Update deployment to us-central1 for Gemini Live Native Audio

- Change region from asia-southeast1 to us-central1
- Update dev.sh to use Vertex AI env vars
- Update setup_infra.sh default region

Author: oadultradeepfield

docs/deployment.md

@@ -2,26 +2,24 @@
 
 This guide covers deploying Jemmie to Google Cloud Run.
 
----
-
 ## Prerequisites
 
 - Google Cloud account with billing enabled
 - `gcloud` CLI installed and authenticated
 - Project ID noted
 
----
-
 ## One-Time Infrastructure Setup
 
-Run the setup script to enable APIs, create Artifact Registry, and configure Firestore:
+Run the setup script to enable APIs, create Artifact Registry, and configure Firestore.
 
 ```bash
 export PROJECT_ID="your-project-id"
+export REGION=us-central1
 ./scripts/setup_infra.sh $PROJECT_ID
 ```
 
 This enables:
+
 - Cloud Run API
 - Artifact Registry API
 - Cloud Build API
@@ -30,8 +28,6 @@ This enables:
 
 And grants Vertex AI permissions to the Cloud Run service account.
 
----
-
 ## GitHub Actions Setup
 
 ### 1. Create Service Account
@@ -65,14 +61,11 @@ rm key.json
 ### 3. Add GitHub Secrets
 
 Navigate to your repository: **Settings → Secrets and variables → Actions**
-
 | Secret | Value |
 |--------|-------|
 | `GCP_PROJECT_ID` | Your GCP project ID |
 | `GCP_SERVICE_ACCOUNT_KEY` | Base64 output from step 2 |
 
----
-
 ## Deploy
 
 Push to main branch:
@@ -82,45 +75,41 @@ git push origin main
 ```
 
 The CI pipeline will:
+
 1. Run lint, type check, and tests
 2. Build Docker image (production target)
 3. Push to Artifact Registry
 4. Deploy to Cloud Run
 
----
-
 ## Cloud Run Configuration
 
-| Setting | Value | Purpose |
-|---------|-------|---------|
-| Session affinity | Enabled | Keeps reconnects on same instance |
-| Timeout | 3600s | Long-lived WebSocket connections |
-| Min instances | 1 | Prevents cold start |
-| Authentication | Unauthenticated | Device-as-identity model |
-
----
+| Setting          | Value           | Purpose                           |
+|------------------|-----------------|-----------------------------------|
+| Session affinity | Enabled         | Keeps reconnects on same instance |
+| Timeout          | 3600s           | Long-lived WebSocket connections  |
+| Min instances    | 1               | Prevent cold start                |
+| Authentication   | Unauthenticated | Device-as-identity model          |
 
 ## Verification
 
 After deployment:
 
 ```bash
 SERVICE_URL=$(gcloud run services describe jemmie-backend \
-    --region=asia-southeast1 \
+    --region=us-central1 \
     --format='value(status.url)')
 
 curl $SERVICE_URL/health
 ```
 
 WebSocket endpoint: `wss://{service-url}/ws/{device_id}`
 
----
-
 ## Troubleshooting
 
 ### Deployment fails with permission error
 
 Ensure the service account has all required roles:
+
 ```bash
 gcloud projects get-iam-policy $PROJECT_ID \
     --flatten="bindings[].members" \
@@ -130,6 +119,7 @@ gcloud projects get-iam-policy $PROJECT_ID \
 ### Cloud Run returns 403 on Vertex AI calls
 
 Grant Vertex AI permissions to the Cloud Run service account:
+
 ```bash
 PROJECT_NUMBER=$(gcloud projects describe $PROJECT_ID --format='value(projectNumber)')
 gcloud projects add-iam-policy-binding $PROJECT_ID \
@@ -140,6 +130,7 @@ gcloud projects add-iam-policy-binding $PROJECT_ID \
 ### WebSocket connections drop
 
 Check Cloud Run timeout and session affinity settings:
+
 ```bash
 gcloud run services describe jemmie-backend \
     --region=us-central1 \

scripts/dev.sh

@@ -1,16 +1,16 @@
 #!/bin/bash
 set -e
 
-if [ -z "$GOOGLE_API_KEY" ]; then
-    echo "ERROR: GOOGLE_API_KEY environment variable not set!"
-    echo ""
-    echo "Get your API key from: https://aistudio.google.com/apikey"
-    echo "Then run:"
-    echo ""
-    echo "  export GOOGLE_API_KEY=your-api-key"
-    echo "  ./scripts/dev.sh"
-    exit 1
-fi
+# For local development with Gemini API
+# Get your API key from: https://aistudio.google.com/apikey
+# Then run:
+#   export GOOGLE_API_KEY=your-api-key
+#   export GOOGLE_GENAI_USE_VERTEXAI=FALSE
+#   ./scripts/dev.sh
+#
+# For production, GOOGLE_GENAI_USE_VERTEXAI=TRUE is set in Cloud Run
+
+# (no API key needed - uses Application Default Credentials)
 
 echo "Starting development environment..."
 echo "WebSocket endpoint: ws://localhost:8080/ws/{device_id}"

scripts/setup_artifact_registry.sh

@@ -2,7 +2,7 @@
 set -e
 
 PROJECT_ID="${1:-$(gcloud config get-value project)}"
-REGION="${REGION:-asia-southeast1}"
+REGION="${REGION:-us-central1}"
 REPO_NAME="jemmie-backend"
 
 echo "Creating Artifact Registry repository..."

scripts/setup_infra.sh

@@ -2,7 +2,7 @@
 set -e
 
 PROJECT_ID="${1:-$(gcloud config get-value project)}"
-REGION="${REGION:-asia-southeast1}"
+REGION="${REGION:-us-central1}"
 
 echo "Setting up infrastructure for project: $PROJECT_ID"
 echo ""