feat: per-user tenancy — BYO provider key, thread ownership, owner-principal sessions

.github/workflows/publish-images-mo4islona.yml

@@ -0,0 +1,120 @@
+name: Publish Images (mo4islona / Docker Hub)
+
+# Personal/dev image publish to Docker Hub under docker.io/mo4islona/centaur-*.
+# Separate from publish-images.yml (which targets the upstream GHCR namespace).
+# Single-arch linux/amd64 only — the target GKE node pools are all amd64, so
+# skipping arm64 + the digest/merge dance keeps this fast.
+#
+# Requires two repo secrets: DOCKERHUB_USERNAME and DOCKERHUB_TOKEN
+# (a Docker Hub access token with Read/Write).
+
+on:
+  workflow_dispatch:
+    inputs:
+      extra_tag:
+        description: "Extra tag to publish alongside latest + sha (e.g. per-user-tenancy)"
+        required: false
+        default: ""
+  push:
+    branches: [feat/per-user-tenancy]
+    paths:
+      - .github/workflows/publish-images-mo4islona.yml
+      - services/**
+      - centaur_sdk/**
+      - packages/**
+      - tools/**
+      - package.json
+      - pnpm-lock.yaml
+      - pnpm-workspace.yaml
+      - .agents/skills/**
+
+concurrency:
+  group: publish-mo4islona-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+env:
+  REGISTRY: docker.io
+  IMAGE_NAMESPACE: mo4islona
+  IMAGE_SOURCE: https://github.com/subsquid/centaur
+  RUST_BUILD_PROFILE: release
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - image: centaur-api-rs
+            context: .
+            dockerfile: services/api-rs/Dockerfile
+            target: ""
+          - image: centaur-slackbotv2
+            context: .
+            dockerfile: services/slackbotv2/Dockerfile
+            target: ""
+          - image: centaur-discordbot
+            context: .
+            dockerfile: services/discordbot/Dockerfile
+            target: ""
+          - image: centaur-agent
+            context: .
+            dockerfile: services/sandbox/Dockerfile
+            target: sandbox
+          - image: centaur-iron-proxy
+            context: .
+            dockerfile: services/iron-proxy/Dockerfile
+            target: ""
+          - image: centaur-console
+            context: services/console
+            dockerfile: services/console/Dockerfile
+            target: ""
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          persist-credentials: false
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f # v3
+
+      - name: Log in to Docker Hub
+        uses: docker/login-action@c94ce9fb468520275223c153574b00df6fe4bcc9 # v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Docker metadata for ${{ matrix.image }}
+        id: meta
+        uses: docker/metadata-action@c299e40c65443455700f0fdfc63efafe5b349051 # v5
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAMESPACE }}/${{ matrix.image }}
+          flavor: |
+            latest=false
+          tags: |
+            type=raw,value=latest
+            type=sha
+            type=raw,value=${{ github.event.inputs.extra_tag }},enable=${{ github.event.inputs.extra_tag != '' }}
+          labels: |
+            org.opencontainers.image.title=${{ matrix.image }}
+            org.opencontainers.image.source=${{ env.IMAGE_SOURCE }}
+
+      - name: Build and push ${{ matrix.image }}
+        uses: docker/build-push-action@10e90e3645eae34f1e60eeb005ba3a3d33f178e8 # v6
+        with:
+          context: ${{ matrix.context }}
+          file: ${{ matrix.dockerfile }}
+          target: ${{ matrix.target }}
+          platforms: linux/amd64
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          build-args: |
+            RUST_BUILD_PROFILE=${{ env.RUST_BUILD_PROFILE }}
+          cache-from: type=gha,scope=mo4islona-${{ matrix.image }}
+          cache-to: type=gha,mode=max,scope=mo4islona-${{ matrix.image }}

contrib/chart/templates/apirs.yaml

@@ -135,6 +135,18 @@ spec:
       serviceAccountName: {{ $apiRsName }}
       {{- with .Values.global.imagePullSecrets }}
       imagePullSecrets:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.apiRs.nodeSelector }}
+      nodeSelector:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.apiRs.affinity }}
+      affinity:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.apiRs.tolerations }}
+      tolerations:
 {{ toYaml . | nindent 8 }}
       {{- end }}
       containers:

contrib/chart/templates/console-worker.yaml

@@ -36,6 +36,18 @@ spec:
       automountServiceAccountToken: false
       {{- with .Values.global.imagePullSecrets }}
       imagePullSecrets:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.nodeSelector }}
+      nodeSelector:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.affinity }}
+      affinity:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.tolerations }}
+      tolerations:
 {{ toYaml . | nindent 8 }}
       {{- end }}
       initContainers:

contrib/chart/templates/console.yaml

@@ -48,6 +48,18 @@ spec:
       automountServiceAccountToken: false
       {{- with .Values.global.imagePullSecrets }}
       imagePullSecrets:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.nodeSelector }}
+      nodeSelector:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.affinity }}
+      affinity:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.tolerations }}
+      tolerations:
 {{ toYaml . | nindent 8 }}
       {{- end }}
       initContainers:

contrib/chart/templates/networkpolicy.yaml

@@ -33,6 +33,22 @@ spec:
           port: 53
         - protocol: TCP
           port: 53
+{{- with .Values.networkPolicy.dnsServerCidrs }}
+    # Extra DNS resolver IPs reachable by ipBlock rather than namespaceSelector.
+    # Needed on GKE with NodeLocal DNSCache, whose resolver runs hostNetwork and
+    # so is not selectable as a kube-system pod (link-local 169.254.20.10 + the
+    # kube-dns ClusterIP). Empty by default (K3s needs nothing here).
+    - to:
+{{- range . }}
+        - ipBlock:
+            cidr: {{ . | quote }}
+{{- end }}
+      ports:
+        - protocol: UDP
+          port: 53
+        - protocol: TCP
+          port: 53
+{{- end }}
 ---
 {{- if .Values.postgres.enabled }}
 apiVersion: networking.k8s.io/v1

contrib/chart/templates/slackbotv2.yaml

@@ -21,6 +21,18 @@ spec:
       automountServiceAccountToken: false
       {{- with .Values.global.imagePullSecrets }}
       imagePullSecrets:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.slackbotv2.nodeSelector }}
+      nodeSelector:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.slackbotv2.affinity }}
+      affinity:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.slackbotv2.tolerations }}
+      tolerations:
 {{ toYaml . | nindent 8 }}
       {{- end }}
       containers:

contrib/chart/templates/workloads.yaml

@@ -25,6 +25,18 @@ spec:
       automountServiceAccountToken: false
       {{- with .Values.global.imagePullSecrets }}
       imagePullSecrets:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.postgres.nodeSelector }}
+      nodeSelector:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.postgres.affinity }}
+      affinity:
+{{ toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.postgres.tolerations }}
+      tolerations:
 {{ toYaml . | nindent 8 }}
       {{- end }}
       containers:

contrib/chart/values.mo4islona-pool.yaml

@@ -0,0 +1,98 @@
+# Pin the whole Centaur stack to a dedicated spot node pool.
+#
+# Pool contract (created out-of-band, see deploy notes):
+#   label:  centaur-pool=true
+#   taint:  centaur=true:NoSchedule   (so nothing else lands on the pool)
+#
+# Layer AFTER the image + env overlays, e.g.:
+#   helm upgrade --install centaur contrib/chart -n centaur --create-namespace \
+#     -f contrib/chart/values.dev.yaml \
+#     -f contrib/chart/values.mo4islona.yaml \
+#     -f contrib/chart/values.mo4islona-pool.yaml
+#
+# Control-plane scheduling (api-rs / slackbotv2 / postgres / repo-cache) is the
+# chart support added in PR feat/chart-node-scheduling. Sandbox runtime pods are
+# pinned via api-rs (SESSION_SANDBOX_NODE_SELECTOR / _TOLERATIONS on extraEnv),
+# which requires the api-rs build carrying that support.
+
+# GKE runs NodeLocal DNSCache (hostNetwork resolver), which the chart's default
+# kube-system namespaceSelector DNS rule does not cover, so pod DNS is denied.
+# Allow the node-local resolver + kube-dns ClusterIP by ipBlock.
+networkPolicy:
+  dnsServerCidrs:
+    - 169.254.20.10/32
+    - 10.0.16.10/32
+
+apiRs:
+  nodeSelector:
+    centaur-pool: "true"
+  tolerations:
+    - key: centaur
+      operator: Equal
+      value: "true"
+      effect: NoSchedule
+  extraEnv:
+    SESSION_SANDBOX_NODE_SELECTOR: "centaur-pool=true"
+    SESSION_SANDBOX_TOLERATIONS: '[{"key":"centaur","operator":"Equal","value":"true","effect":"NoSchedule"}]'
+
+slackbotv2:
+  nodeSelector:
+    centaur-pool: "true"
+  tolerations:
+    - key: centaur
+      operator: Equal
+      value: "true"
+      effect: NoSchedule
+
+postgres:
+  nodeSelector:
+    centaur-pool: "true"
+  tolerations:
+    - key: centaur
+      operator: Equal
+      value: "true"
+      effect: NoSchedule
+  persistence:
+    # premium-rwo = pd-ssd, and WaitForFirstConsumer so the PD is provisioned in
+    # the pod's zone. The default `standard` class is HDD + Immediate-binding
+    # (wrong zone), which fails to attach to the single-zone pool.
+    storageClassName: premium-rwo
+    # PGDATA must be a subdirectory of the mount: a freshly-formatted PD mounted
+    # directly at /var/lib/postgresql/data carries an ext4 lost+found, and initdb
+    # refuses a non-empty data dir.
+    subPath: pgdata
+
+# repo-cache hostPath DaemonSet: confine it to the pool so it runs only on the
+# pool node(s) (not every cluster node) and colocates with the sandbox pods that
+# read its hostPath.
+repoCache:
+  nodeSelector:
+    centaur-pool: "true"
+  tolerations:
+    - key: centaur
+      operator: Equal
+      value: "true"
+      effect: NoSchedule
+
+# console = the iron-control plane api-rs binds to (IRON_CONTROL_URL). Required
+# when apiRs.ironProxy.mode != disabled. Reads its config from centaur-infra-env
+# (IRON_CONTROL_* keys seeded by bootstrap) and self-provisions its DBs.
+console:
+  enabled: true
+  nodeSelector:
+    centaur-pool: "true"
+  tolerations:
+    - key: centaur
+      operator: Equal
+      value: "true"
+      effect: NoSchedule
+
+# agent-sandbox controller (subchart) — pin to the pool too.
+agentSandbox:
+  nodeSelector:
+    centaur-pool: "true"
+  tolerations:
+    - key: centaur
+      operator: Equal
+      value: "true"
+      effect: NoSchedule

contrib/chart/values.mo4islona.yaml

@@ -0,0 +1,39 @@
+# Image overrides pointing the chart at the personal Docker Hub namespace
+# (docker.io/mo4islona/centaur-*), built by .github/workflows/publish-images-mo4islona.yml.
+#
+# Layer this AFTER the environment values file, e.g.:
+#   helm upgrade --install centaur contrib/chart -n centaur --create-namespace \
+#     -f contrib/chart/values.dev.yaml -f contrib/chart/values.mo4islona.yaml
+#
+# tag stays "latest" (set in the base chart); pullPolicy: Always so the mutable
+# latest tag is re-pulled on each rollout.
+
+apiRs:
+  image:
+    repository: docker.io/mo4islona/centaur-api-rs
+    pullPolicy: Always
+
+slackbotv2:
+  image:
+    repository: docker.io/mo4islona/centaur-slackbotv2
+    pullPolicy: Always
+
+discordbot:
+  image:
+    repository: docker.io/mo4islona/centaur-discordbot
+    pullPolicy: Always
+
+sandbox:
+  image:
+    repository: docker.io/mo4islona/centaur-agent
+    pullPolicy: Always
+
+ironProxy:
+  image:
+    repository: docker.io/mo4islona/centaur-iron-proxy
+    pullPolicy: Always
+
+console:
+  image:
+    repository: docker.io/mo4islona/centaur-console
+    pullPolicy: Always

services/api-rs/crates/centaur-iron-control/src/error.rs

@@ -27,6 +27,12 @@ pub enum IronControlError {
         #[source]
         source: reqwest::Error,
     },
+    /// A caller supplied an explicit principal foreign_id that isn't a valid
+    /// URL-safe slug, so it can't be registered. Rejected rather than silently
+    /// falling back to a thread-derived principal (which would run the session
+    /// under the wrong identity / key).
+    #[error("invalid principal foreign_id {foreign_id:?}: must be URL-safe (A-Za-z0-9-._~)")]
+    InvalidPrincipalForeignId { foreign_id: String },
 }
 
 pub type Result<T> = std::result::Result<T, IronControlError>;