atlas-iac/services/hermes/chat-statefulset.yaml

# services/hermes/chat-statefulset.yaml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: hermes-chat-tenant
  namespace: hermes
  labels:
    app: hermes-chat-tenant
  annotations:
    kustomize.toolkit.fluxcd.io/force: enabled
spec:
  serviceName: hermes-chat-tenant
  replicas: 4
  podManagementPolicy: Parallel
  persistentVolumeClaimRetentionPolicy:
    whenDeleted: Retain
    whenScaled: Retain
  updateStrategy:
    type: RollingUpdate
  selector:
    matchLabels:
      app: hermes-chat-tenant
  template:
    metadata:
      labels:
        app: hermes-chat-tenant
      annotations:
        ai.bstein.dev/role: isolated-user-chat
        ai.bstein.dev/router-wire-contract: ollama-numeric-keepalive
        ai.bstein.dev/isolation: one Hermes process and PVC per Keycloak subject
        ai.bstein.dev/model-policy: uniform automatic policy with per-user overrides
        ai.bstein.dev/config-rev: "20260816-telegram-topics"
        ai.bstein.dev/hux-config-rev: "20260824-hux-v1"
        vault.hashicorp.com/agent-inject: "true"
        vault.hashicorp.com/role: hermes-chat
        vault.hashicorp.com/agent-inject-secret-chat-relay-key: kv/data/atlas/hermes/chat-telegram
        vault.hashicorp.com/agent-inject-template-chat-relay-key: |
          {{- with secret "kv/data/atlas/hermes/chat-telegram" -}}
          {{ .Data.data.relay_key }}
          {{- end }}
        vault.hashicorp.com/agent-pre-populate-only: "true"
        vault.hashicorp.com/agent-init-first: "true"
        vault.hashicorp.com/agent-requests-cpu: 25m
        vault.hashicorp.com/agent-requests-mem: 32Mi
        vault.hashicorp.com/agent-limits-cpu: 100m
        vault.hashicorp.com/agent-limits-mem: 128Mi
    spec:
      serviceAccountName: hermes-chat
      automountServiceAccountToken: true
      securityContext:
        fsGroup: 10000
        fsGroupChangePolicy: OnRootMismatch
        seccompProfile:
          type: RuntimeDefault
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: kubernetes.io/arch
                    operator: In
                    values: [arm64]
                  - key: node-role.kubernetes.io/worker
                    operator: In
                    values: ["true"]
                  - key: node-role.kubernetes.io/storage-backbone
                    operator: DoesNotExist
                  - key: kubernetes.io/hostname
                    operator: NotIn
                    values: [titan-05, titan-13, titan-14, titan-17, titan-18, titan-19]
          preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 100
              preference:
                matchExpressions:
                  - key: node-role.kubernetes.io/storage-backbone
                    operator: DoesNotExist
            - weight: 100
              preference:
                matchExpressions:
                  - key: hardware
                    operator: In
                    values: [rpi5]
            - weight: 40
              preference:
                matchExpressions:
                  - key: hardware
                    operator: In
                    values: [rpi4]
        podAntiAffinity:
          preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 100
              podAffinityTerm:
                labelSelector:
                  matchLabels:
                    app: hermes-chat-tenant
                topologyKey: kubernetes.io/hostname
      initContainers:
        - name: init-config
          image: busybox:1.37
          imagePullPolicy: IfNotPresent
          command:
            - sh
            - -c
            - |
              set -eu
              mkdir -p /opt/data/home/.local/bin /opt/data/logs /opt/data/workspace
              if [ ! -e /opt/data/workspace/.hermes-workspace-v1 ]; then
                if [ -d /legacy-home/workspace ]; then
                  cp -a /legacy-home/workspace/. /opt/data/workspace/
                fi
                touch /opt/data/workspace/.hermes-workspace-v1
              fi
              cp /config/config.yaml /opt/data/config.yaml
              cp /config/SOUL.md /opt/data/SOUL.md
              cp /config/AGENTS.md /opt/data/workspace/AGENTS.md
              touch /opt/data/.env
              for key in ANTHROPIC_API_KEY API_SERVER_KEY CLAUDE_API_KEY CLAUDE_CODE_OAUTH_TOKEN GITEA_TOKEN GITEA_USERNAME HERMES_IMAGE_BROKER_KEY OPENAI_API_KEY; do
                grep -v "^${key}=" /opt/data/.env > /opt/data/.env.tmp || true
                mv /opt/data/.env.tmp /opt/data/.env
              done
              rm -f /opt/data/auth.json
              chmod 0600 /opt/data/.env
              chown 10000:10000 \
                /opt/data \
                /opt/data/home \
                /opt/data/home/.local \
                /opt/data/home/.local/bin \
                /opt/data/logs \
                /opt/data/workspace \
                /opt/data/config.yaml \
                /opt/data/SOUL.md \
                /opt/data/workspace/AGENTS.md \
                /opt/data/.env
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 0
            runAsGroup: 0
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: home, mountPath: /opt/data}
            - {name: home, mountPath: /legacy-home, readOnly: true}
            - {name: workspace, mountPath: /opt/data/workspace}
            - {name: config, mountPath: /config, readOnly: true}
          resources:
            requests: {cpu: 25m, memory: 32Mi}
            limits: {cpu: 100m, memory: 64Mi}
        - name: init-hux-runtime
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              umask 077
              tenant_root="/hux-data/${HOSTNAME}"
              export HUX_INIT_ROOT="${tenant_root}"
              mkdir -p \
                "${tenant_root}/binding" \
                "${tenant_root}/context" \
                "${tenant_root}/store" \
                /hux-relay \
                /hux-worker
              chown 10000:10000 \
                "${tenant_root}" \
                "${tenant_root}/binding" \
                "${tenant_root}/context" \
                "${tenant_root}/store" \
                /hux-relay \
                /hux-worker
              chmod 0700 \
                "${tenant_root}" \
                "${tenant_root}/binding" \
                "${tenant_root}/context" \
                "${tenant_root}/store" \
                /hux-relay \
                /hux-worker
              if [ ! -e "${tenant_root}/context/context-key" ]; then
                dd if=/dev/urandom of="${tenant_root}/context/.context-key.tmp" bs=32 count=1 2>/dev/null
                chown 10000:10000 "${tenant_root}/context/.context-key.tmp"
                chmod 0600 "${tenant_root}/context/.context-key.tmp"
                mv "${tenant_root}/context/.context-key.tmp" "${tenant_root}/context/context-key"
              fi
              test "$(wc -c < "${tenant_root}/context/context-key")" -eq 32
              chown 10000:10000 "${tenant_root}/context/context-key"
              chmod 0600 "${tenant_root}/context/context-key"
              ordinal="${HOSTNAME##*-}"
              HUX_INIT_SLOT="slot-${ordinal}" \
                /opt/hermes/.venv/bin/python - <<'PY'
              import hashlib
              import hmac
              import os
              import stat
              from pathlib import Path

              root = Path(os.environ["HUX_INIT_ROOT"])
              key = (root / "context/context-key").read_bytes()
              slot = os.environ["HUX_INIT_SLOT"]
              subject = "usr_" + hmac.new(
                  key,
                  b"hux.subject.id.v1\0" + slot.encode("ascii"),
                  hashlib.sha256,
              ).hexdigest()
              target = root / "binding/subject"
              expected = (subject + "\n").encode("ascii")
              flags = os.O_WRONLY | os.O_CREAT | os.O_EXCL | getattr(os, "O_NOFOLLOW", 0)
              try:
                  descriptor = os.open(target, flags, 0o440)
              except FileExistsError:
                  info = target.lstat()
                  if (
                      not stat.S_ISREG(info.st_mode)
                      or info.st_uid != 10000
                      or stat.S_IMODE(info.st_mode) != 0o440
                      or info.st_nlink != 1
                      or target.read_bytes() != expected
                  ):
                      raise SystemExit("persistent HUX subject binding is unsafe")
              else:
                  try:
                      os.write(descriptor, expected)
                      os.fchown(descriptor, 10000, 10000)
                      os.fchmod(descriptor, 0o440)
                      os.fsync(descriptor)
                  finally:
                      os.close(descriptor)
              PY
              if [ ! -e "${tenant_root}/context/redaction-canary" ]; then
                dd if=/dev/urandom bs=32 count=1 2>/dev/null \
                  | sha256sum | cut -d ' ' -f 1 \
                  > "${tenant_root}/context/.redaction-canary.tmp"
                chown 10000:10000 "${tenant_root}/context/.redaction-canary.tmp"
                chmod 0400 "${tenant_root}/context/.redaction-canary.tmp"
                mv "${tenant_root}/context/.redaction-canary.tmp" "${tenant_root}/context/redaction-canary"
              fi
              for target in /hux-relay/relay-key /hux-worker/worker-key; do
                if [ ! -e "${target}" ]; then
                  dd if=/dev/urandom bs=32 count=1 2>/dev/null \
                    | sha256sum | cut -d ' ' -f 1 > "${target}.tmp"
                  chown 10000:10000 "${target}.tmp"
                  chmod 0400 "${target}.tmp"
                  mv "${target}.tmp" "${target}"
                fi
                test "$(wc -c < "${target}")" -eq 65
                chown 10000:10000 "${target}"
                chmod 0400 "${target}"
              done
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop: [ALL]
              add: [CHOWN, DAC_OVERRIDE, FOWNER]
            runAsUser: 0
            runAsGroup: 0
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: hux-data, mountPath: /hux-data}
            - {name: hux-relay-key, mountPath: /hux-relay}
            - {name: hux-worker-key, mountPath: /hux-worker}
          resources:
            requests: {cpu: 10m, memory: 16Mi}
            limits: {cpu: 50m, memory: 32Mi}
        - name: stage-hux-evidence
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              # HUX-12 enablement is fail-closed: until the Vault annotation
              # projects hux-evidence-key, nothing is staged and the evidence
              # capability stays off. The key is staged only for the hux
              # service and producer containers, never for hermes or webui.
              if [ -s /vault/secrets/hux-evidence-key ]; then
                umask 077
                tr -d '\r\n' < /vault/secrets/hux-evidence-key > /hux-evidence/.evidence-key.tmp
                chown 10000:10000 /hux-evidence/.evidence-key.tmp
                chmod 0400 /hux-evidence/.evidence-key.tmp
                mv /hux-evidence/.evidence-key.tmp /hux-evidence/evidence-key
              fi
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop: [ALL]
              add: [CHOWN, DAC_OVERRIDE, FOWNER]
            runAsUser: 0
            runAsGroup: 0
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: hux-evidence-key, mountPath: /hux-evidence}
          resources:
            requests: {cpu: 10m, memory: 16Mi}
            limits: {cpu: 50m, memory: 32Mi}
        - name: stage-runtime-access
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command:
            - /opt/hermes/.venv/bin/python
            - /opt/coordinator/stage_runtime_access.py
            - chat
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 0
            runAsGroup: 0
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: runtime-access, mountPath: /runtime-access}
          resources:
            requests: {cpu: 25m, memory: 32Mi}
            limits: {cpu: 100m, memory: 64Mi}
        - name: patch-auth
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command:
            - /opt/hermes/.venv/bin/python
            - /opt/coordinator/patch_hermes_auth.py
            - /opt/hermes/hermes_cli/auth.py
            - /patched/auth.py
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: auth-patch, mountPath: /patched}
          resources:
            requests: {cpu: 25m, memory: 64Mi}
            limits: {cpu: 100m, memory: 128Mi}
        - name: patch-stream-recovery
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command:
            - /opt/hermes/.venv/bin/python
            - /opt/coordinator/patch_stream_recovery.py
            - /opt/hermes/agent/conversation_loop.py
            - /patched/conversation_loop.py
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: stream-recovery-patch, mountPath: /patched}
          resources:
            requests: {cpu: 25m, memory: 64Mi}
            limits: {cpu: 100m, memory: 128Mi}
        - name: patch-api-server-sessions
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              /opt/hermes/.venv/bin/python /opt/coordinator/patch_api_server_sessions.py \
                /opt/hermes/gateway/platforms/api_server.py /patched/api_server.py
              grep -Fq 'conversation_history = compact_telegram_history(conversation_history)' /patched/api_server.py
              grep -Fq 'full_history = compact_telegram_history(full_history)' /patched/api_server.py
              grep -Fq 'conversation_history_snapshot = compact_telegram_history(' /patched/api_server.py
              /opt/hermes/.venv/bin/python /opt/coordinator/migrate_telegram_api_sessions.py \
                /opt/data/state.db /opt/data/response_store.db
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: home, mountPath: /opt/data}
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: api-server-patch, mountPath: /patched}
          resources:
            requests: {cpu: 25m, memory: 64Mi}
            limits: {cpu: 100m, memory: 128Mi}
        - name: patch-subprocess-secret-boundary
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command:
            - /bin/sh
            - -ec
            - |
              /opt/hermes/.venv/bin/python /opt/coordinator/patch_subprocess_secret_boundary.py \
                /opt/hermes/tools/environments/local.py /patched/local.py
              /opt/hermes/.venv/bin/python /opt/coordinator/patch_process_output_redaction.py \
                /opt/hermes/tools/process_registry.py /patched/process_registry.py
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          volumeMounts:
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: subprocess-secret-patch, mountPath: /patched}
          resources:
            requests: {cpu: 25m, memory: 64Mi}
            limits: {cpu: 100m, memory: 128Mi}
      containers:
        - name: hermes
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              ordinal="${HOSTNAME##*-}"
              export HERMES_CODE_SANDBOX_URL="http://hermes-chat-sandbox-${ordinal}.hermes-chat-sandbox.hermes.svc.cluster.local:9080/v1/execute"
              set -a
              . /opt/data/.env
              set +a
              API_SERVER_KEY="$(tr -d '\r\n' < /runtime-access/chat-relay-key)"
              test -n "${API_SERVER_KEY}"
              export API_SERVER_KEY
              export HUX_TENANT_SLOT="slot-${ordinal}"
              exec /opt/hermes/.venv/bin/hermes gateway run
          ports:
            - {name: api, containerPort: 8642, protocol: TCP}
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - {name: HERMES_HOME, value: /opt/data}
            - {name: HERMES_AUTH_FILE, value: /runtime-access/hermes-auth.json}
            - {name: HOME, value: /opt/data/home}
            - {name: TERMINAL_CWD, value: /opt/data/workspace}
            - {name: HERMES_WRITE_SAFE_ROOT, value: /opt/data/workspace}
            - {name: PATH, value: '/opt/data/home/.local/bin:/opt/hermes/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin'}
            - {name: PLAYWRIGHT_BROWSERS_PATH, value: /opt/hermes/.playwright}
            - {name: AGENT_BROWSER_EXECUTABLE_PATH, value: /opt/hermes/.playwright/chromium_headless_shell-1228/chrome-linux/headless_shell}
            - {name: AGENT_BROWSER_ARGS, value: "--no-sandbox,--disable-dev-shm-usage"}
            - {name: HERMES_DASHBOARD, value: "0"}
            - {name: API_SERVER_ENABLED, value: "true"}
            - {name: API_SERVER_HOST, value: 0.0.0.0}
            - {name: API_SERVER_PORT, value: "8642"}
            - {name: API_SERVER_CORS_ORIGINS, value: 'https://chat.bstein.dev'}
            - {name: HERMES_IMAGE_BROKER_URL, value: 'http://hermes-image-broker.hermes.svc.cluster.local:9002'}
            - {name: HERMES_IMAGE_BROKER_KEY_FILE, value: /runtime-access/chat-relay-key}
            - {name: HERMES_AUTO_ROUTER_PROFILE, value: chat}
            - {name: HUX_BASE_URL, value: 'http://127.0.0.1:8790'}
            - {name: HUX_RUNTIME_ENABLED, value: "1"}
            # First rollout is observe-only until approval parking/resume is
            # connected to the upstream tool loop and proven live.
            - {name: HUX_TOOL_ENFORCEMENT, value: "0"}
            - {name: HUX_WORKER_KEY_FILE, value: /run/hermes-hux-worker/worker-key}
            - {name: HUX_SUBJECT_FILE, value: /run/hermes-hux-subject/subject}
            - {name: HUX_CONTEXT_KEY_FILE, value: /run/hermes-hux-context/context-key}
            - {name: HUX_PROJECT_SOURCE, value: 'profile:default'}
            - {name: HUX_TIMEOUT_SECONDS, value: "3"}
          volumeMounts:
            - {name: home, mountPath: /opt/data}
            - {name: workspace, mountPath: /opt/data/workspace}
            - {name: runtime-access, mountPath: /runtime-access}
            - {name: auth-patch, mountPath: /opt/hermes/hermes_cli/auth.py, subPath: auth.py}
            - {name: api-server-patch, mountPath: /opt/hermes/gateway/platforms/api_server.py, subPath: api_server.py}
            - {name: coordinator, mountPath: /opt/hermes/gateway/platforms/telegram_continuity.py, subPath: migrate_telegram_api_sessions.py, readOnly: true}
            - {name: stream-recovery-patch, mountPath: /opt/hermes/agent/conversation_loop.py, subPath: conversation_loop.py}
            - {name: subprocess-secret-patch, mountPath: /opt/hermes/tools/environments/local.py, subPath: local.py}
            - {name: subprocess-secret-patch, mountPath: /opt/hermes/tools/process_registry.py, subPath: process_registry.py}
            - {name: image-plugin, mountPath: /opt/hermes/plugins/image_gen/atlas-broker, readOnly: true}
            - {name: auto-router-plugin, mountPath: /opt/data/plugins/auto-router, readOnly: true}
            - {name: hux-runtime-plugin, mountPath: /opt/data/plugins/hux-runtime, readOnly: true}
            - {name: cluster-read-plugin, mountPath: /opt/data/plugins/cluster-read, readOnly: true}
            - {name: hux-worker-key, mountPath: /run/hermes-hux-worker, readOnly: true}
            - {name: hux-data, mountPath: /run/hermes-hux-context, subPathExpr: $(POD_NAME)/context, readOnly: true}
            - {name: hux-data, mountPath: /run/hermes-hux-subject, subPathExpr: $(POD_NAME)/binding, readOnly: true}
          readinessProbe:
            tcpSocket: {port: api}
            initialDelaySeconds: 30
            periodSeconds: 10
            timeoutSeconds: 5
          livenessProbe:
            tcpSocket: {port: api}
            initialDelaySeconds: 90
            periodSeconds: 30
            timeoutSeconds: 10
          securityContext:
            allowPrivilegeEscalation: false
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          resources:
            requests: {cpu: 250m, memory: 256Mi}
            limits: {cpu: "1", memory: 2Gi}
        - name: webui
          image: registry.bstein.dev/bstein/hermes-webui:git-5b1f8320729ebd1514ce83b7e695a02938e5cef8-build-36-release@sha256:dfdacf006af93add1bd811637c30f3d4fd16b6f7e24d81ddc56d2989aa31210a # {"$imagepolicy": "hermes:hermes-webui-release"}
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              api_key="$(tr -d '\r\n' < /runtime-access/chat-relay-key)"
              test -n "${api_key}"
              export API_SERVER_KEY="${api_key}"
              export HERMES_WEBUI_GATEWAY_API_KEY="${api_key}"
              exec /opt/hermes/.venv/bin/python /opt/hermes-webui/server.py
          ports:
            - {name: webui, containerPort: 8787, protocol: TCP}
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - {name: HERMES_HOME, value: /opt/data}
            - {name: HERMES_AUTH_FILE, value: /runtime-access/hermes-auth.json}
            - {name: HOME, value: /opt/data/home}
            - {name: HERMES_WEBUI_AGENT_DIR, value: /opt/hermes}
            - {name: HERMES_WEBUI_HOST, value: 0.0.0.0}
            - {name: HERMES_WEBUI_PORT, value: "8787"}
            - {name: HERMES_WEBUI_STATE_DIR, value: /opt/data/webui}
            - {name: HERMES_WEBUI_DEFAULT_WORKSPACE, value: /opt/data/workspace}
            - {name: HERMES_WEBUI_CHAT_BACKEND, value: gateway}
            - {name: HERMES_WEBUI_GATEWAY_BASE_URL, value: 'http://127.0.0.1:8642'}
            - {name: HERMES_WEBUI_GATEWAY_USE_RUNS_API, value: "true"}
            - {name: HERMES_WEBUI_SKIP_ONBOARDING, value: "1"}
            - {name: HERMES_WEBUI_SECURE, value: "1"}
            - {name: HERMES_WEBUI_COOKIE_NAME, value: hermes_chat_session}
            - {name: HERMES_WEBUI_PROFILE_COOKIE_NAME, value: hermes_chat_profile}
            - {name: HERMES_WEBUI_TRUSTED_AUTH_HEADER, value: X-Hermes-Tenant-Identity}
            # NetworkPolicy admits this port only from hermes-chat-router; the
            # CIDR lets the WebUI validate that router's changing pod address.
            - {name: HERMES_WEBUI_TRUSTED_PROXY_CIDRS, value: 10.42.0.0/16}
            - {name: HERMES_WEBUI_ALLOWED_ORIGINS, value: 'https://chat.bstein.dev'}
            - {name: HERMES_WEBUI_TRUST_FORWARDED_HOST, value: "1"}
            - {name: HERMES_WEBUI_TRUST_FORWARDED_PROTO, value: "1"}
            - {name: HERMES_ROUTER_PROFILE, value: chat}
            - {name: HERMES_STT_URL, value: 'http://hermes-stt.hermes.svc.cluster.local:9000/v1/audio/transcriptions'}
            - {name: HERMES_WEBUI_ATLAS_STT_STREAM_URL, value: 'http://hermes-stt.hermes.svc.cluster.local:9000/v1/audio/transcriptions/stream'}
            - {name: HERMES_LOCAL_STT_COMMAND, value: "/opt/hermes/.venv/bin/python /opt/coordinator/hermes_stt_client.py {input_path} --output-dir {output_dir} --language {language} --model {model}"}
            - {name: HERMES_WEBUI_ATLAS_TTS_URL, value: 'http://hermes-tts.hermes.svc.cluster.local:9001/v1/audio/speech'}
            - {name: HERMES_WEBUI_ATLAS_TTS_STREAM_URL, value: 'http://hermes-tts.hermes.svc.cluster.local:9001/v1/audio/speech/stream'}
            - {name: HUX_CONTEXT_KEY_FILE, value: /run/hermes-hux-context/context-key}
            - {name: HUX_PROJECT_SOURCE, value: 'profile:default'}
          volumeMounts:
            - {name: home, mountPath: /opt/data}
            - {name: workspace, mountPath: /opt/data/workspace}
            - {name: runtime-access, mountPath: /runtime-access, readOnly: true}
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: tmp, mountPath: /tmp}
            - {name: hux-relay-key, mountPath: /run/hermes-webui-hux, readOnly: true}
            - {name: hux-data, mountPath: /run/hermes-hux-context, subPathExpr: $(POD_NAME)/context, readOnly: true}
          readinessProbe:
            httpGet: {path: /health, port: webui}
            initialDelaySeconds: 10
            periodSeconds: 10
            timeoutSeconds: 5
          livenessProbe:
            httpGet: {path: /health, port: webui}
            initialDelaySeconds: 30
            periodSeconds: 30
            timeoutSeconds: 10
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop: [ALL]
            readOnlyRootFilesystem: true
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          resources:
            requests: {cpu: 100m, memory: 224Mi}
            limits: {cpu: 750m, memory: 1Gi}
        - name: hux
          image: registry.bstein.dev/bstein/hermes-webui:git-5b1f8320729ebd1514ce83b7e695a02938e5cef8-build-36-release@sha256:dfdacf006af93add1bd811637c30f3d4fd16b6f7e24d81ddc56d2989aa31210a # {"$imagepolicy": "hermes:hermes-webui-release"}
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              ordinal="${HOSTNAME##*-}"
              export HUX_TENANT_SLOT="slot-${ordinal}"
              exec /opt/hermes/.venv/bin/python -m hux.server
          ports:
            - {name: hux-loopback, containerPort: 8790, protocol: TCP}
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - {name: PYTHONPATH, value: /opt/hermes-hux}
            - {name: PYTHONDONTWRITEBYTECODE, value: "1"}
            - {name: HOME, value: /tmp}
            - {name: HUX_BIND, value: 127.0.0.1}
            - {name: HUX_PORT, value: "8790"}
            - {name: HUX_DATA_ROOT, value: /var/lib/hux/store}
            - {name: HUX_FLAGS, value: 'hux.foundation,hux.activity_timeline,hux.projects,hux.privacy,hux.memory_control,hux.artifacts,hux.research,hux.friendly_modes,hux.multimodal,hux.onboarding,hux.autonomy'}
            - {name: HUX_RELAY_KEY_FILE, value: /run/hermes-webui-hux/relay-key}
            - {name: HUX_WORKER_KEY_FILE, value: /run/hermes-hux-worker/worker-key}
            - {name: HUX_SUBJECT_BINDING_FILE, value: /var/lib/hux/binding/subject}
            - {name: HUX_CONTEXT_KEY_FILE, value: /var/lib/hux/context/context-key}
            - {name: HUX_CANARY_FILE, value: /var/lib/hux/context/redaction-canary}
            - name: HUX_IMAGE_TAG
              value: git-5b1f8320729ebd1514ce83b7e695a02938e5cef8-build-36-release # {"$imagepolicy": "hermes:hermes-webui-release:tag"}
            - name: HUX_IMAGE_DIGEST
              value: sha256:dfdacf006af93add1bd811637c30f3d4fd16b6f7e24d81ddc56d2989aa31210a # {"$imagepolicy": "hermes:hermes-webui-release:digest"}
            - {name: HUX_SWITCHYARD_ROUTE_CATALOG, value: 'atlas/manual/codex/luna,atlas/manual/codex/terra,atlas/manual/codex/sol,atlas/manual/claude/haiku,atlas/manual/claude/fable,atlas/manual/claude/sonnet,atlas/manual/claude/opus,atlas/manual/local/qwen-14b'}
            - {name: HUX_RELEASE_EVIDENCE_KEY_FILE, value: /run/hermes-hux-evidence/evidence-key}
            - {name: HUX_RELEASE_EVIDENCE_POLICY_FILE, value: /etc/hux-evidence/policy.json}
            - {name: HUX_READS_PER_MINUTE, value: "600"}
            - {name: HUX_WRITES_PER_MINUTE, value: "120"}
            - {name: HUX_REQUEST_TIMEOUT_SECONDS, value: "10"}
          volumeMounts:
            - {name: hux-data, mountPath: /var/lib/hux, subPathExpr: $(POD_NAME)}
            - {name: hux-relay-key, mountPath: /run/hermes-webui-hux, readOnly: true}
            - {name: hux-worker-key, mountPath: /run/hermes-hux-worker, readOnly: true}
            - {name: hux-tmp, mountPath: /tmp}
            - {name: hux-evidence-key, mountPath: /run/hermes-hux-evidence, readOnly: true}
            - {name: hux-evidence-policy, mountPath: /etc/hux-evidence, readOnly: true}
          readinessProbe:
            exec:
              command:
                - /opt/hermes/.venv/bin/python
                - -c
                - "import json,urllib.request; body=json.load(urllib.request.urlopen('http://127.0.0.1:8790/healthz', timeout=2)); assert body['status']=='ok'"
            initialDelaySeconds: 2
            periodSeconds: 5
            timeoutSeconds: 5
            failureThreshold: 12
          livenessProbe:
            exec:
              command:
                - /opt/hermes/.venv/bin/python
                - -c
                - "import json,urllib.request; body=json.load(urllib.request.urlopen('http://127.0.0.1:8790/healthz', timeout=2)); assert body['status']=='ok'"
            initialDelaySeconds: 20
            periodSeconds: 20
            timeoutSeconds: 5
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop: [ALL]
            readOnlyRootFilesystem: true
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          resources:
            requests: {cpu: 25m, memory: 64Mi}
            limits: {cpu: 250m, memory: 256Mi}
        - name: hux-evidence-producer
          image: registry.bstein.dev/bstein/hermes-webui:git-5b1f8320729ebd1514ce83b7e695a02938e5cef8-build-36-release@sha256:dfdacf006af93add1bd811637c30f3d4fd16b6f7e24d81ddc56d2989aa31210a # {"$imagepolicy": "hermes:hermes-webui-release"}
          imagePullPolicy: IfNotPresent
          command: [/bin/sh, -ec]
          args:
            - |
              ordinal="${HOSTNAME##*-}"
              export HUX_TENANT_SLOT="slot-${ordinal}"
              while true; do
                if [ -s /run/hermes-hux-evidence/evidence-key ] \
                  && [ -s /run/hermes-hux-subject/subject ] \
                  && [ -n "${HUX_PRODUCER_PROJECT_ID:-}" ] \
                  && [ -n "${HUX_PRODUCER_CONVERSATION_ID:-}" ]; then
                  HUX_PRODUCER_SUBJECT="$(tr -d '\r\n' < /run/hermes-hux-subject/subject)" \
                    /opt/hermes/.venv/bin/python -c \
                    'from hux_producer import run_once; run_once()' || true
                fi
                sleep 60
              done
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - {name: PYTHONPATH, value: /opt/hermes-hux}
            - {name: PYTHONDONTWRITEBYTECODE, value: "1"}
            - {name: HOME, value: /tmp}
            - {name: HUX_BASE_URL, value: 'http://127.0.0.1:8790'}
            - {name: HUX_PRODUCER_WORKLOAD, value: hermes-webui}
            - {name: HUX_PRODUCER_NAMESPACE, value: hermes}
            - {name: HUX_PRODUCER_POD_SELECTOR, value: 'app=hermes-chat-tenant'}
            - {name: HUX_PRODUCER_WORKLOAD_KIND, value: statefulset}
            - {name: HUX_PRODUCER_WORKLOAD_NAME, value: hermes-chat-tenant}
            - {name: HUX_PRODUCER_TIMEOUT_SECONDS, value: "10"}
            - {name: HUX_RELEASE_EVIDENCE_KEY_FILE, value: /run/hermes-hux-evidence/evidence-key}
            - {name: HUX_RELEASE_EVIDENCE_POLICY_FILE, value: /etc/hux-evidence/policy.json}
          envFrom:
            - configMapRef:
                name: hermes-hux-evidence-scope
                optional: true
          volumeMounts:
            - {name: hux-evidence-key, mountPath: /run/hermes-hux-evidence, readOnly: true}
            - {name: hux-evidence-policy, mountPath: /etc/hux-evidence, readOnly: true}
            - {name: hux-data, mountPath: /run/hermes-hux-subject, subPathExpr: $(POD_NAME)/binding, readOnly: true}
            - {name: hux-tmp, mountPath: /tmp}
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop: [ALL]
            readOnlyRootFilesystem: true
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          resources:
            requests: {cpu: 10m, memory: 48Mi}
            limits: {cpu: 100m, memory: 128Mi}
        - name: telegram-media
          image: registry.bstein.dev/bstein/hermes-agent@sha256:81970563e542f0720773e72297810b3a844b83e381e278f25c0916c78d930107
          imagePullPolicy: IfNotPresent
          command: [/opt/hermes/.venv/bin/python, /opt/coordinator/telegram_media_server.py]
          ports:
            - {name: telegram-media, containerPort: 8788, protocol: TCP}
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - {name: HERMES_MEDIA_RELAY_KEY_FILE, value: /runtime-access/chat-relay-key}
          volumeMounts:
            - {name: home, mountPath: /opt/data, readOnly: true}
            - {name: workspace, mountPath: /opt/data/workspace, readOnly: true}
            - {name: coordinator, mountPath: /opt/coordinator, readOnly: true}
            - {name: runtime-access, mountPath: /runtime-access, readOnly: true}
          readinessProbe:
            httpGet: {path: /healthz, port: telegram-media}
            initialDelaySeconds: 2
            periodSeconds: 10
          livenessProbe:
            httpGet: {path: /healthz, port: telegram-media}
            initialDelaySeconds: 10
            periodSeconds: 20
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop: [ALL]
            readOnlyRootFilesystem: true
            runAsUser: 10000
            runAsGroup: 10000
            seccompProfile:
              type: RuntimeDefault
          resources:
            requests: {cpu: 10m, memory: 24Mi}
            limits: {cpu: 100m, memory: 64Mi}
      volumes:
        - name: runtime-access
          emptyDir:
            medium: Memory
            sizeLimit: 2Mi
        - name: config
          configMap:
            name: hermes-chat-config
        - name: coordinator
          configMap:
            name: hermes-coordinator
            defaultMode: 0555
        - name: auth-patch
          emptyDir: {}
        - name: api-server-patch
          emptyDir: {}
        - name: stream-recovery-patch
          emptyDir: {}
        - name: subprocess-secret-patch
          emptyDir: {}
        - name: auto-router-plugin
          configMap:
            name: hermes-auto-router-plugin
        - name: image-plugin
          configMap:
            name: hermes-chat-image-plugin
        - name: cluster-read-plugin
          configMap:
            name: hermes-cluster-read-plugin
        - name: hux-runtime-plugin
          configMap:
            name: hermes-hux-runtime-plugin
            items:
              - {key: __init__.py, path: __init__.py}
              - {key: context_ids.py, path: context_ids.py}
              - {key: emitters.py, path: emitters.py}
              - {key: runtime.py, path: runtime.py}
              - {key: tool_policy.py, path: tool_policy.py}
              - {key: plugin.yaml, path: plugin.yaml}
              - {key: hux-hook-init.py, path: hux_hook/__init__.py}
              - {key: hux-hook-client.py, path: hux_hook/client.py}
              - {key: hux-hook-hooks.py, path: hux_hook/hooks.py}
        - name: tmp
          emptyDir:
            sizeLimit: 256Mi
        - name: hux-relay-key
          emptyDir:
            medium: Memory
            sizeLimit: 1Mi
        - name: hux-worker-key
          emptyDir:
            medium: Memory
            sizeLimit: 1Mi
        - name: hux-tmp
          emptyDir:
            sizeLimit: 64Mi
        - name: hux-evidence-key
          emptyDir:
            medium: Memory
            sizeLimit: 1Mi
        - name: hux-evidence-policy
          configMap:
            name: hermes-hux-evidence-policy
            defaultMode: 0444
        - name: hux-data
          persistentVolumeClaim:
            claimName: hermes-chat-hux-data
  volumeClaimTemplates:
    - metadata:
        name: home
        labels:
          app: hermes-chat-tenant
      spec:
        accessModes: [ReadWriteOnce]
        storageClassName: astreae
        resources:
          requests:
            storage: 10Gi
    - metadata:
        name: workspace
        labels:
          app: hermes-chat-tenant
          ai.bstein.dev/data: user-workspace
      spec:
        accessModes: [ReadWriteMany]
        storageClassName: astreae
        resources:
          requests:
            storage: 10Gi