add DocuSeal to GKE — manifests, Terraform, Argo CD app; update ci-base Dockerfile

migrate Terraform backend to GCS, remove dead Vault GCE nodes
Switch gcp/ backend from Cloudflare R2 (S3-compatible) to GCS — uses Application Default Credentials automatically, no extra env vars needed. State has been migrated to gs://neuron-785695-terraform-state/gcp/. Remove GCE Vault node resources (vault-node-1/2/3) from vault-nodes.tf and Terraform state. GKE Vault (Workload Identity + vault-unseal SA) has been serving traffic since the NEG cutover; GCE nodes are confirmed dead. Removed from state and config: - google_compute_instance.vault_node (x3) - google_compute_disk.vault_data (x3) - google_compute_instance_group.vault (x3) - google_compute_firewall.vault_iap_ssh / vault_raft / vault_api (GCE-targeted) - google_storage_bucket_object.vault_startup - google_storage_bucket_iam_member.vault_node_bucket_read - google_project_iam_member.vault_node_{log_writer,metric_writer,compute_viewer} Kept: KMS, vault-unseal SA, WI binding, NEG data sources, GKE firewall, global LB (address, cert, backend service, url maps, proxies, fwd rules).
2026-05-05 10:38:45 -05:00 · 2026-05-05 10:38:22 -05:00 · 2026-05-05 10:27:45 -05:00 · 2026-05-05 10:22:21 -05:00 · 2026-05-05 10:22:02 -05:00 · 2026-05-05 10:16:49 -05:00
67 changed files with 1883 additions and 489 deletions
@@ -148,3 +148,118 @@ resource "google_secret_manager_secret" "license_admin_token" {
    ignore_changes = [id]
  }
 }
+
+# ── Gitea — database + user on neuron-prod-pg15 ───────────────────────────────
+# Gitea on GKE uses the existing Cloud SQL instance with a Cloud SQL Auth Proxy
+# sidecar. Connection via unix socket — no public IP exposure.
+
+resource "google_sql_database" "gitea" {
+  name     = "gitea"
+  instance = google_sql_database_instance.main.name
+  project  = var.project_id
+}
+
+resource "random_password" "gitea_db" {
+  length  = 32
+  special = false # Cloud SQL unix socket path DSN; keep alphanumeric for simplicity
+}
+
+resource "google_sql_user" "gitea" {
+  name     = "gitea"
+  instance = google_sql_database_instance.main.name
+  project  = var.project_id
+  password = random_password.gitea_db.result
+}
+
+# ── Gitea service account (for Workload Identity → Cloud SQL) ─────────────────
+
+resource "google_service_account" "gitea" {
+  account_id   = "gitea-gke"
+  display_name = "Gitea GKE SA"
+  description  = "Service account for the Gitea pod on GKE. Used by the Cloud SQL Auth Proxy sidecar via Workload Identity."
+  project      = var.project_id
+}
+
+resource "google_project_iam_member" "gitea_sql_client" {
+  project = var.project_id
+  role    = "roles/cloudsql.client"
+  member  = "serviceAccount:${google_service_account.gitea.email}"
+}
+
+# ── Secret Manager — Gitea database URL ───────────────────────────────────────
+# Full DSN using the Cloud SQL Auth Proxy unix socket path.
+# The proxy sidecar mounts the socket at /cloudsql/<connection_name>.
+
+resource "google_secret_manager_secret" "gitea_database_url" {
+  secret_id = "gitea-database-url"
+  project   = var.project_id
+
+  replication {
+    auto {}
+  }
+}
+
+resource "google_secret_manager_secret_version" "gitea_database_url" {
+  secret      = google_secret_manager_secret.gitea_database_url.id
+  secret_data = "host=/cloudsql/${google_sql_database_instance.main.connection_name} user=gitea password=${random_password.gitea_db.result} dbname=gitea sslmode=disable"
+}
+
+# gitea-db-password — the raw password only, for use in Gitea's GITEA__database__PASSWD.
+# The Cloud SQL Auth Proxy provides the unix socket; Gitea uses standard postgres
+# password auth through the proxy socket.
+resource "google_secret_manager_secret" "gitea_db_password" {
+  secret_id = "gitea-db-password"
+  project   = var.project_id
+
+  replication {
+    auto {}
+  }
+}
+
+resource "google_secret_manager_secret_version" "gitea_db_password" {
+  secret      = google_secret_manager_secret.gitea_db_password.id
+  secret_data = random_password.gitea_db.result
+}
+
+# Allow the Gitea GCP SA to access its secrets
+resource "google_secret_manager_secret_iam_member" "gitea_database_url_accessor" {
+  project   = var.project_id
+  secret_id = google_secret_manager_secret.gitea_database_url.secret_id
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.gitea.email}"
+}
+
+resource "google_secret_manager_secret_iam_member" "gitea_db_password_accessor" {
+  project   = var.project_id
+  secret_id = google_secret_manager_secret.gitea_db_password.secret_id
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.gitea.email}"
+}
+
+# ESO Workload Identity — allows the external-secrets controller SA on GKE
+# to impersonate the gitea GCP SA for Secret Manager access.
+resource "google_service_account_iam_member" "eso_gitea_wi" {
+  service_account_id = google_service_account.gitea.name
+  role               = "roles/iam.workloadIdentityUser"
+  member             = "serviceAccount:${var.project_id}.svc.id.goog[external-secrets/external-secrets]"
+
+  depends_on = [google_container_cluster.neuron_platform]
+}
+
+resource "google_service_account_iam_member" "eso_gitea_token_creator" {
+  service_account_id = google_service_account.gitea.name
+  role               = "roles/iam.serviceAccountTokenCreator"
+  member             = "serviceAccount:${var.project_id}.svc.id.goog[external-secrets/external-secrets]"
+
+  depends_on = [google_container_cluster.neuron_platform]
+}
+
+output "gitea_service_account_email" {
+  description = "Gitea GKE SA email — used in the Workload Identity annotation"
+  value       = google_service_account.gitea.email
+}
+
+output "gitea_database_secret_id" {
+  description = "Secret Manager secret ID for the Gitea database URL"
+  value       = google_secret_manager_secret.gitea_database_url.secret_id
+}
@@ -0,0 +1,68 @@
+# ── DocuSeal GKE service account and Workload Identity ───────────────────────
+# DocuSeal runs in the `docuseal` namespace on GKE. It accesses GCP Secret
+# Manager via Workload Identity — no JSON key files.
+#
+# Secret flow: Vault → GCP Secret Manager → ExternalSecret → k8s Secret → pod
+
+resource "google_service_account" "docuseal_gke" {
+  account_id   = "docuseal-gke"
+  display_name = "DocuSeal GKE"
+  description  = "Identity for DocuSeal pod on GKE. Accesses Secret Manager via Workload Identity."
+  project      = var.project_id
+}
+
+# ── Secret Manager access ─────────────────────────────────────────────────────
+
+resource "google_secret_manager_secret_iam_member" "docuseal_secret_key_base" {
+  project   = var.project_id
+  secret_id = "docuseal-secret-key-base"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.docuseal_gke.email}"
+}
+
+resource "google_secret_manager_secret_iam_member" "docuseal_smtp_host" {
+  project   = var.project_id
+  secret_id = "docuseal-smtp-host"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.docuseal_gke.email}"
+}
+
+resource "google_secret_manager_secret_iam_member" "docuseal_smtp_port" {
+  project   = var.project_id
+  secret_id = "docuseal-smtp-port"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.docuseal_gke.email}"
+}
+
+resource "google_secret_manager_secret_iam_member" "docuseal_smtp_username" {
+  project   = var.project_id
+  secret_id = "docuseal-smtp-username"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.docuseal_gke.email}"
+}
+
+resource "google_secret_manager_secret_iam_member" "docuseal_smtp_password" {
+  project   = var.project_id
+  secret_id = "docuseal-smtp-password"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.docuseal_gke.email}"
+}
+
+resource "google_secret_manager_secret_iam_member" "docuseal_smtp_from" {
+  project   = var.project_id
+  secret_id = "docuseal-smtp-from"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.docuseal_gke.email}"
+}
+
+# ── Workload Identity binding ─────────────────────────────────────────────────
+# The k8s SA `docuseal` in the `docuseal` namespace can impersonate this GCP SA.
+# ESO uses this to authenticate to Secret Manager without a key file.
+
+resource "google_service_account_iam_member" "docuseal_workload_identity" {
+  service_account_id = google_service_account.docuseal_gke.name
+  role               = "roles/iam.workloadIdentityUser"
+  member             = "serviceAccount:${var.project_id}.svc.id.goog[docuseal/docuseal]"
+
+  depends_on = [google_container_cluster.neuron_platform]
+}
@@ -254,6 +254,46 @@ resource "google_service_account_iam_member" "ci_pusher_wif_neuron_web" {
  member             = "principalSet://iam.googleapis.com/${google_iam_workload_identity_pool.gitea.name}/attribute.repository/neuron-technologies/neuron-web"
 }

+# ── Gitea Actions runner on GKE ───────────────────────────────────────────────
+# A k8s Deployment in the `ci` namespace on GKE runs act_runner with a DinD
+# sidecar. The pod uses this GCP SA via Workload Identity to pull the runner
+# registration token from Secret Manager at startup.
+#
+# See servers/gcp/k8s/gitea-runner/ for the k8s manifests.
+
+resource "google_service_account" "gitea_runner_gke" {
+  account_id   = "gitea-runner-gke"
+  display_name = "Gitea Actions runner (GKE identity)"
+  description  = "Workload Identity SA for the GKE-hosted Gitea Actions runner. Read-only access to the registration token."
+  project      = var.project_id
+}
+
+# Allow the GKE SA to read the runner registration token from Secret Manager.
+resource "google_secret_manager_secret_iam_member" "runner_gke_token_access" {
+  project   = var.project_id
+  secret_id = "gitea-runner-token"
+  role      = "roles/secretmanager.secretAccessor"
+  member    = "serviceAccount:${google_service_account.gitea_runner_gke.email}"
+}
+
+# Workload Identity binding — the k8s SA `gitea-runner` in the `ci` namespace
+# can impersonate this GCP SA without a JSON key file.
+resource "google_service_account_iam_member" "gitea_runner_gke_workload_identity" {
+  service_account_id = google_service_account.gitea_runner_gke.name
+  role               = "roles/iam.workloadIdentityUser"
+  member             = "serviceAccount:${var.project_id}.svc.id.goog[ci/gitea-runner]"
+
+  depends_on = [google_container_cluster.neuron_platform]
+}
+
+# Allow the runner to pull images from Artifact Registry (needed for build jobs
+# that pull from neuron-* repos, and for the runner's own ci-base image).
+resource "google_project_iam_member" "gitea_runner_gke_ar_reader" {
+  project = var.project_id
+  role    = "roles/artifactregistry.reader"
+  member  = "serviceAccount:${google_service_account.gitea_runner_gke.email}"
+}
+
 # ── Outputs ───────────────────────────────────────────────────────────────────

 output "gitea_runner_vm_name" {
@@ -0,0 +1,94 @@
+# ── GKE Autopilot Cluster — neuron-platform ───────────────────────────────────
+#
+# Regional Autopilot cluster in us-central1. Autopilot manages the node pool
+# automatically — no node pools to configure, no VM sizes to choose.
+# Pods spread across a/b/c zones automatically by GKE.
+#
+# Workload Identity is required so Vault pods can access KMS (auto-unseal)
+# and Gitea pods can use Cloud SQL Auth Proxy without key files.
+#
+# After `terraform apply`, register the cluster with Legion Argo CD:
+#   1. Get the endpoint: terraform output gke_cluster_endpoint
+#   2. Get credentials: gcloud container clusters get-credentials neuron-platform \
+#        --region us-central1 --project neuron-785695
+#   3. Register with Argo CD:
+#        argocd cluster add <context-name> --name gke-neuron-platform
+#   4. Update servers/gcp/k8s/argocd-apps/*.yaml with the actual cluster endpoint.
+
+resource "google_container_cluster" "neuron_platform" {
+  provider = google-beta
+
+  name     = "neuron-platform"
+  location = "us-central1"
+  project  = var.project_id
+
+  # Autopilot — GKE manages node pools, scaling, and node security
+  enable_autopilot = true
+
+  release_channel {
+    channel = "REGULAR"
+  }
+
+  # Workload Identity — required for Vault KMS and Gitea Cloud SQL proxy
+  workload_identity_config {
+    workload_pool = "${var.project_id}.svc.id.goog"
+  }
+
+  # Deletion protection — prevent accidental cluster destruction
+  deletion_protection = true
+
+  # Cluster networking — uses default VPC (same as Vault GCE nodes)
+  ip_allocation_policy {}
+
+  # Disable basic auth and client cert (Workload Identity handles auth)
+  master_auth {
+    client_certificate_config {
+      issue_client_certificate = false
+    }
+  }
+}
+
+# ── Workload Identity bindings ────────────────────────────────────────────────
+#
+# Each GCP SA gets a binding that allows its corresponding k8s ServiceAccount
+# (in the GKE cluster) to impersonate it via Workload Identity.
+# The k8s SA must also have the annotation:
+#   iam.gke.io/gcp-service-account: <gcp-sa-email>
+
+# Vault pods (namespace: vault, k8s SA: vault) → vault-unseal GCP SA
+# This lets the Vault StatefulSet access KMS for auto-unseal without key files.
+resource "google_service_account_iam_member" "vault_workload_identity" {
+  service_account_id = google_service_account.vault_unseal.name
+  role               = "roles/iam.workloadIdentityUser"
+  member             = "serviceAccount:${var.project_id}.svc.id.goog[vault/vault]"
+
+  depends_on = [google_container_cluster.neuron_platform]
+}
+
+# Gitea pod (namespace: gitea, k8s SA: gitea) → gitea GCP SA
+# This lets the Cloud SQL Auth Proxy sidecar connect to neuron-prod-pg15.
+resource "google_service_account_iam_member" "gitea_workload_identity" {
+  service_account_id = google_service_account.gitea.name
+  role               = "roles/iam.workloadIdentityUser"
+  member             = "serviceAccount:${var.project_id}.svc.id.goog[gitea/gitea]"
+
+  depends_on = [google_container_cluster.neuron_platform]
+}
+
+# ── Outputs ───────────────────────────────────────────────────────────────────
+
+output "gke_cluster_endpoint" {
+  description = "GKE cluster API endpoint — use as destination.server in Argo CD Applications"
+  value       = "https://${google_container_cluster.neuron_platform.endpoint}"
+  sensitive   = true
+}
+
+output "gke_cluster_name" {
+  description = "GKE cluster name"
+  value       = google_container_cluster.neuron_platform.name
+}
+
+output "gke_workload_pool" {
+  description = "Workload Identity pool — used in Workload Identity bindings"
+  value       = "${var.project_id}.svc.id.goog"
+}
@@ -0,0 +1,26 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: docuseal-gke
+  namespace: argocd
+spec:
+  project: default
+  source:
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
+    targetRevision: main
+    path: servers/gcp/k8s/docuseal
+  destination:
+    server: https://34.63.89.52
+    namespace: docuseal
+  ignoreDifferences:
+    - group: apps
+      kind: Deployment
+      jsonPointers:
+        - /status
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+      - CreateNamespace=true
+      - ServerSideApply=true
@@ -0,0 +1,46 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: gitea-gke
+  namespace: argocd
+  annotations:
+    # Apply to the Legion Argo CD instance after registering the GKE cluster.
+    # See vault-gke.yaml for the cluster registration steps.
+    #
+    # Migration checklist (Gitea from Legion to GKE):
+    #   1. terraform apply (creates GKE cluster, Cloud SQL gitea DB, secrets)
+    #   2. Register GKE cluster with Legion Argo CD (see vault-gke.yaml)
+    #   3. Install ESO on GKE: helm install external-secrets external-secrets/external-secrets
+    #        --namespace external-secrets --create-namespace
+    #   4. Apply this Application to Legion Argo CD
+    #   5. Verify gitea pod is running on GKE with DB connectivity
+    #   6. Take a tar of /data from the Legion Gitea pod and restore to GKE PVC
+    #   7. Update Cloudflare Tunnel on Legion: remove git.neuralplatform.ai route
+    #   8. Add GKE ingress / GCP LB rule for git.neuralplatform.ai
+    #   9. Decommission Gitea on Legion (remove gitea*.yaml from servers/legion/apps/)
+spec:
+  project: default
+  source:
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
+    targetRevision: main
+    path: servers/gcp/k8s/gitea
+  destination:
+    # Replace GKE_CLUSTER_ENDPOINT after `terraform apply`:
+    #   terraform -chdir=servers/gcp output -raw gke_cluster_endpoint
+    server: https://34.63.89.52
+    namespace: gitea
+  # terminatingReplicas is a newer Deployment status field that Argo CD's
+  # bundled schema doesn't know about — causes ComparisonError during diff.
+  # Ignoring /status entirely is safe: Argo CD never manages status fields.
+  ignoreDifferences:
+    - group: apps
+      kind: Deployment
+      jsonPointers:
+        - /status
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+      - CreateNamespace=true
+      - ServerSideApply=true
@@ -0,0 +1,26 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: gitea-runner-gke
+  namespace: argocd
+spec:
+  project: default
+  source:
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
+    targetRevision: main
+    path: servers/gcp/k8s/gitea-runner
+  destination:
+    server: https://34.63.89.52
+    namespace: ci
+  ignoreDifferences:
+    - group: apps
+      kind: Deployment
+      jsonPointers:
+        - /status
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+      - CreateNamespace=true
+      - ServerSideApply=true
@@ -0,0 +1,33 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: vault-gke
+  namespace: argocd
+  annotations:
+    # Syncs plain k8s manifests for Vault to the GKE cluster.
+    # The Vault Helm release itself is in vault-helm-gke.yaml.
+    #
+    # Apply to Legion Argo CD after registering the GKE cluster:
+    #   1. gcloud container clusters get-credentials neuron-platform \
+    #        --region us-central1 --project neuron-785695
+    #   2. argocd cluster add <context-name> --name gke-neuron-platform
+    #   3. Update GKE_CLUSTER_ENDPOINT in all argocd-apps/*.yaml:
+    #        terraform -chdir=servers/gcp output -raw gke_cluster_endpoint
+spec:
+  project: default
+  source:
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
+    targetRevision: main
+    path: servers/gcp/k8s/vault
+  destination:
+    # Replace GKE_CLUSTER_ENDPOINT after `terraform apply`:
+    #   terraform -chdir=servers/gcp output -raw gke_cluster_endpoint
+    server: https://34.63.89.52
+    namespace: vault
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+      - CreateNamespace=true
+      - ServerSideApply=true
@@ -0,0 +1,175 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: vault-helm-gke
+  namespace: argocd
+  annotations:
+    # Deploys the Vault Helm chart to the GKE cluster via Legion Argo CD.
+    # The destination.server must be updated after `terraform apply`:
+    #   terraform -chdir=servers/gcp output -raw gke_cluster_endpoint
+spec:
+  project: default
+  source:
+    repoURL: https://helm.releases.hashicorp.com
+    chart: vault
+    targetRevision: "0.29.1"
+    helm:
+      values: |
+        global:
+          enabled: true
+
+        injector:
+          enabled: false
+
+        ui:
+          enabled: true
+
+        server:
+          image:
+            repository: hashicorp/vault
+            tag: "1.19.2"
+
+          # Workload Identity — Vault pod k8s SA impersonates vault-unseal GCP SA
+          # for KMS auto-unseal. Binding is in servers/gcp/gke.tf.
+          serviceAccount:
+            create: true
+            name: vault
+            annotations:
+              iam.gke.io/gcp-service-account: vault-unseal@neuron-785695.iam.gserviceaccount.com
+
+          # GKE Autopilot: no privileged containers. Vault doesn't need privilege.
+          # Request IPC_LOCK so Vault can lock memory (prevents secrets swap).
+          securityContext:
+            capabilities:
+              add:
+                - IPC_LOCK
+
+          # Single replica for now — raft standby unsealing requires the barrier
+          # key to be shared over port 8201 (cluster port), but 8201 only opens
+          # after unsealing (deadlock). Running solo on pod 0 until we resolve
+          # the standby bootstrap path. Scale back to 3 once fixed.
+          ha:
+            enabled: true
+            replicas: 1
+            raft:
+              enabled: true
+              setNodeId: true
+              config: |
+                ui = true
+
+                listener "tcp" {
+                  tls_disable     = 1
+                  address         = "[::]:8200"
+                  cluster_address = "[::]:8201"
+                }
+
+                storage "raft" {
+                  path = "/vault/data"
+
+                  retry_join {
+                    leader_api_addr = "http://vault-helm-gke-0.vault-helm-gke-internal:8200"
+                  }
+                  retry_join {
+                    leader_api_addr = "http://vault-helm-gke-1.vault-helm-gke-internal:8200"
+                  }
+                  retry_join {
+                    leader_api_addr = "http://vault-helm-gke-2.vault-helm-gke-internal:8200"
+                  }
+                }
+
+                seal "gcpckms" {
+                  project    = "neuron-785695"
+                  region     = "global"
+                  key_ring   = "vault"
+                  crypto_key = "vault-unseal"
+                }
+
+                telemetry {
+                  prometheus_retention_time = "30s"
+                  disable_hostname          = false
+                }
+
+          # Spread pods across GKE zones
+          topologySpreadConstraints:
+            - maxSkew: 1
+              topologyKey: topology.kubernetes.io/zone
+              whenUnsatisfiable: ScheduleAnyway
+              labelSelector:
+                matchLabels:
+                  app.kubernetes.io/name: vault
+                  component: server
+
+          # 10Gi SSD per pod — premium-rwo = pd-ssd on GKE Autopilot
+          dataStorage:
+            enabled: true
+            size: 10Gi
+            storageClass: standard-rwo
+            accessMode: ReadWriteOnce
+
+          # Annotations on the pod template — bumping rollme triggers a
+          # StatefulSet rolling restart (pods 2→1→0) to pick up the
+          # corrected raft retry_join addresses.
+          podAnnotations:
+            rollme: "2026-05-05-raft-join-fix"
+
+          readinessProbe:
+            enabled: true
+            # standbyok=true: standby pods pass (they serve reads).
+            # sealedok removed: sealed/uninit pods fail readiness so they
+            # leave the service endpoints and stop receiving external traffic.
+            path: "/v1/sys/health?standbyok=true"
+            initialDelaySeconds: 5
+            periodSeconds: 5
+            failureThreshold: 3
+
+          livenessProbe:
+            enabled: true
+            # Liveness keeps sealedok=true so sealed pods aren't killed —
+            # they need time to join raft and auto-unseal.
+            path: "/v1/sys/health?standbyok=true&sealedok=true&uninitcode=200"
+            initialDelaySeconds: 60
+            periodSeconds: 10
+            failureThreshold: 3
+
+          # GKE Autopilot requires resource requests on all containers
+          resources:
+            requests:
+              memory: 256Mi
+              cpu: 500m
+            limits:
+              memory: 512Mi
+              cpu: 1000m
+
+          service:
+            enabled: true
+            type: ClusterIP
+            port: 8200
+            targetPort: 8200
+            annotations:
+              # Container-native NEG for GCP Global HTTPS LB backend cutover.
+              # GKE will create a zonal NEG named k8s1-<hash>-vault-vault-helm-gke-8200-<hash>
+              # in each zone where Vault pods are scheduled.
+              # After this syncs, list NEGs:
+              #   gcloud compute network-endpoint-groups list --filter="name~vault" --project neuron-785695
+              # Then reference them in servers/gcp/vault-gke-lb.tf.
+              cloud.google.com/neg: '{"exposed_ports":{"8200":{}}}'
+
+          # Ingress disabled — Vault is exposed via GCP HTTPS LB.
+          # After migration, update the existing LB backend (vault-nodes.tf)
+          # to target a GKE NEG instead of the GCE instance groups.
+          # See: https://cloud.google.com/kubernetes-engine/docs/how-to/standalone-neg
+          ingress:
+            enabled: false
+
+  destination:
+    # Replace GKE_CLUSTER_ENDPOINT after `terraform apply`:
+    #   terraform -chdir=servers/gcp output -raw gke_cluster_endpoint
+    server: https://34.63.89.52
+    namespace: vault
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+      - CreateNamespace=true
+      - ServerSideApply=true
@@ -0,0 +1,101 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: docuseal
+  namespace: docuseal
+  labels:
+    app: docuseal
+spec:
+  replicas: 1
+  # Recreate — only one pod can hold the RWO PVC at a time.
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app: docuseal
+  template:
+    metadata:
+      labels:
+        app: docuseal
+    spec:
+      serviceAccountName: docuseal
+      containers:
+        - name: docuseal
+          image: ee.docuseal.com/lip5viwf/ds-ee:latest
+          ports:
+            - containerPort: 3000
+          env:
+            - name: SECRET_KEY_BASE
+              valueFrom:
+                secretKeyRef:
+                  name: docuseal-secrets
+                  key: secret_key_base
+            - name: FORCE_SSL
+              value: "false"
+            - name: HOST
+              value: "sign.neurontechnologies.ai"
+            - name: SMTP_ADDRESS
+              valueFrom:
+                secretKeyRef:
+                  name: docuseal-secrets
+                  key: smtp_host
+            - name: SMTP_PORT
+              valueFrom:
+                secretKeyRef:
+                  name: docuseal-secrets
+                  key: smtp_port
+            - name: SMTP_USERNAME
+              valueFrom:
+                secretKeyRef:
+                  name: docuseal-secrets
+                  key: smtp_username
+            - name: SMTP_PASSWORD
+              valueFrom:
+                secretKeyRef:
+                  name: docuseal-secrets
+                  key: smtp_password
+            - name: SMTP_FROM
+              valueFrom:
+                secretKeyRef:
+                  name: docuseal-secrets
+                  key: smtp_from
+          volumeMounts:
+            - name: data
+              mountPath: /data
+          readinessProbe:
+            httpGet:
+              path: /
+              port: 3000
+            initialDelaySeconds: 15
+            periodSeconds: 10
+          resources:
+            requests:
+              cpu: 100m
+              memory: 256Mi
+            limits:
+              cpu: 500m
+              memory: 512Mi
+      volumes:
+        - name: data
+          persistentVolumeClaim:
+            claimName: docuseal-data
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: docuseal
+  namespace: docuseal
+  annotations:
+    # External GCP Network LB — Cloudflare proxies in front, provides TLS.
+    # After provisioning, get the external IP:
+    #   kubectl -n docuseal get svc docuseal
+    # Update Cloudflare DNS A record for sign.neurontechnologies.ai to this IP.
+    cloud.google.com/load-balancer-type: "External"
+spec:
+  selector:
+    app: docuseal
+  ports:
+    - name: http
+      port: 80
+      targetPort: 3000
+  type: LoadBalancer
@@ -0,0 +1,56 @@
+---
+# SecretStore for GKE — uses GCP Secret Manager directly via Workload Identity.
+# The DocuSeal GCP SA has secretmanager.secretAccessor on its own secrets
+# (see servers/gcp/docuseal.tf).
+apiVersion: external-secrets.io/v1
+kind: SecretStore
+metadata:
+  name: gcp-secretmanager
+  namespace: docuseal
+spec:
+  provider:
+    gcpsm:
+      projectID: neuron-785695
+      # Workload Identity — ESO impersonates the docuseal-gke GCP SA to access
+      # Secret Manager. No JSON key file required.
+      auth:
+        workloadIdentity:
+          clusterLocation: us-central1
+          clusterName: neuron-platform
+          serviceAccountRef:
+            name: docuseal
+            namespace: docuseal
+---
+# Pull all DocuSeal secrets from GCP Secret Manager into a single k8s Secret.
+apiVersion: external-secrets.io/v1
+kind: ExternalSecret
+metadata:
+  name: docuseal-secrets
+  namespace: docuseal
+spec:
+  refreshInterval: 1h
+  secretStoreRef:
+    name: gcp-secretmanager
+    kind: SecretStore
+  target:
+    name: docuseal-secrets
+    creationPolicy: Owner
+  data:
+    - secretKey: secret_key_base
+      remoteRef:
+        key: docuseal-secret-key-base
+    - secretKey: smtp_host
+      remoteRef:
+        key: docuseal-smtp-host
+    - secretKey: smtp_port
+      remoteRef:
+        key: docuseal-smtp-port
+    - secretKey: smtp_username
+      remoteRef:
+        key: docuseal-smtp-username
+    - secretKey: smtp_password
+      remoteRef:
+        key: docuseal-smtp-password
+    - secretKey: smtp_from
+      remoteRef:
+        key: docuseal-smtp-from
@@ -0,0 +1,6 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: docuseal
+  labels:
+    app.kubernetes.io/name: docuseal
@@ -0,0 +1,12 @@
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: docuseal-data
+  namespace: docuseal
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 5Gi
+  storageClassName: standard-rwo
@@ -0,0 +1,10 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: docuseal
+  namespace: docuseal
+  annotations:
+    # Workload Identity — allows ESO to authenticate to GCP Secret Manager
+    # as the docuseal-gke GCP SA without a JSON key file.
+    # The GCP SA binding is in servers/gcp/docuseal.tf (docuseal_workload_identity).
+    iam.gke.io/gcp-service-account: docuseal-gke@neuron-785695.iam.gserviceaccount.com
@@ -0,0 +1,116 @@
+---
+# Gitea Actions runner on GKE Autopilot.
+#
+# Architecture:
+#   - init container registers the runner with Gitea (--no-interactive, idempotent)
+#   - main container runs act_runner daemon
+#
+# LIMITATION: GKE Autopilot does not allow privileged containers, so Docker-in-Docker
+# (DinD) is not available. The runner uses "host" label mode, which runs steps
+# directly inside the runner pod container rather than spawning Docker containers.
+# forgejo-runner v11 still requires a Docker daemon to start — this deployment
+# will fail until one of the following is resolved:
+#
+#   Option A (recommended): Switch to a GKE Standard node pool for the ci namespace.
+#     Add a standard node pool with sandbox.config.sandboxType=gvisor or allow
+#     privileged pods on specific node pools.
+#
+#   Option B: Use Cloud Build for docker build steps; use this runner for shell steps.
+#     Requires forking all workflow files to remove docker steps.
+#
+#   Option C: Point docker_host at an external Docker TCP daemon (e.g., on the GCE runner VM).
+#
+# Until resolved, the existing GCE VM runner (gitea-runner-1 in us-central1-a)
+# handles CI jobs. This deployment is a placeholder for when privileged pods are available.
+#
+# The ci namespace has pod-security.kubernetes.io/enforce: privileged set
+# (see namespace.yaml) but Autopilot's Warden enforces this at the node level.
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: gitea-runner
+  namespace: ci
+  labels:
+    app: gitea-runner
+spec:
+  # Set to 0 until the DinD/privileged-pod blocker is resolved (see comments above).
+  # Change to 1 when a Docker daemon is available (Standard node pool or external TCP).
+  replicas: 0
+  selector:
+    matchLabels:
+      app: gitea-runner
+  template:
+    metadata:
+      labels:
+        app: gitea-runner
+    spec:
+      serviceAccountName: gitea-runner
+      securityContext:
+        runAsNonRoot: false
+      initContainers:
+        - name: register
+          image: us-central1-docker.pkg.dev/neuron-785695/neuron-ci/ci-base:latest
+          command: ["/bin/sh", "-c"]
+          args:
+            - |
+              # Idempotent registration — skip if already registered (.runner file exists).
+              if [ -f /data/.runner ]; then
+                echo "Runner already registered, skipping."
+              else
+                act_runner register \
+                  --instance "${GITEA_INSTANCE_URL}" \
+                  --token "${GITEA_RUNNER_REGISTRATION_TOKEN}" \
+                  --name "gke-runner-$(hostname)" \
+                  --labels "ubuntu-latest:host,ubuntu-24.04:host,linux,x64" \
+                  --no-interactive
+              fi
+              cat > /data/config.yaml << 'CONFIGEOF'
+              runner:
+                capacity: 2
+                timeout: 3h
+                envs:
+                  BASH_ENV: "/usr/local/bin/git-ssh-init.sh"
+              container:
+                # GKE Autopilot blocks privileged DinD — steps run directly
+                # in the runner container (host mode). Labels use :host suffix.
+                network: host
+                docker_host: "-"
+                force_pull: false
+                valid_volumes: []
+              CONFIGEOF
+          env:
+            # Use the internal Gitea service to avoid Cloudflare Access auth.
+            # Gitea is in the same GKE cluster — use the ClusterIP service.
+            - name: GITEA_INSTANCE_URL
+              value: "http://gitea.gitea.svc.cluster.local:80"
+          envFrom:
+            - secretRef:
+                name: gitea-runner-token
+          volumeMounts:
+            - mountPath: /data
+              name: data
+          workingDir: /data
+      containers:
+        - name: runner
+          image: us-central1-docker.pkg.dev/neuron-785695/neuron-ci/ci-base:latest
+          command: ["act_runner", "daemon", "--config", "/data/config.yaml"]
+          env:
+            - name: GITEA_INSTANCE_URL
+              value: "http://gitea.gitea.svc.cluster.local:80"
+          envFrom:
+            - secretRef:
+                name: gitea-runner-token
+          resources:
+            requests:
+              cpu: "500m"
+              memory: "512Mi"
+            limits:
+              cpu: "4"
+              memory: "4Gi"
+          volumeMounts:
+            - mountPath: /data
+              name: data
+          workingDir: /data
+      volumes:
+        - name: data
+          emptyDir: {}
@@ -0,0 +1,39 @@
+---
+# SecretStore for the CI namespace — uses GCP Secret Manager via Workload Identity.
+# The gitea-runner-gke GCP SA has secretmanager.secretAccessor on gitea-runner-token
+# (see servers/gcp/gitea-runner.tf).
+apiVersion: external-secrets.io/v1
+kind: SecretStore
+metadata:
+  name: gcp-secretmanager
+  namespace: ci
+spec:
+  provider:
+    gcpsm:
+      projectID: neuron-785695
+      auth:
+        workloadIdentity:
+          clusterLocation: us-central1
+          clusterName: neuron-platform
+          serviceAccountRef:
+            name: gitea-runner
+            namespace: ci
+---
+# Pull the Gitea runner registration token from Secret Manager into a k8s Secret.
+apiVersion: external-secrets.io/v1
+kind: ExternalSecret
+metadata:
+  name: gitea-runner-token
+  namespace: ci
+spec:
+  refreshInterval: 1h
+  secretStoreRef:
+    name: gcp-secretmanager
+    kind: SecretStore
+  target:
+    name: gitea-runner-token
+    creationPolicy: Owner
+  data:
+    - secretKey: GITEA_RUNNER_REGISTRATION_TOKEN
+      remoteRef:
+        key: gitea-runner-token
@@ -0,0 +1,8 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: ci
+  labels:
+    app.kubernetes.io/name: ci
+    # DinD (Docker-in-Docker) requires privileged pods.
+    pod-security.kubernetes.io/enforce: privileged
@@ -0,0 +1,11 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: gitea-runner
+  namespace: ci
+  annotations:
+    # Workload Identity — allows ESO (and optionally the runner pod) to
+    # authenticate to GCP Secret Manager as the gitea-runner-gke GCP SA
+    # without a JSON key file.
+    # The GCP SA binding is in servers/gcp/gitea-runner.tf (gitea_runner_gke_workload_identity).
+    iam.gke.io/gcp-service-account: gitea-runner-gke@neuron-785695.iam.gserviceaccount.com
@@ -0,0 +1,104 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: gitea-custom-css
+  namespace: gitea
+data:
+  header.tmpl: |
+    <link rel="stylesheet" href="/assets/css/custom.css">
+  custom.css: |
+    /* ── Typography & base ── */
+    :root {
+      --color-primary: #6366f1;
+      --color-primary-dark: #4f46e5;
+      --color-secondary: #8b5cf6;
+      --color-bg: #0f0f17;
+      --color-surface: #16161f;
+      --color-surface-2: #1e1e2e;
+      --color-border: #2a2a3d;
+      --color-text: #e2e2f0;
+      --color-text-muted: #8888aa;
+      --color-green: #22d3a5;
+      --radius: 10px;
+    }
+
+    body {
+      background: var(--color-bg) !important;
+      color: var(--color-text) !important;
+      font-family: "Inter", -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif !important;
+    }
+
+    /* ── Top navbar ── */
+    .navbar, #navbar {
+      background: var(--color-surface) !important;
+      border-bottom: 1px solid var(--color-border) !important;
+      box-shadow: 0 1px 12px rgba(0,0,0,0.4) !important;
+    }
+
+    .navbar .brand svg, .navbar .brand img { filter: brightness(1.2); }
+
+    /* ── Sidebar & panels ── */
+    .repository, .ui.container, .ui.segment,
+    .ui.card, .ui.cards > .card {
+      background: var(--color-surface) !important;
+      border: 1px solid var(--color-border) !important;
+      border-radius: var(--radius) !important;
+    }
+
+    /* ── Buttons ── */
+    .ui.primary.button, .ui.green.button {
+      background: var(--color-primary) !important;
+      border: none !important;
+      border-radius: 8px !important;
+    }
+    .ui.primary.button:hover { background: var(--color-primary-dark) !important; }
+
+    /* ── Inputs ── */
+    input, textarea, select,
+    .ui.input > input, .ui.dropdown {
+      background: var(--color-surface-2) !important;
+      border: 1px solid var(--color-border) !important;
+      color: var(--color-text) !important;
+      border-radius: 8px !important;
+    }
+
+    /* ── Sign-in page ── */
+    .user.signin .ui.segment,
+    .user.signup .ui.segment {
+      background: var(--color-surface) !important;
+      border: 1px solid var(--color-border) !important;
+      border-radius: 16px !important;
+      box-shadow: 0 8px 32px rgba(0,0,0,0.5) !important;
+      padding: 2.5rem !important;
+    }
+
+    /* ── Repo file tree ── */
+    .repository.file.list .file-list {
+      background: var(--color-surface) !important;
+      border-radius: var(--radius) !important;
+      border: 1px solid var(--color-border) !important;
+    }
+
+    .repository.file.list .file-list tr:hover td {
+      background: var(--color-surface-2) !important;
+    }
+
+    /* ── Labels & badges ── */
+    .ui.label { border-radius: 6px !important; }
+
+    /* ── Dashboard activity feed ── */
+    .feeds .news { border-bottom: 1px solid var(--color-border) !important; }
+
+    /* ── Code blocks ── */
+    pre, code {
+      background: var(--color-surface-2) !important;
+      border: 1px solid var(--color-border) !important;
+      border-radius: 6px !important;
+    }
+
+    /* ── Muted footer ── */
+    #footer {
+      background: var(--color-surface) !important;
+      border-top: 1px solid var(--color-border) !important;
+      color: var(--color-text-muted) !important;
+    }
@@ -0,0 +1,163 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: gitea
+  namespace: gitea
+  labels:
+    app: gitea
+spec:
+  replicas: 1
+  # Recreate — only one pod can hold the RWO PVC at a time.
+  # Scale to multiple replicas only after adding shared storage (e.g. Filestore).
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app: gitea
+  template:
+    metadata:
+      labels:
+        app: gitea
+    spec:
+      serviceAccountName: gitea
+
+      containers:
+        - name: gitea
+          image: gitea/gitea:1.25.5
+          ports:
+            - name: http
+              containerPort: 3000
+            - name: ssh
+              containerPort: 22
+          env:
+            # Database — connect through the Cloud SQL Auth Proxy unix socket
+            - name: GITEA__database__DB_TYPE
+              value: postgres
+            # Unix socket path used by the Cloud SQL Auth Proxy sidecar
+            - name: GITEA__database__HOST
+              value: /cloudsql/neuron-785695:us-central1:neuron-prod-pg15
+            - name: GITEA__database__NAME
+              value: gitea
+            - name: GITEA__database__USER
+              value: gitea
+            - name: GITEA__database__PASSWD
+              valueFrom:
+                secretKeyRef:
+                  name: gitea-db
+                  key: password
+            # Server
+            - name: GITEA__server__DOMAIN
+              value: git.neuralplatform.ai
+            - name: GITEA__server__ROOT_URL
+              value: https://git.neuralplatform.ai
+            - name: GITEA__server__SSH_DOMAIN
+              value: git.neuralplatform.ai
+            - name: GITEA__server__SSH_PORT
+              value: "22"
+            - name: GITEA__server__START_SSH_SERVER
+              value: "false"
+            # Service
+            - name: GITEA__service__DISABLE_REGISTRATION
+              value: "true"
+            - name: GITEA__service__REQUIRE_SIGNIN_VIEW
+              value: "false"
+            # Security
+            - name: GITEA__security__INSTALL_LOCK
+              value: "true"
+            # Packages
+            - name: GITEA__packages__ENABLED
+              value: "true"
+            # Webhooks — allow calls back into the cluster and GKE VPC
+            - name: GITEA__webhook__ALLOWED_HOST_LIST
+              value: 10.0.0.0/8,172.16.0.0/12,192.168.0.0/16
+            # Actions
+            - name: GITEA__actions__ENABLED
+              value: "true"
+            - name: GITEA__actions__DEFAULT_ACTIONS_URL
+              value: https://code.forgejo.org
+          volumeMounts:
+            - name: data
+              mountPath: /data
+            - name: cloudsql-socket
+              mountPath: /cloudsql
+            - name: custom-css
+              mountPath: /data/gitea/custom/public/assets/css/custom.css
+              subPath: custom.css
+            - name: custom-css
+              mountPath: /data/gitea/custom/templates/custom/header.tmpl
+              subPath: header.tmpl
+          resources:
+            requests:
+              memory: 256Mi
+              cpu: 100m
+            limits:
+              memory: 512Mi
+              cpu: 500m
+          readinessProbe:
+            httpGet:
+              path: /
+              port: 3000
+            initialDelaySeconds: 10
+            periodSeconds: 10
+            failureThreshold: 3
+          livenessProbe:
+            httpGet:
+              path: /
+              port: 3000
+            initialDelaySeconds: 30
+            periodSeconds: 15
+            failureThreshold: 3
+
+        # Cloud SQL Auth Proxy sidecar — provides unix socket at /cloudsql/
+        # Authenticates to Cloud SQL using Workload Identity (no key file).
+        - name: cloud-sql-proxy
+          image: gcr.io/cloud-sql-connectors/cloud-sql-proxy:2
+          args:
+            - "--structured-logs"
+            - "--unix-socket=/cloudsql"
+            - "neuron-785695:us-central1:neuron-prod-pg15"
+          securityContext:
+            runAsNonRoot: true
+          volumeMounts:
+            - name: cloudsql-socket
+              mountPath: /cloudsql
+          resources:
+            requests:
+              memory: 32Mi
+              cpu: 10m
+            limits:
+              memory: 128Mi
+              cpu: 100m
+
+      volumes:
+        - name: data
+          persistentVolumeClaim:
+            claimName: gitea-data
+        - name: cloudsql-socket
+          emptyDir: {}
+        - name: custom-css
+          configMap:
+            name: gitea-custom-css
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: gitea
+  namespace: gitea
+  annotations:
+    # External GCP Network LB — Cloudflare proxies in front, provides TLS.
+    # After provisioning, get the external IP:
+    #   kubectl --context=gke_neuron-785695_us-central1_neuron-platform -n gitea get svc gitea
+    # Update Cloudflare DNS A record for git.neuralplatform.ai to this IP.
+    cloud.google.com/load-balancer-type: "External"
+spec:
+  selector:
+    app: gitea
+  ports:
+    - name: http
+      port: 80
+      targetPort: 3000
+    - name: ssh
+      port: 22
+      targetPort: 22
+  type: LoadBalancer
@@ -0,0 +1,75 @@
+---
+# SecretStore for GKE — uses GCP Secret Manager directly via Workload Identity.
+# On GKE we use the GCP provider instead of a Vault-backed store, since
+# Vault itself may be in the process of being migrated.
+# The Gitea GCP SA has secretmanager.secretAccessor on its own secret (see cloud-sql.tf).
+#
+# Pre-requisite: install ESO on GKE before applying this:
+#   helm install external-secrets external-secrets/external-secrets \
+#     --namespace external-secrets --create-namespace
+apiVersion: external-secrets.io/v1
+kind: SecretStore
+metadata:
+  name: gcp-secretmanager
+  namespace: gitea
+spec:
+  provider:
+    gcpsm:
+      projectID: neuron-785695
+      # Workload Identity — ESO impersonates the gitea-gke GCP SA to access
+      # Secret Manager. The gitea SA has secretAccessor on gitea-db-password.
+      auth:
+        workloadIdentity:
+          clusterLocation: us-central1
+          clusterName: neuron-platform
+          serviceAccountRef:
+            name: gitea
+            namespace: gitea
+---
+# gitea-db — Gitea database password, pulled from GCP Secret Manager.
+# The Secret Manager secret "gitea-database-url" stores the full DSN, but we
+# extract just the password field for use in GITEA__database__PASSWD.
+#
+# The full DSN format from Terraform:
+#   host=/cloudsql/<conn> user=gitea password=<pw> dbname=gitea sslmode=disable
+#
+# ESO extracts the raw secret value. Since GCP Secret Manager stores the full
+# DSN as a single string, we store the password separately as "gitea-db-password"
+# so Gitea can receive it as a discrete env var.
+#
+# Bootstrap: after `terraform apply`, run:
+#   PASSWORD=$(gcloud secrets versions access latest --secret=gitea-database-url \
+#     | grep -oP '(?<=password=)\S+')
+#   echo -n "$PASSWORD" | gcloud secrets create gitea-db-password \
+#     --data-file=- --project=neuron-785695
+#
+# Or simpler — let Terraform write it directly. The gitea-db-password secret
+# is managed by the gitea-database-url secret version output. Use the full DSN
+# secret and parse in-pod, or store password separately.
+#
+# For simplicity: pull the full DSN and use it as GITEA__database__PASSWD
+# is wrong (it's a DSN, not a password). Instead, use the Cloud SQL proxy
+# unix socket and no password — configure Gitea to use peer auth.
+#
+# ACTUAL APPROACH: ExternalSecret pulls the full DSN string into a k8s Secret
+# key "dsn". A separate gitea-db secret provides just the password field.
+# Terraform outputs both; add a gitea-db-password Secret Manager secret in cloud-sql.tf.
+apiVersion: external-secrets.io/v1
+kind: ExternalSecret
+metadata:
+  name: gitea-db
+  namespace: gitea
+spec:
+  refreshInterval: 1h
+  secretStoreRef:
+    name: gcp-secretmanager
+    kind: SecretStore
+  target:
+    name: gitea-db
+    creationPolicy: Owner
+  data:
+    - secretKey: password
+      remoteRef:
+        # This secret is populated by Terraform (gitea-db-password in cloud-sql.tf)
+        # It contains just the raw database password (no DSN prefix).
+        key: gitea-db-password
@@ -0,0 +1,6 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: gitea
+  labels:
+    app.kubernetes.io/name: gitea
@@ -0,0 +1,14 @@
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: gitea-data
+  namespace: gitea
+spec:
+  # standard-rwo = pd-balanced on GKE Autopilot (ReadWriteOnce)
+  # Use premium-rwo (pd-ssd) if repo performance becomes a bottleneck.
+  storageClassName: standard-rwo
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 50Gi
@@ -0,0 +1,10 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: gitea
+  namespace: gitea
+  annotations:
+    # Workload Identity — allows the Cloud SQL Auth Proxy sidecar to authenticate
+    # to Cloud SQL as the gitea GCP SA without a JSON key file.
+    # The GCP SA binding is in servers/gcp/gke.tf (gitea_workload_identity).
+    iam.gke.io/gcp-service-account: gitea-gke@neuron-785695.iam.gserviceaccount.com
@@ -0,0 +1,6 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: vault
+  labels:
+    app.kubernetes.io/name: vault
@@ -20,17 +20,9 @@ terraform {
    }
  }

-  backend "s3" {
-    bucket   = "legion-terraform-state"
-    key      = "gcp/terraform.tfstate"
-    region   = "auto"
-    endpoint = "https://651161e0a3d321561b4c90b5bcd5f15b.r2.cloudflarestorage.com"
-
-    # R2 is S3-compatible but not AWS — skip AWS-specific checks
-    skip_credentials_validation = true
-    skip_metadata_api_check     = true
-    skip_region_validation      = true
-    force_path_style            = true
+  backend "gcs" {
+    bucket = "neuron-785695-terraform-state"
+    prefix = "gcp"
  }
 }

@@ -15,9 +15,41 @@ set -euxo pipefail
 exec > >(tee /var/log/runner-bootstrap.log) 2>&1

 apt-get update
-# nodejs/npm needed for JavaScript actions like actions/checkout and
-# google-github-actions/auth. python3 is for our inline label-rewrite below.
-apt-get install -y curl ca-certificates docker.io git jq nodejs npm python3
+
+# Core system tools and C build dependencies required by CI jobs.
+# libcurl4-openssl-dev + build-essential are needed by the El compiler C build;
+# libssl-dev/libsqlite3-dev/libpq-dev for downstream projects.
+# python3 is for the inline label-rewrite step below.
+apt-get install -y --no-install-recommends \
+    curl \
+    ca-certificates \
+    docker.io \
+    git \
+    jq \
+    python3 \
+    wget \
+    unzip \
+    zip \
+    xz-utils \
+    rsync \
+    file \
+    sudo \
+    make \
+    build-essential \
+    pkg-config \
+    gcc \
+    libcurl4-openssl-dev \
+    libssl-dev \
+    libsqlite3-dev \
+    libpq-dev \
+    libffi-dev \
+    zlib1g-dev
+
+# Node.js 20 LTS via NodeSource — Ubuntu 24.04's bundled nodejs is 18.x
+# which works but 20 LTS matches what our other CI images use.
+curl -fsSL https://deb.nodesource.com/setup_20.x | bash -
+apt-get install -y --no-install-recommends nodejs
+npm install -g yarn

 # Make docker usable by the unprivileged runner user
 systemctl enable --now docker
@@ -25,6 +57,10 @@ systemctl enable --now docker
 useradd -m -s /bin/bash runner || true
 usermod -aG docker runner

+# Allow the runner user to install packages and run system commands
+# in CI workflow steps without a password prompt.
+echo "runner ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers
+
 # act_runner — pinned to a known-good release. Bump RUNNER_VERSION when
 # upgrading. The project moved from gitea/act_runner to gitea/runner around
 # the 0.6.x series; the binary inside the asset is still called act_runner.
@@ -1,204 +1,49 @@
-# ── Vault HA Cluster — GCE-based Raft ────────────────────────────────────────
+# ── Vault LB — GCP Global HTTPS LB ───────────────────────────────────────────
 #
-# Three GCE e2-small VMs across us-central1-{a,b,c} running HashiCorp Vault
-# in Raft HA mode. Auto-unseal via the existing GCP KMS key (vault-kms.tf).
+# LB frontend is unchanged: forwarding rule → target HTTPS proxy → url map → backend.
+# Backend uses GKE container-native NEGs (cut over from GCE instance groups).
 #
-# Architecture:
-#   - vault-node-1 (us-central1-a) — bootstrapped first, others join via retry_join
-#   - vault-node-2 (us-central1-b)
-#   - vault-node-3 (us-central1-c)
-#   - Each VM gets the vault-node SA attached for KMS auto-unseal + ADC
-#   - Internal traffic: VMs talk Raft over port 8201 on GCP internal IPs
-#   - External access: HTTPS regional LB → port 8200
-#     (vault.neuralplatform.ai Cloudflare DNS → GCP LB IP)
+# Architecture (GKE):
+#   - Vault Helm chart runs 3 pods in namespace: vault on GKE neuron-platform cluster
+#   - GKE auto-creates zonal container-native NEGs via Service annotation:
+#       cloud.google.com/neg: '{"exposed_ports":{"8200":{}}}'
+#   - NEG backends replace GCE instance group backends in the Global HTTPS LB
+#   - DNS unchanged: vault.neuralplatform.ai → 34.54.164.21 (same GCP Global LB IP)
 #
-# Ops SSH:
-#   gcloud compute ssh vault-node-1 --zone=us-central1-a --tunnel-through-iap
-#
-# After first boot, initialize Vault on node-1:
-#   vault operator init   (save the root token + recovery keys securely)
-# Nodes 2 and 3 auto-join via retry_join once they boot.
+# GCE nodes (vault-node-1/2/3) and their instance groups have been removed.
+# GKE Vault (via Workload Identity + vault-unseal SA) is the sole Vault backend.

 locals {
-  vault_version = "1.19.2"
-
-  vault_nodes = {
-    "vault-node-1" = { zone = "us-central1-a", node_id = "vault-node-1" }
-    "vault-node-2" = { zone = "us-central1-b", node_id = "vault-node-2" }
-    "vault-node-3" = { zone = "us-central1-c", node_id = "vault-node-3" }
+  # Container-native NEG names created by GKE from the Vault Service annotation:
+  #   cloud.google.com/neg: '{"exposed_ports":{"8200":{}}}'
+  # GKE creates one NEG per zone where Vault pods are scheduled.
+  # Confirmed with: gcloud compute network-endpoint-groups list --project neuron-785695
+  vault_neg_names = {
+    "us-central1-b" = "k8s1-bfbeff02-vault-vault-helm-gke-8200-db1a474f"
+    "us-central1-c" = "k8s1-bfbeff02-vault-vault-helm-gke-8200-db1a474f"
+    "us-central1-f" = "k8s1-bfbeff02-vault-vault-helm-gke-8200-db1a474f"
  }
 }

-# ── Service Account for Vault nodes ──────────────────────────────────────────
-# Reuses the vault-unseal SA from vault-kms.tf for KMS access.
-# Additional roles: logging + metrics from ops agent.
+# ── GKE container-native NEG data sources ────────────────────────────────────
+# GKE auto-creates these when the Vault Service annotation is applied.

-resource "google_project_iam_member" "vault_node_log_writer" {
-  project = var.project_id
-  role    = "roles/logging.logWriter"
-  member  = "serviceAccount:${google_service_account.vault_unseal.email}"
-}
-
-resource "google_project_iam_member" "vault_node_metric_writer" {
-  project = var.project_id
-  role    = "roles/monitoring.metricWriter"
-  member  = "serviceAccount:${google_service_account.vault_unseal.email}"
-}
-
-# Allow Vault nodes to read their own instance metadata (needed for GCP auth method later)
-resource "google_project_iam_member" "vault_node_compute_viewer" {
-  project = var.project_id
-  role    = "roles/compute.viewer"
-  member  = "serviceAccount:${google_service_account.vault_unseal.email}"
-}
-
-# ── Startup script staged in GCS ─────────────────────────────────────────────
-# Stored in the existing runner-assets bucket (reuse infrastructure).
-# The script installs Vault, writes the config, and starts the systemd unit.
-
-resource "google_storage_bucket_object" "vault_startup" {
-  name   = "vault/startup.sh"
-  bucket = google_storage_bucket.runner_assets.name
-  source = "${path.module}/vault/startup.sh"
-
-  metadata = {
-    sha256 = filesha256("${path.module}/vault/startup.sh")
-  }
-}
-
-resource "google_storage_bucket_iam_member" "vault_node_bucket_read" {
-  bucket = google_storage_bucket.runner_assets.name
-  role   = "roles/storage.objectViewer"
-  member = "serviceAccount:${google_service_account.vault_unseal.email}"
-}
-
-# ── Vault node VMs ────────────────────────────────────────────────────────────
-
-resource "google_compute_instance" "vault_node" {
-  for_each     = local.vault_nodes
-  name         = each.key
-  machine_type = "e2-small"
-  zone         = each.value.zone
-  project      = var.project_id
-
-  tags = ["vault-node", "allow-iap-ssh"]
-
-  boot_disk {
-    initialize_params {
-      image = "projects/debian-cloud/global/images/family/debian-12"
-      size  = 20
-      type  = "pd-balanced"
-    }
-  }
-
-  # Separate persistent disk for Raft data — survives VM recreation
-  attached_disk {
-    source      = google_compute_disk.vault_data[each.key].self_link
-    device_name = "vault-data"
-    mode        = "READ_WRITE"
-  }
-
-  network_interface {
-    network = "default"
-    # No external IP — accessed via IAP SSH and the internal LB
-    # Vault API is published externally via the regional LB below
-    access_config {}
-  }
-
-  service_account {
-    email  = google_service_account.vault_unseal.email
-    scopes = ["cloud-platform"]
-  }
-
-  metadata = {
-    # Pull and execute the real startup script from GCS
-    startup-script = <<-EOT
-      #!/usr/bin/env bash
-      set -euxo pipefail
-      apt-get update -y
-      apt-get install -y curl ca-certificates apt-transport-https gnupg google-cloud-cli
-      gsutil cat gs://${google_storage_bucket.runner_assets.name}/${google_storage_bucket_object.vault_startup.name} \
-        > /tmp/vault-startup.sh
-      chmod +x /tmp/vault-startup.sh
-      VAULT_NODE_ID="${each.value.node_id}" \
-      VAULT_VERSION="${local.vault_version}" \
-        /tmp/vault-startup.sh
-    EOT
-
-    enable-oslogin = "TRUE"
-  }
-
-  allow_stopping_for_update = true
-
-  depends_on = [
-    google_compute_disk.vault_data,
-    google_storage_bucket_object.vault_startup,
-    google_storage_bucket_iam_member.vault_node_bucket_read,
-  ]
-}
-
-# ── Persistent data disks ─────────────────────────────────────────────────────
-# 10 GiB per node. Kept separate from the boot disk so Raft data
-# survives a full VM deletion and recreation.
-
-resource "google_compute_disk" "vault_data" {
-  for_each = local.vault_nodes
-  name     = "${each.key}-data"
-  zone     = each.value.zone
+data "google_compute_network_endpoint_group" "vault_gke" {
+  for_each = local.vault_neg_names
+  name     = each.value
+  zone     = each.key
  project  = var.project_id
-  type     = "pd-ssd"
-  size     = 10
-
-  labels = {
-    managed-by = "terraform"
-    service    = "vault"
-    node       = each.key
-  }
-
-  lifecycle {
-    prevent_destroy = true
-  }
 }

-# ── Firewall rules ────────────────────────────────────────────────────────────
-
-# IAP SSH — ops access without a public SSH port
-resource "google_compute_firewall" "vault_iap_ssh" {
-  name      = "vault-iap-ssh"
-  network   = "default"
-  project   = var.project_id
-  direction = "INGRESS"
-  priority  = 1000
-
-  allow {
-    protocol = "tcp"
-    ports    = ["22"]
-  }
-
-  source_ranges = ["35.235.240.0/20"] # GCP IAP CIDR
-  target_tags   = ["vault-node"]
-}
-
-# Raft cluster traffic — node-to-node port 8201 (internal only)
-resource "google_compute_firewall" "vault_raft" {
-  name      = "vault-raft-internal"
-  network   = "default"
-  project   = var.project_id
-  direction = "INGRESS"
-  priority  = 1000
-
-  allow {
-    protocol = "tcp"
-    ports    = ["8201"]
-  }
-
-  source_tags = ["vault-node"]
-  target_tags = ["vault-node"]
-}
-
-# Vault API — allow from the GCP health check ranges and the LB
-resource "google_compute_firewall" "vault_api" {
-  name      = "vault-api-from-lb"
+# ── Firewall — GKE health checks ─────────────────────────────────────────────
+# Allow GCP health check probers to reach pod IPs on port 8200.
+# Container-native NEGs direct health checks to pod IPs (not node IPs).
+# GKE Autopilot pod CIDR: 10.45.128.0/22 (from cluster ip_allocation_policy).
+# Without this rule, GCP LB health checks fail → "no healthy upstream".
+# No target_tags — applies to all instances/pods in the default network.
+# Safe: GCP health check ranges (130.211.0.0/22, 35.191.0.0/16) are GCP-internal only.
+resource "google_compute_firewall" "vault_api_gke" {
+  name      = "vault-api-from-lb-gke"
  network   = "default"
  project   = var.project_id
  direction = "INGRESS"
@@ -209,35 +54,15 @@ resource "google_compute_firewall" "vault_api" {
    ports    = ["8200"]
  }

-  # GCP health check ranges (130.211.0.0/22, 35.191.0.0/16)
-  # and RFC1918 for any internal service access
-  source_ranges = ["130.211.0.0/22", "35.191.0.0/16", "10.0.0.0/8", "172.16.0.0/12", "192.168.0.0/16"]
-  target_tags   = ["vault-node"]
-}
-
-# ── Instance Groups (one unmanaged group per zone for the LB) ─────────────────
-
-resource "google_compute_instance_group" "vault" {
-  for_each  = local.vault_nodes
-  name      = "vault-${each.key}"
-  zone      = each.value.zone
-  project   = var.project_id
-  instances = [google_compute_instance.vault_node[each.key].self_link]
-
-  named_port {
-    name = "vault-api"
-    port = 8200
-  }
+  # GCP health check source ranges only — restricted port, low risk
+  source_ranges = ["130.211.0.0/22", "35.191.0.0/16"]
 }

 # ── Regional HTTPS LB for vault.neuralplatform.ai ─────────────────────────────
-# We use a global external HTTPS LB (same scheme as the prod marketing LB)
-# so we can attach a Google-managed cert for vault.neuralplatform.ai.
+# Global external HTTPS LB with a Google-managed cert for vault.neuralplatform.ai.
 #
 # TLS is terminated at the LB. Vault listens on plain 8200 internally.
-# The Cloudflare DNS A record for vault.neuralplatform.ai (neuralplatform zone)
-# must point to vault_lb_ip output below — add it in Cloudflare dashboard
-# or in the legion Terraform if you bring that zone under TF management.
+# Cloudflare DNS A record for vault.neuralplatform.ai → vault_lb_ip output below.

 resource "google_compute_global_address" "vault" {
  name    = "vault-ip"
@@ -275,16 +100,17 @@ resource "google_compute_backend_service" "vault" {
  load_balancing_scheme = "EXTERNAL_MANAGED"
  protocol              = "HTTP"   # Vault serves plain HTTP; TLS terminates at the LB
  timeout_sec           = 30
-  port_name             = "vault-api"

  health_checks = [google_compute_health_check.vault.self_link]

+  # GKE container-native NEG backends — one per zone where Vault pods are scheduled.
+  # RATE balancing mode is required for NEGs with EXTERNAL_MANAGED load balancers.
  dynamic "backend" {
-    for_each = local.vault_nodes
+    for_each = data.google_compute_network_endpoint_group.vault_gke
    content {
-      group           = google_compute_instance_group.vault[backend.key].self_link
-      balancing_mode  = "UTILIZATION"
-      max_utilization = 0.8
+      group                 = backend.value.self_link
+      balancing_mode        = "RATE"
+      max_rate_per_endpoint = 100
    }
  }

@@ -348,10 +174,3 @@ output "vault_lb_ip" {
  description = "Global IP for vault.neuralplatform.ai — set as DNS A record in Cloudflare (neuralplatform.ai zone)"
  value       = google_compute_global_address.vault.address
 }
-
-output "vault_node_zones" {
-  description = "Zone placement of each Vault node"
-  value = {
-    for k, v in local.vault_nodes : k => v.zone
-  }
-}
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/backup
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/github-runner
  destination:
@@ -27,7 +27,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/gitea-runner
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: neuron-prod
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/dharma
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/external-secrets
  destination:
@@ -7,7 +7,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/fornax/coordinator
  destination:
@@ -29,7 +29,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/fornax/ui
  destination:
@@ -51,7 +51,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/fornax/grafana
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/gitea
  destination:
@@ -1,165 +0,0 @@
---
-# Gitea CI runner — general-purpose (legion)
-# Uses host Docker socket for container management and docker build/push.
-apiVersion: apps/v1
-kind: Deployment
-metadata:
-  name: gitea-runner
-  namespace: ci
-  labels:
-    app: gitea-runner
-spec:
-  replicas: 1
-  selector:
-    matchLabels:
-      app: gitea-runner
-  template:
-    metadata:
-      labels:
-        app: gitea-runner
-      annotations:
-        config-version: "2026-04-27-containerd-sock"
-    spec:
-      securityContext:
-        runAsNonRoot: false  # act_runner needs root for container management
-      initContainers:
-        - name: register
-          image: registry.neuralplatform.ai/ci-base:latest
-          workingDir: /data
-          command: ["/bin/sh", "-c"]
-          args:
-            - |
-              act_runner register \
-                --instance "$GITEA_INSTANCE_URL" \
-                --token "$GITEA_RUNNER_REGISTRATION_TOKEN" \
-                --name legion \
-                --labels "self-hosted:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-latest:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-24.04:docker://registry.neuralplatform.ai/ci-base:latest,linux,x64" \
-                --no-interactive
-              cat > /data/config.yaml << 'EOF'
-              runner:
-                capacity: 2
-                timeout: 3h
-              container:
-                network: host
-                docker_host: "unix:///run/k3s/containerd/containerd.sock"
-                force_pull: false
-                valid_volumes: []
-                default_image: "registry.neuralplatform.ai/ci-base:latest"
-                extra_hosts:
-                  - "gitea.git.svc.cluster.local:10.43.1.53"
-              EOF
-          envFrom:
-            - secretRef:
-                name: gitea-runner-secret
-          volumeMounts:
-            - name: data
-              mountPath: /data
-      containers:
-        - name: runner
-          image: registry.neuralplatform.ai/ci-base:latest
-          workingDir: /data
-          command: ["act_runner", "daemon", "--config", "/data/config.yaml"]
-          envFrom:
-            - secretRef:
-                name: gitea-runner-secret
-          volumeMounts:
-            - name: data
-              mountPath: /data
-            - name: docker-sock
-              mountPath: /var/run/docker.sock
-          resources:
-            requests:
-              memory: 512Mi
-              cpu: 250m
-            limits:
-              memory: 4Gi
-              cpu: "4"
-      volumes:
-        - name: data
-          emptyDir: {}
-        - name: docker-sock
-          hostPath:
-            path: /run/k3s/containerd/containerd.sock
-            type: Socket
---
-# Neuron Technologies CI runner
-apiVersion: apps/v1
-kind: Deployment
-metadata:
-  name: neuron-technologies-runner
-  namespace: ci
-  labels:
-    app: neuron-technologies-runner
-spec:
-  replicas: 1
-  selector:
-    matchLabels:
-      app: neuron-technologies-runner
-  template:
-    metadata:
-      labels:
-        app: neuron-technologies-runner
-      annotations:
-        config-version: "2026-04-27-containerd-sock"
-    spec:
-      securityContext:
-        runAsNonRoot: false
-      initContainers:
-        - name: register
-          image: registry.neuralplatform.ai/ci-base:latest
-          workingDir: /data
-          command: ["/bin/sh", "-c"]
-          args:
-            - |
-              act_runner register \
-                --instance "$GITEA_INSTANCE_URL" \
-                --token "$GITEA_RUNNER_REGISTRATION_TOKEN" \
-                --name neuron-technologies \
-                --labels "self-hosted:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-latest:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-24.04:docker://registry.neuralplatform.ai/ci-base:latest,linux,x64" \
-                --no-interactive
-              cat > /data/config.yaml << 'EOF'
-              runner:
-                capacity: 2
-                timeout: 3h
-              container:
-                network: host
-                docker_host: "unix:///run/k3s/containerd/containerd.sock"
-                force_pull: false
-                valid_volumes: []
-                default_image: "registry.neuralplatform.ai/ci-base:latest"
-                extra_hosts:
-                  - "gitea.git.svc.cluster.local:10.43.1.53"
-              EOF
-          envFrom:
-            - secretRef:
-                name: neuron-technologies-runner-secret
-          volumeMounts:
-            - name: data
-              mountPath: /data
-      containers:
-        - name: runner
-          image: registry.neuralplatform.ai/ci-base:latest
-          workingDir: /data
-          command: ["act_runner", "daemon", "--config", "/data/config.yaml"]
-          envFrom:
-            - secretRef:
-                name: neuron-technologies-runner-secret
-          volumeMounts:
-            - name: data
-              mountPath: /data
-            - name: docker-sock
-              mountPath: /var/run/docker.sock
-          resources:
-            requests:
-              memory: 512Mi
-              cpu: 250m
-            limits:
-              memory: 4Gi
-              cpu: "4"
-      volumes:
-        - name: data
-          emptyDir: {}
-        - name: docker-sock
-          hostPath:
-            path: /run/k3s/containerd/containerd.sock
-            type: Socket
@@ -6,7 +6,7 @@ metadata:
  labels:
    app: gitea
 spec:
-  replicas: 1
+  replicas: 0  # Scaled down — Gitea has been migrated to GKE. DNS now points at GKE LB.
  strategy:
    type: Recreate
  selector:
@@ -0,0 +1,22 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: gke-apps
+  namespace: argocd
+spec:
+  project: default
+  source:
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
+    targetRevision: main
+    path: servers/gcp/k8s/argocd-apps
+  destination:
+    # This Application lives on Legion Argo CD — it creates child Applications
+    # there that target the GKE cluster.
+    server: https://kubernetes.default.svc
+    namespace: argocd
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+      - CreateNamespace=true
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/monitoring
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/mudcraft
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/daemon
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/dev
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/marketing
  destination:
@@ -15,7 +15,7 @@ spec:

  # Only allow deployments sourced from the infrastructure repo.
  sourceRepos:
-    - "http://gitea.git.svc.cluster.local:3000/will/infrastructure.git"
+    - "https://git.neuralplatform.ai/will/infrastructure.git"
    - "https://code.forgejo.org"
    - "registry.neuralplatform.ai"

@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/prod
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/stage
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/swarm
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/tim
  destination:
@@ -12,7 +12,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/neuron-technologies/web
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/ollama
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/packages
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/postgres
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/registry
  destination:
@@ -6,7 +6,7 @@ metadata:
 spec:
  project: default
  source:
-    repoURL: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+    repoURL: https://git.neuralplatform.ai/will/infrastructure.git
    targetRevision: main
    path: servers/legion/k8s/vault
  destination:
@@ -36,7 +36,7 @@ resource "helm_release" "argocd" {
        server.insecure: true  # TLS terminated at Traefik
      repositories:
        gitea-infrastructure:
-          url: http://gitea.git.svc.cluster.local:3000/will/infrastructure.git
+          url: https://git.neuralplatform.ai/will/infrastructure.git
          name: infrastructure
          type: git
  YAML
@@ -96,7 +96,7 @@ resource "kubernetes_secret" "argocd_gitea_repo" {

  data = {
    type     = "git"
-    url      = "http://gitea.git.svc.cluster.local:3000/will/infrastructure.git"
+    url      = "https://git.neuralplatform.ai/will/infrastructure.git"
    username = "will"
    password = var.gitea_api_token
  }
@@ -136,7 +136,7 @@ resource "kubernetes_manifest" "argocd_root_app" {
    spec = {
      project = "default"
      source = {
-        repoURL        = "http://gitea.git.svc.cluster.local:3000/will/infrastructure.git"
+        repoURL        = "https://git.neuralplatform.ai/will/infrastructure.git"
        targetRevision = "main"
        path           = "servers/legion/apps"
      }
@@ -23,3 +23,31 @@ resource "cloudflare_record" "np_web_stage" {
  proxied = true
  ttl     = 1
 }
+
+# vault.neuralplatform.ai — GCP Global HTTPS LB with managed cert.
+# DNS-only (not proxied) — GCP managed TLS cert terminates at the LB.
+# Backend is now GKE container-native NEGs (cut over from GCE instance groups).
+# IP: terraform output vault_lb_ip from servers/gcp workspace = 34.54.164.21
+resource "cloudflare_record" "np_vault" {
+  zone_id = local.zone_neuralplatform_ai
+  name    = "vault"
+  type    = "A"
+  content = "34.54.164.21"
+  proxied = false
+  ttl     = 60
+}
+
+# git.neuralplatform.ai — Gitea on GKE via GCP Network LB.
+# Cloudflare proxied (provides TLS termination). Gitea listens on port 80 internally.
+# IP: external IP of the GKE LoadBalancer service in namespace gitea.
+# Get it with:
+#   kubectl --context=gke_neuron-785695_us-central1_neuron-platform -n gitea get svc gitea
+# Update content below with the actual IP, then terraform apply.
+resource "cloudflare_record" "np_gitea" {
+  zone_id = local.zone_neuralplatform_ai
+  name    = "git"
+  type    = "A"
+  content = "34.31.145.131"  # GKE LoadBalancer external IP — gitea svc in namespace gitea
+  proxied = true
+  ttl     = 1
+}
@@ -0,0 +1,42 @@
+# Cloudflare Zero Trust Access — git.neuralplatform.ai (Gitea)
+#
+# The Gitea Access application itself is currently managed in the Cloudflare
+# dashboard, NOT in Terraform. This file only manages the *service token* the
+# Gitea Actions runners use to authenticate through CF Access while still
+# keeping the human Google-OAuth gate for browser users.
+#
+# Why not import the application here?
+#   - Importing the existing dashboard app risks drifting the human-auth
+#     policy (Google IdP, allowed emails) which is settled and working.
+#   - Service tokens can be added to a dashboard-managed app without
+#     importing the app itself; the token resource lives at the account
+#     level and is referenced from a policy.
+#   - We pay only the cost we need to. If we later want all Access apps
+#     in TF we can do a focused import pass.
+#
+# After `terraform apply` produces the token id/secret, Will must:
+#   1. Run `vault kv put secret/gitea-runner-cf-access ...` (see outputs).
+#   2. In the Cloudflare dashboard, edit the existing "Gitea" Access
+#      application's policies and add a new policy:
+#         Action:  Service Auth (decision = non_identity)
+#         Include: Service Token = "gitea-runner"
+#      This grants the service token bypass through CF Access on
+#      git.neuralplatform.ai without changing the human-auth flow.
+
+resource "cloudflare_zero_trust_access_service_token" "gitea_runner" {
+  account_id = var.cloudflare_account_id
+  name       = "gitea-runner"
+  # Default duration is "8760h" (1 year). Rotate via re-apply when needed.
+  duration = "forever"
+}
+
+output "gitea_runner_cf_access_client_id" {
+  description = "CF Access service token client ID for the Gitea Actions runner. Store in Vault at secret/gitea-runner-cf-access."
+  value       = cloudflare_zero_trust_access_service_token.gitea_runner.client_id
+}
+
+output "gitea_runner_cf_access_client_secret" {
+  description = "CF Access service token client secret. Store in Vault at secret/gitea-runner-cf-access. Only emitted at creation time."
+  value       = cloudflare_zero_trust_access_service_token.gitea_runner.client_secret
+  sensitive   = true
+}
@@ -6,7 +6,7 @@ ENV DEBIAN_FRONTEND=noninteractive

 COPY --from=runner-bin /bin/forgejo-runner /usr/local/bin/act_runner

-# Core system tools
+# Core system tools + C build deps needed by El compiler and other CI jobs
 RUN apt-get update && apt-get install -y --no-install-recommends \
    ca-certificates \
    curl \
@@ -27,8 +27,19 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
    pkg-config \
    software-properties-common \
    zstd \
+    sudo \
+    libcurl4-openssl-dev \
+    libssl-dev \
+    libsqlite3-dev \
+    libpq-dev \
+    libffi-dev \
+    zlib1g-dev \
+    dpkg-dev \
    && rm -rf /var/lib/apt/lists/*

+# Allow any user to run sudo without a password (CI containers need apt-get etc.)
+RUN echo "ALL ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers
+
 # Node.js 20 LTS via binary tarball (nodesource apt repo is unreliable on Ubuntu 24.04)
 RUN NODE_VERSION=20.19.1 \
    && curl -fsSL "https://nodejs.org/dist/v${NODE_VERSION}/node-v${NODE_VERSION}-linux-x64.tar.xz" \
@@ -85,3 +96,20 @@ RUN curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg \
    && apt-get update \
    && apt-get install -y --no-install-recommends gh \
    && rm -rf /var/lib/apt/lists/*
+
+# Cloudflare Access bootstrap for git clones to git.neuralplatform.ai.
+# This script is sourced by bash in build containers via BASH_ENV (set by
+# act_runner's container.env in deployment.yaml) so it runs before every
+# step. It configures git insteadOf + CF Access extraHeaders from
+# CF_ACCESS_CLIENT_ID / CF_ACCESS_CLIENT_SECRET env vars.
+#
+# We deliberately don't set ENTRYPOINT / CMD here — act_runner spawns
+# build containers with its own entrypoint to keep them alive between
+# steps, and overriding it breaks job execution.
+
+# SSH-based git clone init script.
+# Sourced before every CI step via BASH_ENV (set in deployment.yaml).
+# Writes GITEA_SSH_PRIVATE_KEY to ~/.ssh/gitea_key and rewrites HTTPS
+# Gitea URLs to SSH so actions/checkout and git clone both use SSH auth.
+COPY git-ssh-init.sh /usr/local/bin/git-ssh-init.sh
+RUN chmod +x /usr/local/bin/git-ssh-init.sh
@@ -8,7 +8,7 @@ metadata:
  labels:
    app: gitea-runner
  annotations:
-    config-version: "2026-05-04-docker-sock-fix"
+    config-version: "2026-05-05-extra-hosts"
 spec:
  replicas: 1
  selector:
@@ -19,7 +19,7 @@ spec:
      labels:
        app: gitea-runner
      annotations:
-        config-version: "2026-05-04-docker-sock-fix"
+        config-version: "2026-05-05-extra-hosts"
    spec:
      securityContext:
        runAsNonRoot: false
@@ -35,18 +35,25 @@ spec:
                --name legion \
                --labels "self-hosted:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-latest:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-24.04:docker://registry.neuralplatform.ai/ci-base:latest,linux,x64" \
                --no-interactive
-              cat > /data/config.yaml << 'EOF'
+              cat > /data/config.yaml << EOF
              runner:
                capacity: 2
                timeout: 3h
              container:
                network: host
                docker_host: "unix:///var/run/docker.sock"
-                force_pull: false
+                force_pull: true
                valid_volumes: []
                default_image: "registry.neuralplatform.ai/ci-base:latest"
+                env:
+                  GITEA_SSH_PRIVATE_KEY: "${GITEA_SSH_PRIVATE_KEY}"
+                  BASH_ENV: "/usr/local/bin/git-ssh-init.sh"
+                # Build containers run with network: host, so k8s DNS names
+                # don't resolve. Inject the gitea-proxy ClusterIP so that
+                # actions/checkout can reach gitea-proxy.ci.svc.cluster.local
+                # (the URL Gitea passes as GITHUB_SERVER_URL to the runner).
                extra_hosts:
-                  - "gitea.git.svc.cluster.local:10.43.1.53"
+                  - "gitea-proxy.ci.svc.cluster.local:10.43.88.7"
              EOF
          envFrom:
            - secretRef:
@@ -92,7 +99,7 @@ metadata:
  labels:
    app: neuron-technologies-runner
  annotations:
-    config-version: "2026-05-04-docker-sock-fix"
+    config-version: "2026-05-05-extra-hosts"
 spec:
  replicas: 2
  selector:
@@ -103,7 +110,7 @@ spec:
      labels:
        app: neuron-technologies-runner
      annotations:
-        config-version: "2026-05-04-docker-sock-fix"
+        config-version: "2026-05-05-extra-hosts"
    spec:
      securityContext:
        runAsNonRoot: false
@@ -119,18 +126,25 @@ spec:
                --name "legion-nt-$(hostname)" \
                --labels "self-hosted:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-latest:docker://registry.neuralplatform.ai/ci-base:latest,ubuntu-24.04:docker://registry.neuralplatform.ai/ci-base:latest,linux,x64" \
                --no-interactive
-              cat > /data/config.yaml << 'EOF'
+              cat > /data/config.yaml << EOF
              runner:
                capacity: 2
                timeout: 3h
              container:
                network: host
                docker_host: "unix:///var/run/docker.sock"
-                force_pull: false
+                force_pull: true
                valid_volumes: []
                default_image: "registry.neuralplatform.ai/ci-base:latest"
+                # Build containers run with network: host, so k8s DNS names
+                # don't resolve. Inject the gitea-proxy ClusterIP so that
+                # actions/checkout can reach gitea-proxy.ci.svc.cluster.local
+                # (the URL Gitea passes as GITHUB_SERVER_URL to the runner).
+                env:
+                  GITEA_SSH_PRIVATE_KEY: "${GITEA_SSH_PRIVATE_KEY}"
+                  BASH_ENV: "/usr/local/bin/git-ssh-init.sh"
                extra_hosts:
-                  - "gitea.git.svc.cluster.local:10.43.1.53"
+                  - "gitea-proxy.ci.svc.cluster.local:10.43.88.7"
              EOF
          envFrom:
            - secretRef:
@@ -1,12 +1,19 @@
 ---
 # gitea-runner-secret — neural-platform org runner token
+#
+# GITEA_INSTANCE_URL: act_runner daemon polls git.neuralplatform.ai via HTTPS.
+# GITEA_SSH_PRIVATE_KEY: ED25519 deploy key for SSH-based git clones inside
+#   build containers. git-ssh-init.sh (sourced via BASH_ENV) writes this key
+#   to ~/.ssh/gitea_key and rewrites HTTPS Gitea URLs to SSH so that
+#   actions/checkout and plain `git clone` both use SSH auth without any
+#   HTTPS credential or CF Access token.
 apiVersion: external-secrets.io/v1beta1
 kind: ExternalSecret
 metadata:
  name: gitea-runner-secret
  namespace: ci
  annotations:
-    force-sync: "2026-04-23"
+    force-sync: "2026-05-05-gitea-proxy"
 spec:
  refreshInterval: 1h
  secretStoreRef:
@@ -17,13 +24,18 @@ spec:
    creationPolicy: Owner
    template:
      data:
-        GITEA_INSTANCE_URL: "http://gitea.git.svc.cluster.local:3000"
+        GITEA_INSTANCE_URL: "http://gitea-proxy.ci.svc.cluster.local:3000"
        GITEA_RUNNER_REGISTRATION_TOKEN: "{{ .runner_token }}"
+        GITEA_SSH_PRIVATE_KEY: "{{ .ci_ssh_private_key }}"
  data:
    - secretKey: runner_token
      remoteRef:
        key: secret/data/gitea
        property: runner_token
+    - secretKey: ci_ssh_private_key
+      remoteRef:
+        key: secret/data/gitea
+        property: ci_ssh_private_key
 ---
 # neuron-technologies-runner-secret — neuron-technologies org runner token
 apiVersion: external-secrets.io/v1beta1
@@ -31,6 +43,8 @@ kind: ExternalSecret
 metadata:
  name: neuron-technologies-runner-secret
  namespace: ci
+  annotations:
+    force-sync: "2026-05-05-gitea-proxy"
 spec:
  refreshInterval: 1h
  secretStoreRef:
@@ -41,10 +55,15 @@ spec:
    creationPolicy: Owner
    template:
      data:
-        GITEA_INSTANCE_URL: "http://gitea.git.svc.cluster.local:3000"
+        GITEA_INSTANCE_URL: "http://gitea-proxy.ci.svc.cluster.local:3000"
        GITEA_RUNNER_REGISTRATION_TOKEN: "{{ .runner_token }}"
+        GITEA_SSH_PRIVATE_KEY: "{{ .ci_ssh_private_key }}"
  data:
    - secretKey: runner_token
      remoteRef:
        key: secret/data/gitea
        property: neuron_technologies_runner_token
+    - secretKey: ci_ssh_private_key
+      remoteRef:
+        key: secret/data/gitea
+        property: ci_ssh_private_key
@@ -0,0 +1,50 @@
+#!/bin/sh
+# git-cf-access-init.sh
+#
+# Configures git so any clone/fetch from Gitea ends up going to
+# git.neuralplatform.ai with the runner's Cloudflare Access service-token
+# headers attached.
+#
+# How this gets invoked:
+#   The forgejo-runner job execution path runs each step via a
+#   non-interactive bash invocation inside the build container. Setting
+#   BASH_ENV=/usr/local/bin/git-cf-access-init.sh in act_runner's
+#   container.env causes bash to source this script before any step's
+#   commands run. (See servers/legion/k8s/gitea-runner/deployment.yaml.)
+#
+# What it does:
+#   1. Rewrites http://gitea.git.svc.cluster.local:3000/ → https://git.neuralplatform.ai/
+#      via insteadOf. The runner registered against the in-cluster URL (no
+#      CF Access on the daemon's polling loop), so act_runner advertises
+#      that URL to the build container as github.server_url. Build
+#      containers run with network: host and can't resolve
+#      *.svc.cluster.local, so we need to redirect to the public URL.
+#   2. Adds the CF Access service-token headers to outbound requests to
+#      git.neuralplatform.ai so the clone authenticates through CF Access.
+#
+# Idempotent — re-runs replace any prior config keys without accumulating
+# duplicate header entries.
+#
+# Known limitation: actions/checkout sets an Authorization extraheader
+# keyed to the server URL it was given (the in-cluster URL). After
+# insteadOf substitution the request goes to the public URL where git
+# matches http.<public>.extraheader, and the in-cluster-keyed
+# Authorization header is dropped. For public repos this is fine. For
+# private repos the per-job token will not be sent — see the PR
+# description for the follow-up plan if dharma-el's CI needs that token.
+
+if [ -n "${CF_ACCESS_CLIENT_ID:-}" ] && [ -n "${CF_ACCESS_CLIENT_SECRET:-}" ]; then
+    git config --global --replace-all \
+        url."https://git.neuralplatform.ai/".insteadOf \
+        "http://gitea.git.svc.cluster.local:3000/" 2>/dev/null || true
+
+    # Reset extraHeader on the public URL, then add both CF Access headers.
+    git config --global --unset-all \
+        http."https://git.neuralplatform.ai/".extraHeader 2>/dev/null || true
+    git config --global --add \
+        http."https://git.neuralplatform.ai/".extraHeader \
+        "CF-Access-Client-Id: ${CF_ACCESS_CLIENT_ID}" 2>/dev/null || true
+    git config --global --add \
+        http."https://git.neuralplatform.ai/".extraHeader \
+        "CF-Access-Client-Secret: ${CF_ACCESS_CLIENT_SECRET}" 2>/dev/null || true
+fi
@@ -0,0 +1,48 @@
+#!/bin/sh
+# git-ssh-init.sh
+#
+# Sets up SSH authentication for git operations inside CI build containers.
+#
+# How this gets invoked:
+#   act_runner runs each step via a non-interactive bash invocation.
+#   Setting BASH_ENV=/usr/local/bin/git-ssh-init.sh in act_runner's
+#   container.env causes bash to source this before any step's commands.
+#   (See servers/legion/k8s/gitea-runner/deployment.yaml.)
+#
+# What it does:
+#   1. Writes GITEA_SSH_PRIVATE_KEY (from the runner secret) to ~/.ssh/gitea_key
+#   2. Creates an ~/.ssh/config entry so git uses that key for git.neuralplatform.ai
+#   3. Sets a git insteadOf rule to rewrite HTTPS Gitea URLs to SSH,
+#      so `actions/checkout` and any direct `git clone https://...` also use SSH
+#
+# Idempotent — safe to re-run on every step.
+
+if [ -n "${GITEA_SSH_PRIVATE_KEY:-}" ]; then
+    mkdir -p ~/.ssh
+    chmod 700 ~/.ssh
+
+    # Write the private key
+    printf '%s\n' "${GITEA_SSH_PRIVATE_KEY}" > ~/.ssh/gitea_key
+    chmod 600 ~/.ssh/gitea_key
+
+    # SSH config — use this key for git.neuralplatform.ai, skip host key checking
+    # (build containers are ephemeral; we don't persist a known_hosts file)
+    cat > ~/.ssh/gitea_config << 'EOF'
+Host git.neuralplatform.ai
+  HostName git.neuralplatform.ai
+  User git
+  IdentityFile ~/.ssh/gitea_key
+  StrictHostKeyChecking no
+  UserKnownHostsFile /dev/null
+EOF
+    chmod 600 ~/.ssh/gitea_config
+
+    # Point SSH at our per-job config (merge with any existing config if present)
+    export GIT_SSH_COMMAND="ssh -F ~/.ssh/gitea_config"
+
+    # NOTE: Do NOT add url.insteadOf SSH rewrite here.
+    # Gitea's built-in SSH server is disabled (START_SSH_SERVER=false) so
+    # SSH git clones would fail. HTTPS git operations work directly from
+    # the build container host network — Cloudflare bypasses the CF Access
+    # gate for git smart-HTTP and release asset paths.
+fi
@@ -0,0 +1,100 @@
+---
+# gitea-proxy — plain nginx proxy to GKE Gitea LB IP
+#
+# Routes directly to the GKE Network LB IP, bypassing Cloudflare entirely.
+# No CF Access service-token auth needed. The runners can't add custom HTTP
+# headers, so going through Cloudflare Access is not viable for the gRPC
+# Actions ping endpoint (/api/actions/*).
+#
+# Runners use:
+#   GITEA_INSTANCE_URL: http://gitea-proxy.ci.svc.cluster.local:3000
+#
+# The GKE Gitea LB IP (34.31.145.131) is the external IP of the
+# gitea Service in the gitea namespace on the GKE cluster.
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: gitea-proxy-config
+  namespace: ci
+data:
+  default.conf: |
+    server {
+      listen 3000;
+
+      location / {
+        # Direct to GKE Gitea LB IP — bypasses Cloudflare/CF Access entirely.
+        proxy_pass http://34.31.145.131;
+
+        # Tell Gitea which hostname it's being served as so self-referencing
+        # URLs (clone URLs, webhook URLs) are generated correctly.
+        proxy_set_header Host git.neuralplatform.ai;
+
+        # Forward all original request headers (auth tokens, etc).
+        proxy_pass_request_headers on;
+
+        # Rewrite Location headers in redirects back to the proxy URL.
+        proxy_redirect http://34.31.145.131/
+                       http://gitea-proxy.ci.svc.cluster.local:3000/;
+
+        # Standard proxy headers.
+        proxy_set_header X-Real-IP         $remote_addr;
+        proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
+        proxy_set_header X-Forwarded-Proto http;
+
+        # Allow large request bodies (git push payloads).
+        client_max_body_size 512m;
+      }
+    }
+---
+# Deployment
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: gitea-proxy
+  namespace: ci
+  labels:
+    app: gitea-proxy
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: gitea-proxy
+  template:
+    metadata:
+      labels:
+        app: gitea-proxy
+      annotations:
+        config-version: "2026-05-05-direct-lb"
+    spec:
+      containers:
+        - name: nginx
+          image: nginx:alpine
+          ports:
+            - containerPort: 3000
+          volumeMounts:
+            - name: config
+              mountPath: /etc/nginx/conf.d
+          resources:
+            requests:
+              cpu: 50m
+              memory: 32Mi
+            limits:
+              cpu: 200m
+              memory: 64Mi
+      volumes:
+        - name: config
+          configMap:
+            name: gitea-proxy-config
+---
+# Service — ClusterIP reachable by all runner pods as gitea-proxy.ci:3000
+apiVersion: v1
+kind: Service
+metadata:
+  name: gitea-proxy
+  namespace: ci
+spec:
+  selector:
+    app: gitea-proxy
+  ports:
+    - port: 3000
+      targetPort: 3000
@@ -82,7 +82,12 @@ spec:
    apiVersion: apps/v1
    kind: Deployment
    name: neuron-marketing
-  minReplicas: 0
+  # minReplicas=1 to match the file's own convention (see header comment).
+  # Kubernetes only allows minReplicas=0 when at least one Object or External
+  # metric is configured (queue depth, custom signal, etc.); with only a
+  # Resource (CPU) metric, scale-to-zero is rejected and the whole HPA is
+  # invalid — which was blocking neuron-prod's Argo CD sync.
+  minReplicas: 1
  maxReplicas: 8
  metrics:
    - type: Resource
@@ -117,6 +117,32 @@ spec:
            matchLabels:
              kubernetes.io/metadata.name: neuron-prod
 ---
+# ── dharma: accept from Traefik (kube-system) and neuron-prod namespace ──────
+# The dharma pod was healthy and the IngressRoute was correct, but cross-
+# namespace ingress from kube-system (Traefik) was denied by default-deny-all,
+# so every external request landed at Traefik and bounced back as 502. This
+# allow rule mirrors `allow-mcp-ingress` and brings dharma into line with the
+# other neuron-prod services.
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: allow-dharma-ingress
+  namespace: neuron-prod
+spec:
+  podSelector:
+    matchLabels:
+      app: dharma
+  policyTypes:
+    - Ingress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: kube-system
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: neuron-prod
+---
 # ── Egress: all prod pods may reach platform (postgres/redis), vault,
 #            monitoring (alloy OTLP), kube-dns, and the internet (external APIs) ─
 apiVersion: networking.k8s.io/v1
@@ -8,7 +8,7 @@
 # routes still go to the Cloud Run prod URL until we cut over.
 #
 # Vault paths used:
-#   secret/neuron-technologies/anthropic        api_key
+#   secret/neuron-technologies/anthropic        api_key, web_demo_key
 #   secret/neuron-technologies/supabase         anon_key, service_role_key, project_url
 #   secret/neuron-technologies/marketing-test   stripe_*, etc.
 #   secret/neuron-technologies/notifications    resend_api_key
@@ -38,7 +38,7 @@ spec:
    - secretKey: NEURON_LLM_0_KEY
      remoteRef:
        key: secret/data/neuron-technologies/anthropic
-        property: api_key
+        property: web_demo_key
    - secretKey: SUPABASE_ANON_KEY
      remoteRef:
        key: secret/data/neuron-technologies/supabase
@@ -92,37 +92,11 @@ resource "cloudflare_zero_trust_tunnel_cloudflared_config" "legion" {
      }
    }

-    ingress_rule {
-      hostname = "vault.neuralplatform.ai"
-      service  = "https://traefik.kube-system.svc:443"
-      origin_request {
-        no_tls_verify = true
-      }
-    }
+    # vault.neuralplatform.ai — moved to GCP Global HTTPS LB (34.54.164.21)
+    # DNS is now a direct A record (not proxied) in dns-neuralplatform.tf

-    ingress_rule {
-      hostname = "watch.nook.family"
-      service  = "https://traefik.kube-system.svc:443"
-      origin_request {
-        no_tls_verify = true
-      }
-    }
-
-    ingress_rule {
-      hostname = "jellyfin.nook.family"
-      service  = "https://traefik.kube-system.svc:443"
-      origin_request {
-        no_tls_verify = true
-      }
-    }
-
-    ingress_rule {
-      hostname = "bazarr.nook.family"
-      service  = "https://traefik.kube-system.svc:443"
-      origin_request {
-        no_tls_verify = true
-      }
-    }
+    # watch.nook.family, jellyfin.nook.family, bazarr.nook.family — removed
+    # This infrastructure is focused on Neuron; nook.family media stack retired


    # fornax.neuralplatform.ai — Fornax torrent coordinator (qBittorrent API proxy)
@@ -134,16 +108,12 @@ resource "cloudflare_zero_trust_tunnel_cloudflared_config" "legion" {
      }
    }

-    # git.neuralplatform.ai — Gitea web UI (HTTP via Traefik)
-    ingress_rule {
-      hostname = "git.neuralplatform.ai"
-      service  = "https://traefik.kube-system.svc:443"
-      origin_request {
-        no_tls_verify = true
-      }
-    }
+    # git.neuralplatform.ai — REMOVED: Gitea is now on GKE.
+    # DNS is a direct Cloudflare A record (proxied) pointing at the GKE LoadBalancer IP.
+    # See: servers/legion/dns-neuralplatform.tf (cloudflare_record.np_gitea)

    # ssh.git.neuralplatform.ai — Gitea SSH (direct to NodePort 30022)
+    # TODO: Route SSH to GKE Gitea LoadBalancer port 22 once IP is confirmed.
    ingress_rule {
      hostname = "ssh.git.neuralplatform.ai"
      service  = "ssh://localhost:30022"
Author	SHA1	Message	Date
Will Anderson	086a7b0bf0	add DocuSeal to GKE — manifests, Terraform, Argo CD app; update ci-base Dockerfile	2026-05-05 10:38:45 -05:00
Will Anderson	73b9095d83	migrate Terraform backend to GCS, remove dead Vault GCE nodes Switch gcp/ backend from Cloudflare R2 (S3-compatible) to GCS — uses Application Default Credentials automatically, no extra env vars needed. State has been migrated to gs://neuron-785695-terraform-state/gcp/. Remove GCE Vault node resources (vault-node-1/2/3) from vault-nodes.tf and Terraform state. GKE Vault (Workload Identity + vault-unseal SA) has been serving traffic since the NEG cutover; GCE nodes are confirmed dead. Removed from state and config: - google_compute_instance.vault_node (x3) - google_compute_disk.vault_data (x3) - google_compute_instance_group.vault (x3) - google_compute_firewall.vault_iap_ssh / vault_raft / vault_api (GCE-targeted) - google_storage_bucket_object.vault_startup - google_storage_bucket_iam_member.vault_node_bucket_read - google_project_iam_member.vault_node_{log_writer,metric_writer,compute_viewer} Kept: KMS, vault-unseal SA, WI binding, NEG data sources, GKE firewall, global LB (address, cert, backend service, url maps, proxies, fwd rules).	2026-05-05 10:38:22 -05:00
Will Anderson	d54ec38a1c	fix GCE runner: add build tools, libcurl, NOPASSWD sudo, Node.js 20	2026-05-05 10:27:45 -05:00
Will Anderson	422cca6962	scale gitea-runner to 0 while Autopilot DinD blocker is pending Set replicas: 0 to prevent crash-loop restarts. The runner requires a Docker daemon (even in host mode) which Autopilot won't provide. Scale to 1 when a Standard node pool or external Docker TCP endpoint is available.	2026-05-05 10:22:21 -05:00
Will Anderson	9a368986b2	fix gitea-runner deployment for GKE Autopilot constraints Switch to host-mode labels and document the DinD limitation: Autopilot's Warden blocks privileged containers at the node level. Use internal Gitea service URL (gitea.gitea.svc.cluster.local) to bypass Cloudflare Access. Document three resolution paths (Standard node pool, Cloud Build, external Docker TCP). GCE VM runner remains active in the interim.	2026-05-05 10:22:02 -05:00
Will Anderson	c31edc8b83	deploy gitea CI runner to GKE as k8s pod Replaces the GCE VM runner with a k8s Deployment in the ci namespace on neuron-platform GKE. Uses Docker-in-Docker for build isolation since Autopilot doesn't expose the node socket. Runner token pulled from Secret Manager via ESO + Workload Identity. - servers/gcp/k8s/gitea-runner/: namespace, serviceaccount, external-secrets, deployment manifests (ci namespace, dind sidecar, idempotent registration) - servers/gcp/k8s/argocd-apps/gitea-runner-gke.yaml: Argo CD Application - servers/gcp/gitea-runner.tf: gitea-runner-gke GCP SA with secretAccessor on gitea-runner-token, Workload Identity binding for ci/gitea-runner, artifactregistry.reader for pulling ci-base image	2026-05-05 10:16:49 -05:00
Will Anderson	4dc687f2ae	Fix runner extra_hosts: inject gitea-proxy ClusterIP for build containers Build containers run with network: host, so k8s DNS doesn't resolve. Add extra_hosts entry pointing gitea-proxy.ci.svc.cluster.local to its ClusterIP (10.43.88.7) in both the will-org and neuron-technologies runner configs. Without this, actions/checkout fails because Gitea passes the proxy hostname as GITHUB_SERVER_URL and the build container can't reach it. Also update stale CF Access comments that no longer reflect the architecture.	2026-05-05 05:41:59 -05:00
Will Anderson	c94721b70b	Remove SSH insteadOf rewrite from git-ssh-init.sh Gitea's SSH server is disabled (START_SSH_SERVER=false) so the SSH URL rewrite breaks git HTTPS clones in build containers. HTTPS git ops work directly from Legion host network — Cloudflare bypasses CF Access for smart-HTTP and release asset paths.	2026-05-05 05:27:44 -05:00
Will Anderson	a301302138	Route gitea-proxy directly to GKE LB IP, drop CF Access CF Access was blocking /api/actions/* even with the correct service-token headers — the gRPC Actions ping endpoint returns 403 through Cloudflare. Direct connection to the GKE Network LB IP (34.31.145.131) bypasses CF entirely. Verified /api/actions/ping returns 200 from the LB IP directly.	2026-05-05 05:12:48 -05:00
Will Anderson	3462ea37a4	Add ignoreDifferences for Deployment status in gitea-gke Argo CD's bundled schema is missing terminatingReplicas from the Deployment status spec, causing a ComparisonError on every sync. Ignoring /status entirely is safe — Argo CD never manages status fields; they're owned by kube-controller-manager.	2026-05-05 05:07:15 -05:00
Will Anderson	82e05429f1	Fix Argo CD child app repoURL: k3s Gitea → GKE Gitea gitea-gke and vault-gke were pointing at the dead k3s-internal Gitea (http://gitea.git.svc.cluster.local:3000). gke-apps already uses https://git.neuralplatform.ai and syncs successfully — update child apps to match. Once gke-apps picks this up, gitea-gke will sync the pending GITEA__actions__ENABLED=true commit.	2026-05-05 05:04:01 -05:00
Will Anderson	686041946b	Enable Gitea Actions (GITEA__actions__ENABLED=true)	2026-05-05 04:56:29 -05:00
Will Anderson	158689dd0e	Roll runner pods to pick up updated GITEA_INSTANCE_URL	2026-05-05 04:45:38 -05:00
Will Anderson	b5c7125c13	Roll gitea-proxy pod for SNI fix	2026-05-05 04:42:08 -05:00
Will Anderson	464b33a176	Fix nginx SNI for Cloudflare upstream TLS Without proxy_ssl_server_name on, nginx connects to Cloudflare's IP without SNI, triggering handshake_failure (alert 40). Cloudflare requires SNI to route TLS to the correct origin.	2026-05-05 04:41:46 -05:00
Will Anderson	cebaf44db3	Scale GKE Vault to 1 replica while raft standby deadlock is unresolved Pods 1 and 2 can't unseal because port 8201 (cluster port) only opens after barrier initialization, but the barrier key transfer requires 8201 to be reachable — deadlock. Running single-node on pod 0 so vault.neuralplatform.ai reliably routes to the active, unsealed instance.	2026-05-05 04:37:20 -05:00
Will Anderson	939e66bfbb	Route runner daemon through nginx CF Access proxy act_runner cannot inject custom HTTP headers, so CF Access blocks its unauthenticated calls to git.neuralplatform.ai. Add a gitea-proxy deployment (nginx:alpine) in the ci namespace that injects the CF Access service-token headers and proxies to https://git.neuralplatform.ai. Both runner secrets now point GITEA_INSTANCE_URL at the in-cluster proxy (http://gitea-proxy.ci.svc.cluster.local:3000). Build containers still clone via SSH through git-ssh-init.sh — unaffected.	2026-05-05 04:17:52 -05:00
Will Anderson	ae3257525e	wire soul to web_demo_key for Anthropic API isolation	2026-05-05 04:09:41 -05:00
Will Anderson	c2900400a4	ci: switch runner git clones from HTTPS+CF Access to SSH deploy key	2026-05-05 04:07:35 -05:00
Will Anderson	2e5655c583	vault: rolling restart to fix raft join; remove sealedok from readiness probe	2026-05-05 03:59:33 -05:00
Will Anderson	3f08c6dc01	fix(runner): bump config-version to trigger rollout with new Gitea URL	2026-05-05 03:48:15 -05:00
will.anderson	fedf549a92	Merge pull request 'fix(runner): point GITEA_INSTANCE_URL at GKE Gitea public URL' (#1 ) from fix/runner-gitea-url into main	2026-05-05 08:42:19 +00:00
Will Anderson	01507ad8ae	fix(runner): point GITEA_INSTANCE_URL at GKE Gitea public URL Gitea moved to GKE. The old in-cluster URL http://gitea.git.svc.cluster.local:3000 now has 0 pods behind it. The runner daemon gets connection refused on every poll. Switch both ExternalSecrets to https://git.neuralplatform.ai. The public URL does not require CF Access tokens — the runner can connect without them. CF_ACCESS_CLIENT_ID/SECRET remain for git operations inside CI job containers (via BASH_ENV init script).	2026-05-05 03:39:17 -05:00
Will Anderson	479190ef47	migrate all Argo CD repo sources from Legion Gitea to GKE Gitea	2026-05-05 02:25:14 -05:00
Will Anderson	804060c958	fix vault raft join addresses for GKE StatefulSet naming	2026-05-04 23:56:10 -05:00
Will Anderson	872a834989	Cut over Vault and Gitea traffic from GCE/Legion to GKE Vault: - Annotate vault-helm-gke Service with cloud.google.com/neg exposed_ports to create container-native NEGs (k8s1-bfbeff02-vault-...) in 3 zones - Add vault-api-from-lb-gke firewall rule allowing GCP health check ranges (130.211.0.0/22, 35.191.0.0/16) to reach GKE pod IPs on port 8200 - Replace GCE instance group backends in google_compute_backend_service.vault with GKE NEG backends (RATE balancing mode, 100 req/endpoint) - GCP Global HTTPS LB frontend unchanged — DNS stays at 34.54.164.21 - vault.neuralplatform.ai now terminates at GKE pods (all 3 NEGs HEALTHY) Gitea: - Change GKE Gitea Service from ClusterIP to LoadBalancer (external IP: 34.31.145.131) - Add Cloudflare DNS A record for git.neuralplatform.ai → 34.31.145.131 (proxied) - Remove git.neuralplatform.ai route from Legion Cloudflare tunnel config - Add Cloudflare config rule: flexible SSL for git.neuralplatform.ai (origin serves HTTP/3000, CF proxies HTTPS termination) - Scale Legion Gitea deployment to 0 replicas (PVC preserved) - git.neuralplatform.ai now serves from GKE Gitea pod	2026-05-04 23:46:20 -05:00
Will Anderson	010d81d6d9	fix(runner): correct Gitea ClusterIP and enable force_pull extra_hosts had stale IP 10.43.1.53 (Gitea is now at 10.43.15.98). force_pull: true ensures runners always use the latest ci-base image. Applies to both gitea-runner and neuron-technologies-runner.	2026-05-04 23:32:09 -05:00
Will Anderson	1c99931de6	fix(vault-gke): switch vault-1/2 to standard-rwo storage class SSD quota (SSD_TOTAL_GB) is exhausted at 500/500 GB. pd-ssd PVCs for vault-1 and vault-2 cannot provision. Switching to standard-rwo (pd-balanced) unblocks the Vault HA cluster immediately. Quota increase request submitted (Cloud Quotas preference API, preference ID 0aa576b1-5050-4fa7-911d-6f408230df7f, target 1000GB). Once approved, can optionally migrate back to premium-rwo for all-SSD Vault storage.	2026-05-04 22:41:38 -05:00
Will Anderson	3d581368c3	fix(vault-gke): relax topology spread to ScheduleAnyway while SSD quota is exhausted vault-1 and vault-2 are stuck Pending because GCE quota (SSD_TOTAL_GB 500/500) prevents new nodes from provisioning. ScheduleAnyway lets them land in any zone that has capacity while the quota increase request is pending. Also adds ESO Workload Identity IAM bindings to Terraform state (previously applied out-of-band via gcloud; now tracked in cloud-sql.tf).	2026-05-04 22:37:17 -05:00
Will Anderson	8592a1ed74	fix(gitea-gke): configure ESO SecretStore to use workload identity with gitea SA	2026-05-04 22:05:01 -05:00
Will Anderson	9252b069a2	fix(gitea-gke): update ESO API version from v1beta1 to v1	2026-05-04 22:00:21 -05:00
Will Anderson	c4f7dcc185	fix(vault-gke): raise cpu request to 500m for GKE Autopilot anti-affinity minimum	2026-05-04 21:59:51 -05:00
Will Anderson	b572f5720b	infra: wire GKE cluster endpoint and fix gitea SA account_id - Replace GKE_CLUSTER_ENDPOINT placeholder with real endpoint (34.63.89.52) in all three Argo CD Application manifests - Fix gitea GCP SA account_id from 'gitea' (5 chars, too short) to 'gitea-gke' to satisfy GCP's 6-30 char constraint - Update serviceaccount.yaml annotation to match new SA email (gitea-gke@...)	2026-05-04 21:56:06 -05:00
Will Anderson	4ef5e99f31	fix(ci-base): correct Gitea namespace in git insteadOf redirect Gitea lives in the 'git' namespace on Legion k3s, not 'gitea'. The system gitconfig redirect must use gitea.git.svc.cluster.local:3000. Update when Gitea migrates to GKE (namespace will be 'gitea' there).	2026-05-04 20:42:27 -05:00
Will Anderson	9330107fcc	migrate Vault and Gitea to GKE Autopilot cluster Implements Option A: move Vault (3x GCE e2-small) and Gitea (Legion k8s) onto a new GKE Autopilot cluster (neuron-platform, us-central1) managed through Legion Argo CD. Terraform (servers/gcp/): - gke.tf: GKE Autopilot cluster, Workload Identity bindings for Vault (KMS) and Gitea (Cloud SQL) - cloud-sql.tf: gitea database + user on neuron-prod-pg15, gitea GCP SA, gitea-database-url and gitea-db-password Secret Manager secrets - vault-nodes.tf: PENDING DECOMMISSION comment with migration checklist k8s manifests (servers/gcp/k8s/): - vault/: namespace.yaml - gitea/: namespace, serviceaccount (Workload Identity annotation), pvc (50Gi standard-rwo), deployment (Gitea + Cloud SQL Auth Proxy sidecar), service, configmap (custom CSS carried from Legion), external-secrets (GCP SM provider) - argocd-apps/: vault-gke.yaml, vault-helm-gke.yaml (Helm chart, HA Raft 3 replicas, GCP KMS auto-unseal, topologySpread across zones, 10Gi premium-rwo), gitea-gke.yaml — all target GKE_CLUSTER_ENDPOINT placeholder Legion (servers/legion/): - apps/gke-apps.yaml: App-of-Apps entry point on Legion Argo CD that syncs the GKE Application manifests - k8s/gitea-runner/Dockerfile: add system-level git insteadOf so GKE CI runners resolve Gitea in-cluster without Cloudflare Access headers	2026-05-04 20:40:48 -05:00
will.anderson	cc368d2ad1	fix(ci): remove duplicate runner Deployment from apps/	2026-05-04 22:55:55 +00:00
Will Anderson	b4b05bfe40	fix(ci): remove duplicate runner Deployment from apps/ The stale apps/gitea-runner.yaml contained two Deployment manifests that conflicted with the canonical Deployments owned by the gitea-runner-config Argo Application (pointing at k8s/gitea-runner/). Dual ownership caused Argo CD to fight itself — restarting runner pods mid-job and producing the "context canceled" failures on neuron-technologies/dharma-el CI. Canonical Deployments (config-version 2026-05-04-cf-access-public-url, docker.sock, CF Access env, replicas=2 for nt-runner) live in k8s/gitea-runner/deployment.yaml and are managed by gitea-runner-config.	2026-05-04 17:55:28 -05:00
Will Anderson	48106b27ec	vault: cut over to GCE Raft HA cluster, retire nook.family media stack - dns-neuralplatform.tf: add vault.neuralplatform.ai A record → 34.54.164.21 (GCP LB) DNS-only (not proxied) so GCP managed TLS cert can provision correctly - main.tf: remove vault.neuralplatform.ai from Cloudflare tunnel ingress (now served directly via GCP Global HTTPS LB) - main.tf: remove watch.nook.family, jellyfin.nook.family, bazarr.nook.family from tunnel ingress (nook.family media stack retired; infra is Neuron-focused) GCE Vault cluster already initialized and running (3-node Raft, active since 2026-05-04T16:05). Secrets migrated 48/48 from k3s vault. ESO ClusterSecretStore validated against new vault. k3s vault-0 is now superseded.	2026-05-04 16:40:03 -05:00
will.anderson	0006380c27	route runner build container clones via public URL with CF Access (#7 )	2026-05-04 21:37:53 +00:00
will.anderson	7ab97eb88d	add CF Access service token for Gitea Actions runner (#5 )	2026-05-04 21:25:20 +00:00
will.anderson	23fc64e7b7	fix(neuron-prod): set neuron-marketing-hpa minReplicas to 1 (#8 )	2026-05-04 21:21:16 +00:00
will.anderson	a64860064b	fix(neuron-prod): add allow-dharma-ingress NetworkPolicy (#6 )	2026-05-04 21:16:01 +00:00