fix(llama-embed): raise the memory limit to match its own context and batch

#739 raised the context from 2048 to 4096 and the micro-batch from an effective 512 to 1024, both to match how Paperless chunks documents. It left memory at 2Gi/4Gi, which was sized for the smaller values. The pod ran fine interactively and was OOMKilled (exit 137) partway through the first sustained run, around 300 documents into an embedding pass. Nothing in the server log marked it — the process was killed outright, so the last line is a half-finished task and the reason is only visible in the container's lastState.terminated. This would have surfaced when Paperless first indexed the archive, i.e. on the first real use rather than under test. The weights are only ~600 MiB. What sets the footprint is four slots at 4096 context with a 1024-token micro-batch, plus Vulkan allocating the iGPU working set from system RAM where it counts against this cgroup — the same effect that put llama-server at 32Gi for 19 GiB of weights. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-09-02 08:14:19 +02:00
276 changed files with 2804 additions and 13255 deletions
@@ -42,7 +42,7 @@ when:

 steps:
  - name: fetch-alerts
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
+    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-08-10
    environment:
      SOPS_AGE_KEY:
        from_secret: sops_age_key
@@ -101,7 +101,7 @@ steps:
        bash scripts/ci/make-diag-kubeconfig.sh

  - name: alert-triage-agent
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    environment:
      CLAUDE_CODE_OAUTH_TOKEN:
        from_secret: claude_code_oauth_token
@@ -19,13 +19,12 @@
 # the admin account could (#330).
 #
 # The token is still OWNED by gitea_admin, because Gitea scopes packages to
-# a user or org namespace and most of these images live under `gitea_admin/*`.
-# The rebuild-critical ones additionally publish to `konairius/` — an org
-# created for the ghcr mirror path, with gitea_admin as a member so this same
-# token can push there. The full gitea_admin/* -> org migration is a
-# 38-reference change tracked separately in #330 rather than bundled here.
-# Note this pipeline runs no LLM agent, so the exfil path that motivated the
-# other cutovers does not apply to it.
+# a user or org namespace and these images live under `gitea_admin/*` — a
+# different account cannot publish there. Moving them to a dedicated org is
+# the remaining cleanup; it is a 38-reference migration across pipelines and
+# Flux-deployed CronJobs, so it is tracked separately in #330 rather than
+# bundled here. Note this pipeline runs no LLM agent, so the exfil path that
+# motivated the other cutovers does not apply to it.
 when:
  - event: push
    path:
@@ -88,30 +87,6 @@ steps:
          --destination gitea.renner.uno/gitea_admin/ci-k8s:$TAG \
          --destination gitea.renner.uno/gitea_admin/ci-k8s:latest

-  - name: build-vmtest
-    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
-    directory: ci/images/vmtest
-    depends_on: []
-    when:
-      - event: deployment
-      - event: push
-        path: ["ci/images/vmtest/**", ".woodpecker/build-ci-images.yml"]
-    environment:
-      GITEA_USER: gitea_admin
-      GITEA_TOKEN:
-        from_secret: ci_images_token
-    commands:
-      - mkdir -p /kaniko/.docker
-      - AUTH=$(printf '%s:%s' "$GITEA_USER" "$GITEA_TOKEN" | base64 | tr -d '\n')
-      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"}}}' "$AUTH" > /kaniko/.docker/config.json
-      - TAG=$(date +%Y-%m-%d)
-      - |
-        /kaniko/executor \
-          --context . \
-          --dockerfile Containerfile \
-          --destination gitea.renner.uno/gitea_admin/ci-vmtest:$TAG \
-          --destination gitea.renner.uno/gitea_admin/ci-vmtest:latest
-
  - name: build-tofu
    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
    directory: ci/images/tofu
@@ -172,28 +147,6 @@ steps:
  # not for who uses it — its first consumer is the zigbee2mqtt seed-data init
  # container, but nothing about it is zigbee-specific. Exists so tools come from
  # the image rather than an `apk add` on every container start.
-  #
-  # Published THREE times, and the third one is not a typo. Its consumers — the
-  # zigbee2mqtt seed-data init container and the trmnl renderer — start while
-  # the cluster is still being rebuilt, and at that moment gitea.renner.uno does
-  # not exist, because Gitea runs inside the cluster being built. A cold-rebuild
-  # rehearsal caught both in ImagePullBackOff.
-  #
-  # So there is a copy on ghcr.io. But the manifests must NOT point at it:
-  # normal operation pulls from the local registry, and an external reference
-  # would make every steady-state pull leave the network. The recovery source
-  # and the everyday source are different things, exactly as they are for Flux's
-  # git source.
-  #
-  # Talos resolves that with machine.registries.mirrors, which takes an ordered
-  # endpoint list — gitea first, ghcr second (bootstrap.sh). Containerd keeps
-  # the repository path when it falls through, so the two registries have to
-  # agree on it: `gitea_admin/alpine-jq` cannot be served by ghcr, where the
-  # namespace is fixed to the account name. Hence the second Gitea destination
-  # under konairius/, which the manifests reference and which matches the ghcr
-  # path exactly.
-  #
-  # The gitea_admin/ copy stays until nothing references it any more.
  - name: build-alpine-jq
    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
    directory: ci/images/alpine-jq
@@ -206,47 +159,23 @@ steps:
      GITEA_USER: gitea_admin
      GITEA_TOKEN:
        from_secret: ci_images_token
-      GHCR_USER: konairius
-      GHCR_TOKEN:
-        from_secret: ghcr_mirror_token
    commands:
      - mkdir -p /kaniko/.docker
      - AUTH=$(printf '%s:%s' "$GITEA_USER" "$GITEA_TOKEN" | base64 | tr -d '\n')
-      - GH_AUTH=$(printf '%s:%s' "$GHCR_USER" "$GHCR_TOKEN" | base64 | tr -d '\n')
-      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"},"ghcr.io":{"auth":"%s"}}}' "$AUTH" "$GH_AUTH" > /kaniko/.docker/config.json
+      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"}}}' "$AUTH" > /kaniko/.docker/config.json
      - TAG=$(date +%Y-%m-%d)
      - |
        /kaniko/executor \
          --context . \
          --dockerfile Containerfile \
          --destination gitea.renner.uno/gitea_admin/alpine-jq:$TAG \
-          --destination gitea.renner.uno/gitea_admin/alpine-jq:latest \
-          --destination gitea.renner.uno/konairius/alpine-jq:$TAG \
-          --destination gitea.renner.uno/konairius/alpine-jq:latest \
-          --destination ghcr.io/konairius/alpine-jq:$TAG \
-          --destination ghcr.io/konairius/alpine-jq:latest
+          --destination gitea.renner.uno/gitea_admin/alpine-jq:latest

  # Runtime image, not CI tooling: upstream containernetworking/plugins binaries
  # baked in, consumed by the multus cni-plugins-installer DaemonSet. Replaces a
  # node-side fetch from GitHub that had NO checksum verification and sat on the
  # critical path of every node rebuild. The download now happens here, once,
  # where the sha256 is checked — see the Containerfile for the full rationale.
-  #
-  # This one ALSO publishes to ghcr.io. A cold-rebuild rehearsal proved why: the
-  # DaemonSet could not pull from gitea.renner.uno, because Gitea runs inside the
-  # very cluster being rebuilt. Without the plugins there is no macvlan, so
-  # Multus cannot attach ceph-san, so Rook's network canary never starts, so the
-  # CephCluster sits at "Configuring Ceph Mons" — and OpenBao waits forever for a
-  # PVC that will never bind. Moving the download here fixed the integrity gap
-  # but quietly swapped github.com on the critical path for something strictly
-  # worse: a registry that by definition does not exist during a recovery. The
-  # GHCR copy is public, so a bare node can pull it with no credential and no
-  # Gitea.
-  #
-  # The test that proved this also proved the rule is not specific to this
-  # image. alpine-jq and alpine-openssl mirror for the same reason — see the
-  # note on build-alpine-jq. Anything an app needs while the cluster is still
-  # being built cannot live only in that cluster's own registry.
  - name: build-cni-plugins
    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
    directory: ci/images/cni-plugins
@@ -259,36 +188,23 @@ steps:
      GITEA_USER: gitea_admin
      GITEA_TOKEN:
        from_secret: ci_images_token
-      GHCR_USER: konairius
-      GHCR_TOKEN:
-        from_secret: ghcr_mirror_token
    commands:
      - mkdir -p /kaniko/.docker
      - AUTH=$(printf '%s:%s' "$GITEA_USER" "$GITEA_TOKEN" | base64 | tr -d '\n')
-      - GH_AUTH=$(printf '%s:%s' "$GHCR_USER" "$GHCR_TOKEN" | base64 | tr -d '\n')
-      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"},"ghcr.io":{"auth":"%s"}}}' "$AUTH" "$GH_AUTH" > /kaniko/.docker/config.json
+      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"}}}' "$AUTH" > /kaniko/.docker/config.json
      - TAG=$(date +%Y-%m-%d)
      - |
        /kaniko/executor \
          --context . \
          --dockerfile Containerfile \
          --destination gitea.renner.uno/gitea_admin/cni-plugins:$TAG \
-          --destination gitea.renner.uno/gitea_admin/cni-plugins:latest \
-          --destination gitea.renner.uno/konairius/cni-plugins:$TAG \
-          --destination gitea.renner.uno/konairius/cni-plugins:latest \
-          --destination ghcr.io/konairius/cni-plugins:$TAG \
-          --destination ghcr.io/konairius/cni-plugins:latest
+          --destination gitea.renner.uno/gitea_admin/cni-plugins:latest

  # Runtime image, not CI tooling: alpine + the openssl CLI, for shell scripts
  # needing s_client/x509. Named for what it is; first consumer is the paperless
  # fetch-bridge-cert init container. Replaces an `apk add openssl` that ran on
  # every paperless pod start. The openssl CLI is genuinely absent from the alpine
  # base, so this one needs an image rather than a busybox substitute.
-  #
-  # Mirrored to ghcr.io for the reason given on build-alpine-jq: it is an init
-  # container, so paperless does not start without it, and during a rebuild
-  # gitea.renner.uno is not there to serve it. A cold-rebuild rehearsal left
-  # paperless in Init:ErrImagePull on exactly this image.
  - name: build-alpine-openssl
    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
    directory: ci/images/alpine-openssl
@@ -301,25 +217,17 @@ steps:
      GITEA_USER: gitea_admin
      GITEA_TOKEN:
        from_secret: ci_images_token
-      GHCR_USER: konairius
-      GHCR_TOKEN:
-        from_secret: ghcr_mirror_token
    commands:
      - mkdir -p /kaniko/.docker
      - AUTH=$(printf '%s:%s' "$GITEA_USER" "$GITEA_TOKEN" | base64 | tr -d '\n')
-      - GH_AUTH=$(printf '%s:%s' "$GHCR_USER" "$GHCR_TOKEN" | base64 | tr -d '\n')
-      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"},"ghcr.io":{"auth":"%s"}}}' "$AUTH" "$GH_AUTH" > /kaniko/.docker/config.json
+      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"}}}' "$AUTH" > /kaniko/.docker/config.json
      - TAG=$(date +%Y-%m-%d)
      - |
        /kaniko/executor \
          --context . \
          --dockerfile Containerfile \
          --destination gitea.renner.uno/gitea_admin/alpine-openssl:$TAG \
-          --destination gitea.renner.uno/gitea_admin/alpine-openssl:latest \
-          --destination gitea.renner.uno/konairius/alpine-openssl:$TAG \
-          --destination gitea.renner.uno/konairius/alpine-openssl:latest \
-          --destination ghcr.io/konairius/alpine-openssl:$TAG \
-          --destination ghcr.io/konairius/alpine-openssl:latest
+          --destination gitea.renner.uno/gitea_admin/alpine-openssl:latest

  # Heating CAN readers: open3e (Vitocal/VX3) + E3onCAN (E380 meter), one image
  # run as two services on edge-1. Replaces pip-into-venv on the host, where an
@@ -349,56 +257,3 @@ steps:
          --dockerfile Containerfile \
          --destination gitea.renner.uno/gitea_admin/heating-can:$TAG \
          --destination gitea.renner.uno/gitea_admin/heating-can:latest
-
-  # netbird-service-sync: the controller that reconciles NetBird reverse-proxy
-  # services (public Ingress hosts -> TLS passthrough -> Traefik ClusterIP).
-  # Runtime artifact like heating-can; same registry so the audit tracks it.
-  - name: build-netbird-service-sync
-    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
-    directory: ci/images/netbird-service-sync
-    depends_on: []
-    when:
-      - event: deployment
-      - event: push
-        path: ["ci/images/netbird-service-sync/**", ".woodpecker/build-ci-images.yml"]
-    environment:
-      GITEA_USER: gitea_admin
-      GITEA_TOKEN:
-        from_secret: ci_images_token
-    commands:
-      - mkdir -p /kaniko/.docker
-      - AUTH=$(printf '%s:%s' "$GITEA_USER" "$GITEA_TOKEN" | base64 | tr -d '\n')
-      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"}}}' "$AUTH" > /kaniko/.docker/config.json
-      - TAG=$(date +%Y-%m-%d)
-      - |
-        /kaniko/executor \
-          --context . \
-          --dockerfile Containerfile \
-          --destination gitea.renner.uno/gitea_admin/netbird-service-sync:$TAG \
-          --destination gitea.renner.uno/gitea_admin/netbird-service-sync:latest
-
-  # bird: BGP daemon for the pod-network mesh over pod0 (#576). Runtime
-  # artifact like heating-can; same registry so the audit tracks it.
-  - name: build-bird
-    image: registry.gitlab.com/gitlab-ci-utils/container-images/kaniko:v1.25.18-debug
-    directory: ci/images/bird
-    depends_on: []
-    when:
-      - event: deployment
-      - event: push
-        path: ["ci/images/bird/**", ".woodpecker/build-ci-images.yml"]
-    environment:
-      GITEA_USER: gitea_admin
-      GITEA_TOKEN:
-        from_secret: ci_images_token
-    commands:
-      - mkdir -p /kaniko/.docker
-      - AUTH=$(printf '%s:%s' "$GITEA_USER" "$GITEA_TOKEN" | base64 | tr -d '\n')
-      - printf '{"auths":{"gitea.renner.uno":{"auth":"%s"}}}' "$AUTH" > /kaniko/.docker/config.json
-      - TAG=$(date +%Y-%m-%d)
-      - |
-        /kaniko/executor \
-          --context . \
-          --dockerfile Containerfile \
-          --destination gitea.renner.uno/gitea_admin/bird:$TAG \
-          --destination gitea.renner.uno/gitea_admin/bird:latest
@@ -14,8 +14,8 @@
 #                 and  ci-claude:latest.  The Containerfile makes this layer
 #                 byte-reproducible, so rebuilding an unchanged version dedups
 #                 in the registry instead of orphaning a fresh ~260 MB copy.
-#   3. bump-pr  — (only when any pin is off the latest version) rewrite every
-#                 consumer pin to  ci-claude:<version>  and open a PR. A human
+#   3. bump-pr  — (only when the version actually changed) rewrite the 5
+#                 consumer pins to  ci-claude:<version>  and open a PR. A human
 #                 reviews and merges it; nothing auto-merges.
 #
 # TAG SCHEME — deliberately version tags (ci-claude:<claude-version>), NOT the dated
@@ -107,22 +107,31 @@ steps:
          [0-9]*.[0-9]*.[0-9]*) : ;;
          *) echo "npm returned an implausible version '$$NEW' — aborting."; exit 1 ;;
        esac
-        # Distinct tags the consumers currently reference. Divergent pins are
-        # normal (a pipeline bumped on its own); the bump rewrites all of them.
-        CUR=$$(grep -hoE 'ci-claude:[0-9][0-9A-Za-z._-]*' $$CONSUMERS | sed 's/^ci-claude://' | sort -u | tr '\n' ' ' | sed 's/ $$//')
-        [ -n "$$CUR" ] || { echo "No ci-claude pin found in the consumers — aborting."; exit 1; }
-        echo "latest=$$NEW  current=[$$CUR]  event=$$CI_PIPELINE_EVENT"
+        # The tag every consumer currently references. They are kept identical,
+        # so this must resolve to EXACTLY ONE non-empty tag; anything else
+        # (no match, or a partial/disagreeing set) aborts rather than risk a
+        # corrupting sed later. Counting with `grep -c .` (non-empty lines)
+        # instead of `wc -l` is deliberate: an empty result yields 0 here but
+        # 1 under wc -l, which would slip past the guard.
+        CUR=$$(grep -hoE 'ci-claude:[0-9A-Za-z._-]+' $$CONSUMERS | sed 's/^ci-claude://' | sort -u)
+        NTAGS=$$(printf '%s' "$$CUR" | grep -c . || true)
+        if [ -z "$$CUR" ] || [ "$$NTAGS" -ne 1 ]; then
+          echo "Expected exactly one ci-claude tag across the consumers, got [$$CUR]."
+          echo "Refusing to guess — fix the pins by hand."; exit 1
+        fi
+        echo "latest=$$NEW  current=$$CUR  event=$$CI_PIPELINE_EVENT"
        printf '%s' "$$NEW" > "$$CI_WORKSPACE/.ci-refresh/version"
        printf '%s' "$$CUR" > "$$CI_WORKSPACE/.ci-refresh/current"
-        # Build when any pin is off latest, or on a push (Containerfile edit must
-        # ship even if the version is unchanged). Bump when any pin is off latest.
+        # Build when the version changed, or on a push (Containerfile edit must
+        # ship even if the version is unchanged). Bump the pins only when the
+        # version the consumers reference is not the latest.
        if [ "$$CUR" != "$$NEW" ] || [ "$$CI_PIPELINE_EVENT" = "push" ]; then
          touch "$$CI_WORKSPACE/.ci-refresh/do-build"; echo "will build $$NEW"
        else
          echo "already on $$NEW — nothing to build"
        fi
        if [ "$$CUR" != "$$NEW" ]; then
-          touch "$$CI_WORKSPACE/.ci-refresh/do-bump"; echo "will open bump PR [$$CUR] -> $$NEW"
+          touch "$$CI_WORKSPACE/.ci-refresh/do-bump"; echo "will open bump PR $$CUR -> $$NEW"
        else
          echo "consumers already pin $$NEW — no PR"
        fi
@@ -166,9 +175,6 @@ steps:
      - |
        [ -f "$$CI_WORKSPACE/.ci-refresh/do-bump" ] || { echo "no version change — no PR"; exit 0; }
        set -eu
-        # Secret values can carry a trailing newline; it breaks the push URL and the auth header.
-        PUSH_TOKEN=$$(printf '%s' "$$PUSH_TOKEN" | tr -d '\r\n')
-        [ -n "$$PUSH_TOKEN" ] || { echo "PUSH_TOKEN is empty — aborting"; exit 1; }
        NEW=$$(cat "$$CI_WORKSPACE/.ci-refresh/version")
        CUR=$$(cat "$$CI_WORKSPACE/.ci-refresh/current")
        # Both were validated non-empty by the decide step; re-assert before a
@@ -184,11 +190,9 @@ steps:
        # is not a maintained list.
        CONSUMERS=$$(grep -rlE 'ci-claude:[0-9]' .woodpecker)
        [ -n "$$CONSUMERS" ] || { echo "No ci-claude consumers found — refusing to continue."; exit 1; }
-        # NEW goes into a sed replacement: allow only a plain x.y.z version.
-        printf '%s' "$$NEW" | grep -qE '^[0-9]+\.[0-9]+\.[0-9]+$$' || { echo "refusing non-x.y.z version '$$NEW'"; exit 1; }
-        # Every version-pinned tag becomes NEW, whatever it was; consumers only.
+        # Exact-string swap of the current tag for the new one, consumers only.
        for f in $$CONSUMERS; do
-          sed -i "s#ci-claude:[0-9][0-9A-Za-z._-]*#ci-claude:$$NEW#g" "$$f"
+          sed -i "s#ci-claude:$$CUR#ci-claude:$$NEW#g" "$$f"
        done
        # Author is claude-bot (the PAT's account); the commit subject makes
        # the automated-bump nature clear.
@@ -60,7 +60,7 @@ steps:
  # run that never touched them. Tracked in #461.
  - name: kustomize-build
    depends_on: [] # see note above `steps:`
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
+    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-08-10
    commands:
      - |
        set -eu
@@ -158,7 +158,7 @@ steps:
  # whenever it trails the deployed server. If you bump the server, bump here.
  - name: woodpecker-lint
    depends_on: [] # see note above `steps:`
-    image: woodpeckerci/woodpecker-cli:v3.18.0-alpine
+    image: woodpeckerci/woodpecker-cli:v3.17.0-alpine
    commands:
      - |
        set -eu
@@ -246,7 +246,7 @@ steps:
  # this step fails permanently on its own documentation.
  - name: no-digest-pins
    depends_on: [] # see note above `steps:`
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
+    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-08-10
    commands:
      - |
        set -eu
@@ -17,6 +17,6 @@ when:

 steps:
  - name: model-selection-tests
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    commands:
      - sh scripts/ci/test-model-selection.sh
@@ -7,21 +7,7 @@ when:
      # install-netbird-proxy.yml also applies the netbird-proxy-watchdog role,
      # so changes to it must redeploy this host too.
      - "ansible/roles/netbird-proxy-watchdog/**"
-      # install-netbird-proxy.yml also applies the edge bouncer.
-      - "ansible/roles/crowdsec-edge-bouncer/**"
      - "ansible/plays/install-netbird-proxy.yml"
-      # The variables that steer those roles, not just the roles themselves.
-      # Without these two, a change that only touches inventory never deploys
-      # and nothing says so: arming the bouncer lives in host_vars, and its
-      # LAPI key lives in group_vars, so both would sit in git looking applied
-      # while the hosts kept running the previous configuration.
-      #
-      # pve_hosts.yml is deliberately NOT listed. It is the single inventory
-      # file for the whole fleet, so listing it would redeploy the proxies for
-      # any unrelated host edit. The cost is that a change to [netbird_proxy]
-      # group membership needs a manual deployment run.
-      - "ansible/inventory/host_vars/proxy-*.yml"
-      - "ansible/inventory/group_vars/netbird_proxy/**"
      - ".woodpecker/deploy-netbird-proxy.yml"
  - event: deployment
    evaluate: 'CI_PIPELINE_DEPLOY_TARGET == "deploy-netbird-proxy"'
@@ -4,6 +4,9 @@ when:
    branch: master
    path:
      - "ansible/roles/netbird-server/**"
+      # install-netbird-server.yml also applies the netbird-proxy-watchdog role,
+      # so changes to it must redeploy this host too.
+      - "ansible/roles/netbird-proxy-watchdog/**"
      - "ansible/plays/install-netbird-server.yml"
      - ".woodpecker/deploy-netbird-server.yml"
  - event: deployment
@@ -4,7 +4,7 @@ when:

 steps:
  - name: e2e
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
+    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-08-10
    environment:
      SOPS_AGE_KEY:
        from_secret: sops_age_key
@@ -26,12 +26,12 @@ steps:
    # fetch-models.sh response mapping. No cluster, no secrets, no network — the
    # ci-claude image just provides sh + jq. Keeps the flag-injection regression
    # test from rotting (it exercises the exact injection shapes).
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    commands:
      - sh scripts/ci/test-model-selection.sh

  - name: e2e-test
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
+    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-08-10
    environment:
      SOPS_AGE_KEY:
        from_secret: sops_age_key
@@ -49,7 +49,7 @@ steps:
    # (the issue-pick logic needs jq, which only the ci-claude image has);
    # minting a 1h read-only token per run is cheap. Degrades to
    # snapshot-only if the ci-diagnostics ServiceAccount isn't reconciled.
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
+    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-08-10
    environment:
      SOPS_AGE_KEY:
        from_secret: sops_age_key
@@ -62,7 +62,7 @@ steps:
      - bash scripts/ci/make-diag-kubeconfig.sh

  - name: issue-fix-agent
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    environment:
      CLAUDE_CODE_OAUTH_TOKEN:
        from_secret: claude_code_oauth_token
@@ -65,7 +65,7 @@ when:

 steps:
  - name: pr-review-agent
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    environment:
      CLAUDE_CODE_OAUTH_TOKEN:
        from_secret: claude_code_oauth_token
@@ -70,7 +70,7 @@ steps:
    # this step is skipped — otherwise we'd fork-bomb PRs.
    when:
      - branch: master
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    environment:
      CLAUDE_CODE_OAUTH_TOKEN:
        from_secret: claude_code_oauth_token
@@ -30,7 +30,7 @@ when:

 steps:
  - name: update-audit-agent
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
+    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.250
    environment:
      CLAUDE_CODE_OAUTH_TOKEN:
        from_secret: claude_code_oauth_token
@@ -1,198 +0,0 @@
-# Nightly wiki maintenance for the Notes vault (k.renner/notes), replacing the
-# workstation's wiki-maintenance and wiki-consolidate timers: distill the raw/
-# backlog into wiki/ (wiki-distill agent), audit wiki/ with the vault's own
-# script (structural checks, no model), then consolidate (wiki-consolidate
-# agent: verifies a cohort of notes against sources, repos and the live
-# cluster and edits them). The pipeline commits and pushes as Claude Bot; the
-# push webhook then refreshes the wiki service.
-#
-# The agents address the vault as ~/Notes and the repos as ~/source/<repo>, so
-# each agent step links those into the workspace instead of rewriting them.
-when:
-  - event: cron
-    cron: wiki-maintenance
-  # Hand-started runs: `woodpecker pipeline create k.renner/homelab --branch master`.
-  # The typesafe_api_key secret is scoped to cron and manual, not deployment.
-  - event: manual
-
-steps:
-  - name: checkout-notes
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
-    pull: true
-    environment:
-      WIKI_BOT_TOKEN:
-        from_secret: wiki_bot_token
-    commands:
-      - |
-        set -eu
-        # Token only on the wire, never in .git/config. Secrets set from files carry
-        # a trailing newline (run 10220: "could not read Username"); strip CR/LF.
-        TOKEN=$$(printf '%s' "$$WIKI_BOT_TOKEN" | tr -d '\r\n')
-        git -c http.extraheader="AUTHORIZATION: token $$TOKEN" \
-          clone --depth 50 --branch main https://gitea.renner.uno/k.renner/notes.git notes
-        echo "raw backlog: $$(ls notes/raw/*.md | grep -v _index.md | wc -l) file(s)"
-        # Full history for the consolidate agent's `git log` checks; the
-        # workspace clone is shallow and tree-filtered.
-        mkdir -p sources
-        git clone --quiet https://gitea.renner.uno/k.renner/homelab.git sources/homelab
-        git -c http.extraheader="AUTHORIZATION: token $$TOKEN" \
-          clone --quiet https://gitea.renner.uno/k.renner/homeassistant.git sources/homeassistant
-
-  - name: fetch-diag
-    # Read-only live-cluster access for the consolidate agent, same mechanism
-    # as issue-fix.yml: ./.ci-diag/kubectl plus a 1h ci-diagnostics token.
-    image: gitea.renner.uno/gitea_admin/ci-k8s:2026-09-08
-    environment:
-      SOPS_AGE_KEY:
-        from_secret: sops_age_key
-      S3_ACCESS_KEY_ID:
-        from_secret: hetzner_s3_access_key
-      S3_SECRET_ACCESS_KEY:
-        from_secret: hetzner_s3_secret_key
-    commands:
-      - bash scripts/ci/fetch-kubeconfig.sh
-      - bash scripts/ci/make-diag-kubeconfig.sh
-
-  - name: distill
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
-    pull: true
-    environment:
-      CLAUDE_CODE_OAUTH_TOKEN:
-        from_secret: claude_code_oauth_token
-      TYPESAFE_API_KEY:
-        from_secret: typesafe_api_key
-      WIKI_INDEX_BASIC_AUTH:
-        from_secret: wiki_index_basic_auth
-    commands:
-      - |
-        set -eu
-        # An empty queue must not start the agent: its startup context alone
-        # costs more than the whole audit.
-        BACKLOG=$$(ls notes/raw/*.md | grep -v _index.md | wc -l)
-        if [ "$$BACKLOG" -eq 0 ]; then
-          echo "distill: raw/ is empty, skipping the agent"
-          exit 0
-        fi
-        ln -sfn "$$PWD/notes" "$$HOME/Notes"
-        mkdir -p "$$HOME/.config/wiki-index"
-        printf '%s' "$$WIKI_INDEX_BASIC_AUTH" | tr -d '\r\n' > "$$HOME/.config/wiki-index/basic-auth"
-        chmod 600 "$$HOME/.config/wiki-index/basic-auth"
-        "$$HOME/Notes/bin/wiki_rank.py" --probe-index
-        MODEL=$$(sh scripts/ci/pick-model.sh "distill raw source notes into atomic wiki notes: read sources, check the wiki for existing coverage, write or merge notes; careful writing, moderate reasoning")
-        if [ -n "$$MODEL" ]; then echo "distill: selected model $$MODEL"; else echo "distill: using the CLI default model"; fi
-        cd notes
-        cat ../ci/agents/wiki-maintenance/distill.md | claude $${MODEL:+--model $$MODEL} \
-          --agent wiki-distill \
-          --permission-mode acceptEdits \
-          --allowedTools "Read,Edit,Write,Glob,Grep,Bash(/root/Notes/bin/wiki_rank.py:*),Bash(/root/Notes/bin/wiki_ask.py:*),Bash(mv /root/Notes/raw/:*),Bash(mkdir -p /root/Notes/:*),Bash(~/Notes/bin/wiki_rank.py:*),Bash(~/Notes/bin/wiki_ask.py:*),Bash(mv ~/Notes/raw/:*),Bash(mkdir -p ~/Notes/:*)" \
-          --settings ../ci/agents/wiki-maintenance/settings.json \
-          -p
-
-  - name: audit
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
-    pull: true
-    commands:
-      # Findings are report lines in output/, not a pipeline failure.
-      - python3 -B notes/bin/wiki_audit.py --vault "$$PWD/notes"
-
-  - name: consolidate
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
-    pull: true
-    environment:
-      CLAUDE_CODE_OAUTH_TOKEN:
-        from_secret: claude_code_oauth_token
-      TYPESAFE_API_KEY:
-        from_secret: typesafe_api_key
-      WIKI_INDEX_BASIC_AUTH:
-        from_secret: wiki_index_basic_auth
-    commands:
-      - |
-        set -eu
-        ln -sfn "$$PWD/notes" "$$HOME/Notes"
-        ln -sfn "$$PWD/sources" "$$HOME/source"
-        mkdir -p "$$HOME/.config/wiki-index"
-        printf '%s' "$$WIKI_INDEX_BASIC_AUTH" | tr -d '\r\n' > "$$HOME/.config/wiki-index/basic-auth"
-        chmod 600 "$$HOME/.config/wiki-index/basic-auth"
-        if [ -f .ci-diag/kubeconfig ]; then
-          ln -sf "$$PWD/.ci-diag/kubectl" /usr/local/bin/kubectl
-          export KUBECONFIG="$$PWD/.ci-diag/kubeconfig"
-        else
-          echo "consolidate: no ci-diag kubeconfig, live cluster checks unavailable"
-        fi
-        cd notes
-        # JSON output: the agent's report and the run's cost come out of one
-        # record at the end instead of a 45-minute silent log.
-        # One source for the Bash allowlist: the settings file (run 10608: the
-        # agent writes ~/Notes, the rules carried only /root/Notes).
-        BASH_RULES=$$(jq -r '[.permissions.allow[] | select(startswith("Bash("))] | join(",")' ../ci/agents/wiki-maintenance/consolidate-settings.json)
-        rc=0
-        cat ../ci/agents/wiki-maintenance/consolidate.md | claude \
-          --agent wiki-consolidate \
-          --permission-mode acceptEdits \
-          --allowedTools "Read,Edit,Write,Glob,Grep,Agent,$$BASH_RULES" \
-          --settings ../ci/agents/wiki-maintenance/consolidate-settings.json \
-          --output-format json \
-          -p > /tmp/consolidate.json || rc=$$?
-        # A killed or crashed claude leaves no valid JSON; report rc, not a jq error.
-        if [ -s /tmp/consolidate.json ] && jq -e . /tmp/consolidate.json > /dev/null; then
-          jq -r '.result // empty' /tmp/consolidate.json
-          jq -c '{duration_ms, num_turns, total_cost_usd, is_error}' /tmp/consolidate.json
-        else
-          echo "consolidate: no result record (exit $$rc)"; head -c 2000 /tmp/consolidate.json || true
-        fi
-        echo "consolidate: git log --oneline of the run:"
-        git log --oneline origin/main..HEAD
-        exit $$rc
-
-  - name: push-notes
-    image: gitea.renner.uno/gitea_admin/ci-claude:2.1.280
-    pull: true
-    environment:
-      WIKI_BOT_TOKEN:
-        from_secret: wiki_bot_token
-    commands:
-      - |
-        set -eu
-        cd notes
-        git config user.email "claude-bot@renner.uno"
-        git config user.name "Claude Bot"
-        # Only the vault content is committed; anything an agent leaves elsewhere
-        # in the checkout (run 10362: scratch scripts in the root) stays behind.
-        # The consolidate agent commits its own snapshots, so HEAD may already be
-        # ahead of origin/main with a clean tree.
-        git add -A -- wiki raw output .wiki-consolidation-state.json
-        git diff --cached --quiet || git commit -q -m "wiki maintenance $$(date -u +%F): distill + audit + consolidate (CI run $$CI_PIPELINE_NUMBER)"
-        if [ "$$(git rev-list --count origin/main..HEAD)" -eq 0 ]; then
-          echo "wiki-maintenance: no changes"
-          exit 0
-        fi
-        # The workstation pushes main on its own (notes-push.path), so main moves
-        # while the agents run (run 10241: non-fast-forward). The run's work goes
-        # to a branch and a PR that this step merges through Gitea; a conflict
-        # leaves the PR open for a human and fails the run.
-        BRANCH="wiki-maintenance/$$(date -u +%F)-$$CI_PIPELINE_NUMBER"
-        git checkout -q -b "$$BRANCH"
-        TOKEN=$$(printf '%s' "$$WIKI_BOT_TOKEN" | tr -d '\r\n')
-        git -c http.extraheader="AUTHORIZATION: token $$TOKEN" push origin "HEAD:$$BRANCH"
-        API="https://gitea.renner.uno/api/v1/repos/k.renner/notes"
-        jq -n --arg title "wiki maintenance $$(date -u +%F) (CI run $$CI_PIPELINE_NUMBER)" \
-              --arg body "Nightly distill + audit + consolidate. Merged by the pipeline; open only if the merge failed." \
-              --arg head "$$BRANCH" '{title: $$title, body: $$body, head: $$head, base: "main"}' > /tmp/pr.json
-        code=$$(curl -s -o /tmp/pr-resp.json -w '%{http_code}' -X POST \
-          -H "Authorization: token $$TOKEN" -H "Content-Type: application/json" --data @/tmp/pr.json "$$API/pulls")
-        echo "POST pull -> HTTP $$code"
-        [ "$$code" -ge 200 ] && [ "$$code" -lt 300 ] || { cat /tmp/pr-resp.json; exit 1; }
-        NUM=$$(jq -r '.number' /tmp/pr-resp.json)
-        # Merge without delete_branch_after_merge: run 10254 got HTTP 500 from that
-        # call although the merge itself had landed. Verify the PR state instead of
-        # trusting the status code, then drop the branch as a separate, best-effort step.
-        code=$$(curl -s -o /tmp/merge-resp.json -w '%{http_code}' -X POST \
-          -H "Authorization: token $$TOKEN" -H "Content-Type: application/json" \
-          --data '{"Do":"merge"}' "$$API/pulls/$$NUM/merge")
-        echo "POST merge #$$NUM -> HTTP $$code"
-        MERGED=$$(curl -s -H "Authorization: token $$TOKEN" "$$API/pulls/$$NUM" | jq -r '.merged')
-        echo "PR #$$NUM merged: $$MERGED"
-        [ "$$MERGED" = "true" ] || { cat /tmp/merge-resp.json; echo "PR #$$NUM left open"; exit 1; }
-        code=$$(curl -s -o /tmp/del-resp.json -w '%{http_code}' -X DELETE -H "Authorization: token $$TOKEN" \
-          "$$API/branches/$$(printf '%s' "$$BRANCH" | sed 's|/|%2F|g')")
-        echo "DELETE branch $$BRANCH -> HTTP $$code (best effort)"
@@ -1,23 +0,0 @@
---
-# Bouncer key for crowdsec-edge-bouncer, scoped to the proxy hosts.
-#
-# A single encrypted VALUE rather than a whole-file vault like its siblings in
-# group_vars/all and group_vars/edge_host. Deliberate: adding this variable to
-# an existing whole-file vault means decrypting that file first, which exposes
-# every other secret in it for the sake of appending one line. encrypt_string
-# keeps the blast radius at this one value.
-#
-# Same value as OpenBao homelab/infrastructure/crowdsec-edge-bouncer-key, which
-# the crowdsec-edge-bouncer-key ExternalSecret feeds to BOUNCER_KEY_edge on the
-# LAPI. Both sides must move together — and unlike crowdsec-bouncer-key, this
-# one is NOT in the quarterly rotator, because nothing on these hosts re-reads
-# OpenBao on a schedule. Rotating it means updating OpenBao and re-running
-# plays/install-netbird-proxy.yml.
-vault_crowdsec_edge_bouncer_key: !vault |
-          $ANSIBLE_VAULT;1.1;AES256
-          32653764333536316163303838333839616562663862666637626165303837633866643033393239
-          3363376538656236643161386432326437633232643564630a633635306365323537346139626364
-          36613033343938363366306239363638383466396337623333623136313234656366333366623433
-          6437656632373364340a346566343866366433306533353431383064356362346536383136323663
-          39333864643534326433653233303137373162663066633132623931316135323733333863306330
-          6366623463666435636362376238313639666263323163303733
@@ -1,5 +1,5 @@
 ---
-# Deploy the NetBird reverse-proxy HA cluster members (the public edge).
+# Deploy a NetBird reverse-proxy HA cluster member (a second proxy host).
 # Usage: ansible-playbook -i inventory/ plays/install-netbird-proxy.yml
 #
 # Prerequisites:
@@ -7,11 +7,12 @@
 #   listed under [netbird_proxy] in inventory/pve_hosts.yml. Docker is
 #   installed by the (Arch-aware) docker-install role.
 # - DNS A/AAAA for the host's own name (e.g. proxy-1.renner.uno) -> its IP.
-# - Uses the existing vault_netbird_proxy_token (shared by every member) — no
-#   new token to generate; a shared token + same NB_PROXY_DOMAIN forms the
-#   cluster.
-# - AFTER deploy + verify: add this host's A/AAAA to proxy.renner.uno so
-#   traffic round-robins across the cluster.
+# - Uses the existing vault_netbird_proxy_token (shared with VPS-1) — no new
+#   token to generate; a shared token + same NB_PROXY_DOMAIN forms the cluster.
+# - AFTER deploy + verify: add this host's IP to the proxy domain so traffic
+#   round-robins across the cluster. Repoint proxy.renner.uno away from its
+#   CNAME to netbird.renner.uno (management host) to A/AAAA records listing
+#   BOTH proxy IPs (VPS-1 + this host).

 - name: Deploy NetBird reverse-proxy cluster member
  hosts: netbird_proxy
@@ -25,7 +26,3 @@
    - docker-install
    - netbird-proxy
    - netbird-proxy-watchdog
-    # Enforces the cluster's CrowdSec decisions here, where public traffic
-    # enters. Deployed in dry-run: it logs what it would block and creates no
-    # nftables rules until crowdsec_bouncer_mode is flipped per host.
-    - crowdsec-edge-bouncer
@@ -18,3 +18,4 @@
    - common
    - docker-install
    - netbird-server
+    - netbird-proxy-watchdog
@@ -43,10 +43,10 @@
 #
 # WHAT IS NOT, and must not be added:
 #
-#   vault_netbird_proxy_token    LOOKS symmetric — it is in proxy.env on the
-#                                proxies, an Ansible template — but the
-#                                management server does not validate it from
-#                                any file. It is registered
+#   vault_netbird_proxy_token    LOOKS symmetric — it is in proxy.env on both
+#                                the server and the proxies, all Ansible
+#                                templates — but the management server does not
+#                                validate it from any of them. It is registered
 #                                in the management DATABASE. Rotating it changes
 #                                only what the proxies present, so every proxy
 #                                is locked out: "proxy auth failed: too many
@@ -1,56 +0,0 @@
---
-# CrowdSec firewall bouncer on the NetBird proxy hosts.
-#
-# Enforces the decisions the in-cluster CrowdSec already detects, at the point
-# where public traffic actually enters: these hosts terminate TLS for every
-# *.renner.uno name. A drop here never reaches the mesh, the relay or the
-# cluster.
-#
-# Detection stays in the cluster. The reverse-proxy produces no parseable
-# access log and there is no CrowdSec collection for it, while Traefik has a
-# working one — and every scenario in use needs HTTP detail (paths, status
-# codes, user agents) that does not exist at this layer.
-
-# Pinned. A release tarball URL is immutable, and this checksum is of the
-# amd64 asset for exactly this tag.
-crowdsec_bouncer_version: v0.0.36
-crowdsec_bouncer_sha256: f86e4b72693549d99f40a9402abefb894108f047a3fbe6e72fada25ee17ce88b
-
-# The LAPI over the public path these hosts already serve, see
-# kubernetes/infrastructure/crowdsec/lapi-ingress.yaml. The proxies cannot
-# reach it over the mesh: the reverse-proxy container runs non-root, so its
-# embedded NetBird client terminates the tunnel in userspace and the tunnel
-# belongs to that process rather than to the host netns.
-#
-# No ca_cert_path: Traefik serves a Let's Encrypt certificate for this name,
-# and the bouncer adds ca_cert_path to the system pool rather than replacing
-# it — the system pool already trusts LE.
-crowdsec_bouncer_api_url: https://crowdsec.renner.uno/
-crowdsec_bouncer_api_key: "{{ vault_crowdsec_edge_bouncer_key }}"
-
-# Enforcing. The bouncer creates two nftables tables (ip crowdsec, ip6
-# crowdsec6) holding the decisions it pulled, touches no existing chain and
-# sets no chain policy — SSH and :443 stay open unless the source is banned.
-#
-# `dry-run` is the other value: it pulls and logs what it would block without
-# creating a single rule. That is what a new host should be given before it
-# enforces, and what an existing one can be dropped back to, per host through
-# inventory/host_vars, without disarming the rest.
-#
-# Worth knowing before arming anything new: the first bouncer on a path
-# enforces the whole CAPI set at once — around 15k addresses today — so the
-# false-positive window belongs before the switch, not after it. And because
-# this chain sits on the INPUT hook while these hosts run no other firewall,
-# there is no conntrack rule accepting established flows ahead of it: a banned
-# address is dropped even when it is answering our own outbound connection.
-# That is why capi_whitelists.yaml carries the proxies and netbird.renner.uno.
-crowdsec_bouncer_mode: nftables
-
-crowdsec_bouncer_update_frequency: 10s
-crowdsec_bouncer_log_level: info
-
-# Bound to loopback deliberately. These hosts are public and have no metrics
-# collection path at all today — opening :60601 to the internet to gain a
-# scrape target is the wrong trade. Revisit when host metrics exist.
-crowdsec_bouncer_metrics_listen: 127.0.0.1
-crowdsec_bouncer_metrics_port: 60601
@@ -1,10 +0,0 @@
---
-# A restart drops both nftables tables and rebuilds them from the next
-# successful LAPI pull, so there is a short window with no blocklist. That is
-# the documented behaviour of a clean shutdown, not a defect — and it is why
-# retry_initial_connect matters.
- name: Restart crowdsec firewall bouncer
-  ansible.builtin.systemd:
-    name: crowdsec-firewall-bouncer.service
-    state: restarted
-    daemon_reload: true
@@ -1,111 +0,0 @@
---
-# See defaults/main.yml for why this exists and why it starts inert.
-#
-# The static release tarball, not a distribution package: the AUR package
-# `crowdsec-firewall-bouncer-nftables` depends on `crowdsec` and would pull the
-# whole Security Engine onto hosts that only need the bouncer, the group is
-# mixed (proxy-1 Arch, VPS-1 Debian), and the AUR would be a new pattern this
-# repo does not otherwise have. The tarball is the same artefact the AUR
-# package downloads.
-
- name: Ensure nftables is present (the enforcing backend)
-  ansible.builtin.package:
-    name: nftables
-    state: present
-
- name: Ensure the bouncer config directory exists
-  ansible.builtin.file:
-    path: /etc/crowdsec/bouncers
-    state: directory
-    mode: "0755"
-
-# Fetched on the CONTROLLER and pushed, not pulled by the host.
-#
-# proxy-1 is IPv6-only, and neither github.com nor objects.githubusercontent.com
-# publishes an AAAA record — a get_url on the host fails with "Network is
-# unreachable" and no amount of retrying fixes it. Mirroring into
-# gitea.renner.uno is no help either: its public records are the internal LB
-# VIP (10.2.99.86 / fd99::86), so it is unreachable from outside the LAN. And
-# upstream ships no container image for this bouncer, which is expected —
-# it manipulates host netfilter.
-#
-# The CI runner has v4, so it does the fetching. The checksum is verified
-# there, before anything is copied to a host.
- name: Fetch the pinned bouncer release on the controller
-  ansible.builtin.get_url:
-    url: >-
-      https://github.com/crowdsecurity/cs-firewall-bouncer/releases/download/{{
-      crowdsec_bouncer_version }}/crowdsec-firewall-bouncer-linux-amd64.tgz
-    dest: /tmp/crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}.tgz
-    checksum: sha256:{{ crowdsec_bouncer_sha256 }}
-    mode: "0644"
-  delegate_to: localhost
-  become: false
-  run_once: true
-
-# Only the download had to move to the controller — the extraction stays on
-# the host. The ci-ansible image ships a tar that ansible rejects ("detected as
-# tar type None. GNU tar required"), while both host distributions have GNU
-# tar. Shipping the archive rather than the unpacked binary also keeps the
-# artefact that the checksum was verified against intact all the way to the
-# host.
- name: Copy the bouncer archive to the host
-  ansible.builtin.copy:
-    src: /tmp/crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}.tgz
-    dest: /opt/crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}.tgz
-    mode: "0644"
-
-# Only the binary. The tarball also carries install.sh / upgrade.sh /
-# uninstall.sh, which manage systemd and config themselves — that is this
-# role's job, and letting them run would leave two owners for the same files.
- name: Unpack the bouncer binary
-  ansible.builtin.unarchive:
-    src: /opt/crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}.tgz
-    dest: /opt
-    remote_src: true
-    include:
-      - crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}/crowdsec-firewall-bouncer
-    creates: /opt/crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}/crowdsec-firewall-bouncer
-
- name: Install the bouncer binary
-  ansible.builtin.copy:
-    src: /opt/crowdsec-firewall-bouncer-{{ crowdsec_bouncer_version }}/crowdsec-firewall-bouncer
-    dest: /usr/local/bin/crowdsec-firewall-bouncer
-    remote_src: true
-    mode: "0755"
-  notify: Restart crowdsec firewall bouncer
-
-# 0600: this file carries the LAPI key.
- name: Deploy the bouncer configuration
-  ansible.builtin.template:
-    src: crowdsec-firewall-bouncer.yaml.j2
-    dest: /etc/crowdsec/bouncers/crowdsec-firewall-bouncer.yaml
-    mode: "0600"
-  notify: Restart crowdsec firewall bouncer
-
- name: Deploy the bouncer systemd unit
-  ansible.builtin.template:
-    src: crowdsec-firewall-bouncer.service.j2
-    dest: /etc/systemd/system/crowdsec-firewall-bouncer.service
-    mode: "0644"
-  notify: Restart crowdsec firewall bouncer
-
- name: Enable and start the bouncer
-  ansible.builtin.systemd:
-    name: crowdsec-firewall-bouncer.service
-    enabled: true
-    state: started
-    daemon_reload: true
-
- name: Report how to read it
-  ansible.builtin.debug:
-    msg: |
-      crowdsec-firewall-bouncer deployed in mode "{{ crowdsec_bouncer_mode }}".
-
-      What it would block:  journalctl -u crowdsec-firewall-bouncer -f
-      What it does block:   nft list table ip crowdsec ; nft list table ip6 crowdsec6
-      Metrics:              curl -s {{ crowdsec_bouncer_metrics_listen }}:{{ crowdsec_bouncer_metrics_port }}/metrics
-
-      Rollback: systemctl stop crowdsec-firewall-bouncer. A clean stop deletes
-      both nftables tables by itself. Only after a kill -9 do they need
-      removing by hand: nft delete table ip crowdsec ; nft delete table ip6 crowdsec6
@@ -1,24 +0,0 @@
-# Managed by Ansible (crowdsec-edge-bouncer role) — do not edit.
-#
-# Upstream's shipped unit orders itself After=crowdsec.service. There is no
-# local Security Engine on these hosts — the LAPI is in the cluster — so that
-# dependency is dropped; nss-lookup instead, because api_url is a public name
-# that has to resolve before the first connect.
-[Unit]
-Description=CrowdSec firewall bouncer (edge enforcement for the NetBird proxy)
-After=network-online.target nss-lookup.target
-Wants=network-online.target
-
-[Service]
-Type=notify
-# -t validates the config and exits non-zero on a bad one, so a broken config
-# fails the start instead of taking the running rules down with it.
-ExecStartPre=/usr/local/bin/crowdsec-firewall-bouncer -c /etc/crowdsec/bouncers/crowdsec-firewall-bouncer.yaml -t
-ExecStart=/usr/local/bin/crowdsec-firewall-bouncer -c /etc/crowdsec/bouncers/crowdsec-firewall-bouncer.yaml
-Restart=always
-RestartSec=10
-LimitNOFILE=65536
-KillMode=mixed
-
-[Install]
-WantedBy=multi-user.target
@@ -1,55 +0,0 @@
-# Managed by Ansible (crowdsec-edge-bouncer role) — do not edit.
-mode: {{ crowdsec_bouncer_mode }}
-update_frequency: {{ crowdsec_bouncer_update_frequency }}
-
-# stdout, so journald owns the log and there is no second rotation policy to
-# keep in sync. The shipped example writes to /var/log; that only matters for
-# hosts without journald.
-log_mode: stdout
-log_level: {{ crowdsec_bouncer_log_level }}
-
-api_url: {{ crowdsec_bouncer_api_url }}
-api_key: {{ crowdsec_bouncer_api_key }}
-
-# Without this the process EXITS when the LAPI is unreachable at startup —
-# there is no retry on the first connect. And because a clean shutdown deletes
-# the nftables tables, the systemd restart loop would leave the host with no
-# blocklist at all until the first successful connect. Present in the code,
-# absent from the shipped example config.
-retry_initial_connect: true
-
-disable_ipv4: false
-disable_ipv6: false
-deny_action: DROP
-
-supported_decisions_types:
-  - ban
-
-# Separate tables per family; the bouncer never touches an existing table or
-# chain, and never sets a chain policy. Priority -10 puts it ahead of anything
-# at the default filter priority.
-nftables:
-  ipv4:
-    enabled: true
-    set-only: false
-    table: crowdsec
-    chain: crowdsec-chain
-    priority: -10
-  ipv6:
-    enabled: true
-    set-only: false
-    table: crowdsec6
-    chain: crowdsec6-chain
-    priority: -10
-
-# input is the traffic this exists for: the reverse-proxy is host-networked and
-# owns :443, so its packets traverse INPUT. forward covers the bridge-networked
-# http-redirect sidecar, whose packets never do.
-nftables_hooks:
-  - input
-  - forward
-
-prometheus:
-  enabled: true
-  listen_addr: {{ crowdsec_bouncer_metrics_listen }}
-  listen_port: {{ crowdsec_bouncer_metrics_port }}
@@ -26,16 +26,8 @@ edge_open3e_poll_seconds: 30
 # entries when added), so the multi-frame read is cheap at a 30 s poll.
 # Records faults/info events only (NormalOperation, LagDeviceDtcReported,
 # LegionellaProtectionTemperatureReached) — NOT setpoint or program changes.
-# onebase.*: the 29 paired ZigBee slots on the One Base gateway (ECU 0x6c3) --
-# 8 radiator valves, 9 climate sensors, 9 underfloor actuators, 3 floor
-# thermostats. Per-device room temperature, active setpoint, humidity, valve
-# opening/demand, battery and RSSI; the local replacement for the ViCare cloud
-# room sensors. Needs the onebase overlay (files/open3e-onebase.py) -- the stock
-# codec declares 57 bytes and a paired slot answers with 68.
-# Only CurrentValues are polled. The room and device registries (1884.., 2084..)
-# are static configuration and are read on demand, not every 30 s.
 edge_open3e_datapoints: >-
-  vitocal.259,vitocal.260,vitocal.268,vitocal.269,vitocal.271,vitocal.274,vitocal.286,vitocal.288,vitocal.318,vitocal.321,vitocal.324,vitocal.335,vitocal.336,vitocal.396,vitocal.402,vitocal.403,vitocal.476,vitocal.477,vitocal.543,vitocal.548,vitocal.565,vitocal.1007,vitocal.1043,vitocal.1190,vitocal.1211,vitocal.1391,vitocal.1416,vitocal.1417,vitocal.2320,vitocal.2346,vitocal.2350,vitocal.2351,vitocal.2352,vitocal.2369,vitocal.2370,vitocal.2486,vitocal.2487,vitocal.2488,vitocal.2543,vitocal.2560,vitocal.2622,vitocal.2624,vitocal.2625,vitocal.2735,vitocal.3016,vx3.266,vx3.378,vx3.379,vx3.380,vx3.535,vx3.1603,vx3.1664,vx3.1690,vx3.1801,vx3.1802,vx3.1810,vx3.1828,vx3.1830,vx3.1577,vx3.1578,vx3.1579,vx3.1580,vx3.1581,vx3.1582,vx3.1834,vx3.1836,vx3.2214,vx3.2240,vx3.3085,vx3.3086,vx3.3087,vx3.3088,vx3.3089,vx3.3090,vcmu.320,vcmu.322,vcmu.325,vcmu.355,vcmu.389,vcmu.391,vcmu.1772,vcmu.1773,vcmu.1775,vcmu.1776,vcmu.2256,vcmu.2259,vcmu.2333,vcmu.2334,vcmu.2569,vcmu.2571,vitocal.2791,onebase.2086,onebase.2089,onebase.2092,onebase.2095,onebase.2098,onebase.2101,onebase.2104,onebase.2107,onebase.2110,onebase.2113,onebase.2116,onebase.2119,onebase.2122,onebase.2125,onebase.2128,onebase.2131,onebase.2134,onebase.2137,onebase.2140,onebase.2143,onebase.2262,onebase.2265,onebase.2268,onebase.2271,onebase.2274,onebase.2277,onebase.2280,onebase.2283,onebase.2286
+  vitocal.259,vitocal.260,vitocal.268,vitocal.269,vitocal.271,vitocal.274,vitocal.286,vitocal.288,vitocal.318,vitocal.321,vitocal.324,vitocal.335,vitocal.336,vitocal.396,vitocal.402,vitocal.403,vitocal.476,vitocal.477,vitocal.543,vitocal.548,vitocal.565,vitocal.1007,vitocal.1043,vitocal.1190,vitocal.1211,vitocal.1391,vitocal.1416,vitocal.1417,vitocal.2320,vitocal.2346,vitocal.2350,vitocal.2351,vitocal.2352,vitocal.2369,vitocal.2370,vitocal.2486,vitocal.2487,vitocal.2488,vitocal.2543,vitocal.2560,vitocal.2622,vitocal.2624,vitocal.2625,vitocal.2735,vitocal.3016,vx3.266,vx3.378,vx3.379,vx3.380,vx3.535,vx3.1603,vx3.1664,vx3.1690,vx3.1801,vx3.1802,vx3.1810,vx3.1828,vx3.1830,vx3.1577,vx3.1578,vx3.1579,vx3.1580,vx3.1581,vx3.1582,vx3.1834,vx3.1836,vx3.2214,vx3.2240,vx3.3085,vx3.3086,vx3.3087,vx3.3088,vx3.3089,vx3.3090,vcmu.320,vcmu.322,vcmu.325,vcmu.355,vcmu.389,vcmu.391,vcmu.1772,vcmu.1773,vcmu.1775,vcmu.1776,vcmu.2256,vcmu.2259,vcmu.2333,vcmu.2334,vcmu.2569,vcmu.2571,vitocal.2791
 # ECU addresses; -dev selects codec set, tx selects the UDS address.
 # vcmu (refrigerant-circuit controller) has no dedicated codec set —
 # omitting dpList makes open3e fall back to the generic datapoint list,
@@ -46,9 +38,6 @@ edge_open3e_devices:
  # files/open3e-vx3plus.py) — die Stock-vx3-Liste filtert sie sonst weg.
  vx3: { tx: "0x6a1", dpList: "/config/vx3plus.py" }
  vcmu: { tx: "0x68c" }
-  # One Base ZigBee gateway. Overlay re-declares the paired ZigBee slots at
-  # their real 68-byte length (see files/open3e-onebase.py).
-  onebase: { tx: "0x6c3", dpList: "/config/onebase.py" }

 edge_e3oncan_gap_seconds: 10
 edge_e3oncan_mqtt_format: "{device}_{didNumber:04d}_{didName}"
@@ -1,54 +0,0 @@
-# Managed by ansible (roles/heating-can). Datapoint overlay for the One Base
-# ZigBee gateway on ECU 0x6c3 (BACKENDGATEWAY).
-#
-# Upstream declares ZigBee<N>DeviceCurrentValues as 57 bytes. A PAIRED device
-# slot on this gateway answers with 68: the 57 known bytes plus the parent
-# device serial and three trailing bytes. udsoncan reads only the declared
-# length and then treats the next two bytes -- the first half of that serial --
-# as a foreign data identifier, so every paired slot fails to decode with
-# "Server returned values for data identifier 0x.... that was not requested".
-# The bogus identifier differs per device because the serial does.
-#
-# This overlay re-declares the 29 paired slots at 68 bytes, reusing upstream's
-# sub-type list so the per-device-type O3ESwitch (climate sensor / TRV / floor
-# thermostat / actuator, incl. DeviceTemperatureSetpoint) is inherited rather
-# than copied.
-#
-# EMPTY slots answer with 57 and decode against the stock codec, so they are
-# left alone and are not polled. A slot that is later unpaired therefore starts
-# reporting a decode error on its topic instead of a value.
-#
-# Verified on ECU 0x6c3, 2026-09-17: 29 paired slots, all 68 bytes.
-from open3e.Open3Ecodecs import O3EComplexType, RawCodec
-from open3e.Open3Edatapoints import dataIdentifiers as _master
-
-# ZigBee<N>DeviceCurrentValues of the paired slots: devices 1-20 run 2086..2143
-# in steps of 3, devices 21-29 continue at 2262..2286.
-_PAIRED_CURRENT_VALUES = [2086 + 3 * n for n in range(20)] + [
-    2262 + 3 * n for n in range(9)
-]
-
-
-def _with_parent_serial(did):
-    stock = _master["dids"][did]
-    return O3EComplexType(
-        68,
-        stock.id,
-        list(stock.subTypes)
-        + [
-            RawCodec(8, "ParentDeviceSerialnumber"),
-            RawCodec(2, "Unknown_65_66"),
-            RawCodec(1, "Unknown_67"),
-        ],
-        desc=stock.desc,
-        info=stock.info,
-        acc=stock.acc,
-    )
-
-
-dataIdentifiers = {
-    "dids": {
-        **_master["dids"],
-        **{did: _with_parent_serial(did) for did in _PAIRED_CURRENT_VALUES},
-    }
-}
@@ -129,13 +129,6 @@
    mode: "0644"
  notify: Restart heating-can

- name: Deploy open3e One Base datapoint overlay (68-byte ZigBee slots)
-  ansible.builtin.copy:
-    src: open3e-onebase.py
-    dest: "{{ heating_can_base_dir }}/onebase.py"
-    mode: "0644"
-  notify: Restart heating-can
-
 # --- Compose project ---------------------------------------------------------
 - name: Deploy the broker credential env file
  ansible.builtin.template:
@@ -17,7 +17,6 @@ services:
    volumes:
      - ./devices.json:/config/devices.json:ro
      - ./vx3plus.py:/config/vx3plus.py:ro
-      - ./onebase.py:/config/onebase.py:ro
    env_file:
      # Broker credential, 0600 -- see heating-can.env.j2.
      - heating-can.env
@@ -1,5 +1,5 @@
 ---
-# Self-healing watchdog for the NetBird reverse proxy on the proxy hosts.
+# Self-healing watchdog for the NetBird reverse proxy (VPS-1 and proxy-1).
 #
 # Why it exists: the proxy's embedded NetBird client pins each proxied
 # domain's route to a specific routing-peer identity. The netbird-ingress
@@ -20,7 +20,9 @@ netbird_watchdog_probe_hosts:
  - home.renner.uno
  - immich.renner.uno

-# Address the probe connects to. The proxy binds :443 on the host directly.
+# Address the probe connects to. 127.0.0.1 reaches the local proxy on both
+# members: VPS-1 routes it via Traefik's HostSNI(*) TCP passthrough, proxy-1
+# binds :443 directly.
 netbird_watchdog_probe_address: 127.0.0.1

 netbird_watchdog_container: netbird-proxy
@@ -30,13 +32,8 @@ netbird_watchdog_fail_threshold: 3

 # Minimum seconds between watchdog-issued restarts, so a genuinely broken
 # backend can't put the proxy into a restart loop. Shared by both the URL-probe
-# watchdog and the flood watchdog below. One minute: the route-manager wedge
-# recurs on every routing-peer loss, and rolling node reboots lose one peer
-# every few minutes; a suppressed restart there leaves the wedge standing
-# until the URL probe's three rounds fire (3.5 min of 000 on every domain,
-# measured 2026-09-07). A minute still caps a loop at one restart per
-# probe interval.
-netbird_watchdog_cooldown_seconds: 60
+# watchdog and the flood watchdog below.
+netbird_watchdog_cooldown_seconds: 600

 # Log fingerprint of the reverse-proxy route-manager wedge, matched as a plain
 # substring in the proxy's docker logs by the flood watchdog. On this build the
@@ -1,7 +1,7 @@
 ---
 # Deploys the reverse-proxy watchdog (see defaults/main.yml for why).

- name: Ensure curl is installed (probe dependency)
+- name: Ensure curl is installed (probe dependency; proxy-1 is Arch, VPS-1 Debian)
  ansible.builtin.package:
    name: curl
    state: present
@@ -1,8 +1,8 @@
 ---
 # NetBird reverse-proxy HA cluster member (proxy-only host, e.g. proxy-1).
-# Every member uses the same NB_PROXY_DOMAIN + token. NetBird load-balances/
-# fails the cluster over via DNS round-robin on the proxy domain (see
-# plays/install-netbird-proxy.yml).
+# Joins the same proxy cluster as the VPS-1 proxy: identical NB_PROXY_DOMAIN
+# + the same token. NetBird load-balances/fails the cluster over via DNS
+# round-robin on the proxy domain (see plays/install-netbird-proxy.yml).

 # MUST match the netbird-server role's netbird_version — the reverse-proxy
 # release line has to track the management server it enrolls against.
@@ -11,13 +11,11 @@ netbird_version: "0.73.2"
 # Management server this proxy enrolls against (public API, TLS).
 netbird_server_host: netbird.renner.uno

-# Proxy domain — MUST be identical across every cluster member, and must NOT
-# be a subdomain of the management domain (netbird.renner.uno) per NetBird
-# docs (TLS/routing conflicts).
+# Proxy domain — MUST be identical across every cluster member.
 netbird_proxy_domain: "proxy.renner.uno"

-# A shared NB_PROXY_TOKEN across members with the same NB_PROXY_DOMAIN is a
-# valid HA cluster, so one token serves every member.
+# Reuse VPS-1's proxy token — a shared NB_PROXY_TOKEN across members with the
+# same NB_PROXY_DOMAIN is a valid HA cluster, so no separate token is needed.
 netbird_proxy_token: "{{ vault_netbird_proxy_token }}"

 # Relay auth secret — MUST be identical across every relay instance and the
@@ -1,7 +1,8 @@
 # NetBird reverse-proxy — HA cluster member on a proxy-only host.
-# Just the reverse-proxy container, bound directly to :443. No Traefik: the
-# proxy owns the port and terminates TLS + answers its own tls-alpn-01 ACME
-# challenges directly.
+# Just the reverse-proxy container, bound directly to :443. No Traefik:
+# unlike VPS-1 (where the proxy shares :443 with management/signal/relay via
+# SNI routing), this host runs nothing else, so the proxy owns the port and
+# terminates TLS + answers its own tls-alpn-01 ACME challenges directly.
 services:
  proxy:
    image: netbirdio/reverse-proxy:{{ netbird_version }}
@@ -23,3 +23,9 @@ netbird_datastore_key: "{{ vault_netbird_datastore_key }}"

 # Relay secret for WebSocket relay authentication
 netbird_relay_secret: "{{ vault_netbird_relay_secret }}"
+
+# Reverse proxy configuration
+# IMPORTANT: Proxy domain must NOT be a subdomain of the management domain
+# (netbird.renner.uno) per NetBird docs - causes TLS/routing conflicts
+netbird_proxy_domain: "proxy.renner.uno"
+netbird_proxy_token: "{{ vault_netbird_proxy_token }}"
@@ -16,3 +16,7 @@
 - name: Restart netbird coturn
  ansible.builtin.shell: |
    cd /opt/netbird && docker compose up -d --force-recreate coturn
+
+- name: Restart netbird proxy
+  ansible.builtin.shell: |
+    cd /opt/netbird && docker compose up -d --force-recreate proxy
@@ -36,6 +36,7 @@
    - /opt/netbird/management
    - /opt/netbird/signal
    - /opt/netbird/letsencrypt
+    - /opt/netbird/proxy/certs

 - name: Deploy docker-compose.yml
  ansible.builtin.template:
@@ -58,83 +59,23 @@
    mode: "0644"
  notify: Restart netbird coturn

-# The public edge is the proxy hosts (plays/install-netbird-proxy.yml); this
-# host runs no reverse proxy. The tasks below remove the one it used to run,
-# its watchdogs first so they cannot restart it mid-removal, and can go once
-# the role has run against every host that ever carried it.
- name: Check for reverse-proxy watchdog units
-  ansible.builtin.stat:
-    path: "/etc/systemd/system/{{ item }}"
-  loop:
-    - netbird-proxy-watchdog.timer
-    - netbird-proxy-flood-watchdog.service
-  register: _proxy_watchdog_units
-
- name: Stop and disable reverse-proxy watchdogs
-  ansible.builtin.systemd:
-    name: "{{ item.item }}"
-    state: stopped
-    enabled: false
-  loop: "{{ _proxy_watchdog_units.results }}"
-  loop_control:
-    label: "{{ item.item }}"
-  when: item.stat.exists
-
- name: Remove reverse-proxy watchdog units and scripts
-  ansible.builtin.file:
-    path: "{{ item }}"
-    state: absent
-  loop:
-    - /etc/systemd/system/netbird-proxy-watchdog.timer
-    - /etc/systemd/system/netbird-proxy-watchdog.service
-    - /etc/systemd/system/netbird-proxy-flood-watchdog.service
-    - /usr/local/bin/netbird-proxy-watchdog
-    - /usr/local/bin/netbird-proxy-flood-watchdog
-    - /run/netbird-proxy-watchdog
-  register: _proxy_watchdog_files
-
- name: Reload systemd after removing the watchdog units
-  ansible.builtin.systemd:
-    daemon_reload: true
-  when: _proxy_watchdog_files.changed
+- name: Deploy proxy.env
+  ansible.builtin.template:
+    src: proxy.env.j2
+    dest: /opt/netbird/proxy.env
+    mode: "0644"
+  notify: Restart netbird proxy

 - name: Pull NetBird container images
  ansible.builtin.shell: |
    cd /opt/netbird && docker compose pull
  changed_when: false

-# --remove-orphans removes containers of services no longer in the compose
-# file (the reverse proxy).
 - name: Start NetBird containers
  ansible.builtin.shell: |
-    cd /opt/netbird && docker compose up -d --remove-orphans
+    cd /opt/netbird && docker compose up -d
  register: docker_compose_result
-  changed_when: >-
-    'Started' in docker_compose_result.stdout or
-    'Creating' in docker_compose_result.stdout or
-    'Removed' in docker_compose_result.stdout or
-    'Removed' in docker_compose_result.stderr
-
- name: Remove reverse-proxy volumes
-  ansible.builtin.shell: |
-    set -eu
-    for v in netbird_netbird_proxy_certs netbird_netbird_proxy_data netbird_netbird_proxy_state; do
-      if [ -n "$(docker volume ls -q --filter "name=^${v}$")" ]; then
-        docker volume rm "$v"
-        echo "removed $v"
-      fi
-    done
-  register: _proxy_volumes
-  changed_when: "'removed' in _proxy_volumes.stdout"
-
- name: Remove reverse-proxy config and state
-  ansible.builtin.file:
-    path: "{{ item }}"
-    state: absent
-  loop:
-    - /opt/netbird/proxy.env
-    - /opt/netbird/proxy
-    - /opt/netbird/proxy-state
+  changed_when: "'Started' in docker_compose_result.stdout or 'Creating' in docker_compose_result.stdout"

 # The management store has no cluster-side backup. Pulling one put a root
 # key on this host that the cluster held, and this server's own state is
@@ -187,9 +187,10 @@ services:
      - "traefik.http.routers.netbird-management.service=netbird-management"
      - "traefik.http.services.netbird-management.loadbalancer.server.port=33073"
      - "traefik.http.services.netbird-management.loadbalancer.server.scheme=h2c"
-      # gRPC reverse-proxy service — the proxy hosts (proxy-1, proxy-2)
-      # connect here via NB_PROXY_MANAGEMENT_ADDRESS=https://host:443 and get
-      # a 404 without it. Same backend (h2c 33073).
+      # gRPC reverse-proxy service — remote reverse-proxy instances (e.g.
+      # proxy-1) connect here via NB_PROXY_MANAGEMENT_ADDRESS=https://host:443.
+      # The local proxy uses management:33073 directly, so it never needed
+      # this route; remote ones get a 404 without it. Same backend (h2c 33073).
      - "traefik.http.routers.netbird-proxyservice.rule=Host(`{{ netbird_server_host }}`) && PathPrefix(`/management.ProxyService/`)"
      - "traefik.http.routers.netbird-proxyservice.entrypoints=websecure"
      - "traefik.http.routers.netbird-proxyservice.tls=true"
@@ -220,6 +221,37 @@ services:
        max-size: "500m"
        max-file: "2"

+  # Reverse Proxy Service (per docs.netbird.io/selfhosted/migration/enable-reverse-proxy)
+  proxy:
+    image: netbirdio/reverse-proxy:{{ netbird_version }}
+    container_name: netbird-proxy
+    restart: unless-stopped
+    networks:
+      - netbird
+    depends_on:
+      - management
+    env_file:
+      - ./proxy.env
+    volumes:
+      - netbird_proxy_certs:/certs
+      - /opt/netbird/proxy-state:/var/lib/netbird
+    labels:
+      - "traefik.enable=true"
+      - "traefik.tcp.routers.proxy-passthrough.entrypoints=websecure"
+      - "traefik.tcp.routers.proxy-passthrough.rule=HostSNI(`*`)"
+      - "traefik.tcp.routers.proxy-passthrough.tls.passthrough=true"
+      - "traefik.tcp.routers.proxy-passthrough.service=proxy-tls"
+      - "traefik.tcp.routers.proxy-passthrough.priority=1"
+      - "traefik.tcp.services.proxy-tls.loadbalancer.server.port=8443"
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "500m"
+        max-file: "2"
+
 networks:
  netbird:
    driver: bridge
+
+volumes:
+  netbird_proxy_certs:
@@ -0,0 +1,8 @@
+NB_PROXY_DOMAIN={{ netbird_proxy_domain }}
+NB_PROXY_TOKEN={{ netbird_proxy_token }}
+NB_PROXY_MANAGEMENT_ADDRESS=http://management:33073
+NB_PROXY_ALLOW_INSECURE=true
+NB_PROXY_ADDRESS=:8443
+NB_PROXY_ACME_CERTIFICATES=true
+NB_PROXY_ACME_CHALLENGE_TYPE=tls-alpn-01
+NB_PROXY_CERTIFICATE_DIRECTORY=/certs
@@ -1,120 +0,0 @@
-{
-  "//": "Permission scope for the wiki-consolidate agent in CI: read the vault and the",
-  "//2": "source repos cloned under /root/source, read the live cluster through the",
-  "//3": "ci-diagnostics kubeconfig, edit wiki/, output/ and the rotation state only.",
-  "//4": "Deny rules win; git and kubectl are read-only subcommands, no network tools.",
-  "permissions": {
-    "allow": [
-      "Read(//root/Notes/**)",
-      "Read(//root/source/**)",
-      "Edit(//root/Notes/wiki/**)",
-      "Edit(//root/Notes/output/**)",
-      "Edit(//root/Notes/.wiki-consolidation-state.json)",
-      "Bash(/root/Notes/bin/wiki-snapshot.sh:*)",
-      "Bash(/root/Notes/bin/wiki_rank.py:*)",
-      "Bash(/root/Notes/bin/wiki_ask.py:*)",
-      "Bash(~/Notes/bin/wiki-snapshot.sh:*)",
-      "Bash(~/Notes/bin/wiki_rank.py:*)",
-      "Bash(~/Notes/bin/wiki_ask.py:*)",
-      "Bash(~/Notes/bin/wiki-delete.sh:*)",
-      "Bash(git -C /root/Notes log:*)",
-      "Bash(git -C /root/Notes show:*)",
-      "Bash(git -C /root/Notes diff:*)",
-      "Bash(git -C /root/Notes status:*)",
-      "Bash(git -C /root/Notes rev-parse:*)",
-      "Bash(git -C /root/Notes ls-files:*)",
-      "Bash(git -C ~/Notes log:*)",
-      "Bash(git -C ~/Notes show:*)",
-      "Bash(git -C ~/Notes diff:*)",
-      "Bash(git -C ~/Notes status:*)",
-      "Bash(git -C ~/Notes rev-parse:*)",
-      "Bash(git -C ~/Notes ls-files:*)",
-      "Bash(kubectl get:*)",
-      "Bash(kubectl describe:*)",
-      "Bash(kubectl version:*)",
-      "Bash(kubectl api-resources:*)",
-      "Bash(kubectl explain:*)",
-      "Bash(git log:*)",
-      "Bash(git show:*)",
-      "Bash(git status:*)",
-      "Bash(git diff:*)",
-      "Bash(git rev-parse:*)",
-      "Bash(git ls-files:*)",
-      "Bash(git -C /root/source/homelab log:*)",
-      "Bash(git -C /root/source/homelab show:*)",
-      "Bash(git -C /root/source/homelab diff:*)",
-      "Bash(git -C /root/source/homelab status:*)",
-      "Bash(git -C /root/source/homelab rev-parse:*)",
-      "Bash(git -C /root/source/homelab ls-files:*)",
-      "Bash(git -C /root/source/homeassistant log:*)",
-      "Bash(git -C /root/source/homeassistant show:*)",
-      "Bash(git -C /root/source/homeassistant diff:*)",
-      "Bash(git -C /root/source/homeassistant status:*)",
-      "Bash(git -C /root/source/homeassistant rev-parse:*)",
-      "Bash(git -C /root/source/homeassistant ls-files:*)",
-      "Bash(git -C ~/source/homelab log:*)",
-      "Bash(git -C ~/source/homelab show:*)",
-      "Bash(git -C ~/source/homelab diff:*)",
-      "Bash(git -C ~/source/homelab status:*)",
-      "Bash(git -C ~/source/homelab rev-parse:*)",
-      "Bash(git -C ~/source/homelab ls-files:*)",
-      "Bash(git -C ~/source/homeassistant log:*)",
-      "Bash(git -C ~/source/homeassistant show:*)",
-      "Bash(git -C ~/source/homeassistant diff:*)",
-      "Bash(git -C ~/source/homeassistant status:*)",
-      "Bash(git -C ~/source/homeassistant rev-parse:*)",
-      "Bash(git -C ~/source/homeassistant ls-files:*)",
-      "Bash(/root/Notes/bin/wiki-delete.sh:*)",
-      "Bash(ls:*)",
-      "Bash(date:*)",
-      "Bash(wc:*)",
-      "Bash(stat:*)",
-      "Bash(head:*)",
-      "Bash(tail:*)"
-    ],
-    "deny": [
-      "Read(//proc/**)",
-      "Read(//etc/**)",
-      "Read(//root/.config/**)",
-      "Read(//root/.kube/**)",
-      "Read(//woodpecker/**/.ci-diag/**)",
-      "Read(//root/Notes/.git/**)",
-      "Read(//root/source/*/.git/**)",
-      "Read(//root/Notes/_archive/**)",
-      "Edit(//root/Notes/raw/**)",
-      "Edit(//root/Notes/_archive/**)",
-      "Edit(//root/source/**)",
-      "Bash(kubectl apply:*)",
-      "Bash(kubectl delete:*)",
-      "Bash(kubectl patch:*)",
-      "Bash(kubectl edit:*)",
-      "Bash(kubectl scale:*)",
-      "Bash(kubectl rollout:*)",
-      "Bash(kubectl exec:*)",
-      "Bash(kubectl cp:*)",
-      "Bash(kubectl drain:*)",
-      "Bash(kubectl cordon:*)",
-      "Bash(kubectl label:*)",
-      "Bash(kubectl annotate:*)",
-      "Bash(git push:*)",
-      "Bash(git reset:*)",
-      "Bash(git checkout:*)",
-      "Bash(git clean:*)",
-      "Bash(git rebase:*)",
-      "Bash(git config:*)",
-      "Bash(git -C /root/Notes config:*)",
-      "Bash(git -C ~/Notes config:*)",
-      "Bash(git -C /root/source/homelab config:*)",
-      "Bash(git -C /root/source/homeassistant config:*)",
-      "Bash(git -C ~/source/homelab config:*)",
-      "Bash(git -C ~/source/homeassistant config:*)",
-      "Bash(rm:*)",
-      "Bash(rm -rf:*)",
-      "Bash(curl:*)",
-      "Bash(wget:*)",
-      "Bash(ssh:*)",
-      "Bash(tea:*)",
-      "Bash(woodpecker:*)"
-    ]
-  }
-}
@@ -1,5 +0,0 @@
-You are running as the scheduled wiki-maintenance job in CI. The Notes vault is checked out at ~/Notes (the repository k.renner/notes); the homelab and homeassistant repositories are checked out under ~/source/. `kubectl` reads the live cluster with a read-only identity.
-
-Run the wiki-consolidate agent now for today's scheduled consolidation pass. Follow its instructions exactly: snapshot the vault to git first, abort if that fails, select today's rotating cohort of 20 from the state file, verify claims against both their sources and the live systems, act on the verdicts, write the dated report to ~/Notes/output/, update the state file, and snapshot again at the end. Do not push; the pipeline pushes the run's commits.
-
-Report the counts and the coverage progress when you are done.
@@ -1,7 +0,0 @@
-You are running as the scheduled wiki-maintenance job in CI. The Notes vault is checked out at ~/Notes (the repository k.renner/notes). Do exactly this, nothing else:
-
-Glob ~/Notes/raw/*.md (top level only, never ~/Notes/raw/_processed/). Distill every file you find into ~/Notes/wiki/ following your process, including the ranker check for every concept. Move each fully processed source to ~/Notes/raw/_processed/. If raw/ has no top-level files besides _index.md, say so and stop.
-
-Do not commit or push; the pipeline does that after you finish. Do not touch anything outside ~/Notes/wiki/ and ~/Notes/raw/.
-
-End with one line per file: which wiki notes it produced or merged into, or "nothing distillable".
@@ -1,41 +0,0 @@
-{
-  "//": "Permission scope for the wiki-maintenance agent (wiki-distill) in CI.",
-  "//2": "The step's environment holds a Gitea push token, a TypeSafe key and the wiki-index",
-  "//3": "client credential, so the Read tool is kept away from /proc (env dumps), /root outside",
-  "//4": "the vault symlink target, /etc and the checkout's .git/. The vault itself is the work.",
-  "//5": "Mirrors ci/agents/issue-fix/settings.json; absolute globs because the vault is reached",
-  "//6": "through the ~/Notes symlink the pipeline creates.",
-  "permissions": {
-    "allow": [
-      "Read(//root/Notes/**)",
-      "Edit(//root/Notes/wiki/**)",
-      "Edit(//root/Notes/output/**)",
-      "Edit(//root/Notes/raw/**)",
-      "Bash(/root/Notes/bin/wiki_rank.py:*)",
-      "Bash(/root/Notes/bin/wiki_ask.py:*)",
-      "Bash(mv /root/Notes/raw/:*)",
-      "Bash(mkdir -p /root/Notes/:*)",
-      "Bash(~/Notes/bin/wiki_rank.py:*)",
-      "Bash(~/Notes/bin/wiki_ask.py:*)",
-      "Bash(mv ~/Notes/raw/:*)",
-      "Bash(mkdir -p ~/Notes/:*)",
-      "Bash(ls:*)",
-      "Bash(date:*)",
-      "Bash(wc:*)",
-      "Bash(head:*)",
-      "Bash(tail:*)"
-    ],
-    "deny": [
-      "Read(//proc/**)",
-      "Read(//etc/**)",
-      "Read(//root/.config/**)",
-      "Read(//root/Notes/.git/**)",
-      "Read(//root/Notes/_archive/**)",
-      "Edit(//root/Notes/_archive/**)",
-      "Bash(rm:*)",
-      "Bash(git:*)",
-      "Bash(curl:*)",
-      "Bash(wget:*)"
-    ]
-  }
-}
@@ -1,22 +0,0 @@
-FROM alpine:3.22@sha256:14358309a308569c32bdc37e2e0e9694be33a9d99e68afb0f5ff33cc1f695dce
-
-# bird: BGP daemon for the pod-network mesh over pod0. Distro package, no build
-# step — the image is the binary plus its runtime.
-#
-# 3.22 rather than the 3.21 the other images here pin: bird2 first appears in
-# 3.22/community.
-#
-# Config is read from /etc/bird/bird.conf, not the package default
-# /etc/bird.conf, so the mount cannot shadow what the package ships.
-#
-# kubectl renders the init container's per-node config: go-template-file for
-# the templating, a field selector for excluding the node itself, in-cluster
-# config for the auth. From the distro, so there is no version literal to drift
-# and no download to trust. It is only ever used to read core/v1 Nodes.
-RUN apk add --no-cache bird2 kubectl
-
-# Fails the build rather than the first pod if a binary or its path is wrong.
-RUN /usr/sbin/bird --version \
- && kubectl version --client > /dev/null
-
-ENTRYPOINT ["/usr/sbin/bird", "-f", "-c", "/etc/bird/bird.conf"]
@@ -13,9 +13,7 @@ RUN apk add --no-cache \
      git \
      jq \
      openssh-client \
-      openssl \
-      python3 \
-      py3-yaml
+      openssl

 # Install Claude Code, then make the resulting layer byte-stable so that a
 # rebuild of the SAME version DEDUPS in the Gitea registry instead of orphaning
@@ -1,9 +1,9 @@
 FROM alpine:3.21@sha256:48b0309ca019d89d40f670aa1bc06e426dc0931948452e8491e3d65087abc07d

-ARG TALOS_VERSION=v1.13.10
+ARG TALOS_VERSION=v1.12.6
 ARG VELERO_VERSION=v1.16.0
-ARG KUBECTL_VERSION=v1.36.3
-ARG FLUX_VERSION=2.9.5
+ARG KUBECTL_VERSION=v1.36.2
+ARG FLUX_VERSION=2.9.2

 RUN apk add --no-cache \
      aws-cli \
@@ -18,7 +18,7 @@ RUN apk add --no-cache \

 RUN curl -sfL -o /usr/local/bin/talosctl \
      "https://github.com/siderolabs/talos/releases/download/${TALOS_VERSION}/talosctl-linux-amd64" \
- && echo "28e9ccd9aebd690fa03d8cf2582221d87ea38b2a1704f411d75a03589c58f2aa  /usr/local/bin/talosctl" | sha256sum -c - \
+ && echo "4c41a3b10b075292e64b182283d558e9d8e935f43239de1ed4e4b14359593efb  /usr/local/bin/talosctl" | sha256sum -c - \
 && chmod +x /usr/local/bin/talosctl

 RUN curl -sfL -o /usr/local/bin/kubectl \
@@ -31,7 +31,7 @@ RUN curl -sfL -o /usr/local/bin/kubectl \

 RUN curl -sfL -o /tmp/flux.tar.gz \
      "https://github.com/fluxcd/flux2/releases/download/v${FLUX_VERSION}/flux_${FLUX_VERSION}_linux_amd64.tar.gz" \
- && echo "b853df82adfd7736f580692f9f734473d571606307139f8fd20c2a80dd1ff473  /tmp/flux.tar.gz" | sha256sum -c - \
+ && echo "4092b367fd060097976fb7261601f79420ecf7266328417dfd8dfee27de0e6a3  /tmp/flux.tar.gz" | sha256sum -c - \
 && tar -xzf /tmp/flux.tar.gz -C /usr/local/bin flux \
 && chmod +x /usr/local/bin/flux \
 && rm -f /tmp/flux.tar.gz
@@ -1,12 +0,0 @@
-# netbird-service-sync: reconciles NetBird reverse-proxy services (public
-# domains -> TLS passthrough -> Traefik ClusterIP) from Ingresses carrying the
-# public label. Standard library only, so the image is the interpreter plus one
-# file; runs unprivileged as its own user.
-FROM python:3.13-slim
-
-RUN useradd --system --no-create-home --shell /usr/sbin/nologin controller
-COPY controller.py /opt/controller.py
-
-ENV PYTHONUNBUFFERED=1
-USER controller
-ENTRYPOINT ["python3", "/opt/controller.py"]
@@ -1,275 +0,0 @@
-#!/usr/bin/env python3
-"""netbird-service-sync: reconcile NetBird reverse-proxy services from Ingresses.
-
-Source of truth: every Ingress carrying the public label (dns.renner.uno/public).
-For each host on such an Ingress the NetBird management gets one reverse-proxy
-service in TLS-passthrough mode on :443, targeting Traefik's ClusterIP through a
-subnet resource that covers the cluster's service CIDR. SNI and Host reach
-Traefik untouched, the proxy terminates nothing, and no domain resource is
-involved -- the routing peers stay ordinary IP routers. The target sends a
-PROXY protocol v2 header so Traefik (and CrowdSec behind its access log) sees
-the WAN client address instead of the routing peer's pod IP.
-
-Ownership is derived from NetBird itself, not from local state: a service is
-managed iff its mode is "tls" and its target is this controller's subnet
-resource. Services that do not match (the hand-made HTTP/domain ones) are never
-touched; a desired host held by such a service is reported and skipped until
-that service is removed by hand.
-
-Credentials: the NetBird PAT is read from OpenBao on every pass through the
-Kubernetes auth role, exactly like the setup-key rotator; it never lands in a
-Kubernetes Secret or in a log line.
-"""
-
-import json
-import logging
-import os
-import sys
-import time
-import urllib.error
-import urllib.parse
-import urllib.request
-
-LOG = logging.getLogger("netbird-service-sync")
-
-NB = os.environ.get("NB_MANAGEMENT_URL", "https://netbird.renner.uno").rstrip("/")
-BAO_ADDR = os.environ.get("BAO_ADDR", "http://openbao.openbao.svc.cluster.local:8200").rstrip("/")
-BAO_ROLE = os.environ.get("BAO_ROLE", "netbird-service-sync")
-BAO_PAT_PATH = os.environ.get("BAO_PAT_PATH", "homelab/data/infrastructure/secret-rotators-bootstrap/netbird")
-BAO_PAT_KEY = os.environ.get("BAO_PAT_KEY", "pat")
-SERVICE_CIDR = os.environ.get("SERVICE_CIDR", "10.96.0.0/12")
-NETWORK_NAME = os.environ.get("NETWORK_NAME", "k8s public ingress")
-ROUTER_GROUP = os.environ.get("ROUTER_GROUP_NAME", "k8s-ingress-router")
-SOURCE_GROUP = os.environ.get("SOURCE_GROUP_NAME", "All")
-TRAEFIK_NS, TRAEFIK_SVC = os.environ.get("TRAEFIK_SERVICE", "ingress/traefik").split("/", 1)
-PUBLIC_LABEL = os.environ.get("PUBLIC_LABEL", "dns.renner.uno/public")
-LISTEN_PORT = int(os.environ.get("LISTEN_PORT", "443"))
-INTERVAL = int(os.environ.get("INTERVAL_SECONDS", "60"))
-HTTP_TIMEOUT = 20
-
-K8S_API = "https://kubernetes.default.svc"
-K8S_TOKEN_FILE = "/var/run/secrets/kubernetes.io/serviceaccount/token"
-K8S_CA_FILE = "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"
-
-
-class ApiError(Exception):
-    pass
-
-
-def http(method, url, body=None, headers=None, cafile=None):
-    data = None if body is None else json.dumps(body).encode()
-    req = urllib.request.Request(url, data=data, method=method)
-    req.add_header("Accept", "application/json")
-    if data is not None:
-        req.add_header("Content-Type", "application/json")
-    for k, v in (headers or {}).items():
-        req.add_header(k, v)
-    ctx = None
-    if cafile:
-        import ssl
-        ctx = ssl.create_default_context(cafile=cafile)
-    try:
-        with urllib.request.urlopen(req, timeout=HTTP_TIMEOUT, context=ctx) as resp:
-            raw = resp.read()
-            return json.loads(raw) if raw else None
-    except urllib.error.HTTPError as e:
-        detail = e.read().decode(errors="replace")[:300]
-        raise ApiError(f"{method} {url} -> {e.code}: {detail}") from None
-
-
-# --- Kubernetes -------------------------------------------------------------
-
-def k8s(path):
-    with open(K8S_TOKEN_FILE) as f:
-        token = f.read().strip()
-    return http("GET", K8S_API + path, headers={"Authorization": f"Bearer {token}"}, cafile=K8S_CA_FILE)
-
-
-def desired_hosts():
-    sel = urllib.parse.quote(f"{PUBLIC_LABEL}=true")
-    ingresses = k8s(f"/apis/networking.k8s.io/v1/ingresses?labelSelector={sel}")
-    hosts = {}
-    for ing in ingresses.get("items", []):
-        ref = f'{ing["metadata"]["namespace"]}/{ing["metadata"]["name"]}'
-        for rule in ing.get("spec", {}).get("rules", []):
-            host = rule.get("host")
-            if not host or host.startswith("*"):
-                continue
-            hosts.setdefault(host, ref)
-    return hosts
-
-
-def traefik_cluster_ip():
-    svc = k8s(f"/api/v1/namespaces/{TRAEFIK_NS}/services/{TRAEFIK_SVC}")
-    for ip in svc["spec"].get("clusterIPs", []):
-        if ":" not in ip:
-            return ip
-    raise ApiError(f"{TRAEFIK_NS}/{TRAEFIK_SVC} has no IPv4 ClusterIP")
-
-
-# --- OpenBao ----------------------------------------------------------------
-
-def netbird_pat():
-    with open(K8S_TOKEN_FILE) as f:
-        jwt = f.read().strip()
-    login = http("POST", f"{BAO_ADDR}/v1/auth/kubernetes/login", {"role": BAO_ROLE, "jwt": jwt})
-    bao_token = login["auth"]["client_token"]
-    kv = http("GET", f"{BAO_ADDR}/v1/{BAO_PAT_PATH}", headers={"X-Vault-Token": bao_token})
-    pat = kv["data"]["data"].get(BAO_PAT_KEY)
-    if not pat:
-        raise ApiError(f"OpenBao {BAO_PAT_PATH} has no key {BAO_PAT_KEY}")
-    return pat
-
-
-# --- NetBird management -----------------------------------------------------
-
-class NetBird:
-    def __init__(self, pat):
-        self.h = {"Authorization": f"Token {pat}"}
-
-    def get(self, path):
-        return http("GET", f"{NB}/api/{path}", headers=self.h)
-
-    def list(self, path):
-        # An empty collection comes back as an empty body / null, not [].
-        return self.get(path) or []
-
-    def post(self, path, body):
-        return http("POST", f"{NB}/api/{path}", body, headers=self.h)
-
-    def put(self, path, body):
-        return http("PUT", f"{NB}/api/{path}", body, headers=self.h)
-
-    def delete(self, path):
-        return http("DELETE", f"{NB}/api/{path}", headers=self.h)
-
-
-def group_id(nb, name):
-    for g in nb.list("groups"):
-        if g["name"] == name:
-            return g["id"]
-    raise ApiError(f"NetBird group {name!r} not found")
-
-
-def ensure_network(nb):
-    for n in nb.list("networks"):
-        if n["name"] == NETWORK_NAME:
-            return n["id"]
-    n = nb.post("networks", {"name": NETWORK_NAME,
-                             "description": "managed by netbird-service-sync; public ingress via TLS passthrough"})
-    LOG.info("created network %s (%s)", NETWORK_NAME, n["id"])
-    return n["id"]
-
-
-def ensure_resource(nb, net):
-    for r in nb.list(f"networks/{net}/resources"):
-        if r.get("address") == SERVICE_CIDR:
-            return r["id"]
-    r = nb.post(f"networks/{net}/resources", {"name": "cluster service CIDR", "address": SERVICE_CIDR,
-                                              "description": "managed by netbird-service-sync",
-                                              "enabled": True, "groups": []})
-    LOG.info("created subnet resource %s (%s)", SERVICE_CIDR, r["id"])
-    return r["id"]
-
-
-def ensure_router(nb, net, router_group):
-    for r in nb.list(f"networks/{net}/routers"):
-        if router_group in (r.get("peer_groups") or []):
-            return r["id"]
-    r = nb.post(f"networks/{net}/routers", {"peer_groups": [router_group], "metric": 9999,
-                                            "masquerade": True, "enabled": True})
-    LOG.info("created router on group %s (%s)", ROUTER_GROUP, r["id"])
-    return r["id"]
-
-
-def ensure_policy(nb, resource, source_group):
-    name = f"{NETWORK_NAME} access"
-    for p in nb.list("policies"):
-        if p["name"] == name:
-            return p["id"]
-    p = nb.post("policies", {"name": name, "enabled": True, "rules": [{
-        "name": name, "enabled": True, "action": "accept", "protocol": "tcp",
-        "ports": [str(LISTEN_PORT)], "bidirectional": False, "sources": [source_group],
-        "destinationResource": {"id": resource, "type": "subnet"}}]})
-    LOG.info("created policy %s (%s)", name, p["id"])
-    return p["id"]
-
-
-def service_body(host, resource, target_ip):
-    # proxy_protocol: PROXY v2 header carries the WAN client address to
-    # Traefik (proxyProtocol.trustedIPs on websecure); TCP/TLS modes only.
-    return {"name": host, "domain": host, "mode": "tls", "listen_port": LISTEN_PORT, "enabled": True,
-            "targets": [{"target_type": "subnet", "target_id": resource, "protocol": "tcp",
-                         "host": target_ip, "port": 443, "enabled": True,
-                         "options": {"proxy_protocol": True}}]}
-
-
-def is_managed(svc, resource):
-    return svc.get("mode") == "tls" and any(t.get("target_id") == resource for t in svc.get("targets") or [])
-
-
-def target_matches(svc, target_ip):
-    if svc.get("listen_port") != LISTEN_PORT or not svc.get("enabled"):
-        return False
-    ts = svc.get("targets") or []
-    return len(ts) == 1 and ts[0].get("host") == target_ip and ts[0].get("port") == 443 \
-        and ts[0].get("protocol") == "tcp" and ts[0].get("enabled") \
-        and (ts[0].get("options") or {}).get("proxy_protocol") is True
-
-
-def reconcile():
-    hosts = desired_hosts()
-    target_ip = traefik_cluster_ip()
-    nb = NetBird(netbird_pat())
-
-    router_group = group_id(nb, ROUTER_GROUP)
-    source_group = group_id(nb, SOURCE_GROUP)
-    net = ensure_network(nb)
-    resource = ensure_resource(nb, net)
-    ensure_router(nb, net, router_group)
-    ensure_policy(nb, resource, source_group)
-
-    by_domain = {s["domain"]: s for s in nb.list("reverse-proxies/services")}
-    created = updated = deleted = held = 0
-
-    for host, ref in sorted(hosts.items()):
-        svc = by_domain.get(host)
-        if svc is None:
-            nb.post("reverse-proxies/services", service_body(host, resource, target_ip))
-            LOG.info("created %s (from %s) -> %s:443", host, ref, target_ip)
-            created += 1
-        elif not is_managed(svc, resource):
-            LOG.warning("%s is held by unmanaged service %s (mode %s); remove it to migrate",
-                        host, svc["id"], svc.get("mode"))
-            held += 1
-        elif not target_matches(svc, target_ip):
-            nb.put(f"reverse-proxies/services/{svc['id']}", service_body(host, resource, target_ip))
-            LOG.info("updated %s -> %s:443", host, target_ip)
-            updated += 1
-
-    for domain, svc in by_domain.items():
-        if is_managed(svc, resource) and domain not in hosts:
-            nb.delete(f"reverse-proxies/services/{svc['id']}")
-            LOG.info("deleted %s (no public Ingress carries it any more)", domain)
-            deleted += 1
-
-    LOG.info("reconciled: %d hosts desired, %d created, %d updated, %d deleted, %d held by unmanaged services",
-             len(hosts), created, updated, deleted, held)
-
-
-def main():
-    logging.basicConfig(level=os.environ.get("LOG_LEVEL", "INFO"), stream=sys.stdout,
-                        format="%(asctime)s %(levelname)s %(message)s")
-    if os.environ.get("ONCE") == "1":
-        reconcile()
-        return
-    while True:
-        try:
-            reconcile()
-        except Exception as e:  # keep running; the next pass retries
-            LOG.error("reconcile failed: %s", e)
-        time.sleep(INTERVAL)
-
-
-if __name__ == "__main__":
-    main()
@@ -1,84 +0,0 @@
-FROM alpine:3.21@sha256:48b0309ca019d89d40f670aa1bc06e426dc0931948452e8491e3d65087abc07d
-
-# Everything `bootstrap.sh vmtest` needs, so the rebuild test can run as a Job in
-# the cluster instead of on a workstation. The workstation has 31G and the test
-# wants ~14.5G for four VMs — a single run left under 1G headroom, and two
-# overlapping runs OOM-killed the session. The nodes have ~94G with 70G free.
-#
-# Separate from ci-k8s rather than built on it: that image pins talosctl v1.12.6,
-# and this test creates a cluster at the version talos/patches/cluster.yaml
-# installs. A talosctl older than the Talos it provisions is exactly the kind of
-# skew the test exists to catch, so it must not be baked in here.
-#
-# KEEP TALOS_VERSION IN STEP with the installer image tag in
-# kubernetes/talos/patches/cluster.yaml. The phase reads the version from that
-# file at run time and passes it to the provisioner; this pin only decides which
-# client talks to it.
-ARG TALOS_VERSION=v1.13.10
-ARG KUBECTL_VERSION=v1.36.2
-ARG FLUX_VERSION=2.9.2
-ARG KOPIA_VERSION=0.23.1
-
-# qemu-system-x86_64 and qemu-img are the provisioner's actual backend, and ovmf
-# supplies the UEFI firmware those VMs boot from — without it the provisioner
-# stops with "the required flash image was not found in any of the expected
-# paths", having already downloaded the ISO and generated the PKI. Note the
-# firmware comes from ovmf specifically: qemu-system-x86_64 ships an edk2 CODE
-# image but no VARS, which is the half the provisioner needs.
-#
-# iptables is shelled out to by the CNI plugins that build the VM network; the
-# provisioner fails with "error accessing iptables" without it. The rest is
-# what the phase shells out to: helm for the Cilium install, yq to read the
-# HelmRelease and the installer image, jq for OpenBao's seal status.
-#
-# Versions come from the digest-pinned base rather than being pinned one by one,
-# the same way ci-ansible does it.
-RUN apk add --no-cache \
-      bash \
-      ca-certificates \
-      curl \
-      git \
-      helm \
-      iptables \
-      jq \
-      openssl \
-      ovmf \
-      qemu-img \
-      qemu-system-x86_64 \
-      yq
-
-RUN curl -sfL -o /usr/local/bin/talosctl \
-      "https://github.com/siderolabs/talos/releases/download/${TALOS_VERSION}/talosctl-linux-amd64" \
- && echo "28e9ccd9aebd690fa03d8cf2582221d87ea38b2a1704f411d75a03589c58f2aa  /usr/local/bin/talosctl" | sha256sum -c - \
- && chmod +x /usr/local/bin/talosctl
-
-RUN curl -sfL -o /usr/local/bin/kubectl \
-      "https://dl.k8s.io/release/${KUBECTL_VERSION}/bin/linux/amd64/kubectl" \
- && curl -sfL -o /tmp/kubectl.sha256 \
-      "https://dl.k8s.io/release/${KUBECTL_VERSION}/bin/linux/amd64/kubectl.sha256" \
- && echo "$(cat /tmp/kubectl.sha256)  /usr/local/bin/kubectl" | sha256sum -c - \
- && rm -f /tmp/kubectl.sha256 \
- && chmod +x /usr/local/bin/kubectl
-
-RUN curl -sfL -o /tmp/flux.tar.gz \
-      "https://github.com/fluxcd/flux2/releases/download/v${FLUX_VERSION}/flux_${FLUX_VERSION}_linux_amd64.tar.gz" \
- && echo "4092b367fd060097976fb7261601f79420ecf7266328417dfd8dfee27de0e6a3  /tmp/flux.tar.gz" | sha256sum -c - \
- && tar -xzf /tmp/flux.tar.gz -C /usr/local/bin flux \
- && chmod +x /usr/local/bin/flux \
- && rm -f /tmp/flux.tar.gz
-
-# kopia reads the Velero backup repository directly, which is what the OpenBao
-# restore does when there is no Velero yet to do it: the credentials Velero
-# needs live in OpenBao, so OpenBao cannot be restored by Velero. Its absence
-# was found by a rebuild rehearsal reaching the restore for the first time and
-# stopping at "kopia not found".
-#
-# Same checksum-verified download as the tools above; the tarball unpacks into
-# a versioned directory, so the binary is pulled out by name.
-RUN curl -sfL -o /tmp/kopia.tar.gz \
-      "https://github.com/kopia/kopia/releases/download/v${KOPIA_VERSION}/kopia-${KOPIA_VERSION}-linux-x64.tar.gz" \
- && echo "416d0f84a3dbb321a8b2d8f0997b1a0a6e915babe79ee76fa6e4d2bd1e1c5178  /tmp/kopia.tar.gz" | sha256sum -c - \
- && tar -xzf /tmp/kopia.tar.gz -C /tmp \
- && mv "/tmp/kopia-${KOPIA_VERSION}-linux-x64/kopia" /usr/local/bin/kopia \
- && chmod +x /usr/local/bin/kopia \
- && rm -rf /tmp/kopia.tar.gz "/tmp/kopia-${KOPIA_VERSION}-linux-x64"
@@ -1,976 +0,0 @@
-# Paperclip Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Deploy Paperclip (server-local agent runs, option B) into the homelab cluster behind an Authentik gate, with CNPG Postgres and OpenBao-sourced secrets.
-
-**Architecture:** One Flux `Kustomization` (`paperclip`) applies `kubernetes/apps/paperclip/`: a single-replica Deployment on a ceph-block PVC, CNPG `pclp-pg` (PG17) with barman backups to Hetzner, and a Traefik Ingress that backends only the `paperclip-gate` Authentik outpost. Namespace, baseline CNP, ClusterSecretStore allowlist and image automation live in the existing `infrastructure` tree.
-
-**Tech Stack:** Flux v2 (kustomize-controller, image-reflector v1.2.5), CNPG + barman-cloud plugin, External Secrets → OpenBao, Cilium CNP, Traefik, Authentik blueprints.
-
-**Spec:** `docs/superpowers/specs/2026-09-25-paperclip-design.md`
-
-## Global Constraints
-
- Worktree: `/home/krenner/source/homelab-worktrees/paperclip`, branch `feat/paperclip`. Never edit the main checkout.
- Commits as claude-bot: `git -c user.name='Claude Bot' -c user.email='claude-bot@renner.uno' commit …`, message ends with `Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>`.
- Push: `git -c credential.helper='store --file=/home/krenner/.config/claude-bot/creds' push …`. PR via `tea pr create --login claude-bot`. Never self-merge.
- One Bash command per call: no `&&`, `;`, `|` chaining, no `2>/dev/null`, no `python -c`.
- Namespace `paperclip`; hostname `paperclip.renner.uno`; LAN-only (no `dns.renner.uno/public` label).
- Image `ghcr.io/paperclipai/paperclip`, tag `nightly`, digest-pinned, `digestReflectionPolicy: Always`.
- Postgres: CNPG `pclp-pg`, `ghcr.io/cloudnative-pg/postgresql:17`, 2 instances, `ceph-block`.
- PVC `paperclip-data`, `ceph-block`, 20Gi, mounted at `/paperclip`.
- OpenBao key `apps/paperclip` with fields `BETTER_AUTH_SECRET`, `PAPERCLIP_TOOL_ACTION_SIGNING_SECRET`, `PAPERCLIP_SECRETS_MASTER_KEY` (64-char hex), `CLAUDE_CODE_OAUTH_TOKEN`.
- Env: `PAPERCLIP_DEPLOYMENT_MODE=authenticated`, `PAPERCLIP_DEPLOYMENT_EXPOSURE=private`, `PAPERCLIP_PUBLIC_URL=https://paperclip.renner.uno`, `HOST=0.0.0.0`, `PORT=3100`.
- Manifest comments: one terse factual line; rationale goes in commit messages/PR.
- No `kubectl apply`/`exec`/`rollout restart` against the live cluster; Flux applies after merge. Server-side `--dry-run=server` is allowed.
-
-## Review Focus
-
-1. **Pod replaced mid agent run** (image bump, secret rotation via Reloader, node drain) — `Recreate` kills in-flight runs; expected: Paperclip marks them failed/recoverable on restart, not wedged. Covered in Task 9 Step 6.
-2. **Shared OAuth token expired or missing** — expected: agent runs fail with a visible auth error in Paperclip, server stays healthy. Covered in Task 9 Step 7.
-3. **Direct access bypassing the gate** from another namespace pod — expected: connection refused/dropped by CNP. Covered in Task 9 Step 4.
-4. **Unauthenticated LAN request** to `paperclip.renner.uno` — expected: redirect to Authentik, never Paperclip's UI. Covered in Task 9 Step 3.
-5. **Nightly image with a breaking DB migration** — expected: a base backup exists before the first bump lands. Covered in Task 9 Step 5.
-
---
-
-### Task 1: Upstream facts that decide two manifest values
-
-No repo changes; produces two decisions consumed by Task 6.
-
-**Interfaces:**
- Produces: `SIGNUP_DISABLED` (`"true"` or `"false"`), `CONFIG_PARTIAL_OK` (yes/no), `ENTRYPOINT_NEEDS_ROOT` (yes/no).
-
- [ ] **Step 1: Sign-up vs instance claim.** Read upstream `server/src/auth/better-auth.ts`, `server/src/routes/auth.ts`, and the claim flow for `authenticated`+`private` (search `claim` under `server/src`):
-
-Run: `gh api repos/paperclipai/paperclip/contents/server/src/routes/auth.ts -H "Accept: application/vnd.github.raw"`
-
-Decide: if the first-admin claim creates the account through a path that ignores `disableSignUp` (e.g. invite/claim token handler calling the adapter directly), `SIGNUP_DISABLED="true"`. Otherwise `SIGNUP_DISABLED="false"` — the Authentik gate is the access control, and a follow-up PR flips it to `"true"` after the admin is claimed.
-
- [ ] **Step 2: Partial config.json.** Read `server/src/config.ts` and the config-file loader it calls (search `PAPERCLIP_CONFIG` / `readConfigFile`).
-
-Run: `gh api repos/paperclipai/paperclip/contents/server/src/config.ts -H "Accept: application/vnd.github.raw"`
-
-Decide `CONFIG_PARTIAL_OK=yes` if a file containing only `{"telemetry":{"enabled":false}}` parses (every other section optional/defaulted). If not, stop and report to the user — do not invent config fields.
-
- [ ] **Step 3: Entrypoint privilege.**
-
-Run: `gh api repos/paperclipai/paperclip/contents/docker-entrypoint.sh -H "Accept: application/vnd.github.raw"`
-
-`ENTRYPOINT_NEEDS_ROOT=yes` if it calls `usermod`/`groupmod`/`chown`/`gosu`. Record the result for Task 6.
-
- [ ] **Step 4: Anonymous health response.**
-
-Run: `gh api repos/paperclipai/paperclip/contents/server/src/routes/health.ts -H "Accept: application/vnd.github.raw"`
-
-Confirm `GET /api/health` returns HTTP 200 to an anonymous caller when healthy (probes use it). If it returns 401 anonymously, use a `tcpSocket` probe on 3100 in Task 6 instead.
-
---
-
-### Task 2: Namespace, baseline CNP, ClusterSecretStore allowlist
-
-**Files:**
- Modify: `kubernetes/infrastructure/namespaces/namespaces.yaml` (append)
- Create: `kubernetes/infrastructure/network-policies/paperclip-ns.yaml`
- Modify: `kubernetes/infrastructure/network-policies/kustomization.yaml`
- Modify: `kubernetes/infrastructure/post-crds/openbao-clustersecretstore.yaml` (sorted list)
-
-**Interfaces:**
- Produces: namespace `paperclip`; ESO may reference `openbao` from it.
-
- [ ] **Step 1: Append the namespace** to the end of `namespaces.yaml`:
-
-```yaml
---
-apiVersion: v1
-kind: Namespace
-metadata:
-  name: paperclip
-```
-
- [ ] **Step 2: Create `paperclip-ns.yaml`** (copy of `paperless-ns.yaml`'s `default-deny-ingress`, ns renamed, paperless-only CNPs dropped):
-
-```yaml
-# Traefik is not admitted: the only entry is the paperclip-gate outpost (apps/paperclip/network-policy.yaml).
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: default-deny-ingress
-  namespace: paperclip
-spec:
-  endpointSelector: {}
-  ingress:
-    - fromEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": paperclip
-    - fromEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": database
-    - fromEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": monitoring
-    - fromEntities:
-        - host
-        - remote-node
-    - fromEndpoints:
-        - matchExpressions:
-            - key: io.kubernetes.pod.namespace
-              operator: Exists
-      icmps:
-        - fields:
-            - { type: 0,   family: IPv4 }
-            - { type: 3,   family: IPv4 }
-            - { type: 8,   family: IPv4 }
-            - { type: 11,  family: IPv4 }
-            - { type: 12,  family: IPv4 }
-            - { type: 1,   family: IPv6 }
-            - { type: 2,   family: IPv6 }
-            - { type: 3,   family: IPv6 }
-            - { type: 4,   family: IPv6 }
-            - { type: 128, family: IPv6 }
-            - { type: 129, family: IPv6 }
-            - { type: 133, family: IPv6 }
-            - { type: 134, family: IPv6 }
-            - { type: 135, family: IPv6 }
-            - { type: 136, family: IPv6 }
-            - { type: 143, family: IPv6 }
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": paperclip
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": authentik
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": ingress
-            app.kubernetes.io/name: traefik
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": rook-ceph
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": kube-system
-            k8s-app: kube-dns
-      toPorts:
-        - ports:
-            - port: "53"
-              protocol: UDP
-    - toEntities:
-        - host
-        - remote-node
-        - world
-    - toEndpoints:
-        - matchExpressions:
-            - key: io.kubernetes.pod.namespace
-              operator: Exists
-      icmps:
-        - fields:
-            - { type: 0,   family: IPv4 }
-            - { type: 3,   family: IPv4 }
-            - { type: 8,   family: IPv4 }
-            - { type: 11,  family: IPv4 }
-            - { type: 12,  family: IPv4 }
-            - { type: 1,   family: IPv6 }
-            - { type: 2,   family: IPv6 }
-            - { type: 3,   family: IPv6 }
-            - { type: 4,   family: IPv6 }
-            - { type: 128, family: IPv6 }
-            - { type: 129, family: IPv6 }
-            - { type: 133, family: IPv6 }
-            - { type: 134, family: IPv6 }
-            - { type: 135, family: IPv6 }
-            - { type: 136, family: IPv6 }
-            - { type: 143, family: IPv6 }
-```
-
- [ ] **Step 3: Register it** — in `network-policies/kustomization.yaml` add `  - paperclip-ns.yaml` directly after `  - paperless-ns.yaml`.
-
- [ ] **Step 4: Allowlist the namespace** — in `openbao-clustersecretstore.yaml` insert `        - paperclip` between `        - monitoring` and `        - paperless` (list is sorted).
-
- [ ] **Step 5: Render check.**
-
-Run: `kubectl kustomize kubernetes/infrastructure/network-policies`
-Expected: output contains `namespace: paperclip`, exit 0.
-
-Run: `kubectl apply --dry-run=server -f kubernetes/infrastructure/post-crds/openbao-clustersecretstore.yaml`
-Expected: `clustersecretstore.external-secrets.io/openbao configured (server dry run)`.
-
- [ ] **Step 6: Commit** `feat(paperclip): namespace, baseline CNP and OpenBao allowlist`.
-
---
-
-### Task 3: OpenBao secret `apps/paperclip`
-
-Out-of-repo; must exist before merge or the ExternalSecret never becomes Ready.
-
-**Interfaces:**
- Produces: OpenBao KV `apps/paperclip` with the four fields in Global Constraints.
-
- [ ] **Step 1:** Ask wiki-synth for the claude-bot OpenBao write procedure (note `claude-bot-openbao-write-only-identity`): login method, KV mount, exact `bao kv put`/`patch` syntax for an `apps/<name>` key. Use exactly what it returns.
-
- [ ] **Step 2: Write the generated fields** as claude-bot (write-only identity — never read values back). Values generated inline by the command:
-
-`BETTER_AUTH_SECRET` = `openssl rand -hex 32`, `PAPERCLIP_TOOL_ACTION_SIGNING_SECRET` = `openssl rand -hex 32`, `PAPERCLIP_SECRETS_MASTER_KEY` = `openssl rand -hex 32` (64 hex chars = 32 bytes, accepted format).
-
-Generate each with a separate `openssl rand -hex 32` call and pass it directly into the `bao kv put` from Step 1; do not echo or log them.
-
- [ ] **Step 3: Ask the user** to add the shared token themselves (never handle it): "Run `claude setup-token`, then `bao kv patch <mount>/apps/paperclip CLAUDE_CODE_OAUTH_TOKEN=<token>`." Wait for confirmation.
-
- [ ] **Step 4: Verify field names only** (no values): `bao kv metadata get <mount>/apps/paperclip` shows a current version ≥2. If claude-bot cannot read metadata, ask the user to confirm the four keys exist.
-
---
-
-### Task 4: Image automation + initial digest
-
-**Files:**
- Create: `kubernetes/infrastructure/image-automation/paperclip.yaml`
- Modify: `kubernetes/infrastructure/image-automation/kustomization.yaml`
-
-**Interfaces:**
- Produces: ImagePolicy `flux-system:paperclip`; `INITIAL_DIGEST` (`sha256:…`) for Task 6.
-
- [ ] **Step 1: Create `paperclip.yaml`:**
-
-```yaml
-apiVersion: image.toolkit.fluxcd.io/v1
-kind: ImageRepository
-metadata:
-  name: paperclip
-  namespace: flux-system
-spec:
-  image: ghcr.io/paperclipai/paperclip
-  interval: 1h
---
-# Upstream publishes no semver tags; track the nightly tag by digest.
-apiVersion: image.toolkit.fluxcd.io/v1
-kind: ImagePolicy
-metadata:
-  name: paperclip
-  namespace: flux-system
-spec:
-  imageRepositoryRef:
-    name: paperclip
-  digestReflectionPolicy: Always
-  filterTags:
-    pattern: '^nightly$'
-  policy:
-    alphabetical:
-      order: asc
-```
-
- [ ] **Step 2: Register** — add `  - paperclip.yaml` after `  - paperless-ngx.yaml` in `image-automation/kustomization.yaml`.
-
- [ ] **Step 3: Server dry-run.**
-
-Run: `kubectl apply --dry-run=server -f kubernetes/infrastructure/image-automation/paperclip.yaml`
-Expected: both objects `created (server dry run)`; no `unknown field "digestReflectionPolicy"` error.
-
- [ ] **Step 4: Resolve the current nightly digest.**
-
-The workstation has `docker` only (no skopeo/crane/buildx, checked 2026-09-25); `docker manifest inspect` lists per-platform digests but not the index digest, so pull once:
-
-Run: `docker pull ghcr.io/paperclipai/paperclip:nightly`
-Run: `docker image inspect --format '{{index .RepoDigests 0}}' ghcr.io/paperclipai/paperclip:nightly`
-Expected: `ghcr.io/paperclipai/paperclip@sha256:<64 hex>`. Record the `sha256:…` part as `INITIAL_DIGEST`.
-Run: `docker image rm ghcr.io/paperclipai/paperclip:nightly`
-
- [ ] **Step 5: Commit** `feat(paperclip): Flux image automation tracks the nightly digest`.
-
---
-
-### Task 5: Postgres (CNPG) and secrets in the app directory
-
-**Files:**
- Create: `kubernetes/apps/paperclip/cluster.yaml`, `objectstore.yaml`, `scheduled-backup.yaml`, `hetzner-creds-externalsecret.yaml`, `externalsecret.yaml`
-
-**Interfaces:**
- Produces: Secret `pclp-pg-app` (CNPG, key `uri`), Service `pclp-pg-rw`, Secret `paperclip-secrets`.
-
- [ ] **Step 1: `cluster.yaml`:**
-
-```yaml
-apiVersion: postgresql.cnpg.io/v1
-kind: Cluster
-metadata:
-  name: pclp-pg
-  namespace: paperclip
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled # data guard
-spec:
-  instances: 2
-  smartShutdownTimeout: 10
-  stopDelay: 60
-  # PG17: upstream's external Postgres target (docker/docker-compose.yml).
-  imageName: ghcr.io/cloudnative-pg/postgresql:17
-  storage:
-    storageClass: ceph-block
-    size: 10Gi
-  plugins:
-    - name: barman-cloud.cloudnative-pg.io
-      isWALArchiver: true
-      parameters:
-        barmanObjectName: pclp-pg-store
-  bootstrap:
-    initdb:
-      database: paperclip
-      owner: paperclip
-  postgresql:
-    parameters:
-      shared_buffers: "256MB"
-      max_connections: "200"
-      checkpoint_timeout: "15min"
-      max_wal_size: "2GB"
-      wal_compression: "zstd"
-      archive_timeout: "30min"
-    synchronous:
-      method: any
-      number: 1
-      dataDurability: preferred
-  resources:
-    requests:
-      cpu: 200m
-      memory: 512Mi
-    limits:
-      memory: 1Gi
-```
-
- [ ] **Step 2: `objectstore.yaml`:**
-
-```yaml
-apiVersion: barmancloud.cnpg.io/v1
-kind: ObjectStore
-metadata:
-  name: pclp-pg-store
-  namespace: paperclip
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled # backup target
-spec:
-  configuration:
-    destinationPath: s3://renner-uno-k8s-cnpg/pclp-pg
-    endpointURL: https://fsn1.your-objectstorage.com
-    s3Credentials:
-      accessKeyId:
-        name: hetzner-s3-creds
-        key: ACCESS_KEY_ID
-      secretAccessKey:
-        name: hetzner-s3-creds
-        key: ACCESS_SECRET_KEY
-      region:
-        name: hetzner-s3-creds
-        key: REGION
-    wal:
-      compression: gzip
-    data:
-      compression: gzip
-  retentionPolicy: "30d"
-```
-
- [ ] **Step 3: `scheduled-backup.yaml`:**
-
-```yaml
-apiVersion: postgresql.cnpg.io/v1
-kind: ScheduledBackup
-metadata:
-  name: pclp-pg-daily
-  namespace: paperclip
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled # backup schedule
-spec:
-  schedule: "0 45 2 * * *"
-  immediate: true
-  backupOwnerReference: self
-  cluster:
-    name: pclp-pg
-  method: plugin
-  pluginConfiguration:
-    name: barman-cloud.cloudnative-pg.io
-```
-
- [ ] **Step 4: `hetzner-creds-externalsecret.yaml`:**
-
-```yaml
-apiVersion: external-secrets.io/v1
-kind: ExternalSecret
-metadata:
-  name: hetzner-s3-creds
-  namespace: paperclip
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled # barman S3 creds
-spec:
-  refreshInterval: 1h
-  secretStoreRef:
-    kind: ClusterSecretStore
-    name: openbao
-  target:
-    name: hetzner-s3-creds
-  dataFrom:
-    - extract:
-        key: infrastructure/hetzner-s3
-```
-
- [ ] **Step 5: `externalsecret.yaml`:**
-
-```yaml
-apiVersion: external-secrets.io/v1
-kind: ExternalSecret
-metadata:
-  name: paperclip-secrets
-  namespace: paperclip
-spec:
-  refreshInterval: 10m
-  secretStoreRef:
-    kind: ClusterSecretStore
-    name: openbao
-  target:
-    name: paperclip-secrets
-  dataFrom:
-    - extract:
-        key: apps/paperclip
-```
-
- [ ] **Step 6: Commit** `feat(paperclip): CNPG pclp-pg with barman backups and OpenBao secrets`. (Rendered together with Task 6's kustomization.)
-
---
-
-### Task 6: Server workload, storage, ingress, network policy
-
-**Files:**
- Create: `kubernetes/apps/paperclip/pvc.yaml`, `paperclip.yaml`, `gate-service.yaml`, `ingress.yaml`, `network-policy.yaml`, `kustomization.yaml`
-
-**Interfaces:**
- Consumes: `pclp-pg-app` (`uri`), `paperclip-secrets`, ImagePolicy `flux-system:paperclip`, `INITIAL_DIGEST` (Task 4), `SIGNUP_DISABLED` / `ENTRYPOINT_NEEDS_ROOT` / probe choice (Task 1).
- Produces: pods labelled `app.kubernetes.io/name: paperclip` listening on 3100; Service `paperclip:3100`; ExternalName `ak-outpost-paperclip-gate:9000`.
-
- [ ] **Step 1: `pvc.yaml`:**
-
-```yaml
-apiVersion: v1
-kind: PersistentVolumeClaim
-metadata:
-  name: paperclip-data
-  namespace: paperclip
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled # data guard
-spec:
-  accessModes: [ReadWriteOnce]
-  storageClassName: ceph-block
-  resources:
-    requests:
-      storage: 20Gi
-```
-
- [ ] **Step 2: `paperclip.yaml`** — replace `<INITIAL_DIGEST>` with the Task 4 value and `<SIGNUP_DISABLED>` with the Task 1 value. If Task 1 found `ENTRYPOINT_NEEDS_ROOT=no`, replace the container `securityContext` with `runAsNonRoot: true`, `runAsUser: 1000`, `runAsGroup: 1000`, `allowPrivilegeEscalation: false`, `capabilities: {drop: ["ALL"]}`. If Task 1 chose `tcpSocket`, replace each `httpGet` block with `tcpSocket: {port: http}`.
-
-```yaml
-apiVersion: apps/v1
-kind: Deployment
-metadata:
-  name: paperclip
-  namespace: paperclip
-  labels:
-    app.kubernetes.io/name: paperclip
-  annotations:
-    secret.reloader.stakater.com/reload: paperclip-secrets
-spec:
-  replicas: 1
-  strategy:
-    type: Recreate
-  selector:
-    matchLabels:
-      app.kubernetes.io/name: paperclip
-  template:
-    metadata:
-      labels:
-        app.kubernetes.io/name: paperclip
-    spec:
-      enableServiceLinks: false
-      securityContext:
-        fsGroup: 1000
-        fsGroupChangePolicy: OnRootMismatch
-        seccompProfile:
-          type: RuntimeDefault
-      initContainers:
-        # Telemetry is only configurable in config.json.
-        - name: disable-telemetry
-          image: ghcr.io/paperclipai/paperclip:nightly@<INITIAL_DIGEST> # {"$imagepolicy": "flux-system:paperclip"}
-          command:
-            - sh
-            - -c
-            - |
-              set -eu
-              f=/paperclip/instances/default/config.json
-              mkdir -p "$(dirname "$f")"
-              if [ -s "$f" ]; then
-                jq '.telemetry.enabled = false' "$f" > "$f.tmp"
-              else
-                printf '%s\n' '{"telemetry":{"enabled":false}}' > "$f.tmp"
-              fi
-              mv "$f.tmp" "$f"
-          securityContext:
-            runAsNonRoot: true
-            runAsUser: 1000
-            runAsGroup: 1000
-            allowPrivilegeEscalation: false
-            capabilities:
-              drop: ["ALL"]
-          volumeMounts:
-            - name: data
-              mountPath: /paperclip
-      containers:
-        - name: paperclip
-          image: ghcr.io/paperclipai/paperclip:nightly@<INITIAL_DIGEST> # {"$imagepolicy": "flux-system:paperclip"}
-          ports:
-            - name: http
-              containerPort: 3100
-          env:
-            - name: DATABASE_URL
-              valueFrom:
-                secretKeyRef:
-                  name: pclp-pg-app
-                  key: uri
-            - name: HOST
-              value: "0.0.0.0"
-            - name: PORT
-              value: "3100"
-            - name: PAPERCLIP_DEPLOYMENT_MODE
-              value: authenticated
-            - name: PAPERCLIP_DEPLOYMENT_EXPOSURE
-              value: private
-            - name: PAPERCLIP_PUBLIC_URL
-              value: https://paperclip.renner.uno
-            - name: PAPERCLIP_AUTH_DISABLE_SIGN_UP
-              value: "<SIGNUP_DISABLED>"
-          envFrom:
-            - secretRef:
-                name: paperclip-secrets
-          volumeMounts:
-            - name: data
-              mountPath: /paperclip
-          startupProbe:
-            httpGet:
-              path: /api/health
-              port: http
-            periodSeconds: 10
-            timeoutSeconds: 5
-            failureThreshold: 60
-          livenessProbe:
-            httpGet:
-              path: /api/health
-              port: http
-            periodSeconds: 30
-            timeoutSeconds: 10
-          readinessProbe:
-            httpGet:
-              path: /api/health
-              port: http
-            periodSeconds: 10
-            timeoutSeconds: 5
-          resources:
-            requests:
-              cpu: 250m
-              memory: 1Gi
-            limits:
-              memory: 4Gi
-          securityContext:
-            runAsNonRoot: false
-            allowPrivilegeEscalation: true
-            capabilities:
-              drop: ["ALL"]
-              add: ["CHOWN", "DAC_OVERRIDE", "FOWNER", "SETGID", "SETUID"]
-      volumes:
-        - name: data
-          persistentVolumeClaim:
-            claimName: paperclip-data
---
-apiVersion: v1
-kind: Service
-metadata:
-  name: paperclip
-  namespace: paperclip
-spec:
-  selector:
-    app.kubernetes.io/name: paperclip
-  ports:
-    - name: http
-      port: 3100
-      targetPort: http
-```
-
- [ ] **Step 3: `gate-service.yaml`:**
-
-```yaml
-# ExternalName mirror so the Ingress can backend the paperclip-gate outpost in the authentik namespace.
-apiVersion: v1
-kind: Service
-metadata:
-  name: ak-outpost-paperclip-gate
-  namespace: paperclip
-spec:
-  type: ExternalName
-  externalName: ak-outpost-paperclip-gate.authentik.svc.cluster.local
-  ports:
-    - port: 9000
-      targetPort: 9000
-```
-
- [ ] **Step 4: `ingress.yaml`:**
-
-```yaml
-# LAN-only (no dns.renner.uno/public label); backends the gate outpost, never the Service.
-apiVersion: networking.k8s.io/v1
-kind: Ingress
-metadata:
-  name: paperclip
-  namespace: paperclip
-  annotations:
-    cert-manager.io/cluster-issuer: letsencrypt-cluster
-spec:
-  ingressClassName: traefik
-  tls:
-    - hosts:
-        - paperclip.renner.uno
-      secretName: paperclip-tls
-  rules:
-    - host: paperclip.renner.uno
-      http:
-        paths:
-          - path: /
-            pathType: Prefix
-            backend:
-              service:
-                name: ak-outpost-paperclip-gate
-                port:
-                  number: 9000
-```
-
- [ ] **Step 5: `network-policy.yaml`** (Traefik pod-side port is 8443, verified via `kubectl get pods -n ingress -l app.kubernetes.io/name=traefik`):
-
-```yaml
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: paperclip-gate-ingress
-  namespace: paperclip
-spec:
-  endpointSelector:
-    matchLabels:
-      app.kubernetes.io/name: paperclip
-  ingress:
-    - fromEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": authentik
-            goauthentik.io/outpost-name: paperclip-gate
-      toPorts:
-        - ports:
-            - port: "3100"
-              protocol: TCP
---
-# Covers the server and every agent it spawns (same pod).
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: paperclip-egress
-  namespace: paperclip
-spec:
-  endpointSelector:
-    matchLabels:
-      app.kubernetes.io/name: paperclip
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": paperclip
-            cnpg.io/cluster: pclp-pg
-      toPorts:
-        - ports:
-            - port: "5432"
-              protocol: TCP
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": kube-system
-            k8s-app: kube-dns
-      toPorts:
-        - ports:
-            - port: "53"
-              protocol: UDP
-            - port: "53"
-              protocol: TCP
-    # Gitea, Woodpecker, HA, Loki via their *.renner.uno names.
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": ingress
-            app.kubernetes.io/name: traefik
-      toPorts:
-        - ports:
-            - port: "8443"
-              protocol: TCP
-    - toEntities:
-        - kube-apiserver
-      toPorts:
-        - ports:
-            - port: "6443"
-              protocol: TCP
-    - toEntities:
-        - world
-      toPorts:
-        - ports:
-            - port: "443"
-              protocol: TCP
-```
-
- [ ] **Step 6: `kustomization.yaml`:**
-
-```yaml
-apiVersion: kustomize.config.k8s.io/v1beta1
-kind: Kustomization
-resources:
-  - externalsecret.yaml
-  - hetzner-creds-externalsecret.yaml
-  - cluster.yaml
-  - objectstore.yaml
-  - scheduled-backup.yaml
-  - pvc.yaml
-  - paperclip.yaml
-  - gate-service.yaml
-  - ingress.yaml
-  - network-policy.yaml
-```
-
- [ ] **Step 7: Render check.**
-
-Run: `kubectl kustomize kubernetes/apps/paperclip`
-Expected: exit 0; output contains `kind: Cluster`, `kind: Deployment`, `claimName: paperclip-data`, and no literal `<INITIAL_DIGEST>` or `<SIGNUP_DISABLED>`.
-
-Run: `grep -rn "<INITIAL_DIGEST>\|<SIGNUP_DISABLED>" kubernetes/apps/paperclip`
-Expected: no output, exit 1.
-
- [ ] **Step 8: Commit** `feat(paperclip): server Deployment, PVC, gated Ingress and CNPs`.
-
---
-
-### Task 7: Authentik gate
-
-**Files:**
- Modify: `kubernetes/apps/authentik/blueprints-configmap.yaml` (insert before the line `      # navidrome-gate: the Navidrome web UI.`)
- Create: `kubernetes/apps/authentik/paperclip-gate-egress.yaml`
- Modify: `kubernetes/apps/authentik/kustomization.yaml`
-
-**Interfaces:**
- Produces: outpost pods labelled `goauthentik.io/outpost-name: paperclip-gate`, Service `ak-outpost-paperclip-gate.authentik:9000`.
-
- [ ] **Step 1: Insert blueprint entries** (same indentation as the neighbouring `wiki-gate` block):
-
-```yaml
-      # paperclip-gate: Paperclip UI; the CNP admits only this outpost. Paperclip keeps its own login behind it.
-      - id: paperclip-users-group
-        model: authentik_core.group
-        identifiers:
-          name: Paperclip Users
-        state: present
-        attrs:
-          is_superuser: false
-          users:
-            - !Find [authentik_core.user, [username, k.renner]]
-
-      - id: paperclip-gate-provider
-        model: authentik_providers_proxy.proxyprovider
-        identifiers:
-          name: paperclip-gate
-        state: present
-        attrs:
-          authentication_flow: !Find [authentik_flows.flow, [slug, passkey-passwordless-authentication]]
-          authorization_flow: !Find [authentik_flows.flow, [slug, passkey-stepup-authorization]]
-          invalidation_flow: !Find [authentik_flows.flow, [slug, default-provider-invalidation-flow]]
-          mode: proxy
-          internal_host: http://paperclip.paperclip.svc.cluster.local:3100
-          internal_host_ssl_validation: false
-          external_host: https://paperclip.renner.uno
-          cookie_domain: renner.uno
-          access_token_validity: hours=24
-          refresh_token_validity: days=30
-          intercept_header_auth: true
-
-      - model: authentik_core.application
-        identifiers:
-          slug: paperclip-gate
-        state: present
-        attrs:
-          name: Paperclip
-          slug: paperclip-gate
-          provider: !KeyOf paperclip-gate-provider
-          meta_launch_url: https://paperclip.renner.uno
-          meta_description: "Paperclip agent orchestration"
-
-      - model: authentik_policies.policybinding
-        identifiers:
-          target: !Find [authentik_core.application, [slug, paperclip-gate]]
-          group: !KeyOf paperclip-users-group
-          order: 0
-        state: present
-        attrs:
-          enabled: true
-          negate: false
-
-      - model: authentik_outposts.outpost
-        identifiers:
-          name: paperclip-gate
-        state: present
-        attrs:
-          type: proxy
-          service_connection: !KeyOf k8s-conn
-          providers:
-            - !KeyOf paperclip-gate-provider
-          config:
-            log_level: info
-            authentik_host: "http://authentik-server.authentik.svc.cluster.local"
-            authentik_host_insecure: false
-            authentik_host_browser: "https://auth.renner.uno/"
-            refresh_interval: "minutes=5"
-            container_image: null
-            object_naming_template: "ak-outpost-%(name)s"
-            kubernetes_namespace: authentik
-            kubernetes_replicas: 1
-            kubernetes_service_type: ClusterIP
-            kubernetes_disabled_components: ["ingress"]
-            kubernetes_image_pull_secrets: []
-            kubernetes_ingress_annotations: {}
-            kubernetes_ingress_class_name: null
-            kubernetes_ingress_secret_name: "authentik-outpost-tls"
-            kubernetes_json_patches: null
-
-```
-
- [ ] **Step 2: `paperclip-gate-egress.yaml`:**
-
-```yaml
-# Lets the paperclip-gate outpost reach the Paperclip server; scoped to this outpost's pods.
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: paperclip-gate-egress
-  namespace: authentik
-spec:
-  endpointSelector:
-    matchLabels:
-      goauthentik.io/outpost-name: paperclip-gate
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": paperclip
-            app.kubernetes.io/name: paperclip
-      toPorts:
-        - ports:
-            - port: "3100"
-              protocol: TCP
-```
-
- [ ] **Step 3: Register** — add `  - paperclip-gate-egress.yaml` after `  - paperless-gate-egress.yaml` in `apps/authentik/kustomization.yaml`.
-
- [ ] **Step 4: Render + dry-run.**
-
-Run: `kubectl kustomize kubernetes/apps/authentik`
-Expected: exit 0 (valid YAML incl. the blueprint ConfigMap).
-
-Run: `kubectl apply --dry-run=server -f kubernetes/apps/authentik/paperclip-gate-egress.yaml`
-Expected: `ciliumnetworkpolicy.cilium.io/paperclip-gate-egress created (server dry run)`.
-
- [ ] **Step 5: Commit** `feat(authentik): paperclip-gate outpost and Paperclip Users group`.
-
---
-
-### Task 8: Flux Kustomization
-
-**Files:**
- Create: `kubernetes/clusters/homelab/paperclip.yaml`
-
- [ ] **Step 1:**
-
-```yaml
-apiVersion: kustomize.toolkit.fluxcd.io/v1
-kind: Kustomization
-metadata:
-  name: paperclip
-  namespace: flux-system
-spec:
-  interval: 10m
-  retryInterval: 1m
-  timeout: 20m
-  sourceRef:
-    kind: GitRepository
-    name: flux-system
-  path: ./kubernetes/apps/paperclip
-  prune: true
-  wait: true
-  dependsOn:
-    - name: infrastructure
-    - name: infrastructure-crds
-    - name: infrastructure-ceph-cluster
-    - name: infrastructure-cnpg-barman
-  healthCheckExprs:
-    - apiVersion: postgresql.cnpg.io/v1
-      kind: Cluster
-      failed: "has(status.phase) && status.phase.startsWith('Failed')"
-      current: "has(status.phase) && (status.phase == 'Cluster in healthy state' || status.readyInstances > 0)"
-```
-
- [ ] **Step 2: Server dry-run.**
-
-Run: `kubectl apply --dry-run=server -f kubernetes/clusters/homelab/paperclip.yaml`
-Expected: `kustomization.kustomize.toolkit.fluxcd.io/paperclip created (server dry run)`.
-
- [ ] **Step 3: Commit** `feat(paperclip): Flux Kustomization`.
-
---
-
-### Task 9: PR, merge by user, live verification
-
- [ ] **Step 1: Push and open the PR** (Task 3 must be complete).
-
-Run: `git -c credential.helper='store --file=/home/krenner/.config/claude-bot/creds' push -u origin feat/paperclip`
-Run: `tea pr create --login claude-bot --base master --head feat/paperclip --title "feat(paperclip): deploy Paperclip (server-local runs, trial)" --description "<summary of spec decisions + link to spec + bootstrap steps>\n\n🤖 Generated with [Claude Code](https://claude.com/claude-code)"`
-Then wait for the user to merge. Do not merge.
-
- [ ] **Step 2: Reconciliation.** After merge:
-
-Run: `flux get kustomizations paperclip`
-Expected: `Ready True`. If False, read `kubectl describe kustomization paperclip -n flux-system` and the failing object's events; diagnose, don't poke.
-
-Run: `kubectl get cluster -n paperclip pclp-pg`
-Expected: `Cluster in healthy state`, 2 ready instances.
-
-Run: `kubectl get pods -n paperclip -l app.kubernetes.io/name=paperclip`
-Expected: `1/1 Running`.
-
- [ ] **Step 3: Gate enforced (Review Focus 4).**
-
-Run: `curl -sS -o /dev/null -w '%{http_code} %{redirect_url}\n' https://paperclip.renner.uno/`
-Expected: `302` redirecting to `https://paperclip.renner.uno/outpost.goauthentik.io/…` or `auth.renner.uno`, never `200`.
-
- [ ] **Step 4: Gate not bypassable (Review Focus 3).**
-
-`default` has no CNPs (verified 2026-09-25), so its egress is open and only paperclip's ingress policy can block the probe.
-
-Run: `kubectl get ciliumnetworkpolicies -n default`
-Expected: `No resources found` (else pick another CNP-free namespace).
-
-Run: `kubectl run cnp-probe -n default --rm -i --restart=Never --image=curlimages/curl:8.10.1 -- curl -sS -m 5 http://paperclip.paperclip.svc.cluster.local:3100/api/health`
-Expected: timeout (exit 28), no HTTP response.
-
- [ ] **Step 5: First backup (Review Focus 5).**
-
-Run: `kubectl get backup -n paperclip`
-Expected: one `pclp-pg-daily-…` with phase `completed`.
-
- [ ] **Step 6: Image policy.**
-
-Run: `kubectl get imagepolicy -n flux-system paperclip -o jsonpath='{.status.latestRef}'`
-Expected: `nightly` tag with a `sha256:` digest.
-
- [ ] **Step 7: Hand-over to the user (browser steps, theirs):**
-  1. Open `https://paperclip.renner.uno`, pass Authentik, claim the instance as admin.
-  2. Create a test agent with the `claude-local` adapter and a trivial task ("reply with OK"); it must complete with the shared token (Review Focus 2: if it fails, report the run's error text — an expired token must show as an auth error, not a hang).
-  3. If `SIGNUP_DISABLED` was `"false"`, open the follow-up PR flipping `PAPERCLIP_AUTH_DISABLE_SIGN_UP` to `"true"` once the admin is claimed.
-  4. Review Focus 1 is observed on the first nightly bump: check the interrupted-run status in the UI.
-
- [ ] **Step 8: Wiki.** Route the verified Paperclip facts (execution topology, env vars, image tags, digest-reflection usage) through wiki-ingest → wiki-distill once the deployment has proven out.
@@ -1,181 +0,0 @@
-# Paperclip in the homelab cluster — design
-
-Date: 2026-09-25
-Status: draft, awaiting review
-
-## Goal
-
-Run [Paperclip](https://github.com/paperclipai/paperclip) (MIT, agent-orchestration
-server + UI) in the homelab cluster for real use: its agents do actual work against
-homelab systems. This spec covers the **platform** only. Bot identities and their
-credentials are out of scope (see "Deferred").
-
-## Decisions
-
-| Topic | Decision |
-|---|---|
-| Execution model | **B — server-local runs** (upstream default): agent CLIs are child processes of the server, in the server container. Trial only. |
-| Upgrade path | If Paperclip is kept: **C** — `@paperclipai/plugin-kubernetes` sandbox provider + kubernetes-sigs agent-sandbox controller (`agents.x-k8s.io`), per-run Pods and Secrets. Separate spec. |
-| Claude auth | **One shared Claude subscription AI connection**, created once via Paperclip's local sign-in (`claude auth login` into `ai-local-logins/<id>`); Paperclip stores and refreshes it. Server-env `CLAUDE_CODE_OAUTH_TOKEN` is not used: `stripAiAuthBindings` blanks it for managed runs, and upstream accepts no pasted setup-token (the device-code flow needs a Daytona sandbox). |
-| Bot credentials | Out of scope. Future "AIR-Manager" Paperclip role handles permission requests. |
-| Exposure | LAN-only, behind an Authentik forward-auth gate, then Paperclip's own login. |
-
-### Known trade-off of B (accepted for the trial)
-
-Upstream `runChildProcess()` spawns every local adapter with
-`{...sanitizeInheritedPaperclipEnv(process.env), ...opts.env}`; the sanitizer strips only
-`PAPERCLIP_*` keys. Every agent therefore sees `DATABASE_URL`, `BETTER_AUTH_SECRET`,
-the shared filesystem (including every other agent's workspace and the
-`ai-local-logins` credentials) and every agent-scoped secret it can read from disk. Under B, no bot may hold
-a credential we would not tolerate any other bot using. Per-bot isolation is C's job.
-
-## Architecture
-
-```
-LAN client ─▶ Traefik ─▶ ExternalName svc ─▶ ak-outpost-paperclip-gate (authentik ns)
-                                                   │ forward-auth OK
-                                                   ▼
-                                  paperclip Deployment (ns paperclip, 1 replica)
-                                   ├─ server :3100  (+ agent CLIs as child processes)
-                                   ├─ PVC /paperclip  (ceph-block)
-                                   └─▶ pclp-pg-rw (CNPG, PG17, 2 instances) ─▶ barman ▶ Hetzner S3
-```
-
-## Components
-
-All modeled on `kubernetes/apps/paperless/` unless stated.
-
-### Namespace and baseline policy
- `Namespace paperclip` in `kubernetes/infrastructure/namespaces/namespaces.yaml`.
- Baseline default-deny CNP `kubernetes/infrastructure/network-policies/paperclip-ns.yaml`
-  (copy of `paperless-ns.yaml`, renamed).
- Add `paperclip` to the `openbao` ClusterSecretStore namespace allowlist.
-
-### Flux
- `kubernetes/clusters/homelab/paperclip.yaml`: Flux `Kustomization`,
-  `path: ./kubernetes/apps/paperclip`, `prune: true`, `wait: true`, same `dependsOn` and
-  CNPG `healthCheckExprs` as `paperless.yaml`.
- `kubernetes/apps/paperclip/kustomization.yaml`: plain resource list.
-
-### Server Deployment (`paperclip.yaml`)
- Image `ghcr.io/paperclipai/paperclip:nightly@sha256:<digest>` with the Flux
-  image-automation marker for digest updates.
- `replicas: 1`, `strategy: Recreate` (single writer on an RWO PVC).
- Starts as root: upstream `docker-entrypoint.sh` remaps the `node` user, chowns
-  `/paperclip`, then drops to uid 1000 via `gosu`. Caps limited to CHOWN,
-  DAC_OVERRIDE, FOWNER, SETGID, SETUID; `fsGroup: 1000`.
- `automountServiceAccountToken: false` (all agents share the pod filesystem).
- Port 3100; `Service paperclip` ClusterIP :3100.
- Probes: startup and readiness on `GET /api/health` with `Host: localhost` (the
-  private-hostname guard rejects the pod IP); liveness is a TCP check, since
-  `/api/health` runs `SELECT 1` and a DB outage must not restart the server.
- Env:
-  - `DATABASE_URL` assembled from CNPG `pclp-pg-app` Secret (`uri` key).
-  - `HOST=0.0.0.0`, `PORT=3100`
-  - `PAPERCLIP_DEPLOYMENT_MODE=authenticated`
-  - `PAPERCLIP_DEPLOYMENT_EXPOSURE=private`
-  - `PAPERCLIP_PUBLIC_URL=https://paperclip.renner.uno`
-  - `PAPERCLIP_AUTH_DISABLE_SIGN_UP=false` until the first admin is claimed: the
-    claim (`/bootstrap/claim`) requires an already signed-in Better Auth account.
-    A follow-up flips it to `true`.
-  - `PAPERCLIP_TELEMETRY_DISABLED=1` (read by `resolveTelemetryConfig`).
-  - `PAPERCLIP_API_URL=http://paperclip.paperclip.svc.cluster.local:3100` plus
-    `PAPERCLIP_ALLOWED_HOSTNAMES=paperclip.paperclip.svc.cluster.local`: upstream
-    otherwise hands agents the public URL, and the gate (`intercept_header_auth`)
-    rejects their bearer tokens. The Service name also works for sandbox pods under
-    option C; agent JWTs still authenticate each call.
-  - `envFrom: secretRef paperclip-secrets`
- Reloader: `secret.reloader.stakater.com/reload: paperclip-secrets`.
-
-### Storage
- `PVC paperclip-data`, `ceph-block`, 20Gi, mounted at `/paperclip` (`PAPERCLIP_HOME`
-  and `HOME`): instance config, workspaces, repo checkouts, per-agent `claude-runtime`
-  dirs.
-
-### Postgres
- CNPG `Cluster pclp-pg`: 2 instances, `imageName: ghcr.io/cloudnative-pg/postgresql:17`,
-  `ceph-block` 10Gi, `bootstrap.initdb {database: paperclip, owner: paperclip}`,
-  barman-cloud plugin as WAL archiver.
- `ObjectStore pclp-pg-store` → `s3://renner-uno-k8s-cnpg/pclp-pg`, creds via the
-  existing `hetzner-s3-creds` ExternalSecret pattern.
- `ScheduledBackup` nightly, `method: plugin`, `backupOwnerReference: self`.
- `kustomize.toolkit.fluxcd.io/prune: disabled` on Cluster, ObjectStore, ScheduledBackup.
- PG17 deviates from the cluster-wide PG16 convention because upstream's external
-  Postgres target is `postgres:17` (`docker/docker-compose.yml`) and its embedded default
-  is PG18.
-
-### Secrets
-One ExternalSecret `paperclip-secrets` ← OpenBao `apps/paperclip` (whole-object
-extract):
- `BETTER_AUTH_SECRET`
- `PAPERCLIP_TOOL_ACTION_SIGNING_SECRET`
- `PAPERCLIP_SECRETS_MASTER_KEY` (64-char hex). Held in OpenBao, not on the PVC, so
-  encrypted agent secrets in Postgres survive PVC loss.
-
-### Ingress and SSO
- Hostname `paperclip.renner.uno`, Traefik `Ingress`, `cert-manager.io/cluster-issuer:
-  letsencrypt-cluster`, **no** `dns.renner.uno/public` label (LAN-only).
- Authentik gate, per the paperless pattern:
-  - blueprint entries in `kubernetes/apps/authentik/blueprints-configmap.yaml`:
-    proxy provider `paperclip-gate` (`internal_host:
-    http://paperclip.paperclip.svc.cluster.local:3100`, `external_host:
-    https://paperclip.renner.uno`), application, group "Paperclip Users", policy binding;
-  - outpost `ak-outpost-paperclip-gate`;
-  - `kubernetes/apps/authentik/paperclip-gate-egress.yaml` (no `auth.renner.uno`
-    outpost Ingress: like wiki/shelly gates, the outpost serves its own callback on
-    `paperclip.renner.uno`);
-  - `ExternalName` Service in `paperclip` ns pointing at the outpost; the app Ingress
-    targets it on :9000.
- Paperclip's own email/password login stays on (no header/OIDC login upstream).
-
-### Network policy (`kubernetes/apps/paperclip/network-policy.yaml`)
-Covers the server **and every agent** (same pod under B).
- Ingress to the app pod: only from the `paperclip-gate` outpost (by label) on 3100.
- Egress from the app pod:
-  - `pclp-pg` pods :5432
-  - kube-dns :53
-  - `ingress` namespace (Traefik) :8443 (pod-side websecure port) — reaches Gitea, Woodpecker, HA, Loki by their
-    `*.renner.uno` names
-  - `kube-apiserver` entity :6443
-  - `world` TCP :443 (Anthropic API, GitHub, npm)
- No SSH egress; git over HTTPS only. The namespace baseline
-  (`paperclip-ns.yaml`) also restricts `world` to TCP 443 — Cilium unions allows
-  across CNPs, so an unrestricted baseline would override this list.
- Known gap (house baseline): pods in `monitoring`/`database` and other pods in
-  `paperclip` can reach :3100 without the gate; Paperclip's own login still applies.
-
-### Image automation
-`kubernetes/infrastructure/image-automation/paperclip.yaml`: `ImageRepository`
-(`ghcr.io/paperclipai/paperclip`) + `ImagePolicy` filtering tag `^nightly$` with
-`digestReflectionPolicy: Always` (image-reflector-controller v1.2.5 in cluster; field
-available since v0.35.0). First use of digest reflection in this repo.
-
-## Guardrails
- No bot receives cluster-admin or any binding in `flux-system`, `authentik`, `openbao`.
- Git work goes through PRs; bots get no merge rights on protected branches.
- Nothing is granted under B that we would not tolerate any agent misusing.
-
-## Bootstrap (one-time)
-1. Place `apps/paperclip` values in OpenBao (the three generated secrets).
-2. Merge; Flux deploys with sign-up enabled.
-3. Sign up through the gate and claim the instance as admin; then set
-   `PAPERCLIP_AUTH_DISABLE_SIGN_UP=true`.
-4. In onboarding "Connect a model", choose Claude subscription and run the shown login
-   **as the node user** (the entrypoint makes exec sessions root, and root-owned 0600
-   credentials are unreadable to the server):
-   `kubectl exec -it -n paperclip deploy/paperclip -- gosu node sh -c 'export CLAUDE_CONFIG_DIR=<shown path> && mkdir -p "$CLAUDE_CONFIG_DIR" && claude auth login'`
-   Repeat after loss of the PVC or database.
-
-## Verification
- Flux Kustomization `paperclip` Ready; CNPG `pclp-pg` healthy; first ScheduledBackup
-  completes.
- `curl https://paperclip.renner.uno/api/health` through the gate returns healthy after
-  Authentik login; unauthenticated request redirects to `auth.renner.uno`.
- Direct pod access from another namespace is denied (CNP).
- A test agent (claude-local) completes a trivial task using the shared Claude connection.
- Flux ImagePolicy reports a `nightly@sha256:` digest.
-
-## Deferred
- **Option C** (sandboxed runs) — required before any per-bot credential isolation.
- **AIR-Manager** role and all bot identities/credentials.
- Public exposure.
@@ -50,18 +50,15 @@ pass insert homelab/openbao/root-token
 kubectl -n openbao exec -it openbao-0 -- bao operator unseal   # paste key
 ```

-**Unseal is automatic since 2026-09-08** (static seal, PR #825, issue #805):
-the seal key lives in the `openbao-seal-key` Secret (created by `bootstrap.sh`
-from `pass show homelab/openbao/seal-key`) and every restarted replica unseals
-itself. The shamir key above became the **recovery key** in the migration; it
-is kept in `pass` under its old name, `homelab/openbao/unseal-key`. The server
-runs three raft peers (see `infrastructure/openbao/helmrelease.yaml`); a sealed
-one is neither a voter nor a standby, so a replica that stays sealed after a
-restart — wrong or missing seal-key Secret — costs redundancy, two cost
-quorum. The trade-off taken with the static seal: the key that unlocks the
-store sits in the same etcd as the Secrets ESO already syncs there in
-plaintext. Symptom of sealed: ESO stores go NotReady, synced Secrets keep
-their last value (they are real k8s Secrets — pods keep running).
+**Manual unseal is required after every pod restart** (node drain, upgrade,
+crash), **per replica** — the server runs three raft peers (see
+`infrastructure/openbao/helmrelease.yaml`), and a sealed one is neither a
+voter nor a standby, so it stays out of the cluster until unsealed:
+`kubectl -n openbao exec -it openbao-<N> -- bao operator unseal`, same key for
+all three. Three peers mean one sealed replica is survivable; two are not.
+Accepted for now; auto-unseal needs a second trust root (transit/KMS) and is
+explicitly out of scope. Symptom of sealed: ESO stores go NotReady, synced
+Secrets keep their last value (they are real k8s Secrets — pods keep running).

 ## Phase 2 — configure + ClusterSecretStore (follow-up PR)

@@ -237,19 +234,16 @@ one full reconcile + one OpenBao pod restart/unseal cycle:
 - **Backups:** ✅ nightly raft-snapshot CronJob (03:00, keeps 7 on a
  dedicated PVC; see infrastructure/openbao/snapshot-cronjob.yaml) + velero
  `openbao-daily` Schedule (06:15) ships that PVC offsite, 30 d retention.
-  A raft snapshot restore needs the matching seal key — which lives in
+  A raft snapshot restore needs the matching unseal key — which lives in
  `pass`, outside the cluster (GPG key backed up externally 2026-07-14).
  A failed snapshot Job is the sealed-after-restart alarm. Still owed:
  one tested restore before phase 4.
- **Cluster rebuild order:** seal-key Secret from `pass` → Flux brings up
-  openbao (empty) → init → restore raft snapshot → replicas unseal themselves
-  → ClusterSecretStore Ready → ESO materializes Secrets → apps start
-  (`bootstrap.sh flux` does all of it; RUNBOOK-cold-rebuild.md Step 4 by
-  hand). Until then, secret-consuming apps simply stay Pending on missing
-  Secrets — same as any missing dependency.
- **Break-glass:** `pass` holds seal key, recovery key + root token (GPG,
-  offline-capable). With those, everything is recoverable from a raft
-  snapshot alone.
+- **Cluster rebuild order:** Flux brings up openbao (sealed, empty) → restore
+  raft snapshot → unseal (key from `pass`) → ClusterSecretStore Ready → ESO
+  materializes Secrets → apps start. Until unseal, secret-consuming apps
+  simply stay Pending on missing Secrets — same as any missing dependency.
+- **Break-glass:** `pass` holds unseal key + root token (GPG, offline-capable).
+  With those two, everything is recoverable from a raft snapshot alone.
 - **Ansible Vault password:** `ansible/vault-password.sh` reads
  `ANSIBLE_VAULT_PASSWORD` from the environment with no offline backup as of
  2026-07-14. It gates decryption of several Ansible-Vault files (PBS
@@ -1,99 +0,0 @@
-# Traffic classes on the k8s nodes
-
-Issue #576. Five classes, each on its own VLAN.
-
-## Where each class runs
-
-| class | link | mechanism |
-|---|---|---|
-| Ceph replication | `ceph-san`, VLAN 49 | the whole link is handed to the OSD pod (Multus `host-device`) |
-| pod↔pod | `pod0`, VLAN 48 | pod-sourced traffic is routed from table 47; bird puts the peers' pod CIDRs there via `pod0` |
-| world↔pod | `ingress0`, VLAN 47 | BGP session sourced there (next hop of every LoadBalancer VIP); pod egress takes table 47's default |
-| node↔node | `node0`, VLAN 46 | the node InternalIP and etcd live there; so do apiserver↔kubelet, cilium-health and host→remote-pod traffic |
-| world↔node | `mgmt0`, VLAN 40 | API VIP for kubectl, talosctl from outside, node egress (DNS, NTP, image pulls) |
-
-Hard constraint, set by the owner: inter-node and pod traffic never traverses
-the UDM. Everything but world↔node stays on the 10G switch. VLAN 46 and 48 are
-statically addressed, without a gateway; VLAN 46 is isolated in UniFi.
-
-## Pod traffic: one table
-
-Two rules, one per family, send every packet with a pod source to table 47
-(`talos/patches/site-homelab.yaml`, priorities 1000/1001, below main). The
-table decides by longest prefix; no rule carves out exceptions.
-
-bird (`kubernetes/infrastructure/bird/`) owns table 47 and fills it with:
-
- the node's attached networks — `mgmt0`, `ingress0`, `pod0`, `node0` —
-  taken from the interfaces;
- the peers' pod CIDRs, learned over an iBGP mesh on `pod0`, next hop the
-  peer's `pod0` address;
- the node's own pod CIDR via `cilium_host`, originated rather than read back
-  from the kernel (reading it back made BIRD depend on the shape of Cilium's
-  route, and KRT rejects the IPv4 one);
- the egress default via the UDM on `ingress0`, from the node annotations
-  `net.renner.uno/egress-gw-v4/v6`.
-
-`node0` has to be in the table: a pod reaching a node address — the
-`kubernetes` ClusterIP, which socket LB rewrites to the apiserver's advertise
-address, or a hostNetwork pod — would otherwise take the default to the UDM.
-
-The routing decision runs before Cilium's masquerade hook, so the rule sees the
-pod address and the masquerade then rewrites it to the egress link's.
-
-Node-originated traffic does not match the rules and uses main, where Cilium
-routes each peer's pod CIDR via the peer's InternalIP — on `node0`.
-
-No address is written into this repo twice. Each node publishes its `pod0`
-addresses as `nodeAnnotations` beside the `VLANConfig` that assigns them; the
-pod CIDRs come from the Node objects. bird's init container renders both from
-the API and ends with `bird -p`, so BIRD's own parser gates startup.
-
-## Node↔node
-
-`kubelet.nodeIP.validSubnets` and `cluster.etcd.advertisedSubnets` point at
-VLAN 46. Talos picks one InternalIP per family, IPv4 first — the service CIDR
-is IPv4, and the apiserver's advertise address must match it.
-
-The API VIP stays on `mgmt0` for kubectl. No in-cluster client uses it:
-kubelet, scheduler, controller-manager and Cilium go through KubePrism
-(`localhost:7445`), which balances over every address of every control-plane
-member; pods use the ClusterIP.
-
-Talos never updates an existing etcd member's peer URLs
-(siderolabs/talos#12720). Changing `advertisedSubnets` means, per member:
-stage the config, `talosctl etcd forfeit-leadership`, `talosctl etcd leave`,
-reboot. `leave` removes only `/var/lib/etcd`; the node rejoins as a learner
-and is promoted. EPHEMERAL does not need a wipe.
-
-## Moving node IPs: the mixed state
-
-While some nodes have their InternalIP on the new network and some on the old,
-host→remote-pod traffic leaves with a node address that is not an InternalIP:
-the kernel picks the source from the route to the peer's InternalIP. Cilium
-classifies that as `world`, and every ingress policy that admits only
-`host`/`remote-node` (rook-ceph, monitoring, …) drops it. Kernel Ceph clients
-lose the OSDs on the other side; RBD mounts hang.
-
-Bridge it for the duration with a temporary `CiliumClusterwideNetworkPolicy`
-admitting the node subnets of both networks, both families, with
-`enableDefaultDeny: false` so it restricts nothing else, and delete it once
-every node has moved. Keep all OSD nodes schedulable while one is drained:
-draining two takes Ceph below its data availability.
-
-## Failure behaviour
-
-If bird stops, table 47 empties, the rules find nothing and fall through to
-main: pod traffic returns to the node networks, unseparated rather than
-misrouted. The same holds during a cold rebuild until Gitea serves bird's
-image. The default route lives in table 47, not in the Talos config, so a
-default cannot outlive the specific routes and send pod traffic to the UDM.
-
-## Sources
-
-Cilium v1.20 `pkg/datapath/linux/node.go`; BIRD 2.17 user manual (kernel
-protocol, `strict bind`, `direct`); Talos v1.13 `RoutingRuleConfig`,
-`machine.nodeAnnotations`, configuration patching (lists are appended,
-`$patch: delete`), etcd `LeaveCluster`; KubePrism docs;
-siderolabs/talos#12720. PRs #831, #833, #835, #836, #837, #842, #852, #854,
-#855, #857, #858, #918, #919, #922; closed unmerged: #830, #838, #839.
@@ -0,0 +1,194 @@
+# Proposal — separate the four traffic classes onto their own VLANs
+
+Status: draft. The question this started with is **answered** — from the Cilium
+source and flag reference, not from a test — and the answer changes the shape
+of the proposal rather than confirming it.
+
+Supersedes the framing in issue #576, which described this as "give node
+management its own NIC, and Ceph public its own VLAN". That was two of four
+classes and one of them turned out already solved.
+
+## The four classes
+
+| class | what it is | where it goes today |
+|---|---|---|
+| **Ceph replication** | OSD-to-OSD heartbeat, replication, recovery | VLAN 49, its own X710 port — **separated** |
+| **cluster-internal** | pod-to-pod east-west, including Ceph client I/O | pod network, VXLAN over `mgmt0` |
+| **cluster-external** | ingress to LoadBalancer VIPs, egress to LAN and internet | `mgmt0` |
+| **management** | Talos API, etcd, kube-apiserver, kubelet | `mgmt0` |
+
+Three of four share one interface. That is not a design decision — until the
+bare-metal conversion there was one NIC to work with.
+
+Ceph client I/O living on the pod network is correct rather than a leftover:
+that traffic *is* cluster-internal. Ceph is done.
+
+## Principle
+
+**Each class gets its own VLAN. Which physical interface a VLAN lands on is a
+separate, changeable decision.**
+
+Defining classes by VLAN means the separation survives recabling, port failures
+and capacity changes, and it does not require one port per class. Defining them
+by interface would freeze a capacity decision into the architecture.
+
+## The finding that shapes everything else
+
+The original open question was whether `nodePort.directRoutingDevice` moves the
+next-hop of Cilium's direct routes, or only selects an interface. It is
+answerable from the source, and the answer is: **only the interface.**
+
+`installDirectRoute` takes the next-hop as a parameter:
+
+```go
+func installDirectRoute(log *slog.Logger, prefix netip.Prefix, nodeIP net.IP,
+    skipUnreachable bool) (routeSpec *netlink.Route, err error)
+```
+
+and `nodeUpdate` supplies it as `newNode.GetNodeIP(false)`. `GetNodeIP` returns,
+in priority order, `NodeInternalIP` → `NodeExternalIP` → anything else. The
+same `newIP4` is what tunnel mode uses as the VXLAN endpoint.
+
+`NodeInternalIP` comes from the Kubernetes Node object. `--ipv4-node` looks like
+an override but is not one: `initFromConfig()` calls `SetNodeInternalIP` and
+`initFromK8s()` runs **after** it and overwrites the value. In a Kubernetes
+cluster there is no path by which Cilium's node IP differs from the kubelet's
+`InternalIP`.
+
+`--direct-routing-device` is documented as "Device name used to connect nodes in
+direct routing mode" — a device, not an address.
+
+**So cluster-internal traffic is welded to the node IP, and the node IP is the
+kubelet InternalIP.** Two consequences:
+
+- The class cannot be moved off whatever VLAN the node IP lives on.
+- **Routing mode is irrelevant to this problem.** Tunnel mode sources the VXLAN
+  from the node IP; native mode uses it as the direct-route next-hop. Native
+  routing buys nothing here, so the disruptive, non-rolling step this proposal
+  previously contained is simply gone.
+
+## Consequence: draw the line the other way
+
+If cluster-internal cannot leave the node IP, then separation has to come from
+moving *management services* off it — and each one has to be checked
+individually, because they do not share a mechanism.
+
+| management service | can it leave the node IP? | lever |
+|---|---|---|
+| etcd | **yes** | `cluster.etcd.advertisedSubnets` (+ `listenSubnets`, which defaults to it) |
+| kube-apiserver endpoint | **yes** | already a VIP — `Layer2VIPConfig` with a `link:` field |
+| Talos API | **n/a** | `apid` answers on every node address (measured); a client and firewall choice, not a bind setting |
+| kubelet | **no** | it *is* the node IP, by definition |
+
+So four VLANs are reachable, but the boundary does not fall where the class
+names suggest: **kubelet stays with cluster-internal.** That is the honest
+result, and it is worth stating plainly rather than filing kubelet under
+management and hoping.
+
+The remaining classes keep the levers established earlier:
+
+- **cluster-external ingress** — `CiliumBGPNodeConfigOverride.spec.bgpInstances[].peers[].localAddress`.
+  For eBGP the session source is the next-hop, so pinning the session pins
+  where return traffic arrives. The session to the UDM has been running for
+  months.
+- **cluster-external egress** — eBPF masquerading follows the output device, so
+  this is a default-route decision plus `enable-masquerade-to-route-source`
+  (today `false`).
+- **Ceph replication** — Multus macvlan on VLAN 49, done.
+
+## The general escape hatch
+
+Ceph is separated because its OSDs got a second interface and stopped using the
+pod network for replication. That is the *only* mechanism that fully detaches a
+workload from the node IP, and it generalises: anything that must not ride the
+node IP gets its own `NetworkAttachmentDefinition`.
+
+It is also expensive per workload — an attachment, an IPAM pool, a pod spec
+change. Worth it for Ceph replication. Not worth it for east-west traffic in
+general.
+
+## Recommended mapping
+
+Two ways to arrange this, and they differ only in which VLAN gets relabelled:
+
+**A — move the node IP to a new cluster VLAN.** VLAN 40 keeps etcd, the API VIP
+and Talos API. Cleanest match to the class names, but it changes every node's
+identity, which is the one address the whole cluster is administered through.
+
+**B (recommended) — leave the node IP on VLAN 40 and move management off it.**
+VLAN 40 becomes the cluster VLAN; a new VLAN carries etcd and the API VIP.
+Identical end state, and no node ever changes its InternalIP.
+
+Under B:
+
+| VLAN | carries |
+|---|---|
+| 49 | Ceph replication |
+| 40 (relabelled *cluster*) | pod east-west, kubelet |
+| new *management* | etcd, kube-apiserver VIP, Talos API by convention |
+| new *external* | LoadBalancer ingress, egress default route |
+
+The property the user asked for — management carries only management traffic —
+holds under B. The label moves; the property is what matters.
+
+## Migration shape
+
+Every step is rolling and reversible. The routing-mode change is gone.
+
+1. **New VLANs** on the switch, trunked to the X710. Nothing uses them.
+2. **Addresses** on the new VLANs via Talos, one node at a time, `--mode=try`.
+   Still unused; this only proves the L2 works.
+3. **etcd** to the management VLAN via `advertisedSubnets`. One node at a time,
+   waiting for a healthy quorum between each.
+4. **API VIP** to the management VLAN via `Layer2VIPConfig.link`. DNS for
+   `k8s.renner.uno` follows.
+5. **cluster-external**: BGP `localAddress`, default route, masquerade flag.
+   Reversible by reverting the override; failure is visible immediately in
+   `cilium-dbg shell -- bgp/peers`.
+6. Verify VLAN 40 by what is *absent* from it, not by what still works.
+
+Step 3 is the one that deserves care — etcd is the only step where a mistake
+costs quorum rather than connectivity.
+
+## What this does not buy
+
+Management currently shares a 10 Gb link and is not contended. Nothing observed
+today is slow because of this. The case for the split is that traffic classes
+with different failure modes and security postures should not be
+indistinguishable — not that something is broken.
+
+## Sources and measurements
+
+Source, Cilium v1.20.1 (the running version):
+
+- `pkg/datapath/linux/node.go` — `installDirectRoute(…, nodeIP net.IP, …)`,
+  fed from `newIP4 = newNode.GetNodeIP(false)`; `createDirectRouteSpec` rejects
+  a next-hop that is not directly reachable.
+- `pkg/node/types/node.go` — `GetNodeIP` prefers `NodeInternalIP`.
+- `pkg/node/sync/local_node_sync.go` — `initFromConfig()` applies
+  `--ipv4-node`, `initFromK8s()` runs afterwards and overwrites it from the
+  Kubernetes Node.
+- `cilium-agent` flag reference — `--direct-routing-device`: "Device name used
+  to connect nodes in direct routing mode".
+
+Talos:
+
+- Multihoming guide — `cluster.etcd.advertisedSubnets`; `listenSubnets`
+  defaults to it. kube-apiserver and Talos API listen addresses are not
+  covered there.
+- `kubernetes/talos/patches/k8s-{1,2,3}.yaml` — `Layer2VIPConfig` for
+  `10.2.40.60` with `link: mgmt0`.
+
+Measured on the cluster:
+
+- `apid` answered on a node address that is not the InternalIP, so the Talos
+  API is not bound to one address.
+- k8s-1 currently holds one routable v4 address plus the VIP on `mgmt0`; its
+  InternalIP had changed since the last session, which is why nothing here
+  names an address as fixed.
+- Cilium runs `routing-mode: tunnel` with `enable-masquerade-to-route-source`
+  unset and no direct-routing device configured.
+
+Not verified: whether the kube-apiserver's `--advertise-address` follows the
+node IP independently of the VIP. It does not affect the recommendation, since
+clients reach the API through the VIP either way.
@@ -1,315 +0,0 @@
-# Wiki: local web interface and agent API, as a modular .NET solution
-
-## Goal
-
-A web interface for the Notes wiki that runs entirely in the cluster, and an
-HTTP API through which agents ask the wiki typed questions and deposit facts.
-Built as one C# solution with a MudBlazor front end, and built so that every
-external dependency is replaceable: where the notes live, which model embeds
-them, what judges a typed question, what runs an agent, how the vector index
-is stored. The first adapters are what the homelab uses today. The point of
-the design is that the next one is a project and a config line, not a
-rewrite.
-
-## Where we are
-
- `wiki-index` (Python, ns `ai`) syncs `wiki/` from `k.renner/notes` by blob
-  SHA, embeds through `wiki-embed` (bge-m3 on llama-server), answers
-  `POST /rank`. It keeps slug, title and vector, not the text. LAN-only behind
-  Traefik BasicAuth; the Gitea webhook hits the ClusterIP.
- `~/Notes/bin/wiki_rank.py` and `wiki_ask.py` rank through the index, build a
-  document context from a local vault checkout, send it with typed questions
-  to TypeSafe, gate on `covered`. The wiki-synth agent composes answers from
-  those judgements.
- New knowledge enters as a file in `raw/` (facts with provenance, never
-  hypotheses), distilled nightly in CI.
- The house C# project, shelly-manager, sets the conventions: net10.0, an
-  `.slnx` with `src/` and `test/`, xUnit, MudBlazor 9, Blazor Web App with
-  interactive server rendering, minimal APIs, an Authentik proxy-header
-  authentication handler for humans and an API-key handler for machines, a
-  multi-stage Dockerfile, `dotnet test` on every push and a kaniko release
-  pipeline to the Gitea registry. Human-facing apps sit behind an Authentik
-  outpost (`<app>-gate`) with a CNP that admits only that outpost.
-
-## Architecture: ports in the core, adapters around it
-
-```
-                 ┌──────────────────────────────────────────┐
-                 │  Wiki.Host  (ASP.NET Core + Blazor/Mud)  │  composition root,
-                 │  /api/v1, /rank, /webhook, UI pages      │  config, auth
-                 └───────────────┬──────────────────────────┘
-                                 │ uses
-                 ┌───────────────▼──────────────────────────┐
-                 │  Wiki.Application  (use cases)           │  Sync, Rank, Ask,
-                 │                                          │  DepositFact, Queue-
-                 └───────────────┬──────────────────────────┘  Question, Evaluate
-                                 │ depends only on
-                 ┌───────────────▼──────────────────────────┐
-                 │  Wiki.Core  (domain + ports)             │  Note, RawSource,
-                 │  no package references                   │  Report, LinkGraph,
-                 └──────────────────────────────────────────┘  the interfaces
-   implemented by ▲            ▲            ▲            ▲            ▲
-   Adapters.Gitea  Adapters.LlamaCpp  Adapters.TypeSafe  Adapters.FileIndex  Adapters.Anthropic
-   (later: Database, Ollama/Onnx, local Jev, pgvector/Qdrant, OpenAI-compatible answer models)
-```
-
-Rules that keep it modular, enforced by project references and a test:
-
- `Wiki.Core` references nothing. It holds the domain types and the port
-  interfaces. Adapters reference Core only; they never reference each other
-  or the host. `Wiki.Host` is the only project that knows which adapters
-  exist.
- Every port has a **contract test suite** in `test/Wiki.Contracts` that
-  runs against every adapter of that port. A new adapter is done when the
-  suite is green. That is what makes "swap the storage" a safe operation.
- Adapters are selected by configuration and registered by name (keyed
-  services), so several can be active at once. The UI's lab page and the
-  evaluator use that to compare, the API uses the one marked active.
- Nothing in Core or Application knows about HTTP, Gitea, TypeSafe, files or
-  Kubernetes.
-
-### Ports
-
-| port | what it abstracts | first adapter | examples of the next ones |
-|---|---|---|---|
-| `INoteStore` | read notes, raw sources and reports at a head; create a new document (a fact becomes a raw note in the application layer, the store only moves text) | Gitea repository API (`k.renner/notes`, blob SHA) | git checkout on a volume; a database (CNPG Postgres) with the vault imported; an Obsidian sync folder |
-| `IEmbeddingModel` | text → vector, with a model id and dimension | llama-server `/v1/embeddings` (bge-m3 via wiki-embed, the model that won the German-document comparison on accuracy per cost) | the qwen3 embedding variants from that comparison; ONNX in-process; any OpenAI-compatible endpoint |
-| `IVectorIndex` | store vectors per (model id, document), query top-k | in-memory cosine with a file snapshot per model, as today | pgvector in the database; Qdrant |
-| `IChunker` | split a document into addressable units for typed questions | paragraph splitter (headings, fences), as in `wiki_ask.py` | sentence windows; semantic chunks |
-| `IJudge` | typed questions over a state: Choice, Noul, Score, with probabilities | TypeSafe System One (`jev-latest`, cloud; the wiki records no locally hosted Jev today) | a locally hosted Jev with the same request shape, should TypeSafe offer one; a constrained-output local LLM adapter, flagged as uncalibrated |
-| `IAnswerModel` | prose from a question plus evidence: the selected paragraphs with their notes and the judgements; must cite by slug; streaming | Anthropic Messages API (Claude, model from config, Sonnet by default) | any OpenAI-compatible endpoint, including the local llama-server; a different Claude model per question |
-| `IEvaluator` | measure a retrieval or judge configuration against the labelled set (questions and cited sources in `output/`) | recall@k, hit rate, MRR, as `wiki_rank_eval.py` | judge accuracy against hand-labelled paragraphs |
-| `IIdentity` | who is calling, for attribution of facts and saved answers | Authentik proxy headers (humans), API key (machines), as in shelly-manager | OIDC bearer tokens |
-
-Two ports deserve a sentence of scope each. The judge returns judgements
-with probabilities, never prose, and the UI never presents its output as an
-answer, only as ranked, quoted evidence. The answer model turns that
-evidence into words and is allowed nothing else: it sees the selected
-paragraphs and the judgements, not the whole context, and its output is
-shown next to the evidence it was given. Both are single model calls inside
-the request. The service runs no agent: no tool loop, no vault edits. The
-wiki-synth agent in a Claude session remains the path for artifacts and for
-multi-question synthesis, and it can call `/questions` itself.
-
-### Solution layout
-
-```
-Wiki.slnx
-src/
-  Wiki.Core/               domain types, ports, frontmatter and link parsing
-  Wiki.Application/        use cases; depends on Core only
-  Wiki.Adapters.Gitea/     INoteStore (read at a head, create)
-  Wiki.Adapters.LlamaCpp/  IEmbeddingModel (OpenAI-compatible embeddings)
-  Wiki.Adapters.TypeSafe/  IJudge (System One client)
-  Wiki.Adapters.FileIndex/ IVectorIndex (memory + snapshot per model)
-  Wiki.Adapters.Anthropic/ IAnswerModel (Messages API, streaming)
-  Wiki.Host/               ASP.NET Core: API, webhook, MudBlazor UI, auth,
-                           composition root, appsettings
-test/
-  Wiki.Core.Tests/         parsing, link graph, chunker
-  Wiki.Contracts/          one suite per port, parameterised over adapters
-  Wiki.Application.Tests/  use cases with fake ports
-  Wiki.Host.Tests/         API surface, auth handlers
-```
-
-Own repository `k.renner/wiki`, same shape as shelly-manager: `Dockerfile`,
-`.woodpecker/test.yml` on pull requests, `.woodpecker/release.yml` building
-the image with kaniko into the Gitea registry on push to `main`, tagged
-`build.<unix-ts>` plus the commit SHA so Flux image automation can follow
-it. The homelab repo holds only the manifests. `claude-bot` needs write
-access to the new repository.
-
-Two things the shelly-manager history settles in advance: the TypeSafe
-client is written from the documented field contract (endpoint, bearer key,
-`jev-latest`, Noul/Choice/Score, 255-option cap, 32k-token state budget),
-there is no prior C# client to copy; and the UI needs no login page, because
-identity comes from the outpost's headers, which avoids the Blazor-circuit
-sign-in constraint shelly-manager ran into.
-
-### Configuration
-
-Adapters are declared in `appsettings` and selected by name; secrets come
-through environment variables, filled by ExternalSecret, never from the file.
-
-```json
-"Wiki": {
-  "NoteStore": { "Active": "gitea", "gitea": { "Repository": "k.renner/notes", "Branch": "main" } },
-  "Embeddings": {
-    "Active": "bge-m3",
-    "Models": {
-      "bge-m3":   { "Adapter": "LlamaCpp", "Url": "http://wiki-embed.ai.svc:8080", "Model": "bge-m3" },
-      "e5-large": { "Adapter": "Ollama",   "Url": "http://ollama.ai.svc:11434",    "Model": "multilingual-e5-large" }
-    }
-  },
-  "Judges": {
-    "Active": "typesafe",
-    "Models": {
-      "typesafe":  { "Adapter": "TypeSafe", "Url": "https://api.typesafe.ai/v1/systemone", "Model": "jev-latest" },
-      "jev-local": { "Adapter": "TypeSafe", "Url": "http://jev.ai.svc:8080/v1/systemone",  "Model": "jev" }   // illustration: same adapter, other endpoint
-    }
-  },
-  "Index": { "Active": "file", "file": { "Path": "/data/index" } },
-  "AnswerModels": {
-    "Active": "claude-sonnet",
-    "Models": {
-      "claude-sonnet": { "Adapter": "Anthropic", "Model": "claude-sonnet-5" },
-      "qwen-local":    { "Adapter": "OpenAI",    "Url": "http://llama-server.ai.svc:8080/v1", "Model": "qwen3.6-35b-a3b" }
-    }
-  }
-}
-```
-
-Every embedding model gets its own index namespace. Switching the active
-model is a config change after the new index has been built in the
-background; until then both exist and can be compared. Nothing is
-re-embedded in place.
-
-## The web interface (MudBlazor)
-
-Blazor Web App, interactive server rendering, hosted in `Wiki.Host`. Static
-assets vendored into the image, no CDN. Pages:
-
-1. **Search**: a query box, results ranked by the active index with score and
-   tags; for a phrased question, optionally the paragraph-level hits from the
-   judge with their probabilities, shown as quoted evidence.
-2. **Note**: frontmatter, rendered body with `[[wikilinks]]` resolved,
-   backlinks, sources linking to their raw file.
-3. **Browse**: by tag, by `updated`, recently changed, orphans.
-4. **Answers**: a question box. The answer streams in within seconds: rank,
-   judge, compose, all in the request. Below it the evidence: the paragraphs
-   the judge selected with their notes and probabilities, and the `covered`
-   gate; a low gate is shown as "the wiki does not cover this" instead of an
-   answer. A save button writes the answer to `output/` with the wiki-synth
-   frontmatter (`question`, `sources`, `confidence`), so it joins the
-   labelled set. Underneath, the saved answers, newest first, including the
-   ones the wiki-synth agent wrote from a session.
-5. **Maintenance**: the raw queue with ages, the last consolidation and audit
-   reports, index and sync status.
-6. **Lab**: pick any registered embedding model and judge, run the same query
-   through them side by side, and run the evaluator on the labelled set to
-   get recall@k per configuration. This is the page for "play around with
-   different models".
-7. **Settings**: read-only view of the active adapters and their health.
-
-No editing of notes in the browser. The vault stays a git repository edited
-by the consolidate agent and by hand.
-
-## API v1
-
-Reads are authenticated by the hostname's gate (humans) or API key
-(machines); writes always need an API key so a fact carries its author.
-
-| method and path | purpose |
-|---|---|
-| `GET /api/v1/status` | index commit and age, counts, active adapters |
-| `GET /api/v1/notes?q=&tag=&limit=` | list; `q` ranks through the active index |
-| `GET /api/v1/notes/{slug}` | frontmatter, body, links, backlinks, sources |
-| `GET /api/v1/tags` | tags with counts |
-| `POST /api/v1/rank` | `{question, k, model?}` → ranked slugs; also served at `/rank` for the existing scripts |
-| `POST /api/v1/ask` | `{queries, questions, k, judge?, model?}` → context, typed judgements with the answering paragraphs verbatim, `covered`. For agents and code. |
-| `POST /api/v1/questions` | `{question, k?, judge?, model?, answerModel?, save?}` → the answer text with `[[slug]]` citations, the evidence (paragraphs, probabilities), `covered`, model and token usage; streamed as server-sent events when requested; with `save`, the `output/` path and commit |
-| `POST /api/v1/facts` | deposit a raw source: title, source, `kind: fact`, observed_at, content, tags → path and commit |
-| `GET /api/v1/raw` | the distill queue with ages |
-| `GET /api/v1/reports` | `output/` documents by type, newest first |
-| `POST /api/v1/lab/eval` | run the evaluator for a (model, judge) pair |
-| `POST /webhook` | Gitea push, HMAC as today |
-
-`POST /api/v1/facts` is rejected without `source`, with a `kind` other than
-`fact`, or with content under a minimum length. The raw rule is enforced by
-shape: there is no field for a hypothesis. The file gets the vault's
-provenance frontmatter with `captured_by` set from the API key's identity.
-
-`POST /api/v1/ask` is `wiki_ask.py` moved server-side, so the scripts become
-thin clients and agents no longer need a vault checkout. Its result is
-judgements and quotations, never prose.
-
-`POST /api/v1/questions` is `/ask` plus one answer-model call, in process:
-rank the question, build the context, run the default typed questions
-(answered, where, covered), hand the selected paragraphs and judgements to
-the answer model, stream the prose back. Below the `covered` threshold it
-returns the gap instead of an answer. Time is the sum of the parts, about
-0.2 s rank, 1 to 4 s judge, a few seconds of generation. Cost is one model
-call over a few thousand tokens, a few cents on Sonnet, against a quarter
-dollar for a full agent run.
-
-## Auth and network
-
- Humans: `wiki.renner.uno`, LAN-only (no `dns.renner.uno/public` label),
-  Ingress backend `ak-outpost-wiki-gate:9000`, a `wiki-gate-egress` CNP in
-  `authentik`, the proxy-header handler trusts `X-Authentik-*` only from that
-  outpost.
- Machines: `wiki-index.renner.uno` with the API-key handler; the workstation
-  key stays where it is. CI reaches the ClusterIP.
- Secrets (Gitea token, webhook secret, API keys, TypeSafe key, Anthropic
-  key) reach the
-  pod as environment variables from Kubernetes Secrets filled by
-  ExternalSecret from the `openbao` ClusterSecretStore. The pod itself has
-  no OpenBao access. The Python service fetched its secrets from OpenBao
-  directly through its own Kubernetes-auth role; that role, its policy and
-  the `ai` → `openbao` CNP entry are retired with it.
- Egress from the pod: Gitea, wiki-embed, `api.typesafe.ai:443`,
-  `api.anthropic.com:443`. Every
-  further adapter that talks to a new endpoint gets its own CNP line, which
-  is the one place adding an adapter touches the cluster.
-
-## Health endpoints
-
-Three endpoints with three different meanings, wired to the .NET health
-checks with tags so that the probes and the humans see the right thing:
-
-| endpoint | meaning | probe |
-|---|---|---|
-| `GET /health` | the process serves HTTP. Always 200 once Kestrel listens; no dependency checks. | liveness |
-| `GET /ready` | the active note store has synced at least once and the active index is loaded; 503 while the first sync or index build runs. A failed later sync keeps the last good state and stays ready. | readiness, plus a startup probe with enough budget for the first sync (the Python service was killed mid-build when its `/health` returned 503 during the initial sync) |
-| `GET /api/v1/status` | the full picture for humans and the Settings page: index commit and age, counts, queue lengths, and one entry per registered adapter with its last check result and latency. | none |
-
-Adapter checks (Gitea reachable, embeddings endpoint answering, judge
-endpoint answering, answer model answering) are reported in `/status` and
-on the Settings page and are never part of `/ready`: a judge outage must not
-take the reader down. Their failures show up as red entries, not as
-restarts.
-
-## Phases
-
-One PR each, verified before the next. The Python `wiki-index` keeps running
-until phase 3 proves parity.
-
-0. **Skeleton.** Repository, solution, `Wiki.Core` with the ports and domain
-   types, `Wiki.Contracts` with the suites, `Wiki.Host` serving `/health`,
-   `/ready` and `/api/v1/status`. CI: test and release pipelines. Verify:
-   image in the registry, pipeline green, the Deployment's three probes
-   wired and the pod Ready.
-1. **Parity with the Python service.** Gitea, LlamaCpp and FileIndex
-   adapters, the sync use case, `/rank` and `/webhook`. Verify: the 40-query
-   recall eval gives the same numbers as the Python service on the same
-   commit; `wiki_rank.py --probe-index` against the new service passes.
-2. **Read API and UI.** Notes, tags, raw, reports; Search, Note, Browse,
-   Maintenance pages; `wiki-gate`. Verify: a note renders with resolved
-   links; an unauthenticated request is redirected to Authentik.
-3. **Ask.** Chunker, TypeSafe adapter, `/api/v1/ask`, scripts as clients.
-   Verify: `test_wiki_ask.py` passes against the API; a wiki-synth run with
-   the definition unchanged. Then the Python service is retired and the
-   `wiki-index` hostname points at the new one.
-4. **Facts.** API keys as an ExternalSecret, the deposit-fact use case, `/api/v1/facts`. Verify:
-   a fact posted from the workstation shows as a `claude-bot` commit and in
-   the next distill.
-5. **Questions.** Anthropic adapter, `/api/v1/questions` with streaming,
-   the Answers page with save. Verify: a question answered in under ten
-   seconds with citations that resolve to the shown evidence; a question the
-   wiki does not cover returns the gap; a saved answer appears in `output/`
-   with the wiki-synth frontmatter and is picked up by the evaluator.
-6. **Lab.** Evaluator, a second embedding model and a second answer model
-   registered, the Lab page. Verify: two models compared on the labelled set
-   from the UI.
-
-## Not in scope
-
- Editing notes in the browser.
- Replacing Obsidian; the vault stays a git repository and `notes-push.path`
-  keeps pushing.
- Public exposure; LAN and VPN only.
- Running or triggering agents. The service makes single model calls inside
-  a request; it never runs a tool loop or edits the vault. Distill,
-  consolidate and multi-question synthesis stay Claude processes in CI or in
-  a session; the service shows their inputs and outputs and serves their API
-  calls.
@@ -2,12 +2,7 @@

 Converts one MS-01 from "Proxmox host running a Talos VM" to "Talos on bare
 metal", and swaps its Ceph OSD onto a Samsung PM983 at the same time. Written
-after doing jupiter on 2026-08-25.
-
-All three nodes are now on bare metal — `talosctl get systeminformation`
-reports Minisforum MS-01 hardware with distinct chassis serials for k8s-1,
-k8s-2 and k8s-3, with no hypervisor in between. Kept as the procedure for a
-rebuild, not as outstanding work.
+after doing jupiter on 2026-08-25; saturn and neptune remain.

 Everything here that reads like an over-specific warning is something that
 actually happened. The conversion itself is straightforward — the failures were
@@ -432,8 +427,8 @@ is why it belongs here.
  among its devices but not `ens19`, because `ens19` is deliberately
  address-less. Cilium therefore owns `mgmt0` and sees VLAN 40 frames already
  untagged by the 8021q layer, while VLAN 30 frames only ever meet the filter
-  on the shared parent `uplink`, still tagged. Any future address-less VLAN
-  carrying pod traffic will need the same treatment.
+  on the shared parent `enp2s0f0np0`, still tagged. Any future address-less
+  VLAN carrying pod traffic will need the same treatment.

  The symptom does not look like a VLAN problem. A tc-layer drop reaches an
  AF_PACKET sender as ENOBUFS, so the CNI dhcp plugin reports `no buffer space
@@ -67,12 +67,10 @@ These must exist outside the cluster or the rebuild cannot complete:
      `flux` phase creates the `sops-age` Secret, and any not-yet-migrated sops
      files decrypt with it. (Post-OpenBao-phase-4 this may be retired — see
      `MIGRATION-sops-openbao.md`.)
- [ ] **OpenBao seal key, recovery key + root token** — in the GPG `pass`
-      store (`homelab/openbao/seal-key`, `homelab/openbao/unseal-key` — the
-      former shamir key, now the recovery key — and `homelab/openbao/root-token`),
-      and the GPG private key backed up off-machine. The seal key is the one
-      the snapshots are encrypted under: without it, no OpenBao snapshot can be
-      restored and no secret-consuming app can start.
+- [ ] **OpenBao unseal key + root token** — in the GPG `pass` store
+      (`homelab/openbao/unseal-key`, `homelab/openbao/root-token`), and the GPG
+      private key backed up off-machine. Without the unseal key, no OpenBao
+      snapshot can be restored and no secret-consuming app can start.
 - [ ] **OpenBao raft snapshot** reachable — nightly PVC snapshots
      (`infrastructure/openbao/snapshot-cronjob.yaml`) plus the Velero
      `openbao-daily` offsite copy. In a total-loss rebuild the offsite Velero
@@ -107,7 +105,7 @@ config, and the nodes are addressed by name.
 ## Step 1b — boot all three machines from install media

 ```
-https://factory.talos.dev/image/0f6fb8bd8bb53013e9743b7c7b099d96fb9c32fd2464ced32693c2dc0ea5feb9/v1.13.10/metal-amd64.iso
+https://factory.talos.dev/image/0f6fb8bd8bb53013e9743b7c7b099d96fb9c32fd2464ced32693c2dc0ea5feb9/v1.13.9/metal-amd64.iso
 ```

 Write it to a USB stick and boot each MS-01 from it. Each comes up in
@@ -296,65 +294,42 @@ output** — a green step over a still-forming Ceph is a known trap.
 > Steady-state (non-rebuild) bootstrap is just `kubernetes/bootstrap.sh flux`
 > with `GITEA_TOKEN` set; `FLUX_SOURCE` defaults to `gitea`.

-## Step 4 — openbao: restore the snapshot
+## Step 4 — openbao: restore the snapshot + unseal

-`bootstrap.sh flux` does this step itself (`restore_openbao`): it creates the
-`openbao-seal-key` Secret from `pass` before Flux brings OpenBao up, initialises
-the empty server, restores the newest raft snapshot straight out of the Velero
-kopia repository, and waits until every replica has unsealed itself from the
-static seal. What follows is the same procedure by hand, for when the script
-is not an option.
-
-OpenBao unseals itself from the static seal (`openbao-seal-key` Secret, key
-`key`, the 32-byte base64 key from `pass show homelab/openbao/seal-key`). The
-snapshots are encrypted under that key, so it must be the same key the backed
-up cluster used — a different one restores nothing. Until the snapshot is
-restored, the `ClusterSecretStore` stays NotReady, ESO cannot materialize
-Secrets, and every secret-consuming app stays Pending.
+After Step 3, OpenBao comes up **sealed and empty**. Until it is restored and
+unsealed, the `ClusterSecretStore` stays NotReady, ESO cannot materialize
+Secrets, and every secret-consuming app stays Pending. This step is **manual**
+(the unseal key lives in `pass`, off-cluster by design).

 ```sh
-# 1. The seal key, before the pods start (a pod without it never starts).
-kubectl create namespace openbao --dry-run=client -o yaml | kubectl apply -f -
-kubectl -n openbao create secret generic openbao-seal-key \
-  --from-literal=key="$(pass show homelab/openbao/seal-key)"
+# 1. Wait for the OpenBao pods to exist (all three sealed, 0/1 Ready is expected).
+kubectl -n openbao get pods

-# 2. Wait for openbao-0 (uninitialised: 0/1 Ready is expected), then
-#    initialise it. With an auto-seal, init hands out recovery keys and the
-#    server unseals itself. These keys are throwaway: the restore replaces them.
-kubectl -n openbao exec -it openbao-0 -- bao operator init \
-  -recovery-shares=1 -recovery-threshold=1
-
-# 3. Restore the most recent raft snapshot. Source it from the snapshot PVC
+# 2. Restore the most recent raft snapshot. Source it from the snapshot PVC
 #    (infrastructure/openbao/snapshot-cronjob.yaml) or, in a total-loss rebuild,
-#    from the Velero `openbao-daily` offsite copy. Root token from the init
-#    output. No -force: OpenBao checks the snapshot was written under this
-#    server's seal key and refuses otherwise ("different autoseal key") — that
-#    refusal is the signal the Secret is wrong. -force would skip the check and
-#    leave a server that can never unseal.
+#    from the Velero `openbao-daily` offsite copy. Copy it into the pod, then:
 kubectl -n openbao cp <snapshot>.snap openbao-0:/tmp/restore.snap
-kubectl -n openbao exec -it openbao-0 -- \
-  env BAO_TOKEN=<init root token> bao operator raft snapshot restore /tmp/restore.snap
+kubectl -n openbao exec -it openbao-0 -- bao operator raft snapshot restore /tmp/restore.snap

-# 4. The server seals against the snapshot's keyring and unseals itself from
-#    the static seal; openbao-1 and openbao-2 follow (OrderedReady). Nothing
-#    to type. Afterwards the tokens and recovery key from pass apply again.
-kubectl -n openbao get pods -w
+# 3. Unseal with the key from pass (threshold 1) — EVERY replica, same key.
+#    A sealed server is neither a raft voter nor a standby, so unsealing only
+#    openbao-0 leaves the three-peer cluster one node short of quorum.
+kubectl -n openbao exec -it openbao-0 -- bao operator unseal   # paste homelab/openbao/unseal-key
+kubectl -n openbao exec -it openbao-1 -- bao operator unseal   # same key
+kubectl -n openbao exec -it openbao-2 -- bao operator unseal   # same key
 ```

 Verify the store recovers before moving on:

 ```sh
-kubectl -n openbao exec openbao-0 -- bao status   # want: Seal Type static, Sealed false
-kubectl get clustersecretstore openbao            # want: Valid / Ready
-kubectl -n <ns> get externalsecret                # want: SecretSynced
+kubectl get clustersecretstore openbao        # want: Valid / Ready
+kubectl -n <ns> get externalsecret            # want: SecretSynced
 ```

-> A replica that stays **sealed** after a restart could not use the static
-> seal — the Secret is missing or does not match the key the data was
-> written under. Fix the Secret and restart the pod; `bao operator unseal`
-> with the recovery key (`pass show homelab/openbao/unseal-key`) does not
-> unseal an auto-sealed server and is only for recovery-mode operations.
-> Details + rationale: `MIGRATION-sops-openbao.md`.
+> **Manual unseal is required after every OpenBao pod restart**, not just
+> during rebuilds, and for *each* replica that restarts — a rolling node
+> drain seals all three in turn. Details + rationale:
+> `MIGRATION-sops-openbao.md`.

 ## Step 5 — apps: let the rest reconcile

@@ -418,15 +393,14 @@ it: repo → **Settings → Repository → Mirror Settings** in Gitea, or
 - **`bootstrap.sh talos` is dead** — it cannot generate a config (#581), and
  Step 2 no longer uses it. Its `flux` and `wait-apps` phases are still used in
  Steps 3, 5 and 6 and work; do not read this as the whole script being gone.
-  Its `flux` phase carries the OpenBao restore (Step 4); the seal key and the
-  other secrets are read from `pass` or streamed in at run time and never live
-  in CI.
- **OpenBao unseals itself, but only under the seal key the snapshot was
-  written with.** Three raft replicas mean the service survives losing one; a
-  sealed server is neither a voter nor a standby. A pod that stays sealed after
-  a restart has the wrong `openbao-seal-key` Secret — Step 4 covers this; it is
-  repeated here because it is the step most likely to be half-done under
-  pressure.
+  It has also never had an `openbao` phase — Step 4 is manual because the
+  unseal key must not live in CI; see issue #187 on whether to wire a
+  semi-automated OpenBao restore.
+- **OpenBao seals on every restart and each replica must be unsealed
+  individually.** Three raft replicas mean the service survives losing one, but
+  a sealed server is neither a voter nor a standby — unsealing only one leaves
+  the cluster short of quorum. Step 4 covers this; it is repeated here because
+  it is the step most likely to be half-done under pressure.
 - Ceph's cluster network is a Multus attachment on VLAN 49 with whereabouts
  IPAM, not host networking. Nothing in the rebuild needs special handling for
  it — the nodes hold no address on that segment — but a rebuilt cluster will
@@ -21,9 +21,6 @@ metadata:
  namespace: audiobookshelf
  labels:
    app.kubernetes.io/name: audiobookshelf
-  annotations:
-    # envFrom does not restart the pod when the ConfigMap changes; Reloader does.
-    configmap.reloader.stakater.com/reload: audiobookshelf-config
 spec:
  replicas: 1
  strategy:
@@ -70,7 +67,7 @@ spec:
              mountPath: /ca
      containers:
        - name: audiobookshelf
-          image: ghcr.io/advplyr/audiobookshelf:2.36.1 # {"$imagepolicy": "flux-system:audiobookshelf"}
+          image: ghcr.io/advplyr/audiobookshelf:2.36.0 # {"$imagepolicy": "flux-system:audiobookshelf"}
          ports:
            - name: http
              containerPort: 80
@@ -122,7 +119,7 @@ spec:
            claimName: audiobookshelf-metadata
        - name: books
          persistentVolumeClaim:
-            claimName: audiobookshelf-books-bulk
+            claimName: audiobookshelf-books
        - name: le-ca
          emptyDir:
            medium: Memory
@@ -35,14 +35,14 @@ spec:
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
-  name: audiobookshelf-books-bulk
+  name: audiobookshelf-books
  namespace: audiobookshelf
  annotations:
    kustomize.toolkit.fluxcd.io/prune: disabled
 spec:
  accessModes:
    - ReadWriteMany
-  storageClassName: ceph-filesystem-bulk
+  storageClassName: ceph-filesystem
  resources:
    requests:
      storage: 300Gi
@@ -976,280 +976,6 @@ data:
            kubernetes_ingress_secret_name: "authentik-outpost-tls"
            kubernetes_json_patches: null

-      # wiki-gate: the wiki service (ns ai) behind the same passkey flows as
-      # llama-gate. The service trusts X-Authentik-Username only from pod
-      # addresses, and its CNP admits only this outpost, so the gate is the
-      # whole access control for wiki.renner.uno.
-      - id: wiki-gate-provider
-        model: authentik_providers_proxy.proxyprovider
-        identifiers:
-          name: wiki-gate
-        state: present
-        attrs:
-          authentication_flow: !Find [authentik_flows.flow, [slug, passkey-passwordless-authentication]]
-          authorization_flow: !Find [authentik_flows.flow, [slug, passkey-stepup-authorization]]
-          invalidation_flow: !Find [authentik_flows.flow, [slug, default-provider-invalidation-flow]]
-          mode: proxy
-          internal_host: http://wiki.ai.svc.cluster.local:8080
-          internal_host_ssl_validation: false
-          external_host: https://wiki.renner.uno
-          cookie_domain: renner.uno
-          access_token_validity: hours=24
-          refresh_token_validity: days=30
-          intercept_header_auth: true
-
-      - model: authentik_core.application
-        identifiers:
-          slug: wiki-gate
-        state: present
-        attrs:
-          name: Wiki
-          slug: wiki-gate
-          provider: !KeyOf wiki-gate-provider
-          meta_launch_url: https://wiki.renner.uno
-          meta_description: "Notes wiki: search, notes, answers, maintenance"
-
-      - model: authentik_outposts.outpost
-        identifiers:
-          name: wiki-gate
-        state: present
-        attrs:
-          type: proxy
-          service_connection: !KeyOf k8s-conn
-          providers:
-            - !KeyOf wiki-gate-provider
-          config:
-            log_level: info
-            authentik_host: "http://authentik-server.authentik.svc.cluster.local"
-            authentik_host_insecure: false
-            authentik_host_browser: "https://auth.renner.uno/"
-            refresh_interval: "minutes=5"
-            container_image: null
-            object_naming_template: "ak-outpost-%(name)s"
-            kubernetes_namespace: authentik
-            kubernetes_replicas: 1
-            kubernetes_service_type: ClusterIP
-            kubernetes_disabled_components: ["ingress"]
-            kubernetes_image_pull_secrets: []
-            kubernetes_ingress_annotations: {}
-            kubernetes_ingress_class_name: null
-            kubernetes_ingress_secret_name: "authentik-outpost-tls"
-            kubernetes_json_patches: null
-
-      # paperclip-gate: Paperclip UI; the CNP admits only this outpost. Paperclip keeps its own login behind it.
-      - id: paperclip-users-group
-        model: authentik_core.group
-        identifiers:
-          name: Paperclip Users
-        state: present
-        attrs:
-          is_superuser: false
-          users:
-            - !Find [authentik_core.user, [username, k.renner]]
-
-      - id: paperclip-gate-provider
-        model: authentik_providers_proxy.proxyprovider
-        identifiers:
-          name: paperclip-gate
-        state: present
-        attrs:
-          authentication_flow: !Find [authentik_flows.flow, [slug, passkey-passwordless-authentication]]
-          authorization_flow: !Find [authentik_flows.flow, [slug, passkey-stepup-authorization]]
-          invalidation_flow: !Find [authentik_flows.flow, [slug, default-provider-invalidation-flow]]
-          mode: proxy
-          internal_host: http://paperclip.paperclip.svc.cluster.local:3100
-          internal_host_ssl_validation: false
-          external_host: https://paperclip.renner.uno
-          cookie_domain: renner.uno
-          access_token_validity: hours=24
-          refresh_token_validity: days=30
-          intercept_header_auth: true
-
-      - model: authentik_core.application
-        identifiers:
-          slug: paperclip-gate
-        state: present
-        attrs:
-          name: Paperclip
-          slug: paperclip-gate
-          provider: !KeyOf paperclip-gate-provider
-          meta_launch_url: https://paperclip.renner.uno
-          meta_description: "Paperclip agent orchestration"
-
-      - model: authentik_policies.policybinding
-        identifiers:
-          target: !Find [authentik_core.application, [slug, paperclip-gate]]
-          group: !KeyOf paperclip-users-group
-          order: 0
-        state: present
-        attrs:
-          enabled: true
-          negate: false
-
-      - model: authentik_outposts.outpost
-        identifiers:
-          name: paperclip-gate
-        state: present
-        attrs:
-          type: proxy
-          service_connection: !KeyOf k8s-conn
-          providers:
-            - !KeyOf paperclip-gate-provider
-          config:
-            log_level: info
-            authentik_host: "http://authentik-server.authentik.svc.cluster.local"
-            authentik_host_insecure: false
-            authentik_host_browser: "https://auth.renner.uno/"
-            refresh_interval: "minutes=5"
-            container_image: null
-            object_naming_template: "ak-outpost-%(name)s"
-            kubernetes_namespace: authentik
-            kubernetes_replicas: 1
-            kubernetes_service_type: ClusterIP
-            kubernetes_disabled_components: ["ingress"]
-            kubernetes_image_pull_secrets: []
-            kubernetes_ingress_annotations: {}
-            kubernetes_ingress_class_name: null
-            kubernetes_ingress_secret_name: "authentik-outpost-tls"
-            kubernetes_json_patches: null
-
-      # navidrome-gate: the Navidrome web UI. Navidrome auto-creates the
-      # X-Authentik-Username user on first login; the first user is admin.
-      # /rest and /share bypass the gate at the Traefik router.
-      - id: navidrome-gate-provider
-        model: authentik_providers_proxy.proxyprovider
-        identifiers:
-          name: navidrome-gate
-        state: present
-        attrs:
-          authorization_flow: !Find [authentik_flows.flow, [slug, default-provider-authorization-implicit-consent]]
-          invalidation_flow: !Find [authentik_flows.flow, [slug, default-provider-invalidation-flow]]
-          mode: proxy
-          internal_host: http://navidrome.navidrome.svc.cluster.local
-          internal_host_ssl_validation: false
-          external_host: https://navidrome.renner.uno
-          cookie_domain: renner.uno
-          access_token_validity: hours=24
-          refresh_token_validity: days=30
-          intercept_header_auth: true
-
-      - model: authentik_core.application
-        identifiers:
-          slug: navidrome-gate
-        state: present
-        attrs:
-          name: Navidrome
-          slug: navidrome-gate
-          provider: !KeyOf navidrome-gate-provider
-          meta_launch_url: https://navidrome.renner.uno
-          meta_description: "Music streaming (Subsonic)"
-
-      - model: authentik_outposts.outpost
-        identifiers:
-          name: navidrome-gate
-        state: present
-        attrs:
-          type: proxy
-          service_connection: !KeyOf k8s-conn
-          providers:
-            - !KeyOf navidrome-gate-provider
-          config:
-            log_level: info
-            authentik_host: "http://authentik-server.authentik.svc.cluster.local"
-            authentik_host_insecure: false
-            authentik_host_browser: "https://auth.renner.uno/"
-            refresh_interval: "minutes=5"
-            container_image: null
-            object_naming_template: "ak-outpost-%(name)s"
-            kubernetes_namespace: authentik
-            kubernetes_replicas: 1
-            kubernetes_service_type: ClusterIP
-            kubernetes_disabled_components: ["ingress"]
-            kubernetes_image_pull_secrets: []
-            kubernetes_ingress_annotations: {}
-            kubernetes_ingress_class_name: null
-            kubernetes_ingress_secret_name: "authentik-outpost-tls"
-            kubernetes_json_patches: null
-
-      # lidarr-gate: Lidarr runs with AuthenticationMethod=External, so the gate
-      # is its only access control.
-      - id: music-librarians-group
-        model: authentik_core.group
-        identifiers:
-          name: Music Librarians
-        state: present
-        attrs:
-          is_superuser: false
-          # Membership is owned here; blueprint apply reverts UI edits.
-          users:
-            - !Find [authentik_core.user, [username, k.renner]]
-
-      - id: lidarr-gate-provider
-        model: authentik_providers_proxy.proxyprovider
-        identifiers:
-          name: lidarr-gate
-        state: present
-        attrs:
-          authorization_flow: !Find [authentik_flows.flow, [slug, default-provider-authorization-implicit-consent]]
-          invalidation_flow: !Find [authentik_flows.flow, [slug, default-provider-invalidation-flow]]
-          mode: proxy
-          internal_host: http://lidarr.navidrome.svc.cluster.local
-          internal_host_ssl_validation: false
-          external_host: https://lidarr.renner.uno
-          cookie_domain: renner.uno
-          access_token_validity: hours=24
-          refresh_token_validity: days=30
-          intercept_header_auth: true
-
-      - model: authentik_core.application
-        identifiers:
-          slug: lidarr-gate
-        state: present
-        attrs:
-          name: Lidarr
-          slug: lidarr-gate
-          provider: !KeyOf lidarr-gate-provider
-          meta_launch_url: https://lidarr.renner.uno
-          meta_description: "Music library gaps: discographies, Spotify artists"
-
-      - model: authentik_policies.policybinding
-        identifiers:
-          target: !Find [authentik_core.application, [slug, lidarr-gate]]
-          group: !KeyOf music-librarians-group
-          order: 0
-        state: present
-        attrs:
-          enabled: true
-          negate: false
-          failure_result: false
-
-      - model: authentik_outposts.outpost
-        identifiers:
-          name: lidarr-gate
-        state: present
-        attrs:
-          type: proxy
-          service_connection: !KeyOf k8s-conn
-          providers:
-            - !KeyOf lidarr-gate-provider
-          config:
-            log_level: info
-            authentik_host: "http://authentik-server.authentik.svc.cluster.local"
-            authentik_host_insecure: false
-            authentik_host_browser: "https://auth.renner.uno/"
-            refresh_interval: "minutes=5"
-            container_image: null
-            object_naming_template: "ak-outpost-%(name)s"
-            kubernetes_namespace: authentik
-            kubernetes_replicas: 1
-            kubernetes_service_type: ClusterIP
-            kubernetes_disabled_components: ["ingress"]
-            kubernetes_image_pull_secrets: []
-            kubernetes_ingress_annotations: {}
-            kubernetes_ingress_class_name: null
-            kubernetes_ingress_secret_name: "authentik-outpost-tls"
-            kubernetes_json_patches: null
-
      - model: authentik_outposts.outpost
        identifiers:
          name: authentik Embedded Outpost
@@ -120,14 +120,7 @@ spec:
    server:
      resources:
        requests:
-          # 500m, not 200m: the HPA below targets 70% of *this* request, so at
-          # 200m the scale-up threshold sat at 140m — below the pod's own
-          # ordinary peak. Measured over 7d, a warm server pod p99s at ~108m
-          # and peaks at ~190m, and a freshly-started one burns ~300m booting
-          # Django. That made the HPA oscillate 2<->3<->4 all day and pin at
-          # maxReplicas (KubeHpaMaxedOut, issue #761). At 500m the threshold
-          # is 350m/pod, above the boot burst, so only real load scales out.
-          cpu: 500m
+          cpu: 200m
          memory: 1Gi
        limits:
          memory: 2Gi
@@ -161,11 +154,7 @@ spec:
    worker:
      resources:
        requests:
-          # Same reasoning as server above, and the worker is burstier still:
-          # over 7d its per-pod p99 reaches ~336m and peaks hit ~630m while
-          # running blueprint applies and scheduled tasks. 200m understated
-          # that badly enough that routine task batches read as saturation.
-          cpu: 500m
+          cpu: 200m
          memory: 1Gi
        limits:
          memory: 2Gi
@@ -7,7 +7,6 @@ resources:
  - helmrelease.yaml
  - paperless-gate-ingress.yaml
  - paperless-gate-egress.yaml
-  - paperclip-gate-egress.yaml
  - vscode-gate-egress.yaml
  - shelly-manager-gate-egress.yaml
  - esphome-gate-egress.yaml
@@ -17,9 +16,7 @@ resources:
  - prometheus-gate-egress.yaml
  - wordpress-gate-egress.yaml
  - llama-gate-egress.yaml
-  - wiki-gate-egress.yaml
-  - navidrome-gate-egress.yaml
-  - lidarr-gate-egress.yaml
+  - outpost-servicemonitors.yaml
  - ldap-outpost-cert.yaml
  - ldap-outpost-lb.yaml
  - ldap-outpost-cnp.yaml
@@ -40,7 +40,7 @@ metadata:
    # so this Service produces no public record — LDAP stays
    # LAN-internal, which matches the threat model of the cleartext
    # bind on 389.
-    external-dns.kubernetes.io/hostname: ldap.renner.uno
+    external-dns.alpha.kubernetes.io/hostname: ldap.renner.uno
 spec:
  type: LoadBalancer
  externalTrafficPolicy: Local
@@ -1,20 +0,0 @@
-# Lets the lidarr-gate outpost reach Lidarr; the matching ingress rule is in
-# network-policies/navidrome-ns.yaml.
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: lidarr-gate-egress
-  namespace: authentik
-spec:
-  endpointSelector:
-    matchLabels:
-      goauthentik.io/outpost-name: lidarr-gate
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": navidrome
-            app.kubernetes.io/name: lidarr
-      toPorts:
-        - ports:
-            - port: "8686"
-              protocol: TCP
@@ -1,20 +0,0 @@
-# Lets the navidrome-gate outpost reach Navidrome; the matching ingress rule
-# is in network-policies/navidrome-ns.yaml.
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: navidrome-gate-egress
-  namespace: authentik
-spec:
-  endpointSelector:
-    matchLabels:
-      goauthentik.io/outpost-name: navidrome-gate
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": navidrome
-            app.kubernetes.io/name: navidrome
-      toPorts:
-        - ports:
-            - port: "4533"
-              protocol: TCP
@@ -1,19 +0,0 @@
-# Lets the paperclip-gate outpost reach the Paperclip server; scoped to this outpost's pods.
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: paperclip-gate-egress
-  namespace: authentik
-spec:
-  endpointSelector:
-    matchLabels:
-      goauthentik.io/outpost-name: paperclip-gate
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": paperclip
-            app.kubernetes.io/name: paperclip
-      toPorts:
-        - ports:
-            - port: "3100"
-              protocol: TCP
@@ -1,21 +0,0 @@
-# Lets the wiki-gate outpost reach the wiki service. The authentik namespace
-# baseline CNP otherwise stops any authentik pod from entering ai. Scoped to
-# this outpost's pod label; the matching ingress rule is in ai-ns.yaml.
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: wiki-gate-egress
-  namespace: authentik
-spec:
-  endpointSelector:
-    matchLabels:
-      goauthentik.io/outpost-name: wiki-gate
-  egress:
-    - toEndpoints:
-        - matchLabels:
-            "k8s:io.kubernetes.pod.namespace": ai
-            app.kubernetes.io/name: wiki
-      toPorts:
-        - ports:
-            - port: "8080"
-              protocol: TCP
@@ -3,4 +3,3 @@ kind: Kustomization
 resources:
  - externalsecret.yaml
  - helmrelease.yaml
-  - networkpolicy.yaml
@@ -1,22 +0,0 @@
-# The apps namespace is default-deny-ingress, so the ServiceMonitor added in
-# monitoring-rules/external-dns-servicemonitor.yaml scraped into a timeout.
-# The selector matches both releases — the Cloudflare one deployed here and
-# external-dns-unifi from infrastructure/ — since they share the chart's
-# app.kubernetes.io/name and land in the same namespace.
-apiVersion: cilium.io/v2
-kind: CiliumNetworkPolicy
-metadata:
-  name: external-dns-allow-metrics-scrape
-  namespace: apps
-spec:
-  endpointSelector:
-    matchLabels:
-      app.kubernetes.io/name: external-dns
-  ingress:
-    - fromEndpoints:
-        - matchLabels:
-            k8s:io.kubernetes.pod.namespace: monitoring
-      toPorts:
-        - ports:
-            - port: "7979"
-              protocol: TCP
@@ -46,7 +46,7 @@ spec:
          type: RuntimeDefault
      containers:
        - name: gatus
-          image: twinproduction/gatus:v5.37.0 # {"$imagepolicy": "flux-system:gatus"}
+          image: twinproduction/gatus:v5.36.0 # {"$imagepolicy": "flux-system:gatus"}
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
@@ -2,7 +2,7 @@ apiVersion: networking.k8s.io/v1
 kind: Ingress
 metadata:
  name: gitea
-  namespace: gitea
+  namespace: apps
  labels:
    app.kubernetes.io/name: gitea
  annotations:
@@ -2,7 +2,7 @@ apiVersion: traefik.io/v1alpha1
 kind: IngressRouteTCP
 metadata:
  name: gitea-ssh
-  namespace: gitea
+  namespace: apps
  labels:
    app.kubernetes.io/name: gitea
 spec:
@@ -6,6 +6,7 @@ kind: Kustomization
 # and verified. See MIGRATION-gitea-woodpecker.md.
 resources:
  - statefulset.yaml
+  - pdb.yaml
  - service.yaml
  - ingressroute-ssh.yaml
  - ingress.yaml
@@ -1,8 +1,28 @@
-# Prometheus scrape and Traefik are admitted by the namespace baseline in
-# infrastructure/network-policies/gitea-ns.yaml.
-#
+apiVersion: cilium.io/v2
+kind: CiliumNetworkPolicy
+metadata:
+  name: gitea-allow-metrics-scrape
+  namespace: apps
+spec:
+  # Prometheus (monitoring ns) scrapes gitea's /metrics on :3000. The apps
+  # default-deny-ingress only allows same-ns / home-assistant / host, so the
+  # scrape is dropped (context deadline exceeded). This became necessary only
+  # after the LXC->in-cluster cutover — the old static target was external, so
+  # no in-cluster policy applied. Scoped to the gitea pod + :3000.
+  endpointSelector:
+    matchLabels:
+      app.kubernetes.io/name: gitea
+  ingress:
+    - fromEndpoints:
+        - matchLabels:
+            "k8s:io.kubernetes.pod.namespace": monitoring
+      toPorts:
+        - ports:
+            - port: "3000"
+              protocol: TCP
+---
 # Secret rotation: the flux-git rotator (openbao ns) reaches gitea to mint a new
-# flux PAT. It uses the in-cluster Service (gitea.gitea.svc:3000) — the public
+# flux PAT. It uses the in-cluster Service (gitea.apps.svc:3000) — the public
 # gitea.renner.uno name hairpins off the ingress LB from a pod (same failure as
 # pve.renner.uno). Scoped to the gitea pod :3000 and the rotator label; the
 # egress half is rotator-gitea-flux-egress in network-policies/openbao-ns.yaml.
@@ -10,7 +30,7 @@ apiVersion: cilium.io/v2
 kind: CiliumNetworkPolicy
 metadata:
  name: gitea-allow-flux-rotator
-  namespace: gitea
+  namespace: apps
 spec:
  endpointSelector:
    matchLabels:
@@ -34,7 +54,7 @@ apiVersion: cilium.io/v2
 kind: CiliumNetworkPolicy
 metadata:
  name: gitea-allow-review-bot-rotator
-  namespace: gitea
+  namespace: apps
 spec:
  endpointSelector:
    matchLabels:
@@ -0,0 +1,23 @@
+---
+# Eviction protection for the single-replica Gitea StatefulSet.
+#
+# Gitea runs on SQLite (single writer) and its repos + DB live on a single
+# ReadWriteOnce PVC, so it is deliberately replicas:1 and cannot fail over.
+# With no PDB, a node drain / maintenance evicts the pod abruptly, mid-write,
+# with zero coordination. minAvailable:1 on a 1-replica workload makes the
+# eviction API refuse *voluntary* evictions: `kubectl drain` blocks and
+# surfaces this pod, forcing a conscious decision (drain with
+# --disable-eviction, or scale down first) instead of yanking the live DB.
+# It does NOT block involuntary disruptions (node crash, OOM).
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: gitea
+  namespace: apps
+  labels:
+    app.kubernetes.io/name: gitea
+spec:
+  minAvailable: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: gitea
@@ -2,7 +2,7 @@ apiVersion: v1
 kind: Service
 metadata:
  name: gitea
-  namespace: gitea
+  namespace: apps
  labels:
    app.kubernetes.io/name: gitea
 spec:
@@ -2,7 +2,7 @@ apiVersion: apps/v1
 kind: StatefulSet
 metadata:
  name: gitea
-  namespace: gitea
+  namespace: apps
  labels:
    app.kubernetes.io/name: gitea
 spec:
@@ -152,7 +152,7 @@ spec:
          # NOTE: volumeClaimTemplates is immutable on an existing StatefulSet,
          # so bumping this value cannot be `kubectl apply`-ed in place — Flux
          # will reject the whole object. Landing this size change requires a
-          # one-time `kubectl delete statefulset gitea -n gitea --cascade=orphan`
+          # one-time `kubectl delete statefulset gitea -n apps --cascade=orphan`
          # (pod + PVC keep running) so Flux re-creates the STS from this
          # template and re-adopts them. After the cleanup_packages GC reclaims
          # the ~10G of orphaned blobs, real usage is ~9G, so 40Gi is headroom.
@@ -4,9 +4,7 @@ metadata:
  name: gf-pg
  namespace: monitoring
 spec:
-  instances: 2
-  smartShutdownTimeout: 10
-  stopDelay: 60
+  instances: 1
  imageName: ghcr.io/cloudnative-pg/postgresql:16
  storage:
    storageClass: ceph-block
@@ -111,23 +111,9 @@ statistics             max_id=13323904    max_start_ts=1787162400.0
 statistics_short_term  max_id=159205762   max_start_ts=1787166000.0
 ```

-## Phase 2c — cutover: DONE
+## Phase 2c — cutover (REMAINING)

-Verified 2026-09-02 against the live database:
-
-| table | rows | span |
-|---|---|---|
-| `statistics` | 11,092,586 | 2024-03-13 20:00 → current |
-| `statistics_short_term` | 1,883,709 | rolling ~7 days |
-| `statistics_meta` | 1,480 | |
-
-The 2024-03-13 origin is the proof that matters. It is unchanged from the
-pre-migration count in "What is at risk" above, so the history was copied
-across engines rather than started from zero — which is the failure this
-document exists to prevent. `states` carries current rows, so `db_url` points
-at Postgres.
-
-The procedure below is kept as the record of what was done.
+Do not start until phase 1 is verified.

 **Migrate statistics only.** Copying all 18M `states` rows is slow, needs a long
 HA outage, and buys a table that is discarded within 7 days anyway.
@@ -18,7 +18,7 @@ metadata:
    lb-pool: bgp
  annotations:
    lbipam.cilium.io/ips: "10.2.99.87,fd99::87"
-    external-dns.kubernetes.io/hostname: ha.renner.uno
+    external-dns.alpha.kubernetes.io/hostname: ha.renner.uno
 spec:
  type: LoadBalancer
  externalTrafficPolicy: Local
@@ -56,9 +56,10 @@ spec:
        # Re-rendered on every pod start, so a CNPG password rotation takes
        # effect rather than silently leaving a stale URI behind.
        #
-        # Load-bearing: `db_url: !secret recorder_db_url` in the homeassistant
-        # repo's configuration.yaml reads this key, so HA does not start
-        # without it. Removing this init container takes the recorder down.
+        # INERT until the recorder is repointed: nothing reads this key until
+        # `db_url: !secret recorder_db_url` lands in configuration.yaml in the
+        # homeassistant repo. That ordering is required — HA refuses to start
+        # when a referenced secret is missing.
        - name: render-db-url
          image: busybox:1.36
          command:
@@ -95,7 +96,7 @@ spec:
              memory: 32Mi
      containers:
        - name: home-assistant
-          image: ghcr.io/home-assistant/home-assistant:2026.9.3 # {"$imagepolicy": "flux-system:home-assistant"}
+          image: ghcr.io/home-assistant/home-assistant:2026.8.3 # {"$imagepolicy": "flux-system:home-assistant"}
          env:
            - name: TZ
              value: Europe/Berlin
@@ -18,21 +18,19 @@
 # That is the point of the move — it attacks the multiplier rather than trimming
 # what gets recorded, which would only hide the problem.
 #
-# This is the live recorder database. HA writes here; the SQLite file is no
-# longer the system of record. The 2.4 years of long-term statistics were
-# copied across engines rather than regenerated — HA does not migrate data
-# between database engines, so an empty Postgres would have silently started
-# from zero and taken the energy dashboard's history with it.
-# MIGRATION-recorder-postgres.md holds the sequence that was followed.
+# STAGING
+# Phase 1 (this manifest): the cluster comes up EMPTY and HA keeps using SQLite.
+# Nothing depends on it, so a failure here is inert. Verify Ready, WAL archiving
+# to Hetzner, and a successful base backup before going further.
+# Phase 2: migrate the statistics tables and repoint recorder.db_url.
+# See MIGRATION-recorder-postgres.md for the sequence and the rollback.
 apiVersion: postgresql.cnpg.io/v1
 kind: Cluster
 metadata:
  name: ha-pg
  namespace: home-assistant
 spec:
-  instances: 2
-  smartShutdownTimeout: 10
-  stopDelay: 60
+  instances: 1
  imageName: ghcr.io/cloudnative-pg/postgresql:16
  storage:
    storageClass: ceph-block
@@ -126,7 +124,8 @@ spec:
      # padding rather than data.
      #
      # The cost is RPO: up to 30min of WAL is unrecoverable if this cluster and
-      # its volumes are lost outright.
+      # its volume are lost outright. This is a single-instance cluster, so the
+      # archive is the only off-node copy.
      archive_timeout: "30min"
      # Fewer checkpoints => fewer full-page writes, since a page is only
      # fully-logged on its first modification after a checkpoint.
@@ -16,9 +16,6 @@ resources:
  # health broker credential: OpenBao-backed + Class-A rotated (first of the
  # eight users to migrate off mosquitto-bootstrap.yaml).
  - mosquitto-health-credentials-externalsecret.yaml
-  # open3e-int broker credential: open3e-esp32 gateway on the Vitocal internal
-  # CAN; OpenBao-backed + Class-A rotated.
-  - mosquitto-open3e-int-credentials-externalsecret.yaml
  - mosquitto-config.yaml
  - mosquitto-certificate.yaml
  - mosquitto-service.yaml
@@ -63,8 +60,9 @@ resources:
  - zigbee2mqtt-gate-service.yaml
  - zigbee2mqtt-ingress.yaml
  # Postgres backend for the recorder (MIGRATION-recorder-postgres.md).
-  # Live: recorder.db_url in the homeassistant repo points here, and the
-  # long-term statistics carry their full history from 2024.
+  # PHASE 1: this cluster comes up EMPTY and nothing consumes it — HA is still
+  # on SQLite. Deploying it is therefore inert; phase 2 migrates the statistics
+  # tables and repoints recorder.db_url in the homeassistant repo.
  - ha-pg-hetzner-creds-externalsecret.yaml
  - ha-pg-objectstore.yaml
  - ha-pg-cluster.yaml
@@ -60,7 +60,7 @@ spec:
      # Two processes rather than one: socat carries a single address pair per
      # invocation, and the families need different receive and send addresses.
      # They are independent — one crashing does not blind the other — and each
-      # carries its own supervisor loop.
+      # is restarted by the kubelet on its own.
      containers:
        # -u is load-bearing on both: unidirectional, left to right. Without it
        # socat would also carry HA's replies back out, and this must stay a
@@ -77,29 +77,17 @@ spec:
        # UDP4-SENDTO takes its A record and UDP6-SENDTO its AAAA. See
        # mdns-relay-service.yaml for why that indirection is required and why
        # the Service is dual-stack.
-        #
-        # Supervisor loop: socat exits 1 on the first datagram it cannot
-        # deliver, and the ClusterIP has no backend while HA restarts. The
-        # trap forwards SIGTERM so the pod still terminates promptly.
        - name: socat-ipv4
          image: alpine/socat:1.8.0.3@sha256:beb4a68d9e4fe6b0f21ea774a0fde6c31f580dde6368939ed70100c5385b015e
          imagePullPolicy: IfNotPresent
-          command:
-            - /bin/sh
-            - -c
-            - |
-              trap 'kill "$pid"; exit 0' TERM INT
-              while :; do
-                socat -u \
-                  "UDP4-RECV:5353,ip-add-membership=224.0.0.251:net1,so-bindtodevice=net1,reuseaddr" \
-                  "UDP4-SENDTO:home-assistant-mdns.home-assistant.svc.cluster.local:5353" &
-                pid=$!
-                wait "$pid"
-                echo "mdns-relay: socat-ipv4 exited ($?); restarting in 2s" >&2
-                sleep 2
-              done
-          # No probes. The supervisor loop is the liveness story; a probe here
-          # could only re-assert that PID 1 is running.
+          args:
+            - -u
+            - "UDP4-RECV:5353,ip-add-membership=224.0.0.251:net1,so-bindtodevice=net1,reuseaddr"
+            - "UDP4-SENDTO:home-assistant-mdns.home-assistant.svc.cluster.local:5353"
+          # No probes. socat is a single process with no health surface: if it
+          # exits, the container exits and the kubelet restarts it, which is the
+          # entire liveness story. A probe here could only re-assert that PID 1
+          # is running.
          resources:
            requests:
              cpu: 10m
@@ -126,20 +114,10 @@ spec:
        - name: socat-ipv6
          image: alpine/socat:1.8.0.3@sha256:beb4a68d9e4fe6b0f21ea774a0fde6c31f580dde6368939ed70100c5385b015e
          imagePullPolicy: IfNotPresent
-          command:
-            - /bin/sh
-            - -c
-            - |
-              trap 'kill "$pid"; exit 0' TERM INT
-              while :; do
-                socat -u \
-                  "UDP6-RECV:5353,ipv6-join-group=[ff02::fb]:net1,so-bindtodevice=net1,ipv6-v6only=1,reuseaddr" \
-                  "UDP6-SENDTO:home-assistant-mdns.home-assistant.svc.cluster.local:5353" &
-                pid=$!
-                wait "$pid"
-                echo "mdns-relay: socat-ipv6 exited ($?); restarting in 2s" >&2
-                sleep 2
-              done
+          args:
+            - -u
+            - "UDP6-RECV:5353,ipv6-join-group=[ff02::fb]:net1,so-bindtodevice=net1,ipv6-v6only=1,reuseaddr"
+            - "UDP6-SENDTO:home-assistant-mdns.home-assistant.svc.cluster.local:5353"
          resources:
            requests:
              cpu: 10m
@@ -12,7 +12,7 @@ metadata:
    #
    # Listed per-Secret rather than with the blanket auto annotation, so that
    # enrolling another broker user is a visible edit to this line.
-    secret.reloader.stakater.com/reload: mosquitto-health-credentials,mosquitto-open3e-int-credentials
+    secret.reloader.stakater.com/reload: mosquitto-health-credentials
 spec:
  replicas: 1
  strategy:
@@ -78,7 +78,6 @@ spec:
              mosquitto_passwd -b /mosquitto/auth/passwd admin "$ADMIN_PASSWORD"
              mosquitto_passwd -b /mosquitto/auth/passwd sml2mqtt "$SML2MQTT_PASSWORD"
              mosquitto_passwd -b /mosquitto/auth/passwd open3e "$OPEN3E_PASSWORD"
-              mosquitto_passwd -b /mosquitto/auth/passwd open3e-int "$OPEN3E_INT_PASSWORD"
              mosquitto_passwd -b /mosquitto/auth/passwd evcc "$EVCC_PASSWORD"
              mosquitto_passwd -b /mosquitto/auth/passwd health "$HEALTH_PASSWORD"
              mosquitto_passwd -b /mosquitto/auth/passwd zigbee2mqtt "$ZIGBEE2MQTT_PASSWORD"
@@ -125,12 +124,6 @@ spec:
              user open3e
              topic readwrite open3e/#

-              # open3e-esp32 gateway on the Vitocal internal CAN: own subtree
-              # under open3e/ so edge-1's listener on open3e/cmnd never sees
-              # its commands.
-              user open3e-int
-              topic readwrite open3e/int/#
-
              user evcc
              topic read open3e/#
              topic readwrite evcc/#
@@ -194,13 +187,6 @@ spec:
                secretKeyRef:
                  name: mosquitto-open3e-credentials
                  key: password
-            # open3e-int: OpenBao-backed + Class-A rotated (see
-            # mosquitto-open3e-int-credentials-externalsecret.yaml).
-            - name: OPEN3E_INT_PASSWORD
-              valueFrom:
-                secretKeyRef:
-                  name: mosquitto-open3e-int-credentials
-                  key: password
            - name: EVCC_PASSWORD
              valueFrom:
                secretKeyRef:
@@ -19,7 +19,7 @@ metadata:
    # LAN can connect with `mqtt.renner.uno` (matches the broker cert
    # SAN) instead of the bare IP. No Cloudflare label — kept off the
    # public DNS; UniFi serves a split-DNS answer for LAN clients.
-    external-dns.kubernetes.io/hostname: mqtt.renner.uno
+    external-dns.alpha.kubernetes.io/hostname: mqtt.renner.uno
 spec:
  type: LoadBalancer
  externalTrafficPolicy: Local
@@ -1,21 +0,0 @@
-# Projects the mosquitto `open3e-int` broker credential (open3e-esp32 gateway
-# on the Vitocal internal CAN) from OpenBao into the Secret the broker
-# consumes. Rotator: kubernetes/infrastructure/openbao/rotation/mosquitto-open3e-int.yaml.
-#
-# creationPolicy is implicitly Owner: deleting this ExternalSecret deletes the
-# live Secret. Set creationPolicy: Orphan first to detach without destroying.
-apiVersion: external-secrets.io/v1
-kind: ExternalSecret
-metadata:
-  name: mosquitto-open3e-int-credentials
-  namespace: home-assistant
-spec:
-  refreshInterval: 10m
-  secretStoreRef:
-    kind: ClusterSecretStore
-    name: openbao
-  target:
-    name: mosquitto-open3e-int-credentials
-  dataFrom:
-    - extract:
-        key: apps/home-assistant/mosquitto-open3e-int-credentials
@@ -57,14 +57,7 @@ spec:
          # and leaves this one untouched (tags are not deleted), so this keeps
          # resolving — but the image then came from the branch build rather than
          # master. Re-point it at a master-built tag when convenient.
-          #
-          # konairius/, not gitea_admin/: this init container runs during a cold
-          # rebuild, when Gitea does not exist because it runs inside the
-          # cluster being rebuilt. The node then falls back to a ghcr.io copy
-          # (machine.registries.mirrors, applied by bootstrap.sh), and containerd
-          # keeps the repository path when it falls through — so the path has to
-          # exist on both registries. Steady state still pulls from Gitea.
-          image: gitea.renner.uno/konairius/alpine-jq:2026-09-04
+          image: gitea.renner.uno/gitea_admin/alpine-jq:2026-07-30
          command:
            - sh
            - -c
@@ -10,15 +10,15 @@ metadata:
    # Kept permanently as a standing guard.
    kustomize.toolkit.fluxcd.io/prune: disabled
 spec:
-  # A replica is viable because nothing here writes pg_vectors any more. Both
-  # vector indexes run on vchordrq and the vectors extension is gone, so the
-  # long pg_vectors/indexes/<uuid> paths are absent — pg_basebackup's tar
-  # format cannot represent them, and that is what blocks a replica join.
-  instances: 2
-  # See ak-pg cluster.yaml for rationale. The barman-plugin shutdown wedge
-  # (cloudnative-pg#9770) holds an instance up until SIGKILL, so bounding
-  # terminationGracePeriod (stopDelay) is what keeps a failover from waiting
-  # on it. smartShutdownTimeout keeps pod restarts quick.
+  # Single instance during pgvecto.rs → VectorChord migration window;
+  # pg_basebackup tar format chokes on the long pg_vectors/indexes/<uuid>
+  # paths, blocking replica join. Bump back to 2 once Immich's in-app
+  # vchord migration drops the vectors extension.
+  instances: 1
+  # See ak-pg cluster.yaml for rationale. Single-instance, so no failover, but
+  # the barman-plugin shutdown wedge (cloudnative-pg#9770) takes the whole DB
+  # down until SIGKILL — bounding terminationGracePeriod (stopDelay) matters
+  # even without a standby. smartShutdownTimeout keeps pod restarts quick.
  smartShutdownTimeout: 10
  stopDelay: 60
  imageName: ghcr.io/corentingiraud/cnpg-pgvector-vectorchord:16-migration
@@ -36,10 +36,8 @@ spec:
      database: immich
      owner: immich
  postgresql:
-    # vchord only. The image also ships pgvecto.rs (vectors.so), but its
-    # extension is not installed and preloading it would recreate pg_vectors
-    # in PGDATA for nothing. pgvector needs no preloading.
    shared_preload_libraries:
+      - vectors.so
      - vchord.so
    parameters:
      shared_buffers: "512MB"
@@ -59,7 +57,8 @@ spec:
      # padding rather than data.
      #
      # The cost is RPO: up to 30min of WAL is unrecoverable if this cluster and
-      # its volumes are lost outright.
+      # its volume are lost outright. This is a single-instance cluster, so the
+      # archive is the only off-node copy.
      archive_timeout: "30min"
      # Write-reduction tuning. These databases are largely idle, and the
      # Postgres defaults favour throughput over storage that wears out.
@@ -32,7 +32,7 @@ spec:
          type: RuntimeDefault
      containers:
        - name: immich-machine-learning
-          image: ghcr.io/immich-app/immich-machine-learning:v3.2.2 # {"$imagepolicy": "flux-system:immich-machine-learning"}
+          image: ghcr.io/immich-app/immich-machine-learning:v3.1.0 # {"$imagepolicy": "flux-system:immich-machine-learning"}
          ports:
            - name: http
              containerPort: 3003
@@ -28,38 +28,6 @@ spec:
    requests:
      storage: 300Gi
 ---
-# Media root on the EC 2+1 bulk pool; thumbs mount separately below.
-apiVersion: v1
-kind: PersistentVolumeClaim
-metadata:
-  name: immich-library-bulk
-  namespace: immich
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled
-spec:
-  accessModes:
-    - ReadWriteMany
-  storageClassName: ceph-filesystem-bulk
-  resources:
-    requests:
-      storage: 300Gi
---
-# Thumbnails stay replicated: read affinity serves them from the local OSD.
-apiVersion: v1
-kind: PersistentVolumeClaim
-metadata:
-  name: immich-thumbs
-  namespace: immich
-  annotations:
-    kustomize.toolkit.fluxcd.io/prune: disabled
-spec:
-  accessModes:
-    - ReadWriteMany
-  storageClassName: ceph-filesystem
-  resources:
-    requests:
-      storage: 30Gi
---
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
@@ -35,9 +35,6 @@ metadata:
  namespace: immich
  labels:
    app.kubernetes.io/name: immich-server
-  annotations:
-    # envFrom does not restart the pod when the ConfigMap changes; Reloader does.
-    configmap.reloader.stakater.com/reload: immich-config
 spec:
  # Scaled to 0 alongside immich-machine-learning during single-node
  # maintenance; see the note there. The CNPG Postgres cluster stays running,
@@ -87,7 +84,7 @@ spec:
              mountPath: /ca
      containers:
        - name: immich-server
-          image: ghcr.io/immich-app/immich-server:v3.2.2 # {"$imagepolicy": "flux-system:immich-server"}
+          image: ghcr.io/immich-app/immich-server:v3.1.0 # {"$imagepolicy": "flux-system:immich-server"}
          ports:
            - name: http
              containerPort: 3001
@@ -110,8 +107,6 @@ spec:
          volumeMounts:
            - name: library
              mountPath: /data
-            - name: thumbs
-              mountPath: /data/thumbs
            - name: immich-config-file
              mountPath: /config
              readOnly: true
@@ -147,10 +142,7 @@ spec:
      volumes:
        - name: library
          persistentVolumeClaim:
-            claimName: immich-library-bulk
-        - name: thumbs
-          persistentVolumeClaim:
-            claimName: immich-thumbs
+            claimName: immich-library
        - name: immich-config-file
          secret:
            secretName: immich-secrets
@@ -61,9 +61,6 @@ spec:
      enabled: true
      storageClass: ceph-block
      size: 50Gi
-    # A minAvailable:1 PDB on a single replica blocks every node drain.
-    pdb:
-      create: false
    # Bootstrap admin/org/bucket on the fresh PVC. Admin password + token come
    # from the influxdb-bootstrap Secret (OpenBao via ESO), deterministic across
    # rebuilds so the bootstrap token is the live admin token. `influx setup`
@@ -3,8 +3,8 @@ kind: Kustomization
 resources:
  - gatus/
  - external-dns/
+  - gitea/
  - ical-proxy/
-  # gitea/ has its own `gitea` Kustomization (clusters/homelab/gitea.yaml, #878).
  # woodpecker/ moved to its own dedicated `woodpecker` Kustomization
  # (clusters/homelab/woodpecker.yaml) so its wp-pg CNPG cluster's barman
  # dependency + Cluster health-gating don't contaminate this shared apps KS.
@@ -4,7 +4,7 @@
 # the entrypoint exit after one pass instead of looping internally —
 # the CronJob handles scheduling.
 #
-# Books land in /audiobooks (the audiobookshelf-books-bulk PVC, shared
+# Books land in /audiobooks (the audiobookshelf-books PVC, shared
 # with the audiobookshelf Deployment). Config + DB live on
 # libation-config (RWO ceph-block, seeded once from
 # libation-bootstrap secret).
@@ -42,11 +42,9 @@ spec:
              type: RuntimeDefault
          initContainers:
            # Seed the libation-config PVC from the bootstrap secret on
-            # first run only — sentinel file gates re-runs so DB updates
-            # from subsequent liberate passes are preserved. Refreshed
-            # Audible tokens are preserved by the container args, not by
-            # this sentinel: the entrypoint copies AccountsSettings.json
-            # into ephemeral /config-internal rather than symlinking it.
+            # first run only — sentinel file gates re-runs so token
+            # refreshes in AccountsSettings.json and DB updates from
+            # subsequent liberate passes are preserved.
            - name: seed-config
              image: busybox:1.37.0
              command: ["/bin/sh", "-eu", "-c"]
@@ -68,11 +66,6 @@ spec:
                  install -m 660 /bootstrap/AccountsSettings.json /libation-state/AccountsSettings.json
                  install -m 660 /bootstrap/Settings.json          /libation-state/Settings.json
                  install -m 660 /bootstrap/LibationContext.db     /libation-state/LibationContext.db
-                  # Without this key the seeded tokens cannot be decrypted:
-                  # TokenStorageMethod is Encrypted and the container has no
-                  # OS secret store, so a missing key file makes the
-                  # entrypoint mint a throwaway one on every run.
-                  install -m 660 /bootstrap/libation-master.key   /libation-state/libation-master.key
                  touch /libation-state/.bootstrapped
                  ls -la /libation-state
              volumeMounts:
@@ -92,26 +85,7 @@ spec:
                  type: RuntimeDefault
          containers:
            - name: libation
-              # Pinned by digest: Spegel caches mutable tag resolutions, so
-              # :latest stayed on 13.7.11 despite imagePullPolicy Always.
-              image: docker.io/rmcrackan/libation:14.2.1@sha256:6f9d8ba19dc3803dd6ba295ab5a7abfb6ff3bc06286eebf4ecd8ccbe43245888
-              # The entrypoint stages config into ephemeral /config-internal
-              # by copy (only the DB is symlinked to the PVC), so Audible
-              # tokens refreshed during a run are lost at container exit.
-              # Passing args makes the entrypoint set up the environment and
-              # then exec this instead of its own loop; the trap copies the
-              # refreshed file back to the PVC on every exit path.
-              args:
-                - /bin/sh
-                - -c
-                - |
-                  trap 'cp /config-internal/AccountsSettings.json /libation-state/AccountsSettings.json' EXIT
-                  set -e
-                  # Strip the Audible intro/outro; re-liberated books replace their file instead of gaining a " (1)" copy.
-                  jq '.StripAudibleBrandAudio = true | .OverwriteExisting = true' /config-internal/Settings.json > /config-internal/Settings.json.tmp
-                  mv /config-internal/Settings.json.tmp /config-internal/Settings.json
-                  /libation/LibationCli scan
-                  /libation/LibationCli liberate
+              image: docker.io/rmcrackan/libation:latest
              env:
                - name: LIBATION_CONFIG_DIR
                  value: /libation-state
@@ -138,7 +112,7 @@ spec:
                claimName: libation-config
            - name: books
              persistentVolumeClaim:
-                claimName: audiobookshelf-books-bulk
+                claimName: audiobookshelf-books
            - name: bootstrap
              secret:
                secretName: libation-bootstrap
@@ -5,7 +5,7 @@
 # SOPS-encrypted Secret on first boot (sentinel file gate); after that
 # Libation reads and writes it directly so token rotation persists.
 #
-# Books are written to the audiobookshelf-books-bulk PVC (already provisioned
+# Books are written to the audiobookshelf-books PVC (already provisioned
 # by the audiobookshelf seed Kustomization, RWX ceph-filesystem). We
 # share the same namespace so the PVC reference is direct — k8s PVCs
 # are namespace-scoped.
@@ -25,13 +25,6 @@ metadata:
    app.kubernetes.io/name: llama-server
 spec:
  replicas: 1
-  # The startupProbe below sanctions a 40 min cold download, but the default
-  # progressDeadlineSeconds is 600. Left implicit the Deployment declares
-  # ProgressDeadlineExceeded at 10 minutes while the probe is still legitimately
-  # waiting, and KubeDeploymentRolloutStuck fires 15 min after that on a rollout
-  # that is fine. Kept above the startup budget, with room for the image pull
-  # and scheduling that happen before the first probe.
-  progressDeadlineSeconds: 3000
  # A rolling update would load a second copy of the weights onto the same
  # iGPU and into the same node's RAM before releasing the first.
  strategy:
@@ -82,62 +75,11 @@ spec:
            # many small per-expert matmuls that sets the pace, not bandwidth.
            - --threads
            - "4"
-            # 16k is room to spare, not a safety limit. The limit that matters
-            # is not set here at all.
-            #
-            # Isolated by bisection on a scratch pod: this stack breaks on
-            # PROMPT length, and the value of --ctx-size makes no difference to
-            # it. Measured at --ctx-size 65536 with a single slot, so with four
-            # times the headroom:
-            #
-            #     13588 prompt tokens   coherent answer
-            #     15086                 one token, immediate EOS, empty content
-            #     16088                 CJK characters and digit runs
-            #
-            # A single request past that boundary then poisons the slot: the
-            # same 13588-token prompt that answered cleanly a minute earlier
-            # returns one empty token afterwards, and stays broken until the
-            # pod restarts. That is why it looked intermittent, and why a
-            # 650-token classification once came back as 1025 tokens of "!" —
-            # it followed Home Assistant's ~19k prompts through the same slot.
-            #
-            # Nothing reports any of it: finish_reason "stop", a well-formed
-            # OpenAI response, HTTP 200, /health green.
-            #
-            # So the operating rule is about callers, not this flag. Paperless
-            # sits near 5k and is safe; Home Assistant's Assist prompt measured
-            # 18910 and never had a chance. An earlier version of this comment
-            # blamed --ctx-size 32768, which the bisection disproves — 32768
-            # and 65536 both answer correctly as long as the prompt stays under
-            # the boundary.
-            #
-            # The same stack also fails hard at --ubatch-size 2048, which is
-            # the loud version of a limit this one keeps quiet about.
-            #
+            # 16k rather than 8k because the measured prefill falloff is mild
+            # and this model's KV cache is unusually cheap — only 10 of its 40
+            # layers use full attention.
            - --ctx-size
            - "16384"
-            # One slot, so the 16384 belongs to one request at a time.
-            #
-            # Left on auto, llama.cpp opens four slots and — because auto is
-            # also what enables --kv-unified — has them share a single buffer
-            # of this size. Four Paperless requests at once wanted 6209, 6472,
-            # 4303 and 1271 tokens against those 16384, and all four died with
-            # "Context size has been exceeded".
-            #
-            # They were concurrent because a client timeout does not stop the
-            # generation behind it: each abandoned request keeps consuming the
-            # shared budget the next one needs, so one slow document is enough
-            # to start the pile-up. Queueing is the behaviour worth having
-            # here — nothing calls this server in parallel on purpose, and
-            # Paperless's celery worker runs at concurrency 1.
-            #
-            # No liveness probe change goes with this. The server recovered on
-            # its own once the batch drained, so a probe that restarted the pod
-            # would have made the incident worse; with a single slot, one that
-            # asked for a real completion would also queue behind live work and
-            # kill healthy pods during a long document.
-            - --parallel
-            - "1"
            # A micro-batch of 2048 or more makes this Vulkan/Mesa stack fail
            # with ErrorDeviceLost.
            - --ubatch-size
@@ -19,11 +19,6 @@ metadata:
    app.kubernetes.io/name: llama-embed
 spec:
  replicas: 1
-  # Same reasoning as llama-server: the startupProbe budget (600s) must stay
-  # below progressDeadlineSeconds, whose default is also 600 — exactly equal, so
-  # any cold download that uses most of its budget trips
-  # KubeDeploymentRolloutStuck on a rollout that is still progressing normally.
-  progressDeadlineSeconds: 900
  strategy:
    type: Recreate
  selector:
@@ -46,7 +41,7 @@ spec:
          image: ghcr.io/ggml-org/llama.cpp:server-vulkan-b10731
          args:
            - --hf-repo
-            - gpustack/bge-m3-GGUF:Q8_0
+            - Qwen/Qwen3-Embedding-0.6B-GGUF:Q8_0
            # Serves POST /v1/embeddings.
            - --embeddings
            # A short, stable id for consumers. Without it the model is
@@ -55,41 +50,28 @@ spec:
            # One name only: llama-server keeps aliases in a sorted set and
            # reports the first as the id, so a second name can steal it.
            - --alias
-            - bge-m3
-            # bge-m3 is trained for CLS pooling — a different type from the
-            # Qwen family this replaces, which uses last-token. Left unspecified
-            # llama-server takes the GGUF default, which is not visible from
-            # outside and would silently produce vectors the model was not
-            # trained to emit. Read from the model's own 1_Pooling/config.json
-            # rather than assumed; of the candidates measured, three different
-            # pooling types were needed.
+            - qwen3-embedding-0.6b
+            # Qwen3-Embedding is trained for last-token pooling. Left
+            # unspecified llama-server takes the GGUF's default, which is not
+            # visible from outside and would silently produce vectors the model
+            # was not trained to emit. Set before anything is indexed: changing
+            # pooling later invalidates every stored embedding.
            - --pooling
-            - cls
+            - last
            - --n-gpu-layers
            - "99"
            - --threads
            - "2"
-            # Paperless chunks at 1024 tokens, but counts them with its own
-            # tokenizer while this server measures the batch in bge-m3's.
-            # Those disagree: XLM-RoBERTa splits German into more pieces than
-            # llama_index's counter does, so a chunk Paperless considers full
-            # at 1024 arrived here as 1069 and the request failed outright —
-            # "input (1069 tokens) is too large to process".
-            #
-            # The headroom therefore has to absorb a tokenizer ratio, not just
-            # a rounding error, and 1.5x covers far more than the 1.044 that
-            # broke it. The defaults are batch 2048 / ubatch 512, and with
-            # --embeddings llama-server clamps both to the smaller rather than
-            # run them mismatched, so both are set.
-            #
-            # 2048 stays avoided: that is where the chat model met
-            # ErrorDeviceLost on this Vulkan stack. That failure came under a
-            # 20 GB model's memory pressure and may not bind a 600 MB one, but
-            # this does not need to find out.
+            # Paperless chunks documents at 1024 tokens. The defaults are
+            # batch 2048 / ubatch 512, and with --embeddings llama-server
+            # clamps both down to 512 rather than run them mismatched — below
+            # the chunk size. Matching both to 1024 fits a whole chunk in one
+            # pass; 2048 is deliberately avoided, since that is where this
+            # Vulkan stack fails with ErrorDeviceLost.
            - --batch-size
-            - "1536"
+            - "1024"
            - --ubatch-size
-            - "1536"
+            - "1024"
            # Room for a full chunk plus overhead rather than exactly one.
            - --ctx-size
            - "4096"
@@ -109,22 +91,24 @@ spec:
            capabilities:
              drop:
                - ALL
-          # Measured, not estimated: embedding 400 real documents peaked at
-          # 1.6 GiB (Prometheus max_over_time on the working set). The limit is
-          # well above that because the footprint is dominated by the four slots
-          # at 4096 context and the Vulkan working set rather than by the
-          # weights — the Qwen model this replaces peaked at 10.6 GiB on the
-          # same settings for the same 600 MiB of weights, and several
-          # candidates were OOMKilled at limits that looked generous against
-          # their model size.
+          # The weights are only ~600 MiB, but that is not what sets the
+          # footprint: four slots at 4096 context with a 1024-token micro-batch,
+          # and Vulkan allocating the iGPU's working set from system RAM where
+          # it counts against this cgroup.
+          #
+          # 4Gi was sized for the earlier 2048 context and 512 micro-batch. Both
+          # were raised to match Paperless's chunking without raising this, and
+          # the pod then survived interactive use and was OOMKilled (137) partway
+          # through the first sustained run — roughly 300 documents in. Indexing
+          # the archive would have hit it on day one.
          resources:
            requests:
              cpu: 250m
-              memory: 2Gi
+              memory: 4Gi
              gpu.intel.com/i915: "1"
            limits:
              cpu: "2"
-              memory: 8Gi
+              memory: 10Gi
              gpu.intel.com/i915: "1"
          # ~600 MiB rather than ~5 GiB, so a much shorter budget than the chat
          # model needs, but still sized for a cold download.
@@ -6,5 +6,3 @@ resources:
  - embed.yaml
  - gate-service.yaml
  - ingress.yaml
-  - wiki-embed.yaml
-  - wiki-embed-ingress.yaml
@@ -1,20 +1,10 @@
 ---
-# Holds the GGUF weights, downloaded once by llama-server itself via --hf-repo.
-# Prune-disabled so retiring this Kustomization does not throw away tens of GiB
-# of model files that would have to be pulled from Hugging Face again.
+# Holds the GGUF weights, downloaded once by the Deployment's initContainer.
+# Prune-disabled so retiring this Kustomization does not throw away several
+# GiB of model files that would have to be pulled from Hugging Face again.
 #
-# ReadWriteMany so a second replica, a one-off benchmark Job, or a candidate
-# model being evaluated can mount the same weights; the models are read-only at
-# inference time.
-#
-# Sized well beyond the two deployed models. The cache is shared by every
-# candidate under evaluation (scripts/embed-eval-real.sh), and on a slow
-# connection the download is the long pole — a shared cache lets the large
-# candidates pull in the background while the small ones are already being
-# measured, and means a model is never fetched twice across runs.
-#
-# CephFS supports online expansion (allowVolumeExpansion on the storage class),
-# so growing this does not disturb the running pods.
+# ReadWriteMany so a second replica or a one-off benchmark Job can mount the
+# same weights; the models are read-only at inference time.
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
@@ -28,4 +18,4 @@ spec:
  storageClassName: ceph-filesystem
  resources:
    requests:
-      storage: 120Gi
+      storage: 40Gi
@@ -1,57 +0,0 @@
---
-# wiki-embed.renner.uno — LAN-only (no dns.renner.uno/public label), consumed by
-# bin/wiki_rank.py on the workstation and the nightly wiki jobs. Machine client,
-# so Traefik BasicAuth instead of the browser-only Authentik gate; same pattern
-# as kubernetes/apps/monitoring/loki-ingress.yaml.
-apiVersion: traefik.io/v1alpha1
-kind: Middleware
-metadata:
-  name: wiki-embed-basic-auth
-  namespace: ai
-spec:
-  basicAuth:
-    secret: wiki-embed-basic-auth
-    realm: wiki-embed
---
-# OpenBao key apps/ai/wiki-embed-basic-auth, field `users` = htpasswd line.
-apiVersion: external-secrets.io/v1
-kind: ExternalSecret
-metadata:
-  name: wiki-embed-basic-auth
-  namespace: ai
-spec:
-  refreshInterval: 1h
-  secretStoreRef:
-    kind: ClusterSecretStore
-    name: openbao
-  target:
-    name: wiki-embed-basic-auth
-  dataFrom:
-    - extract:
-        key: apps/ai/wiki-embed-basic-auth
---
-apiVersion: networking.k8s.io/v1
-kind: Ingress
-metadata:
-  name: wiki-embed
-  namespace: ai
-  annotations:
-    cert-manager.io/cluster-issuer: letsencrypt-cluster
-    traefik.ingress.kubernetes.io/router.middlewares: ai-wiki-embed-basic-auth@kubernetescrd
-spec:
-  ingressClassName: traefik
-  tls:
-    - hosts:
-        - wiki-embed.renner.uno
-      secretName: wiki-embed-tls
-  rules:
-    - host: wiki-embed.renner.uno
-      http:
-        paths:
-          - path: /
-            pathType: Prefix
-            backend:
-              service:
-                name: wiki-embed
-                port:
-                  number: 8080
@@ -1,117 +0,0 @@
---
-# Embedding endpoint for the ~/Notes wiki ranker (bin/wiki_rank.py). Own
-# Deployment so wiki traffic never queues behind Paperless on llama-embed.
-# Same model and PVC as llama-embed, so no second download.
-apiVersion: apps/v1
-kind: Deployment
-metadata:
-  name: wiki-embed
-  namespace: ai
-  labels:
-    app.kubernetes.io/name: wiki-embed
-spec:
-  replicas: 1
-  # Above the 600s startupProbe budget; see embed.yaml.
-  progressDeadlineSeconds: 900
-  strategy:
-    type: Recreate
-  selector:
-    matchLabels:
-      app.kubernetes.io/name: wiki-embed
-  template:
-    metadata:
-      labels:
-        app.kubernetes.io/name: wiki-embed
-    spec:
-      securityContext:
-        runAsNonRoot: true
-        runAsUser: 1000
-        runAsGroup: 1000
-        fsGroup: 1000
-        seccompProfile:
-          type: RuntimeDefault
-      containers:
-        - name: wiki-embed
-          image: ghcr.io/ggml-org/llama.cpp:server-vulkan-b10731
-          args:
-            - --hf-repo
-            - gpustack/bge-m3-GGUF:Q8_0
-            - --embeddings
-            - --alias
-            - bge-m3
-            # bge-m3 is trained for CLS pooling; the GGUF default is not.
-            - --pooling
-            - cls
-            - --n-gpu-layers
-            - "99"
-            - --threads
-            - "2"
-            # wiki_rank.py caps each note at 1500 chars and sends 2 per request.
-            - --batch-size
-            - "1536"
-            - --ubatch-size
-            - "1536"
-            - --ctx-size
-            - "4096"
-            - --no-webui
-          env:
-            - name: LLAMA_CACHE
-              value: /models
-          ports:
-            - name: http
-              containerPort: 8080
-          securityContext:
-            allowPrivilegeEscalation: false
-            capabilities:
-              drop:
-                - ALL
-          # Same sizing as llama-embed: 1.6 GiB measured peak, headroom for slots.
-          resources:
-            requests:
-              cpu: 250m
-              memory: 2Gi
-              gpu.intel.com/i915: "1"
-            limits:
-              cpu: "2"
-              memory: 8Gi
-              gpu.intel.com/i915: "1"
-          startupProbe:
-            httpGet:
-              path: /health
-              port: http
-            periodSeconds: 10
-            failureThreshold: 60
-          readinessProbe:
-            httpGet:
-              path: /health
-              port: http
-            periodSeconds: 10
-          livenessProbe:
-            httpGet:
-              path: /health
-              port: http
-            periodSeconds: 30
-            timeoutSeconds: 10
-            failureThreshold: 6
-          volumeMounts:
-            - name: models
-              mountPath: /models
-      volumes:
-        - name: models
-          persistentVolumeClaim:
-            claimName: llama-server-models
---
-apiVersion: v1
-kind: Service
-metadata:
-  name: wiki-embed
-  namespace: ai
-  labels:
-    app.kubernetes.io/name: wiki-embed
-spec:
-  selector:
-    app.kubernetes.io/name: wiki-embed
-  ports:
-    - name: http
-      port: 8080
-      targetPort: http
--- a/Show More
+++ b/Show More