gcp-project")
-				os.Setenv("CLOUD_ML_REGION", "us-east5")
+				os.Setenv("CLOUD_ML_REGION", "global")
 
 				// This tests the environment detection logic
 				// We can't easily test the full getAnthropicClient without real credentials
 				// but we can verify the environment variables are read correctly
 				Expect(os.Getenv("USE_VERTEX")).To(Equal("1"))
 				Expect(os.Getenv("ANTHROPIC_VERTEX_PROJECT_ID")).To(Equal("test-gcp-project"))
-				Expect(os.Getenv("CLOUD_ML_REGION")).To(Equal("us-east5"))
+				Expect(os.Getenv("CLOUD_ML_REGION")).To(Equal("global"))
 			})
 
 			It("Should handle missing Vertex configuration gracefully", func() {

components/manifests/overlays/mpp-openshift/ambient-control-plane.yaml

@@ -69,7 +69,7 @@ spec:
             - name: ANTHROPIC_VERTEX_PROJECT_ID
               value: "ambient-code-platform"
             - name: CLOUD_ML_REGION
-              value: "us-east5"
+              value: "global"
             - name: GOOGLE_APPLICATION_CREDENTIALS
               value: "/app/vertex/ambient-code-key.json"
             - name: VERTEX_SECRET_NAME

docs/internal/design/control-plane.guide.md

@@ -726,14 +726,14 @@ acpctl session events <id>                   CLI          open
 | File | Change |
 |------|--------|
 | `runners/ambient_runner/platform/context.py` | Added `current_user_id`, `current_user_name`, `caller_token` fields + `set_current_user()` method to `RunnerContext` |
-| `overlays/mpp-openshift/ambient-control-plane.yaml` | Added `USE_VERTEX=1`, `ANTHROPIC_VERTEX_PROJECT_ID=ambient-code-platform`, `CLOUD_ML_REGION=us-east5`, `GOOGLE_APPLICATION_CREDENTIALS=/app/vertex/ambient-code-key.json`, `VERTEX_SECRET_NAME=ambient-vertex`, `VERTEX_SECRET_NAMESPACE=ambient-code--runtime-int`; added `vertex-credentials` volumeMount + volume from `ambient-vertex` secret |
+| `overlays/mpp-openshift/ambient-control-plane.yaml` | Added `USE_VERTEX=1`, `ANTHROPIC_VERTEX_PROJECT_ID=ambient-code-platform`, `CLOUD_ML_REGION=global`, `GOOGLE_APPLICATION_CREDENTIALS=/app/vertex/ambient-code-key.json`, `VERTEX_SECRET_NAME=ambient-vertex`, `VERTEX_SECRET_NAMESPACE=ambient-code--runtime-int`; added `vertex-credentials` volumeMount + volume from `ambient-vertex` secret |
 
 **Vertex invariants for MPP:**
 - Secret: `secret/ambient-vertex` in `ambient-code--runtime-int` — contains `ambient-code-key.json` (GCP SA key)
 - CP reads it via `nsKube()` (tenant SA) and copies to session namespace via `ensureVertexSecret()`
 - Runner pod mounts it at `/app/vertex/` — set `GOOGLE_APPLICATION_CREDENTIALS=/app/vertex/ambient-code-key.json`
 - `VERTEX_SECRET_NAMESPACE` must be `ambient-code--runtime-int` (not default `ambient-code`)
-- GCP project: `ambient-code-platform`, region: `us-east5`
+- GCP project: `ambient-code-platform`, region: `global`
 
 **acpctl session commands:**
 ```bash

docs/internal/developer/local-development/kind.md

@@ -205,7 +205,7 @@ make kind-up LOCAL_VERTEX=true GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-a
 ```bash
 make kind-up LOCAL_VERTEX=true \
     ANTHROPIC_VERTEX_PROJECT_ID=my-project \
-    CLOUD_ML_REGION=us-east5 \
+    CLOUD_ML_REGION=global \
     GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json
 ```
 

scripts/setup-vertex-kind.sh

@@ -24,14 +24,14 @@
 #
 #   CLOUD_ML_REGION
 #       GCP region for Vertex AI API calls.
-#       Example: us-east5 (Claude is available in limited regions)
+#       Example: global (recommended for cost optimization)
 #
 # SETUP:
 #   Add these to your shell profile (~/.zshrc or ~/.bashrc):
 #
 #     export GOOGLE_APPLICATION_CREDENTIALS="$HOME/.config/gcloud/your-sa-key.json"
 #     export ANTHROPIC_VERTEX_PROJECT_ID="your-gcp-project-id"
-#     export CLOUD_ML_REGION="us-east5"
+#     export CLOUD_ML_REGION="global"
 #
 #   Then reload your shell: source ~/.zshrc
 #
@@ -109,7 +109,7 @@ fi
 
 if [ -z "$CLOUD_ML_REGION" ]; then
     echo "  [MISSING] CLOUD_ML_REGION"
-    echo "            GCP region (e.g., us-east5)"
+    echo "            GCP region (e.g., global)"
     echo ""
     missing_vars=1
 else
@@ -122,7 +122,7 @@ if [ $missing_vars -eq 1 ]; then
     echo ""
     echo '  export GOOGLE_APPLICATION_CREDENTIALS="$HOME/.config/gcloud/your-sa-key.json"'
     echo '  export ANTHROPIC_VERTEX_PROJECT_ID="your-gcp-project-id"'
-    echo '  export CLOUD_ML_REGION="us-east5"'
+    echo '  export CLOUD_ML_REGION="global"'
     echo ""
     echo "Then reload: source ~/.zshrc"
     exit 1