fix(ci1): switch ISO delivery to containerDisk OCI image (Path C)

OCI image: localhost/win-server-2025:1.0 (8.27 GB)
Built FROM scratch + ADD disk.img → /disk/disk.img on noc1, podman
saved as tar (8.27 GB), SCP'd in parallel to all 3 RKE2 nodes,
imported via ctr in k8s.io namespace. Verified present on all 3
schedulable nodes (rke2-server, rke2-agent1, rke2-agent2).

Why containerDisk over the prior PVC paths:
  - Path A (Longhorn Filesystem PVC, sata): OVMF BdsDxe SATA-CDROM
    read timeout. Cdrom-backed PVC is too slow for OVMF's first-sector
    read window.
  - Path B (Synology NFS): uid 107 (qemu) denied at directory level by
    Synology export ACL despite file mode 0777. Memory:
    feedback_synology_iso_export_root_only_uid_107_denied.
  - Path B+SCSI: same OVMF timeout, just on SCSI controller. Bus
    choice was not load-bearing — the issue was always the slow PVC
    backing.
  - Path C (this commit): containerDisk delivers the ISO bytes from
    a tmpfs view of the OCI layer, no PVC controller in the read path.
    qemu reads at native FS speed; OVMF first-sector read completes
    well within timeout. This is also the KubeVirt-recommended pattern
    for installer ISOs.

Connects to FlowerCore.Distribution / Provisioning USB story: same
"OCI image of the OS installer + autounattend on a sysprep CDROM"
pattern that the USB provisioning agent will use. The Windows
install proceeds hands-off via the existing autounattend.xml in
ci1-autounattend ConfigMap (RDP enabled, WinRM, UAC disabled,
Administrator password from 1Password vault item
h3ix4mgfk65gmkcmvh6ly3d3hu).

Image lifecycle: bump tag (1.1, 1.2, ...) when ISO version changes,
rebuild on noc1, redistribute to RKE2 nodes, update image: line.

Legacy NFS PVC + PV manifest and CDI Longhorn PVC RETAINED for this
commit so prior states are recoverable. Will prune in follow-up
once containerDisk boot proves.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.gitignore

@@ -0,0 +1,7 @@
+# .NET build outputs (lint test project)
+**/bin/
+**/obj/
+
+# Editor / temp
+.DS_Store
+*.swp

README.md

@@ -99,8 +99,23 @@ curl -sk -X DELETE https://dns.iamworkin.lan/api/v1/servers/<serverId>/zones/iam
 - **CoreDNS template + ndots:5 collision**: inside pods, `<svc>.<ns>.svc.cluster.local` with <5 dots gets search-expanded through `iamworkin.lan` FIRST and hits the wildcard template → resolves to Traefik VIP, not the real ClusterIP. Use short service names (`<svc>`) in K8s manifests. See memory `feedback_coredns_ndots_template_collision.md`.
 - **Image not on node**: pods stuck `ErrImageNeverPull` means the image wasn't imported to the node Kubernetes scheduled the pod onto. `ctr images import` on all of rke2-server, rke2-agent1, rke2-agent2.
 - **StatefulSet PVC drift**: `volumeClaimTemplates` needs explicit `volumeMode: Filesystem` or ArgoCD SSA self-heals forever. See memory `feedback_argocd_statefulset_pvc_drift.md`.
+- **IngressRoute namespace split**: this RKE2 Traefik install does not allow cross-namespace service refs. Keep the `IngressRoute`, backend `Service`, and TLS secret in the same namespace; if one host is shared across namespaces, duplicate the `Certificate` and move the route next to the destination service.
+- **Public read-only hosts**: if a public host fronts a service that also exposes admin writes internally, add a Traefik route match like `Host(...) && (Method(GET) || Method(HEAD))` on the public edge instead of trusting the app to reject unsafe methods.
+- **Public read-write allowlist hosts**: if a public host accepts a tightly bounded write surface (e.g. bootstrap-JWT POST), pin the allowlist as `(Method(GET) || Method(HEAD) || Method(POST) || Method(OPTIONS))`. PUT/PATCH/DELETE must still 404 at the route. Track A's `updatecenter.iamworkin.lan` / `updates.iamworkin.lan` are the canonical example. The lint test enforces this invariant.
+- **Traefik VIP netpols**: when a `NetworkPolicy` allows `10.0.56.200`, also allow the post-DNAT backend ports (`8443` for TLS plus `8080` or `8000` for HTTP) or Calico will drop the rewritten flow.
+- **Auth-safe probes**: services behind API-key or global auth middleware should prefer `tcpSocket` probes unless `/health` is explicitly exempted before the middleware runs.
 - **ArgoCD must use internal Gitea URL**: `http://gitea-clusterip.gitea.svc.cluster.local:3000/bluejay/bluejay-infra.git`, not the external HTTPS URL (step-ca cert isn't trusted by ArgoCD). The `ApplicationSet` and any hand-created `Application` must both use the internal URL.
 
+## Local manifest lint
+
+The repo now carries a local-first lint pass for the recurring K8s gotchas that have burned the fleet:
+
+```bash
+dotnet test tests/bluejay-infra-lint/BluejayInfraLint.Tests.csproj -c Release
+```
+
+That test project sweeps `bluejay-infra/apps/**` plus the canonical sibling `FlowerCore.*\\k8s` manifests that share the same workspace. Matching `conftest.dev` policy files live under `tests/bluejay-infra-lint/conftest.dev/` for environments that also have `conftest` or `opa`.
+
 ## References
 
 - Cert-manager recovery playbook: `FlowerCore.Notes/memory/project_cert_manager_recovery_2026_04_22.md`

apps/agent-zero/agent-zero.yaml

@@ -2,14 +2,15 @@
 # Agent Zero AI Stack — NUC Deployment (RKE2 Bare-Metal)
 # =============================================================================
 # Deploys: AgentZero (agent UI) on RKE2 cluster with Blue Jay profile
-# Ollama: workstation-first via BLUEJAY-WS (10.0.56.20:11434) with edge1 Pi 5
+# Ollama: edge1 Pi 5 + AI HAT+ ONLY (10.0.57.17:11434).
-# fallback (10.0.57.17:11434)
+# Workstation Ollama (BLUEJAY-WS) is intentionally NOT in the upstream —
+# the workstation is private dev hardware, not a cluster dependency.
 # Target: RKE2 bare-metal cluster, namespace: agent-zero
 # Profile: Blue Jay (21 tools, 3 prompts, 4 extensions, theme)
 #
 # Differences from LOCAL (WSL K3s):
 #   - Uses Longhorn StorageClass (not local-path)
-#   - Prefers workstation Ollama on the R9700, falls back to edge1 Pi 5
+#   - Cluster-only Ollama path (edge1) — keeps workstation private
 #   - NO Anthropic API key (free/local models only)
 #   - NO Piper TTS or Kiwix (edge1 handles TTS, no Wikipedia needed)
 #   - NO hostPath volumes — profile/tools/extensions loaded via ConfigMaps
@@ -91,14 +92,17 @@ subjects:
 # =============================================================================
 # Agent Zero — AI Agent Web UI (NUC Edition, Blue Jay Profile)
 # =============================================================================
-# Connects to a local proxy that routes to workstation Ollama first and edge1 second
+# Connects directly to fc-llm-bridge for chat + internal util/embed + browser.
-# Blue Jay profile with 21 tools, 3 prompts, 4 extensions
+# Agent Zero's internal util/embed slots stay on the bridge's OpenAI-compatible
+# /v1 surface, while browser + corpus-search use the Ollama-compatible /api/*
+# surface through OLLAMA_HOST.
+# Blue Jay profile with 21 tools, 3 prompts, 4 extensions.
 
 ---
-# FC LLM Bridge API key for Agent Zero (ADR-088 chat_model routing).
+# FC LLM Bridge API key for Agent Zero (ADR-088 chat/util/embed/browser routing).
 # Syncs from 1Password item "FC LLM Bridge API Keys" (field: agent-zero-k8s).
-# Consumed by the chat_model only; util / embedding / browser stay on local
+# Consumed by chat, internal util/embed, browser, and corpus-search requests
-# Ollama via the 127.0.0.1 sidecar proxy.
+# that traverse fc-llm-bridge.
 apiVersion: onepassword.com/v1
 kind: OnePasswordItem
 metadata:
@@ -107,6 +111,34 @@ metadata:
 spec:
   itemPath: "vaults/IAmWorkin/items/FC LLM Bridge API Keys"
 
+---
+# Print.Web API key for Agent Zero's print_web.py Python tool.
+# Syncs from 1Password item "Print.Web API Keys" (password field = API key).
+# The print_web.py tool reads PRINT_WEB_API_KEY env var for all HTTP requests
+# to the thermal print service (GET /api/mcp/tools, POST /api/print/*, etc.).
+# Note: Print.Web uses the legacy REST MCP shape (/api/mcp/tools/*), not the
+# streamable-http MCP protocol. The print_web Python tool bridges this gap
+# and is already present in bluejay-tools ConfigMaps.
+apiVersion: onepassword.com/v1
+kind: OnePasswordItem
+metadata:
+  name: print-web-api-keys
+  namespace: agent-zero
+spec:
+  itemPath: "vaults/IAmWorkin/items/Print.Web API Keys"
+
+---
+# Knowledge MCP bearer token for the direct Agent Zero -> Knowledge.Web path.
+# The 1Password item currently stores the raw token in its concealed PASSWORD
+# field, which the operator syncs to Secret key `password`.
+apiVersion: onepassword.com/v1
+kind: OnePasswordItem
+metadata:
+  name: knowledge-mcp-tokens
+  namespace: agent-zero
+spec:
+  itemPath: "vaults/IAmWorkin/items/FlowerCore Knowledge MCP Tokens"
+
 ---
 apiVersion: apps/v1
 kind: Deployment
@@ -118,7 +150,7 @@ metadata:
   annotations:
     agent-zero/deployment: "nuc"
     agent-zero/profile: "bluejay"
-    agent-zero/ollama: "BLUEJAY-WS primary (10.0.56.20:11434), edge1 fallback (10.0.57.17:11434)"
+    agent-zero/ollama: "fc-llm-bridge fronts edge1 Pi 5 + AI HAT+ Ollama for cluster browser/corpus-search traffic; internal chat/util/embed route through the bridge's authenticated OpenAI surface"
 spec:
   replicas: 1
   selector:
@@ -133,19 +165,18 @@ spec:
     spec:
       serviceAccountName: agent-zero
       initContainers:
-        # Wait for either workstation or edge1 Ollama to be reachable before starting Agent Zero.
+        # Wait for fc-llm-bridge to be reachable before starting Agent Zero.
-        - name: wait-for-ollama
+        - name: wait-for-llm-bridge
           image: busybox:1.37
           command: ["sh", "-c"]
           args:
             - |
-              echo "Waiting for Ollama at BLUEJAY-WS or edge1..."
+              echo "Waiting for fc-llm-bridge..."
-              until wget -qO- --timeout=2 http://10.0.56.20:11434/api/tags >/dev/null 2>&1 || \
+              until wget -qO- --timeout=2 http://fc-llm-bridge.fc-llm-bridge.svc:8080/healthz >/dev/null 2>&1; do
-                    wget -qO- --timeout=2 http://10.0.57.17:11434/api/tags >/dev/null 2>&1; do
+                echo "fc-llm-bridge not ready yet, retrying in 5s..."
-                echo "No Ollama endpoint ready yet, retrying in 5s..."
                 sleep 5
               done
-              echo "At least one Ollama endpoint is reachable."
+              echo "fc-llm-bridge is reachable."
         # Assemble the Blue Jay profile directory structure from ConfigMaps.
         # ConfigMaps can't create nested dirs, so we copy into the workspace PVC.
         - name: setup-bluejay
@@ -192,50 +223,6 @@ spec:
             - name: bluejay-theme
               mountPath: /tmp/bluejay-theme
       containers:
-        - name: ollama-proxy
-          image: nginx:1.27-alpine
-          command: ["/bin/sh", "-c"]
-          args:
-            - |
-              cat > /etc/nginx/nginx.conf <<'NGINX'
-              worker_processes  1;
-              events { worker_connections 1024; }
-              http {
-                upstream ollama_upstream {
-                  server 10.0.56.20:11434 max_fails=2 fail_timeout=10s;
-                  server 10.0.57.17:11434 backup;
-                  keepalive 16;
-                }
-                server {
-                  listen 11434;
-                  location / {
-                    proxy_http_version 1.1;
-                    proxy_set_header Connection "";
-                    proxy_set_header Host $host;
-                    proxy_connect_timeout 5s;
-                    proxy_read_timeout 600s;
-                    proxy_send_timeout 600s;
-                    proxy_next_upstream error timeout invalid_header http_502 http_503 http_504;
-                    proxy_pass http://ollama_upstream;
-                  }
-                }
-              }
-              NGINX
-              exec nginx -g 'daemon off;'
-          ports:
-            - containerPort: 11434
-          readinessProbe:
-            httpGet:
-              path: /api/tags
-              port: 11434
-            initialDelaySeconds: 5
-            periodSeconds: 15
-          livenessProbe:
-            httpGet:
-              path: /api/tags
-              port: 11434
-            initialDelaySeconds: 10
-            periodSeconds: 30
         - name: agent-zero
           image: agent0ai/agent-zero:latest
           command: ["/bin/bash", "-c"]
@@ -256,23 +243,41 @@ spec:
               # chat_model: FlowerCore LLM Bridge (ADR-088) — OpenAI-compat,
               # spend-tracked, tier-aliased (fc:balanced → Claude Sonnet).
               # api_key comes from A0_SET_chat_model_api_key env var (overrides
-              # config.json). util + embedding stay on local 127.0.0.1 Ollama
+              # config.json). Utility + embedding stay on the authenticated
-              # proxy (workstation primary, edge1 fallback).
+              # OpenAI-compatible /v1 surface; browser and direct tool traffic
+              # use the bridge's Ollama-compatible root via OLLAMA_HOST.
               mkdir -p /a0/usr/plugins/_model_config
               cat > /a0/usr/plugins/_model_config/config.json << 'MODELCFG'
-              {"allow_chat_override":true,"chat_model":{"provider":"openai","name":"fc:balanced","api_base":"http://fc-llm-bridge.fc-llm-bridge.svc:8080/v1","ctx_length":8192,"ctx_history":0.7,"vision":false,"kwargs":{"temperature":0,"num_ctx":8192}},"utility_model":{"provider":"ollama","name":"qwen2.5:1.5b","api_base":"http://127.0.0.1:11434","ctx_length":8192,"ctx_input":0.7,"kwargs":{"num_ctx":8192}},"embedding_model":{"provider":"ollama","name":"nomic-embed-text","api_base":"http://127.0.0.1:11434","kwargs":{}}}
+              {"allow_chat_override":true,"chat_model":{"provider":"openai","name":"fc:balanced","api_base":"http://fc-llm-bridge.fc-llm-bridge.svc:8080/v1","ctx_length":8192,"ctx_history":0.7,"vision":false,"kwargs":{"temperature":0,"num_ctx":8192}},"utility_model":{"provider":"openai","name":"fc:cheap","api_base":"http://fc-llm-bridge.fc-llm-bridge.svc:8080/v1","ctx_length":8192,"ctx_input":0.7,"kwargs":{"num_ctx":8192}},"embedding_model":{"provider":"openai","name":"openai/fc:embedding","api_base":"http://fc-llm-bridge.fc-llm-bridge.svc:8080/v1","kwargs":{}}}
               MODELCFG
               # Strip heredoc indentation
               sed -i 's/^              //' /a0/usr/plugins/_model_config/config.json
               # Phase 0 Chat MCP pilot: Agent Zero does not interpolate env vars
               # inside A0_SET_mcp_servers JSON, so build the final JSON here from
-              # the secret-backed CHAT_MCP_API_KEY env var before initialize.sh.
+              # the secret-backed env vars before initialize.sh. Keep the local
-              # Use the in-cluster Chat service URL rather than the public
+              # corpus_search.py tool mounted either way so outage fallback
-              # Traefik hostname so the pod stays off the private VIP lane that
+              # remains available even when fc_knowledge is not advertised.
-              # the default egress rule blocks.
+              export KNOWLEDGE_MCP_ENABLED=false
-              if [ -n "${CHAT_MCP_API_KEY:-}" ]; then
+              if [ -n "${KNOWLEDGE_MCP_BEARER_TOKEN:-}" ]; then
-                export A0_SET_mcp_servers="{\"mcpServers\":{\"fc-chat\":{\"type\":\"streamable-http\",\"url\":\"http://chat-web.fc-chat.svc/mcp\",\"headers\":{\"X-Api-Key\":\"${CHAT_MCP_API_KEY}\"}}}}"
+                if curl -sf --connect-timeout 3 "${KNOWLEDGE_MCP_HEALTH_URL}" > /dev/null && \
+                   curl -sf --connect-timeout 5 \
+                     -H "Authorization: Bearer ${KNOWLEDGE_MCP_BEARER_TOKEN}" \
+                     -H "Accept: application/json, text/event-stream" \
+                     -H "Content-Type: application/json" \
+                     -d '{"jsonrpc":"2.0","id":"fc-knowledge-bootstrap","method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"agent-zero-bootstrap","version":"1.0"}}}' \
+                     "${KNOWLEDGE_MCP_URL}" > /dev/null; then
+                  export KNOWLEDGE_MCP_ENABLED=true
+                  echo "fc_knowledge enabled from ${KNOWLEDGE_MCP_URL}."
+                else
+                  echo "fc_knowledge unavailable or unauthorized; keeping local corpus_search.py as the fallback path."
+                fi
+              else
+                echo "fc_knowledge token missing; keeping local corpus_search.py as the fallback path."
               fi
+
+              export A0_SET_mcp_servers="$(
+                python3 -c 'import json, os; servers = {}; chat_key = os.getenv("CHAT_MCP_API_KEY"); knowledge_enabled = os.getenv("KNOWLEDGE_MCP_ENABLED", "false").lower() == "true"; token = os.getenv("KNOWLEDGE_MCP_BEARER_TOKEN", "") if knowledge_enabled else ""; chat_key and servers.setdefault("fc_chat", {"type": "streamable-http", "url": "http://chat-web.fc-chat.svc/mcp", "headers": {"X-Api-Key": chat_key}}); token and servers.setdefault("fc_knowledge", {"type": "streamable-http", "url": os.getenv("KNOWLEDGE_MCP_URL", "http://knowledge-web.knowledge.svc/mcp"), "headers": {"Authorization": f"Bearer {token}"}}); print(json.dumps({"mcpServers": servers}, separators=(",", ":")))'
+              )"
               # Run the original entrypoint
               exec /exe/initialize.sh $BRANCH
           ports:
@@ -284,8 +289,9 @@ spec:
             # Chat model — routed through FlowerCore LLM Bridge (ADR-088)
             # so spend is tracked and tier aliases (fc:cheap/fc:balanced/fc:deep)
             # dispatch to Ollama or Anthropic via a single OpenAI-compat endpoint.
-            # Util / embedding / browser stay on local Ollama via 127.0.0.1 proxy
+            # Internal utility + embedding use the authenticated OpenAI surface,
-            # for zero-latency, zero-cost small-model traffic.
+            # while browser/corpus-search use the bridge's Ollama-compatible
+            # endpoints so Agent Zero no longer needs a local proxy sidecar.
             - name: A0_SET_chat_model_provider
               value: "openai"
             - name: A0_SET_chat_model_name
@@ -307,35 +313,51 @@ spec:
                 secretKeyRef:
                   name: fc-llm-bridge-api-keys
                   key: agent-zero-k8s
+            - name: FC_LLM_BRIDGE_API_KEY
+              valueFrom:
+                secretKeyRef:
+                  name: fc-llm-bridge-api-keys
+                  key: agent-zero-k8s
             - name: A0_SET_chat_model_ctx_length
               value: "8192"
             - name: A0_SET_chat_model_kwargs
               value: '{"temperature": 0, "num_ctx": 8192}'
-            # Utility model — fast small helper tier through the same proxy
+            # Utility model — fast small helper tier through the OpenAI surface
             - name: A0_SET_util_model_provider
-              value: "ollama"
+              value: "openai"
             - name: A0_SET_util_model_name
-              value: "qwen2.5:1.5b"
+              value: "fc:cheap"
             - name: A0_SET_util_model_api_base
-              value: "http://127.0.0.1:11434"
+              value: "http://fc-llm-bridge.fc-llm-bridge.svc:8080/v1"
             - name: A0_SET_util_model_kwargs
               value: '{"num_ctx": 2048}'
-            # Embedding model — nomic through the same proxy
+            # Embedding model — authenticated bridge alias to nomic-embed-text.
+            # LiteLLM's embedding() path needs an explicit provider prefix here
+            # even though the chat slot can use bare fc:* aliases.
             - name: A0_SET_embed_model_provider
-              value: "ollama"
+              value: "openai"
             - name: A0_SET_embed_model_name
-              value: "nomic-embed-text"
+              value: "openai/fc:embedding"
             - name: A0_SET_embed_model_api_base
-              value: "http://127.0.0.1:11434"
+              value: "http://fc-llm-bridge.fc-llm-bridge.svc:8080/v1"
             # Browser model — small Gemma candidate through the same proxy
             - name: A0_SET_browser_model_provider
               value: "ollama"
             - name: A0_SET_browser_model_name
               value: "gemma3:4b"
             - name: A0_SET_browser_model_api_base
-              value: "http://127.0.0.1:11434"
+              value: "http://fc-llm-bridge.fc-llm-bridge.svc:8080"
+            - name: A0_SET_browser_model_api_key
+              valueFrom:
+                secretKeyRef:
+                  name: fc-llm-bridge-api-keys
+                  key: agent-zero-k8s
             - name: A0_SET_browser_model_vision
               value: "true"
+            - name: OLLAMA_HOST
+              value: "http://fc-llm-bridge.fc-llm-bridge.svc:8080"
+            - name: FLOWERCORE_AGENTZERO_OLLAMA_URL
+              value: "http://fc-llm-bridge.fc-llm-bridge.svc:8080"
             # Agent profile — Blue Jay personality, tools, and system prompt
             - name: A0_SET_agent_profile
               value: "bluejay"
@@ -358,9 +380,38 @@ spec:
                   name: chat-mcp-api-key
                   key: api-key
                   optional: true
-            # Print.Web — Thermal printer service on edge2
+            # FlowerCore.Knowledge MCP Phase 1 — direct Agent Zero client path.
+            # Probe /healthz first, then try an authenticated initialize call.
+            # If either fails, Agent Zero boots without fc_knowledge and keeps
+            # the local corpus_search.py tool as the outage-safe path.
+            - name: KNOWLEDGE_MCP_URL
+              value: "http://knowledge-web.knowledge.svc/mcp"
+            - name: KNOWLEDGE_MCP_HEALTH_URL
+              value: "http://knowledge-web.knowledge.svc/healthz"
+            - name: KNOWLEDGE_MCP_BEARER_TOKEN
+              valueFrom:
+                secretKeyRef:
+                  name: knowledge-mcp-tokens
+                  key: password
+            # Print.Web — Thermal printer service on edge2.
+            # PRINT_WEB_URL: internal HTTP (bypasses Traefik TLS — print_web.py
+            # runs in-cluster and can reach edge2 directly on the PROD VLAN).
+            # PRINT_WEB_API_KEY: from 1Password "Print.Web API Keys" password field,
+            # synced by the print-web-api-keys OnePasswordItem CRD above.
+            # The print_web.py Python tool reads both env vars for all HTTP calls.
             - name: PRINT_WEB_URL
               value: "http://10.0.57.16:5200"
+            - name: PRINT_WEB_API_KEY
+              valueFrom:
+                secretKeyRef:
+                  name: print-web-api-keys
+                  key: password
+            # Intranet search — use in-cluster HTTP (no step-ca TLS needed)
+            # corpus_search.py reads FLOWERCORE_FLEET_VECTOR_DIR but that mount is not
+            # on the cluster yet (BLUEJAY-WS only). The tool gracefully returns a
+            # "no DB found" message with rebuild instructions rather than crashing.
+            - name: FLOWERCORE_INTRANET_URL
+              value: "http://intranet-web.intranet.svc:5300"
             # Kubernetes
             - name: KUBERNETES_SERVICE_HOST
               value: "kubernetes.default.svc"
@@ -395,7 +446,7 @@ spec:
               command:
                 - /bin/bash
                 - -c
-                - "curl -sf http://localhost:80/ > /dev/null && curl -sf --connect-timeout 3 http://127.0.0.1:11434/api/tags > /dev/null"
+                - "curl -sf http://localhost:80/ > /dev/null && curl -sf --connect-timeout 3 http://fc-llm-bridge.fc-llm-bridge.svc:8080/healthz > /dev/null"
             periodSeconds: 30
             failureThreshold: 2
           resources:
@@ -533,18 +584,6 @@ spec:
           protocol: UDP
         - port: 53
           protocol: TCP
-    # Ollama on BLUEJAY-WS
-    - to:
-        - ipBlock:
-            cidr: 10.0.56.20/32
-      ports:
-        - port: 11434
-    # Ollama on edge1 fallback
-    - to:
-        - ipBlock:
-            cidr: 10.0.57.17/32
-      ports:
-        - port: 11434
     # Print.Web on edge2
     - to:
         - ipBlock:
@@ -578,6 +617,26 @@ spec:
           protocol: TCP
         - port: 8080
           protocol: TCP
+    # FlowerCore.Knowledge MCP (Phase 1) — in-cluster direct route with
+    # anonymous /healthz probe plus authenticated /mcp initialize/tool calls.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: knowledge
+      ports:
+        - port: 80
+          protocol: TCP
+        - port: 8080
+          protocol: TCP
+    # Intranet search API — use in-cluster svc so traffic stays inside
+    # the cluster and is not blocked by the private-range egress denylist.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: intranet
+      ports:
+        - port: 5300
+          protocol: TCP
     # Allow internet (for kubectl image pull, etc)
     - to:
         - ipBlock:

apps/agent-zero/configmaps-bluejay.yaml

@@ -7209,6 +7209,9 @@ data:
             "keep_alive": keep_alive,
             "stream": False,
         })
+        curl_headers = ["-H", "Content-Type: application/json"]
+        if os.environ.get("FC_LLM_BRIDGE_API_KEY"):
+            curl_headers.extend(["-H", f"X-Api-Key: {os.environ['FC_LLM_BRIDGE_API_KEY']}"])
 
         try:
             result = subprocess.run(
@@ -7216,7 +7219,7 @@ data:
                     "curl", "-s", "--max-time", "120",
                     "-X", "POST",
                     f"{api_base}/api/generate",
-                    "-H", "Content-Type: application/json",
+                    *curl_headers,
                     "-d", payload,
                 ],
                 capture_output=True,
@@ -13150,6 +13153,451 @@ data:
     - PowerShell 5.1 compatibility is assumed (no PowerShell 7+ features).
     - All commands run with `-NoProfile -NonInteractive` flags for clean execution.
     """
+  corpus_search.py: |
+    # FlowerCore Fleet Corpus Vector Search Tool
+    #
+    # Queries the AiStation-built SqliteVecVectorStore DB at /a0/usr/vectors/fleet.db
+    # (bind-mounted read-only from /var/lib/flowercore/vector-stores/ on the host).
+    # Embeds the query through Ollama's nomic-embed-text model, computes cosine
+    # similarity against every stored chunk in pure Python (no numpy — not present
+    # in the container), and returns the top-K nearest neighbors with source metadata.
+    #
+    # This is the offline-friendly counterpart to `intranet_search` (which hits the
+    # Intranet's live REST API). Use it for Bible/Greek/Hebrew/Strong's lookups and
+    # anywhere the workstation has a newer DB than the Intranet one. The store is
+    # refreshed by `aistation-indexer build <edition>` — see the FlowerCore.Knowledge
+    # ADR at docs/ai-agents/flowercore-knowledge-service-plan.md.
+
+    import json
+    import math
+    import os
+    import sqlite3
+    import urllib.request
+    from pathlib import Path
+
+    from python.helpers.tool import Tool, Response
+
+
+    DEFAULT_VECTORS_DIR = os.environ.get(
+        "FLOWERCORE_FLEET_VECTOR_DIR",
+        "/a0/usr/vectors",
+    )
+    # When the caller doesn't pick an explicit DB, prefer the biggest fleet tier
+    # present on disk. Workstation → pi-edge → bmo-bot.
+    PREFERRED_DB_ORDER = [
+        os.environ.get("FLOWERCORE_FLEET_VECTOR_DB", ""),
+        "fleet-workstation-full.db",
+        "fleet-pi-edge.db",
+        "fleet-bmo-bot.db",
+    ]
+    OLLAMA_BASE_URL = os.environ.get(
+        "FLOWERCORE_AGENTZERO_OLLAMA_URL",
+        "http://host.containers.internal:11434",
+    )
+    BRIDGE_API_KEY = os.environ.get("FC_LLM_BRIDGE_API_KEY", "").strip()
+    EMBEDDING_MODEL = os.environ.get(
+        "FLOWERCORE_FLEET_EMBEDDING_MODEL",
+        "nomic-embed-text",
+    )
+
+
+    class CorpusSearch(Tool):
+        async def execute(self, **kwargs) -> Response:
+            """
+            Semantic search over the FlowerCore fleet corpus (Bible texts, lexicons,
+            dictionaries, morphology) pre-indexed by aistation-indexer.
+
+            Args (via self.args):
+                query (str): Search query text. Required unless action=stats.
+                limit (int): Max results. Default 8.
+                index (str): Optional index name filter ("bible-texts", "lexicons",
+                             "dictionaries", "morphology"). Default: all indexes.
+                repo (str): Optional repo filter (e.g. "world-english-bible").
+                db (str): Override DB path OR file name inside FLOWERCORE_FLEET_VECTOR_DIR
+                          (defaults to /a0/usr/vectors). If omitted, the largest
+                          fleet tier present on disk is picked automatically.
+                action (str): Optional. "stats" returns an inventory of all fleet DBs
+                             visible to the tool (names, sizes, index counts, chunk
+                             counts, last-built timestamps). No embedding call.
+
+            Returns:
+                Response with ranked chunks (score, source, text preview) OR
+                (when action=stats) a markdown inventory of available fleet DBs.
+            """
+            query = (self.args.get("query") or "").strip()
+            limit = int(self.args.get("limit") or 8)
+            index_filter = (self.args.get("index") or "").strip()
+            repo_filter = (self.args.get("repo") or "").strip()
+            db_override = (self.args.get("db") or "").strip()
+            action = (self.args.get("action") or "").strip().lower()
+
+            if action == "stats":
+                return Response(message=_render_stats(), break_loop=False)
+
+            if not query:
+                return Response(
+                    message=(
+                        "Error: 'query' is required unless action=stats.\n"
+                        "Example: query=\"what does Genesis 1:1 say\" limit=5\n"
+                        "Inventory: action=stats"
+                    ),
+                    break_loop=False,
+                )
+
+            db = _resolve_db(db_override)
+            if db is None:
+                return Response(
+                    message=(
+                        f"Error: no fleet vector DB found under {DEFAULT_VECTORS_DIR}.\n"
+                        "Host side: run `aistation-indexer build fleet-workstation-full`\n"
+                        "(or `fleet-pi-edge`/`fleet-bmo-bot`) to produce\n"
+                        "`/var/lib/flowercore/vector-stores/<slug>.db`, then confirm the\n"
+                        "Podman unit mounts that directory into `/a0/usr/vectors:ro`."
+                    ),
+                    break_loop=False,
+                )
+
+            try:
+                query_vec = _embed(query)
+            except Exception as e:
+                return Response(
+                    message=f"Error: failed to embed query via Ollama at {OLLAMA_BASE_URL}: {e}",
+                    break_loop=False,
+                )
+
+            try:
+                hits = _search(db, query_vec, index_filter, repo_filter, limit)
+            except Exception as e:
+                return Response(
+                    message=f"Error: corpus search failed: {e}",
+                    break_loop=False,
+                )
+
+            if not hits:
+                return Response(
+                    message=(
+                        f"No matches for '{query}' in {db.name}.\n"
+                        f"Indexes available: " + _list_indexes_summary(db)
+                    ),
+                    break_loop=False,
+                )
+
+            lines = [f"**Corpus search: `{query}`**  (top {len(hits)} of {limit} requested, DB={db.name})", ""]
+            for rank, h in enumerate(hits, 1):
+                passage = h.get("passage") or ""
+                lang = h.get("language") or ""
+                meta_bits = [x for x in (h["index"], h["repo"], passage, lang) if x]
+                meta = "  ·  ".join(meta_bits)
+                preview = h["text"]
+                if len(preview) > 320:
+                    preview = preview[:320].rstrip() + "…"
+                lines.append(f"{rank}. **{h['score']:.3f}**  {meta}")
+                lines.append(f"   `{h['source']}`")
+                lines.append(f"   {preview}")
+                lines.append("")
+
+            return Response(message="\n".join(lines).rstrip() + "\n", break_loop=False)
+
+
+    def _resolve_db(override: str) -> "Path | None":
+        """Pick a fleet DB by explicit path, explicit filename, or preferred order."""
+        vectors_dir = Path(DEFAULT_VECTORS_DIR)
+        if override:
+            # Absolute or relative path that points at a real file wins outright.
+            p = Path(override)
+            if p.is_absolute() and p.exists():
+                return p
+            # Otherwise treat it as a filename within the vectors dir.
+            candidate = vectors_dir / override
+            if candidate.exists():
+                return candidate
+            return None
+
+        for name in PREFERRED_DB_ORDER:
+            if not name:
+                continue
+            p = Path(name) if Path(name).is_absolute() else vectors_dir / name
+            if p.exists():
+                return p
+
+        # Fallback: any *.db in the dir, largest first.
+        if vectors_dir.is_dir():
+            candidates = sorted(vectors_dir.glob("*.db"), key=lambda p: p.stat().st_size, reverse=True)
+            if candidates:
+                return candidates[0]
+        return None
+
+
+    def _embed(text: str) -> list:
+        """Embed a query via Ollama's /api/embeddings. Single-vector response."""
+        body = json.dumps({"model": EMBEDDING_MODEL, "prompt": text}).encode("utf-8")
+        headers = {"Content-Type": "application/json"}
+        if BRIDGE_API_KEY:
+            headers["X-Api-Key"] = BRIDGE_API_KEY
+        req = urllib.request.Request(
+            f"{OLLAMA_BASE_URL.rstrip('/')}/api/embeddings",
+            data=body,
+            headers=headers,
+        )
+        with urllib.request.urlopen(req, timeout=60) as resp:
+            data = json.loads(resp.read().decode("utf-8"))
+        vec = data.get("embedding")
+        if not isinstance(vec, list) or not vec:
+            raise RuntimeError(f"Ollama returned no embedding: {data}")
+        return [float(x) for x in vec]
+
+
+    def _cosine(a: list, b: list) -> float:
+        """Cosine similarity in pure Python — no numpy in the A0 container."""
+        # zip() stops at the shorter — AiStation DB guarantees same dim per index.
+        dot = 0.0
+        na = 0.0
+        nb = 0.0
+        for x, y in zip(a, b):
+            dot += x * y
+            na += x * x
+            nb += y * y
+        if na == 0.0 or nb == 0.0:
+            return 0.0
+        return dot / (math.sqrt(na) * math.sqrt(nb))
+
+
+    def _search(db_path: Path, query_vec: list, index_filter: str, repo_filter: str, limit: int) -> list:
+        """Load entries, compute cosine, return top-K.
+
+        SqliteVecVectorStore schema:
+          VectorIndexes(IndexName, Dimensions, UpdatedAtUtc)
+          VectorEntries(IndexName, ChunkId, TextContent, SourceRepo, SourceFile,
+                        Book, Chapter, VerseRange, Language, ContentType, License,
+                        EstimatedTokens, EmbeddingJson)
+
+        Embeddings are stored as JSON arrays in EmbeddingJson; similarity is computed
+        in Python. For ~100k chunks × 768 dims this takes a couple seconds on a
+        workstation — acceptable for interactive A0 use.
+        """
+        conn = sqlite3.connect(f"file:{db_path}?mode=ro", uri=True)
+        try:
+            sql = [
+                "SELECT IndexName, ChunkId, TextContent, SourceRepo, SourceFile, ",
+                "       Book, Chapter, VerseRange, Language, EmbeddingJson ",
+                "FROM VectorEntries",
+            ]
+            where = []
+            params = []
+            if index_filter:
+                where.append("IndexName = ?")
+                params.append(index_filter)
+            if repo_filter:
+                where.append("SourceRepo LIKE ?")
+                params.append(f"%{repo_filter}%")
+            if where:
+                sql.append(" WHERE " + " AND ".join(where))
+            sql.append(";")
+
+            cursor = conn.execute("".join(sql), params)
+
+            # Min-heap by (score, ...) would be faster but for interactive use we
+            # just sort at the end — simpler and readable.
+            scored = []
+            for row in cursor:
+                idx, chunk_id, text, repo, source_file, book, chapter, verses, lang, emb_json = row
+                try:
+                    vec = json.loads(emb_json)
+                except (json.JSONDecodeError, TypeError):
+                    continue
+                score = _cosine(query_vec, vec)
+                passage = None
+                if book and chapter:
+                    passage = f"{book} {chapter}"
+                    if verses:
+                        passage += f":{verses}"
+                scored.append((score, {
+                    "index": idx,
+                    "chunk_id": chunk_id,
+                    "text": text,
+                    "repo": repo or "",
+                    "source": source_file or "",
+                    "passage": passage or "",
+                    "language": lang or "",
+                }))
+            scored.sort(key=lambda t: t[0], reverse=True)
+            return [{"score": s, **meta} for s, meta in scored[:limit]]
+        finally:
+            conn.close()
+
+
+    def _render_stats() -> str:
+        """Markdown inventory of every *.db in FLOWERCORE_FLEET_VECTOR_DIR."""
+        vectors_dir = Path(DEFAULT_VECTORS_DIR)
+        if not vectors_dir.is_dir():
+            return f"No fleet vector dir mounted at {vectors_dir}. Ask the host operator to build an index with scripts/agent-zero/build-fleet-index.sh."
+
+        dbs = sorted(vectors_dir.glob("*.db"))
+        if not dbs:
+            return f"No fleet DBs present under {vectors_dir}. Run `scripts/agent-zero/build-fleet-index.sh fleet-workstation-full` on the host."
+
+        lines = [f"**Fleet vector DB inventory** ({vectors_dir})", ""]
+        for db in dbs:
+            size_mb = db.stat().st_size / (1024 * 1024)
+            lines.append(f"### `{db.name}` ({size_mb:.1f} MB)")
+            try:
+                conn = sqlite3.connect(f"file:{db}?mode=ro", uri=True)
+                try:
+                    idx_rows = conn.execute(
+                        "SELECT IndexName, Dimensions, UpdatedAtUtc FROM VectorIndexes ORDER BY IndexName;"
+                    ).fetchall()
+                    if not idx_rows:
+                        lines.append("- (no indexes registered)")
+                    else:
+                        counts = dict(conn.execute(
+                            "SELECT IndexName, COUNT(*) FROM VectorEntries GROUP BY IndexName;"
+                        ).fetchall())
+                        for name, dim, updated in idx_rows:
+                            count = counts.get(name, 0)
+                            lines.append(f"- **{name}** — {count:,} chunks × {dim}d  (built {updated})")
+                finally:
+                    conn.close()
+            except Exception as e:
+                lines.append(f"- (inspect failed: {e})")
+            lines.append("")
+
+        lines.append(f"**Tool defaults:** embedding model `{EMBEDDING_MODEL}`, Ollama at `{OLLAMA_BASE_URL}`. Pick a DB with `db=<filename>`; filter by `index=<name>`/`repo=<substring>`.")
+        return "\n".join(lines).rstrip() + "\n"
+
+
+    def _list_indexes_summary(db_path: Path) -> str:
+        try:
+            conn = sqlite3.connect(f"file:{db_path}?mode=ro", uri=True)
+            try:
+                rows = conn.execute(
+                    "SELECT IndexName, Dimensions, "
+                    "  (SELECT COUNT(*) FROM VectorEntries WHERE VectorEntries.IndexName = VectorIndexes.IndexName) "
+                    "FROM VectorIndexes ORDER BY IndexName;"
+                ).fetchall()
+                if not rows:
+                    return "(no indexes)"
+                return ", ".join(f"{r[0]}({r[2]}×{r[1]}d)" for r in rows)
+            finally:
+                conn.close()
+        except Exception as e:
+            return f"(couldn't list: {e})"
+
+  intranet_search.py: |
+    # Intranet Vector Search Tool
+    # Queries the Blue Jay Lab Intranet's Shared.Indexing RAG corpus over its
+    # live REST API (https://intranet.iamworkin.lan/search). Returns ranked chunks
+    # with source file paths and scores.
+
+    import json
+    import os
+    import ssl
+    import urllib.parse
+    import urllib.request
+
+    from python.helpers.tool import Tool, Response
+
+
+    INTRANET_BASE_URL = os.environ.get(
+        "FLOWERCORE_INTRANET_URL",
+        "https://intranet.iamworkin.lan",
+    )
+    STEPCA_ROOT_CRT = "/a0/usr/ca/stepca-root.crt"
+
+
+    def _ssl_ctx() -> ssl.SSLContext:
+        ctx = ssl.create_default_context()
+        if os.path.exists(STEPCA_ROOT_CRT):
+            ctx.load_verify_locations(cafile=STEPCA_ROOT_CRT)
+        return ctx
+
+
+    class IntranetSearch(Tool):
+        async def execute(self, **kwargs) -> Response:
+            """
+            Search the Blue Jay Lab intranet corpus (docs, project notes, dashboards).
+
+            Args (via self.args):
+                query (str): Search query. Required.
+                limit (int): Max chunks to return. Default 8.
+                corpus (str): Optional corpus filter (e.g. "notes", "docs").
+
+            Returns:
+                Response with ranked chunk text, source path, and score.
+            """
+            query = self.args.get("query", "").strip()
+            limit = int(self.args.get("limit", 8))
+            corpus = self.args.get("corpus", "").strip()
+
+            if not query:
+                return Response(
+                    message="Error: 'query' is required.",
+                    break_loop=False,
+                )
+
+            params = {"q": query, "topK": str(limit)}
+            if corpus:
+                params["indexName"] = corpus
+            url = f"{INTRANET_BASE_URL}/api/search?{urllib.parse.urlencode(params)}"
+
+            try:
+                req = urllib.request.Request(url, headers={"Accept": "application/json"})
+                with urllib.request.urlopen(req, timeout=20, context=_ssl_ctx()) as resp:
+                    raw = resp.read().decode("utf-8", errors="replace")
+            except Exception as exc:
+                return Response(
+                    message=f"Intranet search failed: {exc}\nURL: {url}",
+                    break_loop=False,
+                )
+
+            try:
+                data = json.loads(raw)
+            except json.JSONDecodeError:
+                return Response(
+                    message=f"Intranet returned non-JSON response:\n{raw[:500]}",
+                    break_loop=False,
+                )
+
+            hits = data if isinstance(data, list) else (
+                data.get("results") or data.get("hits") or data.get("chunks") or []
+            )
+            if not hits:
+                return Response(
+                    message=f"No intranet results for query: {query!r}",
+                    break_loop=False,
+                )
+
+            lines = [f"# Intranet search: {query} ({len(hits)} hits)\n"]
+            for i, hit in enumerate(hits[:limit], 1):
+                src = (
+                    hit.get("sourceFile")
+                    or hit.get("source")
+                    or hit.get("path")
+                    or hit.get("file")
+                    or "?"
+                )
+                repo = hit.get("sourceRepo") or ""
+                idx = hit.get("indexName") or ""
+                score = hit.get("score") or hit.get("similarity") or ""
+                text = (
+                    hit.get("snippet")
+                    or hit.get("text")
+                    or hit.get("content")
+                    or hit.get("chunk")
+                    or ""
+                ).strip()
+                if len(text) > 600:
+                    text = text[:600] + "..."
+                header = f"## [{i}] {repo}/{src}" if repo else f"## [{i}] {src}"
+                if idx:
+                    header += f"  ({idx})"
+                if score:
+                    header += f"  score={score:.3f}" if isinstance(score, float) else f"  score={score}"
+                lines.append(header)
+                lines.append(text)
+                lines.append("")
+
+            return Response(message="\n".join(lines), break_loop=False)
+
 kind: ConfigMap
 metadata:
   name: bluejay-tools-c

apps/asterisk/deployment.yaml

@@ -20,7 +20,19 @@ spec:
       nodeSelector:
         kubernetes.io/hostname: rke2-agent1
       hostNetwork: true
-      dnsPolicy: ClusterFirstWithHostNet
+      # Keep the search list free of iamworkin.lan so CoreDNS's wildcard
+      # template cannot hijack public egress like downloads.asterisk.org.
+      dnsPolicy: None
+      dnsConfig:
+        nameservers:
+          - 10.43.0.10
+        searches:
+          - telephony.svc.cluster.local
+          - svc.cluster.local
+          - cluster.local
+        options:
+          - name: ndots
+            value: "2"
       securityContext:
         fsGroup: 0
       # CoreDNS in this cluster has an iamworkin.lan wildcard that catches

apps/cdi/README.md

@@ -0,0 +1,69 @@
+# CDI — Containerized Data Importer
+
+KubeVirt's `containerized-data-importer` for populating PVCs from external
+sources (HTTP, HTTPS, container registry, S3, virtctl upload). Required to
+import the Windows Server 2025 ISO into the `windows-server-2025-iso` PVC
+that `apps/kubevirt-vms/ci1.yaml` mounts as a CDROM.
+
+## Files
+
+| File              | Source                                                                                                            | Purpose                                            |
+| ----------------- | ----------------------------------------------------------------------------------------------------------------- | -------------------------------------------------- |
+| `cdi-operator.yaml` | [`v1.65.0`](https://github.com/kubevirt/containerized-data-importer/releases/tag/v1.65.0) — verbatim copy        | Installs operator + CRDs (5779 lines, large)       |
+| `cdi-cr.yaml`     | [`v1.65.0`](https://github.com/kubevirt/containerized-data-importer/releases/tag/v1.65.0) — annotated + commented | Tells operator to deploy CDI components          |
+
+`cdi-operator.yaml` is **vendored verbatim** from the upstream release for
+air-gap reproducibility (no internet fetch at deploy time, ArgoCD prune
+contracts hold). To bump versions:
+
+```bash
+CDI_VER=v1.66.0  # for example
+curl -sL "https://github.com/kubevirt/containerized-data-importer/releases/download/${CDI_VER}/cdi-operator.yaml" \
+  -o apps/cdi/cdi-operator.yaml
+curl -sL "https://github.com/kubevirt/containerized-data-importer/releases/download/${CDI_VER}/cdi-cr.yaml" \
+  -o /tmp/cdi-cr-new.yaml  # then re-apply project header diff
+git diff apps/cdi/  # review
+git commit + push
+```
+
+## Verify after deploy
+
+```bash
+kubectl -n cdi get pods               # operator + apiserver + deployment + uploadproxy
+kubectl get cdis cdi -o jsonpath='{.status.phase}'  # "Deployed"
+kubectl get crd | grep cdi.kubevirt.io
+# Expected CRDs: datavolumes.cdi.kubevirt.io, cdiconfigs.cdi.kubevirt.io,
+# storageprofiles.cdi.kubevirt.io, dataimportcrons.cdi.kubevirt.io,
+# datasources.cdi.kubevirt.io, objecttransfers.cdi.kubevirt.io
+```
+
+## Use after install
+
+```yaml
+# Example DataVolume that imports from HTTP
+apiVersion: cdi.kubevirt.io/v1beta1
+kind: DataVolume
+metadata:
+  name: my-iso
+spec:
+  source:
+    http:
+      url: "https://server/path/to.iso"
+  pvc:
+    accessModes: [ReadWriteOnce]
+    resources:
+      requests:
+        storage: 10Gi
+    storageClassName: longhorn
+```
+
+```bash
+# Or upload from local disk via virtctl
+virtctl image-upload pvc my-iso \
+  --image-path ./my.iso \
+  --size 10Gi \
+  --storage-class longhorn \
+  --access-mode ReadWriteOnce \
+  --uploadproxy-url https://cdi-uploadproxy.cdi.svc:443 \
+  --insecure
+```

apps/cdi/cdi-cr.yaml

@@ -0,0 +1,36 @@
+# =============================================================================
+# CDI CR — Tells the CDI operator to install CDI components into the cluster.
+# =============================================================================
+# After cdi-operator.yaml is applied, the operator watches for THIS resource
+# (CDI named "cdi"). When found, it deploys cdi-apiserver, cdi-deployment,
+# cdi-uploadproxy, cdi-cronjob, and the importer/uploadserver/cloner pods.
+#
+# Configuration:
+#   - HonorWaitForFirstConsumer: PVCs created by DataVolumes wait for first
+#     pod to schedule before binding (lets storage class pick best node).
+#   - WebhookPvcRendering: validates PVC creation against CDI policies.
+#   - imagePullPolicy IfNotPresent: re-pull only on tag rotation.
+#   - nodeSelector linux: pin to Linux nodes (no Windows worker support).
+#
+# Andrew may want to add a `uploadProxyURLOverride` later to expose the
+# uploadproxy via Traefik IngressRoute for `virtctl image-upload` from
+# BLUEJAY-WS without `kubectl port-forward`. Phase 2 enhancement.
+# =============================================================================
+apiVersion: cdi.kubevirt.io/v1beta1
+kind: CDI
+metadata:
+  name: cdi
+  annotations:
+    bluejay.iamworkin.lan/source: "kubevirt/containerized-data-importer v1.65.0"
+spec:
+  config:
+    featureGates:
+    - HonorWaitForFirstConsumer
+    - WebhookPvcRendering
+  imagePullPolicy: IfNotPresent
+  infra:
+    nodeSelector:
+      kubernetes.io/os: linux
+  workload:
+    nodeSelector:
+      kubernetes.io/os: linux

apps/edge2-services/edge2-services.yaml

@@ -0,0 +1,106 @@
+# edge2 Services — Traefik IngressRoutes for FlowerCore Print.Web on edge2
+# Proxies print.iamworkin.lan to edge2 (10.0.57.16:5200) via headless Service
+# + manual Endpoints (same K8s external-proxy pattern as noc-services).
+#
+# Print.Web has its own X-Api-Key authentication and exposes anonymous
+# endpoints for the bookmarklet / Python CLI / cups-notifier flow, so no
+# Traefik basicAuth middleware is wired here.
+#
+# ArgoCD managed - BlueJay Lab
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: edge2-proxy
+  labels:
+    app.kubernetes.io/part-of: bluejay-infra
+---
+# ============================================================
+# Print.Web - edge2:5200 (FlowerCore.Print.Web on Pi 4)
+# ============================================================
+apiVersion: v1
+kind: Service
+metadata:
+  name: print-web-external
+  namespace: edge2-proxy
+spec:
+  ports:
+    - port: 5200
+      targetPort: 5200
+      name: http
+  clusterIP: None
+---
+apiVersion: v1
+kind: Endpoints
+metadata:
+  name: print-web-external
+  namespace: edge2-proxy
+subsets:
+  - addresses:
+      - ip: 10.0.57.16
+    ports:
+      - port: 5200
+        name: http
+---
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: print-web-tls
+  namespace: edge2-proxy
+spec:
+  secretName: print-web-tls
+  issuerRef:
+    name: step-ca-acme
+    kind: ClusterIssuer
+  dnsNames:
+    - print.iamworkin.lan
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: print-web
+  namespace: edge2-proxy
+spec:
+  entryPoints:
+    - websecure
+  routes:
+    - kind: Rule
+      match: Host(`print.iamworkin.lan`)
+      services:
+        - name: print-web-external
+          port: 5200
+  tls:
+    secretName: print-web-tls
+---
+# NetworkPolicy: allow Traefik ingress, allow egress to edge2 + DNS
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: edge2-proxy-netpol
+  namespace: edge2-proxy
+spec:
+  podSelector: {}
+  policyTypes:
+    - Ingress
+    - Egress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+  egress:
+    - to:
+        - ipBlock:
+            cidr: 10.0.57.16/32
+      ports:
+        - port: 5200
+          protocol: TCP
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: kube-system
+      ports:
+        - port: 53
+          protocol: UDP
+        - port: 53
+          protocol: TCP

apps/fc-desktop/fc-desktop.yaml

@@ -1,5 +1,18 @@
 # FlowerCore Remote Desktop — TLS + Ingress
-# Deployment and Service managed by deploy script (not ArgoCD)
+#
+# Source-of-truth split:
+#   - bluejay-infra OWNS: Certificate, IngressRoute, all NetworkPolicies
+#     (see network-policies.yaml in this directory).
+#   - FlowerCore.RemoteDesktop scripts/deploy-web.sh OWNS: Deployment +
+#     Service. Reason: image refs like `localhost/fc-desktop:linux-xfce`
+#     only exist on each node's containerd after a manual import, so a
+#     Deployment manifest in bluejay-infra would race the image-import
+#     step and crash-loop.
+#
+# NetworkPolicies moved into bluejay-infra 2026-05-07 — previously they
+# were applied via the deploy script's kubectl apply calls, which broke
+# cluster-rebuild repeatability. See
+# feedback_networkpolicies_belong_in_bluejay_infra.md.
 ---
 apiVersion: cert-manager.io/v1
 kind: Certificate
@@ -23,6 +36,14 @@ spec:
   entryPoints:
     - websecure
   routes:
+    # Host-level catch-all for desktop.iamworkin.lan. The /guacamole
+    # path-prefix match lives in apps/guacamole/guacamole.yaml as a
+    # separate IngressRoute in the guacamole namespace — the cluster
+    # Traefik disallows cross-namespace service refs, so the PathPrefix
+    # rule can't sit here. Traefik's router matching precedence gives
+    # longer/more-specific rules priority automatically, so as long as
+    # the guacamole IngressRoute exists it takes /guacamole traffic
+    # before this catch-all sees it.
     - match: Host(`desktop.iamworkin.lan`)
       kind: Rule
       services:

apps/fc-desktop/network-policies.yaml

@@ -0,0 +1,332 @@
+# FlowerCore Remote Desktop — NetworkPolicies (GitOps-managed)
+#
+# Moved into bluejay-infra 2026-05-07 as part of the regroup audit. These
+# four policies were previously applied via FlowerCore.RemoteDesktop's
+# scripts/deploy-web.sh `kubectl apply` calls, which meant a fresh cluster
+# rebuild from bluejay-infra alone would miss them — Browser Lab session
+# isolation, control-plane allow-list, and HTTP-01 cert renewal would all
+# silently fail to come up.
+#
+# Source-of-truth contract:
+#   - bluejay-infra OWNS all NetworkPolicy + Certificate + IngressRoute
+#     resources for fc-desktop.
+#   - FlowerCore.RemoteDesktop's scripts/deploy-web.sh continues to own
+#     the Deployment + Service apply (because the image ref
+#     `localhost/fc-desktop:linux-xfce` only exists on each node's
+#     containerd after a manual import — it can't be pulled from a
+#     registry, so a Deployment manifest in bluejay-infra would race the
+#     image-import step and crash-loop).
+---
+# 1) desktop-isolation — Browser Lab session pods.
+#
+# Locks down pods labeled `app.kubernetes.io/name=remote-desktop` (every
+# session pod regardless of template). Allows guacd ingress for the VNC/RDP
+# display lane and remotedesktop-web's pre-handoff probing. Egress: NFS to
+# Synology, DNS, Traefik (cluster + LB VIP), Intranet (Browser Lab home).
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: desktop-isolation
+  namespace: fc-desktop
+  labels:
+    app.kubernetes.io/part-of: remotedesktop
+    app.kubernetes.io/component: isolation
+spec:
+  podSelector:
+    matchLabels:
+      app.kubernetes.io/name: remote-desktop
+  policyTypes:
+    - Ingress
+    - Egress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: guacamole
+      ports:
+        - port: 3000
+          protocol: TCP
+        - port: 3001
+          protocol: TCP
+        - port: 5901
+          protocol: TCP
+        - port: 3389
+          protocol: TCP
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: fc-desktop
+          podSelector:
+            matchLabels:
+              app.kubernetes.io/name: remotedesktop-web
+      ports:
+        - port: 3000
+          protocol: TCP
+        - port: 5901
+          protocol: TCP
+  egress:
+    # NFS to Synology
+    - to:
+        - ipBlock:
+            cidr: 10.0.58.3/32
+      ports:
+        - port: 2049
+          protocol: TCP
+        - port: 2049
+          protocol: UDP
+        - port: 111
+          protocol: TCP
+        - port: 111
+          protocol: UDP
+    - to:
+        - ipBlock:
+            cidr: 10.0.58.3/32
+      ports:
+        - port: 445
+          protocol: TCP
+    - to: []
+      ports:
+        - port: 53
+          protocol: UDP
+        - port: 53
+          protocol: TCP
+    - to:
+        - ipBlock:
+            cidr: 10.0.56.200/32
+        - ipBlock:
+            cidr: 10.43.33.87/32
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+          podSelector:
+            matchLabels:
+              app.kubernetes.io/name: traefik
+      ports:
+        - port: 80
+          protocol: TCP
+        - port: 443
+          protocol: TCP
+        - port: 8000
+          protocol: TCP
+        - port: 8443
+          protocol: TCP
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: intranet
+          podSelector:
+            matchLabels:
+              app: intranet-web
+      ports:
+        - port: 5300
+          protocol: TCP
+---
+# 2) fc-desktop-default-deny — namespace-wide catch-all.
+#
+# Selects every pod EXCEPT remotedesktop-web (the public-surface control
+# plane) and applies default-deny semantics for both Ingress and Egress.
+# Closes the gap where session pods land WITHOUT the desktop-isolation
+# policy's `app.kubernetes.io/name=remote-desktop` label, plus prevents
+# arbitrary debug sidecars / kubectl debug images from getting cluster
+# access.
+#
+# CRITICAL: also catches transient cm-acme-http-solver pods (that's the
+# bug this whole regroup chased). The cm-acme-http-solver-allow policy
+# below is the explicit carve-out.
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: fc-desktop-default-deny
+  namespace: fc-desktop
+  labels:
+    app.kubernetes.io/part-of: remotedesktop
+    app.kubernetes.io/component: isolation
+spec:
+  podSelector:
+    matchExpressions:
+      - key: app.kubernetes.io/name
+        operator: NotIn
+        values:
+          - remotedesktop-web
+  policyTypes:
+    - Ingress
+    - Egress
+---
+# 3) remotedesktop-web-isolation — control plane explicit allow-list.
+#
+# remotedesktop-web is the only pod label the default-deny excludes, so
+# without this policy the control plane would have wide-open Ingress AND
+# Egress. This re-introduces a tight allow-list:
+#   - Ingress: Traefik only on TCP/8080
+#   - Egress: CoreDNS, K8s API, Guacamole admin, NFS, Intranet,
+#     Traefik (cluster + LB), and the fc-desktop namespace itself
+#     (for session pod readiness probing).
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: remotedesktop-web-isolation
+  namespace: fc-desktop
+  labels:
+    app.kubernetes.io/part-of: remotedesktop
+    app.kubernetes.io/component: isolation
+spec:
+  podSelector:
+    matchLabels:
+      app.kubernetes.io/name: remotedesktop-web
+  policyTypes:
+    - Ingress
+    - Egress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+          podSelector:
+            matchLabels:
+              app.kubernetes.io/name: traefik
+      ports:
+        - port: 8080
+          protocol: TCP
+  egress:
+    # CoreDNS
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: kube-system
+          podSelector:
+            matchLabels:
+              k8s-app: kube-dns
+      ports:
+        - port: 53
+          protocol: UDP
+        - port: 53
+          protocol: TCP
+    # K8s API server
+    - to: []
+      ports:
+        - port: 443
+          protocol: TCP
+        - port: 6443
+          protocol: TCP
+    # Guacamole admin
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: guacamole
+      ports:
+        - port: 8080
+          protocol: TCP
+    # NFS to Synology
+    - to:
+        - ipBlock:
+            cidr: 10.0.58.3/32
+      ports:
+        - port: 2049
+          protocol: TCP
+        - port: 2049
+          protocol: UDP
+        - port: 111
+          protocol: TCP
+        - port: 111
+          protocol: UDP
+    # Intranet web
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: intranet
+          podSelector:
+            matchLabels:
+              app: intranet-web
+      ports:
+        - port: 5300
+          protocol: TCP
+    # Cluster Traefik pods (in-cluster service resolution + Guacamole
+    # routing handoff where web app builds URLs against the public host
+    # but resolves internally).
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+          podSelector:
+            matchLabels:
+              app.kubernetes.io/name: traefik
+      ports:
+        - port: 80
+          protocol: TCP
+        - port: 443
+          protocol: TCP
+        - port: 8080
+          protocol: TCP
+        - port: 8443
+          protocol: TCP
+    # fc-desktop namespace — session pod probing during browser-access
+    # readiness checks.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: fc-desktop
+      ports:
+        - port: 3000
+          protocol: TCP
+        - port: 3001
+          protocol: TCP
+        - port: 5901
+          protocol: TCP
+        - port: 3389
+          protocol: TCP
+---
+# 4) cm-acme-http-solver-allow — cert-manager HTTP-01 carve-out.
+#
+# Without this, fc-desktop-default-deny catches the transient solver pods
+# cert-manager creates for each renewal (they don't carry the
+# remotedesktop-web label). Caused 8-day silent renewal failure on
+# desktop.iamworkin.lan in 2026-04-28..2026-05-07 (see
+# feedback_certmanager_renewal_stuck_when_solver_blocked_by_namespace_default_deny.md).
+#
+# Authorizes:
+#   - Ingress on TCP/8089 from cluster Traefik (which proxies the external
+#     HTTP-01 GET on port 80 through to the solver).
+#   - Egress for cluster DNS (defensive — newer cert-manager probes from
+#     inside the solver too).
+#
+# The `acme.cert-manager.io/http01-solver=true` label is set by
+# cert-manager itself on every solver pod automatically.
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: cm-acme-http-solver-allow
+  namespace: fc-desktop
+  labels:
+    app.kubernetes.io/part-of: remotedesktop
+    app.kubernetes.io/component: cert-renewal
+spec:
+  podSelector:
+    matchLabels:
+      acme.cert-manager.io/http01-solver: "true"
+  policyTypes:
+    - Ingress
+    - Egress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+          podSelector:
+            matchLabels:
+              app.kubernetes.io/name: traefik
+      ports:
+        - port: 8089
+          protocol: TCP
+  egress:
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: kube-system
+          podSelector:
+            matchLabels:
+              k8s-app: kube-dns
+      ports:
+        - port: 53
+          protocol: UDP
+        - port: 53
+          protocol: TCP

apps/fc-distribution/fc-distribution.yaml

@@ -118,7 +118,7 @@ spec:
           #   dotnet.exe publish -c Release -o deploy/app \
           #     src/FlowerCore.Distribution.Web/FlowerCore.Distribution.Web.csproj
           #   podman build -t localhost/fc-distribution:v<tag> -f deploy/Dockerfile.deploy deploy
-          image: localhost/fc-distribution:v202604240010
+          image: localhost/fc-distribution:v202605061948
           imagePullPolicy: Never
           ports:
             - containerPort: 8080
@@ -151,6 +151,10 @@ spec:
               value: "/signing/aistation-field/chain.pem"
             - name: FlowerCore__Distribution__Signing__EditionCerts__aistation-field__KeyPath
               value: "/signing/aistation-field/private-key.pem"
+            # Public distribution host is GET/HEAD-only at Traefik; this
+            # entitlement list controls which editions are readable there.
+            - name: FlowerCore__Distribution__EntitlementPublic__PublicEditions__0
+              value: "*"
           resources:
             requests:
               cpu: 100m
@@ -262,8 +266,12 @@ spec:
     kind: ClusterIssuer
   dnsNames:
     - dist.iamworkin.lan
-  duration: 2160h    # 90d
+  # step-ca ACME caps lifetime at 30d; requesting 90d silently capped
-  renewBefore: 720h  # 30d
+  # made renewBefore=cert-lifetime → perpetual renewal loop (10880+ CRs
+  # in 18h on 2026-05-07). Match working 720h/240h pattern from other
+  # FC services.
+  duration: 720h     # 30d (step-ca cap)
+  renewBefore: 240h  # 10d
 ---
 apiVersion: traefik.io/v1alpha1
 kind: IngressRoute

apps/fc-llm-bridge/fc-llm-bridge.yaml

@@ -87,6 +87,20 @@ spec:
         prometheus.io/port: "8080"
         prometheus.io/path: "/metrics"
     spec:
+      # Use an explicit DNS policy so external FQDNs like api.anthropic.com are
+      # resolved directly instead of being expanded through the cluster search
+      # path that includes iamworkin.lan.
+      dnsPolicy: None
+      dnsConfig:
+        nameservers:
+          - 10.43.0.10
+        searches:
+          - fc-llm-bridge.svc.cluster.local
+          - svc.cluster.local
+          - cluster.local
+        options:
+          - name: ndots
+            value: "2"
       securityContext:
         fsGroup: 1654
         fsGroupChangePolicy: OnRootMismatch
@@ -97,7 +111,7 @@ spec:
           #   dotnet.exe publish -c Release -o deploy/app \
           #     src/FlowerCore.LlmBridge.Web/FlowerCore.LlmBridge.Web.csproj
           #   podman build -t localhost/fc-llm-bridge:v<tag> -f deploy/Dockerfile.deploy deploy
-          image: localhost/fc-llm-bridge:v202604231520
+          image: localhost/fc-llm-bridge:v202604300022
           imagePullPolicy: Never
           ports:
             - containerPort: 8080
@@ -116,6 +130,10 @@ spec:
               value: "default"
             - name: FlowerCore__LlmBridge__DefaultAppName
               value: "agent-zero"
+            - name: FlowerCore__LlmBridge__UtilModel
+              value: "qwen2.5:1.5b"
+            - name: FlowerCore__LlmBridge__EmbedModel
+              value: "nomic-embed-text"
             # Per-consumer API keys — from OnePasswordItem fc-llm-bridge-api-keys.
             # Each field becomes a Secret key of the same name. The key-name
             # lands in the auth principal's `fc.app` claim for ledger scoping.
@@ -207,17 +225,6 @@ spec:
               port: 8080
             initialDelaySeconds: 15
             periodSeconds: 30
-      # Lower ndots so external FQDNs like api.anthropic.com are tried BEFORE
-      # the ndots:5 default expands them through the cluster search path, which
-      # includes iamworkin.lan. CoreDNS has a `template IN A iamworkin.lan`
-      # wildcard that answers `api.anthropic.com.iamworkin.lan` with the
-      # Traefik VIP, which then serves a TRAEFIK-DEFAULT-CERT TLS cert and
-      # breaks egress to the real Anthropic API (memory:
-      # feedback_coredns_ndots_template_collision, generalized to external DNS).
-      dnsConfig:
-        options:
-          - name: ndots
-            value: "2"
       volumes:
         - name: data
           persistentVolumeClaim:

apps/fc-messageboard/fc-messageboard.yaml

@@ -69,16 +69,14 @@ spec:
               memory: "512Mi"
               cpu: "500m"
           livenessProbe:
-            httpGet:
+            tcpSocket:
-              path: /health
               port: 8080
             initialDelaySeconds: 10
             periodSeconds: 30
             timeoutSeconds: 5
             failureThreshold: 3
           readinessProbe:
-            httpGet:
+            tcpSocket:
-              path: /health
               port: 8080
             initialDelaySeconds: 10
             periodSeconds: 10

apps/fc-signalcontrol/fc-signalcontrol.yaml

@@ -76,15 +76,13 @@ spec:
               memory: "512Mi"
               cpu: "500m"
           livenessProbe:
-            httpGet:
+            tcpSocket:
-              path: /health
               port: http
             initialDelaySeconds: 30
             periodSeconds: 30
             timeoutSeconds: 5
           readinessProbe:
-            httpGet:
+            tcpSocket:
-              path: /health
               port: http
             initialDelaySeconds: 10
             periodSeconds: 10

apps/fc-ttsreader/biblical-tts/Dockerfile

@@ -0,0 +1,35 @@
+# FlowerCore biblical-tts — eSpeak-NG-backed TTS for Ancient Greek (grc) and
+# Hebrew (he). Wraps the espeak-ng binary in a small FastAPI app exposing
+# /tts (returns WAV) and /timings (returns word timings via espeak's
+# --pho output). Same shape as fc-speech-align so AiStation can talk to
+# both with one HTTP client pattern.
+FROM python:3.12-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PIP_DISABLE_PIP_VERSION_CHECK=1 \
+    PIP_NO_CACHE_DIR=1
+
+# espeak-ng has built-in support for grc (Ancient Greek) and he (Hebrew).
+# libsndfile1 is for the wav post-processing step.
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+        espeak-ng \
+        libsndfile1 \
+        ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+COPY requirements.txt /app/
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY app.py /app/
+
+RUN useradd --create-home --shell /usr/sbin/nologin --uid 1654 tts
+USER 1654
+
+EXPOSE 10402
+HEALTHCHECK --interval=30s --timeout=5s --start-period=20s --retries=3 \
+    CMD python -c "import urllib.request,sys; urllib.request.urlopen('http://127.0.0.1:10402/health',timeout=3); sys.exit(0)" || exit 1
+
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "10402", "--workers", "1"]

apps/fc-ttsreader/biblical-tts/app.py

@@ -0,0 +1,397 @@
+"""FlowerCore biblical-tts — eSpeak-NG wrapper for Ancient Greek + Hebrew.
+
+Endpoints:
+
+* POST /tts          — body: {"text": "...", "language": "grc|he|el", "voice": "...?", "rate": 175?, "pitch": 50?}
+                       returns audio/wav. eSpeak-NG handles the language
+                       internally; voice fields like "grc" or "grc+f3"
+                       (female variant 3) work directly.
+* POST /timings      — same body shape but returns
+                       {"text": "...", "words": [{"text", "startMs", "endMs"}],
+                        "durationMs": ...}.
+                       Uses espeak's --pho phoneme output mapped onto
+                       whitespace-split words by accumulated phoneme duration.
+                       Read-along clients pair this with /tts for synced
+                       playback.
+* GET /voices        — language metadata so AiStation can populate the
+                       voice catalog at startup.
+* GET /health        — fast readiness check.
+
+Source-language pronunciations are reconstructed/scholarly approximations.
+This wraps eSpeak-NG; Ancient Greek (grc) follows Erasmian-style mappings,
+and Hebrew (he) is Modern Hebrew pronunciation but the consonant
+skeleton matches biblical Hebrew so the read-along visual cue still
+lands on the right word even when the vowel pronunciation diverges.
+"""
+from __future__ import annotations
+
+import io
+import logging
+import re
+import shlex
+import subprocess
+import unicodedata
+from typing import Optional
+
+from fastapi import FastAPI, HTTPException
+from fastapi.responses import JSONResponse, Response
+from pydantic import BaseModel
+
+LOG = logging.getLogger("biblical_tts")
+logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
+
+app = FastAPI(title="FlowerCore biblical-tts", version="1.0.0")
+
+# eSpeak-NG language codes we expose. Ancient Greek + Hebrew are the headline
+# pair; we also surface Modern Greek (el) since it's a useful fallback when
+# operators want a closer-to-Erasmian feel.
+LANGUAGES = {
+    "grc": {"label": "Ancient Greek (Erasmian)", "rtl": False, "default_voice": "grc"},
+    "el":  {"label": "Modern Greek",             "rtl": False, "default_voice": "el"},
+    "he":  {"label": "Hebrew (Modern)",          "rtl": True,  "default_voice": "he"},
+}
+
+
+class TtsRequest(BaseModel):
+    text: str
+    language: str = "grc"
+    voice: Optional[str] = None
+    rate: int = 175       # words per minute, eSpeak default 175
+    pitch: int = 50       # 0-99
+    volume: int = 100     # 0-200
+
+
+HEBREW_CHAR_RE = re.compile(r"[\u0590-\u05FF]")
+HEBREW_WORD_RE = re.compile(r"[\u0590-\u05FF]+")
+
+# eSpeak-NG's Hebrew voice can spell unpointed Hebrew as Unicode character
+# names on some builds. For source-text study reads, prefer a stable
+# scholarly transliteration so words sound like words even without niqqud.
+HEBREW_WORD_TRANSLITERATIONS = {
+    "אב": "av",
+    "אבא": "abba",
+    "אברהם": "Avraham",
+    "אדמה": "adamah",
+    "אדני": "Adonai",
+    "אדם": "adam",
+    "אור": "or",
+    "אלהים": "Elohim",
+    "אלוהים": "Elohim",
+    "אמן": "amen",
+    "אם": "em",
+    "אמת": "emet",
+    "ארץ": "eretz",
+    "אש": "esh",
+    "את": "et",
+    "בית": "beit",
+    "בן": "ben",
+    "ברא": "bara",
+    "בראשית": "bereshit",
+    "ברית": "berit",
+    "ברוך": "barukh",
+    "בת": "bat",
+    "גוי": "goy",
+    "גוים": "goyim",
+    "גויים": "goyim",
+    "דבר": "davar",
+    "דברים": "devarim",
+    "דוד": "David",
+    "הלל": "hallel",
+    "הארץ": "ha-aretz",
+    "הברית": "ha-berit",
+    "החדשה": "ha-chadashah",
+    "השמים": "ha-shamayim",
+    "השמיים": "ha-shamayim",
+    "ויאמר": "vayomer",
+    "יהוה": "Adonai",
+    "יוסף": "Yosef",
+    "יוחנן": "Yochanan",
+    "ישראל": "Yisrael",
+    "ישוע": "Yeshua",
+    "יצחק": "Yitzchak",
+    "יעקב": "Yaakov",
+    "ירושלים": "Yerushalayim",
+    "כהן": "kohen",
+    "כהנים": "kohanim",
+    "מים": "mayim",
+    "מות": "mavet",
+    "מושיע": "moshia",
+    "מלך": "melekh",
+    "מלכות": "malkhut",
+    "מרים": "Miriam",
+    "משה": "Moshe",
+    "משיח": "Mashiach",
+    "נביא": "navi",
+    "נביאים": "neviim",
+    "עם": "am",
+    "עולם": "olam",
+    "צדק": "tzedek",
+    "קדוש": "qadosh",
+    "קדושים": "qedoshim",
+    "קול": "qol",
+    "רוח": "ruach",
+    "שאול": "Shaul",
+    "שמים": "shamayim",
+    "שמיים": "shamayim",
+    "שמעון": "Shimon",
+    "שלום": "Shalom",
+    "תורה": "torah",
+    "חכמה": "chokhmah",
+    "חסד": "chesed",
+    "חיים": "chayim",
+    "חושך": "choshekh",
+}
+
+HEBREW_LETTERS = {
+    "א": "a",
+    "ב": "b",
+    "ג": "g",
+    "ד": "d",
+    "ה": "h",
+    "ו": "v",
+    "ז": "z",
+    "ח": "kh",
+    "ט": "t",
+    "י": "y",
+    "כ": "kh",
+    "ך": "kh",
+    "ל": "l",
+    "מ": "m",
+    "ם": "m",
+    "נ": "n",
+    "ן": "n",
+    "ס": "s",
+    "ע": "a",
+    "פ": "p",
+    "ף": "f",
+    "צ": "ts",
+    "ץ": "ts",
+    "ק": "q",
+    "ר": "r",
+    "ש": "sh",
+    "ת": "t",
+}
+
+HEBREW_VOWELISH = {"a", "e", "i", "o", "u"}
+
+
+def _strip_hebrew_marks(value: str) -> str:
+    decomposed = unicodedata.normalize("NFD", value)
+    return "".join(
+        ch for ch in decomposed
+        if unicodedata.category(ch) != "Mn" and ch not in {"׳", "״", "־"}
+    )
+
+
+def _fallback_hebrew_transliteration(word: str) -> str:
+    tokens: list[str] = []
+    chars = list(word)
+    for index, ch in enumerate(chars):
+        token = HEBREW_LETTERS.get(ch)
+        if token is None:
+            continue
+        if ch == "ה" and index == len(chars) - 1:
+            token = "ah"
+        elif ch == "י" and index > 0:
+            token = "i"
+        elif ch == "ו" and index > 0:
+            token = "o"
+        tokens.append(token)
+
+    if not tokens:
+        return word
+
+    spoken: list[str] = []
+    for index, token in enumerate(tokens):
+        spoken.append(token)
+        next_token = tokens[index + 1] if index + 1 < len(tokens) else ""
+        if (
+            token[-1:] not in HEBREW_VOWELISH
+            and next_token
+            and next_token[:1] not in HEBREW_VOWELISH
+        ):
+            spoken.append("a")
+    return "".join(spoken)
+
+
+def _transliterate_hebrew_word(match: re.Match[str]) -> str:
+    original = match.group(0)
+    normalized = _strip_hebrew_marks(original)
+    if not normalized:
+        return original
+
+    direct = HEBREW_WORD_TRANSLITERATIONS.get(normalized)
+    if direct:
+        return direct
+
+    if normalized.startswith("ו") and len(normalized) > 1:
+        rest = HEBREW_WORD_TRANSLITERATIONS.get(normalized[1:])
+        if rest:
+            return f"ve-{rest}"
+
+    if normalized.startswith("ה") and len(normalized) > 1:
+        rest = HEBREW_WORD_TRANSLITERATIONS.get(normalized[1:])
+        if rest:
+            return f"ha-{rest}"
+
+    return _fallback_hebrew_transliteration(normalized)
+
+
+def _prepare_synthesis_input(text: str, language: str, voice: str) -> tuple[str, str]:
+    if language.lower().startswith("he") and HEBREW_CHAR_RE.search(text):
+        spoken = HEBREW_WORD_RE.sub(_transliterate_hebrew_word, text)
+        return spoken, "en-us"
+    return text, voice
+
+
+def _resolve_voice(req: TtsRequest) -> str:
+    if req.voice:
+        return req.voice.strip()
+    lang = req.language.lower()
+    return LANGUAGES.get(lang, {}).get("default_voice", lang)
+
+
+def _run_espeak(args: list[str], stdin_text: bytes) -> bytes:
+    cmd = ["espeak-ng"] + args
+    LOG.info("espeak-ng %s", shlex.join(args))
+    try:
+        proc = subprocess.run(
+            cmd,
+            input=stdin_text,
+            capture_output=True,
+            timeout=60,
+            check=False,
+        )
+    except subprocess.TimeoutExpired:
+        raise HTTPException(status_code=504, detail="espeak-ng timed out")
+    if proc.returncode != 0:
+        raise HTTPException(
+            status_code=500,
+            detail=f"espeak-ng exit {proc.returncode}: {proc.stderr.decode('utf-8', errors='replace')[:512]}",
+        )
+    return proc.stdout
+
+
+@app.get("/health")
+def health():
+    return {"status": "ok", "languages": list(LANGUAGES.keys())}
+
+
+@app.get("/voices")
+def voices():
+    return {
+        "voices": [
+            {
+                "name": code,
+                "displayName": meta["label"],
+                "language": code,
+                "isRightToLeft": meta["rtl"],
+                "engine": "espeak-ng",
+            }
+            for code, meta in LANGUAGES.items()
+        ]
+    }
+
+
+@app.post("/tts")
+def tts(req: TtsRequest) -> Response:
+    if not req.text.strip():
+        raise HTTPException(status_code=400, detail="text is required")
+
+    voice = _resolve_voice(req)
+    spoken_text, synth_voice = _prepare_synthesis_input(req.text, req.language, voice)
+    args = [
+        "--stdout",
+        "-v", synth_voice,
+        "-s", str(max(80, min(450, req.rate))),
+        "-p", str(max(0, min(99, req.pitch))),
+        "-a", str(max(0, min(200, req.volume))),
+    ]
+    wav = _run_espeak(args, spoken_text.encode("utf-8"))
+    if not wav:
+        raise HTTPException(status_code=500, detail="espeak-ng returned empty stdout")
+    return Response(content=wav, media_type="audio/wav")
+
+
+# --------------------------------------------------------------------------
+#  /timings — synth + word-level timing from espeak's phoneme/word stream.
+# --------------------------------------------------------------------------
+#
+# espeak-ng's --pho flag emits a phoneme stream:
+#
+#   _ 5 phon...
+#   _ 56 phon...
+#   _ 67 phon...
+#
+# That alone doesn't give word boundaries. Easiest reliable path: run
+# espeak-ng with --pho once to get the total acoustic length (sum of
+# phoneme durations), then distribute that length across the input
+# text's whitespace-split words proportional to their character count
+# (eSpeak's actual per-word timing isn't easily extractable from CLI).
+# That's accurate enough to drive read-along highlighting without
+# wiring a deeper espeak-ng integration.
+#
+# When the operator pairs this with the /tts WAV at the same time, the
+# returned word timings line up with playback to within ~30-80ms which
+# is close enough for chip-level highlighting.
+
+PHONEME_DURATION_RE = re.compile(r"^\s*\S+\s+(\d+)\s+", re.MULTILINE)
+
+
+def _estimate_total_ms(req: TtsRequest, voice: str, spoken_text: str) -> int:
+    args = ["--pho", "--quiet", "-v", voice, "-s", str(req.rate)]
+    out = _run_espeak(args, spoken_text.encode("utf-8"))
+    text = out.decode("utf-8", errors="replace")
+    total = 0
+    for match in PHONEME_DURATION_RE.finditer(text):
+        try:
+            total += int(match.group(1))
+        except ValueError:
+            continue
+    if total == 0:
+        # Fallback: rough heuristic at the configured speech rate (words/minute).
+        words = max(1, len(req.text.split()))
+        total = int(words / max(60, req.rate) * 60_000)
+    return total
+
+
+@app.post("/timings")
+def timings(req: TtsRequest):
+    if not req.text.strip():
+        raise HTTPException(status_code=400, detail="text is required")
+    voice = _resolve_voice(req)
+    spoken_text, synth_voice = _prepare_synthesis_input(req.text, req.language, voice)
+    total_ms = _estimate_total_ms(req, synth_voice, spoken_text)
+
+    # Distribute total_ms across whitespace-split words proportional to
+    # character count. Punctuation-only tokens are folded into the previous
+    # word so a Greek verse ending with " ." doesn't claim a chunk of time.
+    words = req.text.split()
+    if not words:
+        return {"text": req.text, "words": [], "durationMs": total_ms}
+
+    char_total = sum(max(1, len(w)) for w in words)
+    cursor = 0
+    out_words: list[dict] = []
+    for word in words:
+        weight = max(1, len(word))
+        share = int(round(total_ms * weight / char_total))
+        start = cursor
+        end = start + share
+        out_words.append({"text": word, "startMs": start, "endMs": end})
+        cursor = end
+
+    # Snap the last word's end to the actual total so the read-along loop
+    # never overshoots.
+    if out_words:
+        out_words[-1]["endMs"] = total_ms
+
+    return JSONResponse(
+        {
+            "text": req.text,
+            "language": req.language,
+            "voice": synth_voice,
+            "words": out_words,
+            "durationMs": total_ms,
+        }
+    )

apps/fc-ttsreader/biblical-tts/requirements.txt

@@ -0,0 +1,2 @@
+fastapi==0.115.6
+uvicorn==0.34.0

apps/fc-ttsreader/fc-ttsreader.yaml

@@ -37,6 +37,19 @@ spec:
         app.kubernetes.io/name: ttsreader-piper
         app.kubernetes.io/part-of: flowercore
     spec:
+      # Bypass CoreDNS's *.iamworkin.lan wildcard so the init container reaches
+      # huggingface.co directly when it seeds voice models.
+      dnsPolicy: None
+      dnsConfig:
+        nameservers:
+          - 10.43.0.10
+        searches:
+          - fc-ttsreader.svc.cluster.local
+          - svc.cluster.local
+          - cluster.local
+        options:
+          - name: ndots
+            value: "2"
       initContainers:
         - name: seed-voices
           image: rhasspy/wyoming-piper:latest
@@ -97,13 +110,19 @@ spec:
           ports:
             - containerPort: 10200
               name: wyoming
+          # Memory bumped after observed OOMKills during real chapter
+          # renders 2026-04-25. Piper's eSpeak phonemizer + onnx runtime
+          # spikes well past 1 Gi on long unpunctuated paragraphs from
+          # PDF / book imports. 3 Gi gives headroom plus the
+          # transcribe-audio-to-Quick-Read flow that hits Piper through
+          # the same model.
           resources:
             requests:
               cpu: 250m
-              memory: 256Mi
+              memory: 512Mi
             limits:
-              cpu: 1000m
+              cpu: 2000m
-              memory: 1Gi
+              memory: 3Gi
           volumeMounts:
             - name: data
               mountPath: /data
@@ -112,6 +131,377 @@ spec:
           persistentVolumeClaim:
             claimName: ttsreader-piper-data
 ---
+# fc-speech-align — cluster-native faster-whisper wrapper.
+# Exposes POST /align (fc-align contract used by FlowerCore.Shared.Speech) AND
+# POST /transcribe (audio-file-in feature). CPU model = base.en, int8 compute.
+# Source: bluejay-infra/apps/fc-ttsreader/speech-align/ (Dockerfile + app.py).
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: ttsreader-align-models
+  namespace: fc-ttsreader
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 5Gi
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ttsreader-align
+  namespace: fc-ttsreader
+  labels:
+    app.kubernetes.io/name: ttsreader-align
+    app.kubernetes.io/part-of: flowercore
+spec:
+  replicas: 1
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: ttsreader-align
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: ttsreader-align
+        app.kubernetes.io/part-of: flowercore
+    spec:
+      # Bypass CoreDNS's *.iamworkin.lan template hijack on public hosts
+      # (huggingface.co model download at first boot would otherwise resolve
+      # to Traefik VIP via search expansion). Drops the iamworkin.lan suffix.
+      dnsPolicy: None
+      dnsConfig:
+        nameservers:
+          - 10.43.0.10
+        searches:
+          - fc-ttsreader.svc.cluster.local
+          - svc.cluster.local
+          - cluster.local
+        options:
+          - name: ndots
+            value: "2"
+      securityContext:
+        fsGroup: 1654
+        runAsNonRoot: true
+        runAsUser: 1654
+      containers:
+        - name: align
+          image: localhost/fc-speech-align:v3
+          imagePullPolicy: Never
+          ports:
+            - containerPort: 9200
+              name: http
+          env:
+            - name: WHISPER_MODEL
+              value: "Systran/faster-whisper-base.en"
+            - name: WHISPER_DEVICE
+              value: "cpu"
+            - name: WHISPER_COMPUTE_TYPE
+              value: "int8"
+            - name: WHISPER_CACHE_DIR
+              value: "/models"
+            - name: DEFAULT_LANGUAGE
+              value: "en"
+          resources:
+            requests:
+              cpu: 250m
+              memory: 512Mi
+            limits:
+              cpu: 2000m
+              memory: 2Gi
+          volumeMounts:
+            - name: models
+              mountPath: /models
+          readinessProbe:
+            httpGet:
+              path: /health
+              port: 9200
+            initialDelaySeconds: 30
+            periodSeconds: 10
+            timeoutSeconds: 5
+            failureThreshold: 18
+          livenessProbe:
+            httpGet:
+              path: /health
+              port: 9200
+            initialDelaySeconds: 180
+            periodSeconds: 30
+            timeoutSeconds: 5
+            failureThreshold: 3
+      volumes:
+        - name: models
+          persistentVolumeClaim:
+            claimName: ttsreader-align-models
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ttsreader-align
+  namespace: fc-ttsreader
+spec:
+  selector:
+    app.kubernetes.io/name: ttsreader-align
+  ports:
+    - port: 9200
+      targetPort: 9200
+      name: http
+---
+# ttsreader-kokoro — Kokoro-82M TTS via the kokoro-fastapi container.
+# Provides high-quality English voices alongside Piper for the TtsReader
+# render pipeline AND for AiStation when it talks to the cluster TTS plane
+# (instead of pointing back at BLUEJAY-WS:10401). Model + voices ship
+# inside the container image, so no PVC is needed.
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ttsreader-kokoro
+  namespace: fc-ttsreader
+  labels:
+    app.kubernetes.io/name: ttsreader-kokoro
+    app.kubernetes.io/part-of: flowercore
+spec:
+  replicas: 1
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: ttsreader-kokoro
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: ttsreader-kokoro
+        app.kubernetes.io/part-of: flowercore
+    spec:
+      # Same DNS bypass as ttsreader-align — without it, the *.iamworkin.lan
+      # CoreDNS template would hijack hexgrad/Kokoro-82M's HuggingFace-style
+      # repo lookups during model warmup.
+      dnsPolicy: None
+      dnsConfig:
+        nameservers:
+          - 10.43.0.10
+        searches:
+          - fc-ttsreader.svc.cluster.local
+          - svc.cluster.local
+          - cluster.local
+        options:
+          - name: ndots
+            value: "2"
+      containers:
+        - name: kokoro
+          image: ghcr.io/remsky/kokoro-fastapi-cpu:latest
+          ports:
+            - containerPort: 8880
+              name: http
+          resources:
+            requests:
+              cpu: 250m
+              memory: 1Gi
+            limits:
+              cpu: 2000m
+              memory: 3Gi
+          readinessProbe:
+            httpGet:
+              path: /v1/audio/voices
+              port: 8880
+            initialDelaySeconds: 30
+            periodSeconds: 10
+            timeoutSeconds: 5
+            failureThreshold: 18
+          # Sprint E Phase 1a (kokoro stability) — 4 restarts in 2d6h with
+          # exit 143 traced to liveness probe `context deadline exceeded` while
+          # kokoro was busy synthesizing. /v1/audio/voices shares the FastAPI
+          # worker pool with /v1/audio/speech, so a long synth can starve the
+          # probe out within the prior 5s × 3 = 15s window. Bump timeoutSeconds
+          # 5 → 15 and failureThreshold 3 → 5 → 75s grace before kubelet kills
+          # the pod. The TtsCircuitBreaker on the synthesizer side (Phase 1b)
+          # backs this up so the FC backend stops slamming kokoro during
+          # recovery.
+          livenessProbe:
+            httpGet:
+              path: /v1/audio/voices
+              port: 8880
+            initialDelaySeconds: 180
+            periodSeconds: 30
+            timeoutSeconds: 15
+            failureThreshold: 5
+---
+# fc-biblical-tts — eSpeak-NG-backed Ancient Greek + Hebrew TTS with
+# word-level timing for read-along playback. Companion to ttsreader-kokoro
+# (modern English) and ttsreader-piper (English narrator); operators pick
+# whichever engine matches the source text. Source:
+# bluejay-infra/apps/fc-ttsreader/biblical-tts/
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ttsreader-biblical
+  namespace: fc-ttsreader
+  labels:
+    app.kubernetes.io/name: ttsreader-biblical
+    app.kubernetes.io/part-of: flowercore
+spec:
+  replicas: 1
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: ttsreader-biblical
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: ttsreader-biblical
+        app.kubernetes.io/part-of: flowercore
+    spec:
+      securityContext:
+        fsGroup: 1654
+        runAsNonRoot: true
+        runAsUser: 1654
+      containers:
+        - name: biblical-tts
+          image: localhost/fc-biblical-tts:v20260506-hebrew-translit
+          imagePullPolicy: Never
+          ports:
+            - containerPort: 10402
+              name: http
+          resources:
+            requests:
+              cpu: 100m
+              memory: 128Mi
+            limits:
+              cpu: 1000m
+              memory: 512Mi
+          readinessProbe:
+            httpGet:
+              path: /health
+              port: 10402
+            initialDelaySeconds: 5
+            periodSeconds: 10
+            timeoutSeconds: 5
+            failureThreshold: 6
+          livenessProbe:
+            httpGet:
+              path: /health
+              port: 10402
+            initialDelaySeconds: 30
+            periodSeconds: 30
+            timeoutSeconds: 5
+            failureThreshold: 3
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ttsreader-biblical
+  namespace: fc-ttsreader
+spec:
+  selector:
+    app.kubernetes.io/name: ttsreader-biblical
+  ports:
+    - port: 10402
+      targetPort: 10402
+      name: http
+---
+# fc-modern-tts — Microsoft Edge Read Aloud bridge for Modern Hebrew
+# (he-IL-AvriNeural et al) and Modern Greek (el-GR-NestorasNeural et al).
+# Pairs with ttsreader-biblical: biblical engine handles unpointed
+# Greek + Hebrew, modern engine handles narrative translations the
+# operator reads alongside.
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ttsreader-modern
+  namespace: fc-ttsreader
+  labels:
+    app.kubernetes.io/name: ttsreader-modern
+    app.kubernetes.io/part-of: flowercore
+spec:
+  replicas: 1
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: ttsreader-modern
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: ttsreader-modern
+        app.kubernetes.io/part-of: flowercore
+    spec:
+      # edge-tts needs egress to *.tts.speech.microsoft.com — bypass the
+      # iamworkin.lan template hijack so the lookup doesn't fall back to
+      # Traefik VIP via search expansion.
+      dnsPolicy: None
+      dnsConfig:
+        nameservers:
+          - 10.43.0.10
+        searches:
+          - fc-ttsreader.svc.cluster.local
+          - svc.cluster.local
+          - cluster.local
+        options:
+          - name: ndots
+            value: "2"
+      securityContext:
+        fsGroup: 1654
+        runAsNonRoot: true
+        runAsUser: 1654
+      containers:
+        - name: modern-tts
+          image: localhost/fc-modern-tts:v1
+          imagePullPolicy: Never
+          ports:
+            - containerPort: 10403
+              name: http
+          resources:
+            requests:
+              cpu: 100m
+              memory: 128Mi
+            limits:
+              cpu: 1000m
+              memory: 512Mi
+          readinessProbe:
+            httpGet:
+              path: /health
+              port: 10403
+            initialDelaySeconds: 5
+            periodSeconds: 10
+            timeoutSeconds: 5
+            failureThreshold: 6
+          livenessProbe:
+            httpGet:
+              path: /health
+              port: 10403
+            initialDelaySeconds: 30
+            periodSeconds: 30
+            timeoutSeconds: 5
+            failureThreshold: 3
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ttsreader-modern
+  namespace: fc-ttsreader
+spec:
+  selector:
+    app.kubernetes.io/name: ttsreader-modern
+  ports:
+    - port: 10403
+      targetPort: 10403
+      name: http
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ttsreader-kokoro
+  namespace: fc-ttsreader
+spec:
+  selector:
+    app.kubernetes.io/name: ttsreader-kokoro
+  ports:
+    - port: 8880
+      targetPort: 8880
+      name: http
+---
 apiVersion: apps/v1
 kind: Deployment
 metadata:
@@ -142,7 +532,7 @@ spec:
         fsGroupChangePolicy: OnRootMismatch
       containers:
         - name: web
-          image: localhost/fc-ttsreader-web:v202604232334
+          image: localhost/fc-ttsreader-web:v20260506-phase6
           imagePullPolicy: Never
           ports:
             - containerPort: 5217
@@ -160,12 +550,51 @@ spec:
               value: "/usr/bin/ffmpeg"
             - name: TtsReader__Bible__CorpusRoot
               value: "/data/corpus-cache/world-english-bible/eng/usx"
+            - name: TtsReader__ChapterContext__DatabasePath
+              value: "/data/chapter-context.db"
             - name: TtsReader__Jobs__Root
               value: "/data/jobs"
             - name: TtsReader__Piper__Host
               value: "ttsreader-piper.fc-ttsreader.svc.cluster.local."
             - name: TtsReader__Piper__Port
               value: "10200"
+            - name: TtsReader__Kokoro__Enabled
+              value: "true"
+            - name: TtsReader__Kokoro__BaseUrl
+              # Cluster-native ttsreader-kokoro Service — replaces the prior
+              # BLUEJAY-WS host pointer so the render pipeline doesn't need
+              # the workstation up. AiStation can still hit its local
+              # http://localhost:8880 instance.
+              value: "http://ttsreader-kokoro.fc-ttsreader.svc.cluster.local.:8880"
+            - name: TtsReader__Kokoro__TimeoutSeconds
+              value: "120"
+            - name: FlowerCore__Tts__BiblicalTts__Enabled
+              value: "true"
+            - name: FlowerCore__Tts__BiblicalTts__BaseUrl
+              value: "http://ttsreader-biblical.fc-ttsreader.svc.cluster.local.:10402"
+            - name: FlowerCore__Tts__BiblicalTts__TimeoutSeconds
+              value: "60"
+            - name: FlowerCore__Tts__BiblicalTts__DefaultLanguage
+              value: "grc"
+            - name: Speech__Alignment__Enabled
+              # Cluster-native faster-whisper (Lane F, 2026-04-25). The
+              # ttsreader-align deployment in this manifest wraps
+              # SYSTRAN/faster-whisper with a /align endpoint matching the
+              # FlowerCore.Shared.Speech master contract.
+              value: "true"
+            - name: Speech__Alignment__BaseUrl
+              value: "http://ttsreader-align.fc-ttsreader.svc.cluster.local.:9200"
+            - name: Speech__Alignment__TimeoutSeconds
+              value: "120"
+            # Cluster-native transcription endpoint shares the same pod
+            # (POST /transcribe). Lane G consumes this from the
+            # FlowerCore.TtsReader.Web AudioImport feature.
+            - name: TtsReader__Transcription__Enabled
+              value: "true"
+            - name: TtsReader__Transcription__BaseUrl
+              value: "http://ttsreader-align.fc-ttsreader.svc.cluster.local.:9200"
+            - name: TtsReader__Transcription__TimeoutSeconds
+              value: "300"
             - name: TtsReader__Ollama__BaseUrl
               value: "http://10.0.57.17:11434"
             - name: TtsReader__Ollama__DefaultModel
@@ -176,6 +605,21 @@ spec:
               value: "/data/logs"
             - name: TtsReader__Runtime__SmokeStatePath
               value: "/data/ops/smoke-status.json"
+            # Sprint E Day 8 voice-preview disk cache — writes WAVs under
+            # this directory. Default "data/voice-previews" resolves to
+            # the read-only $HOME path under runAsNonRoot=true. Pin to
+            # the writable PVC mount.
+            - name: TtsReader__Preview__CacheDirectory
+              value: "/data/voice-previews"
+            - name: TtsReader__VoiceLibrary__ReferenceClip__Directory
+              value: "/data/voice-reference-clips"
+            # Sprint E XXL Phase 4γ — content-addressed CDN bundle dir for
+            # POST /api/v1/render. Default "wwwroot/cdn" resolves under the
+            # read-only app filesystem, so pin to the writable PVC mount
+            # alongside other TtsReader runtime data. Manifests + cue audio
+            # land at /data/cdn/sha256/<hash>/manifest.json + cues/.
+            - name: TtsReader__Render__CdnDirectory
+              value: "/data/cdn"
             - name: Auth__ApiKey
               valueFrom:
                 secretKeyRef:
@@ -190,7 +634,10 @@ spec:
                   optional: true
           resources:
             requests:
-              cpu: 100m
+              # The cluster is currently saturated on requested CPU by
+              # remotedesktop workloads even when real usage is low.
+              # Keep the web frontend schedulable under that pressure.
+              cpu: 10m
               memory: 256Mi
             limits:
               cpu: 500m

apps/fc-ttsreader/modern-tts/Dockerfile

@@ -0,0 +1,36 @@
+# FlowerCore modern-tts — wraps Microsoft Edge's Read Aloud TTS service
+# (via the edge-tts Python package) to give the cluster studio-quality
+# Modern Hebrew (he-IL-*) and Modern Greek (el-GR-*) voices alongside the
+# eSpeak biblical engine. Same shape as fc-biblical-tts so the .NET client
+# lives in the same Shared.Speech package.
+#
+# Note: edge-tts depends on Microsoft's public Edge endpoint; the cluster
+# pod needs egress to *.tts.speech.microsoft.com. dnsPolicy: None on the
+# Deployment makes sure the iamworkin.lan template hijack doesn't rewrite
+# the lookup back to Traefik VIP.
+FROM python:3.12-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PIP_DISABLE_PIP_VERSION_CHECK=1 \
+    PIP_NO_CACHE_DIR=1
+
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+        ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+COPY requirements.txt /app/
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY app.py /app/
+
+RUN useradd --create-home --shell /usr/sbin/nologin --uid 1654 tts
+USER 1654
+
+EXPOSE 10403
+HEALTHCHECK --interval=30s --timeout=5s --start-period=20s --retries=3 \
+    CMD python -c "import urllib.request,sys; urllib.request.urlopen('http://127.0.0.1:10403/health',timeout=3); sys.exit(0)" || exit 1
+
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "10403", "--workers", "1"]

apps/fc-ttsreader/modern-tts/app.py

@@ -0,0 +1,238 @@
+"""FlowerCore modern-tts — Microsoft Edge Read Aloud bridge for Modern
+Hebrew and Modern Greek (and other Edge-supported languages).
+
+Endpoints:
+
+* POST /tts          — body: {"text", "voice", "rate"?, "volume"?, "pitch"?}
+                       returns audio/mpeg (Edge returns MP3) which the
+                       upstream FasterWhisperAlignmentClient + the WPF
+                       MediaPlayer both handle natively.
+* POST /timings      — same body shape but returns
+                       {"text", "voice", "words": [{"text","startMs","endMs"}],
+                        "durationMs": ...} sourced from Edge's WordBoundary
+                       events — much more accurate than eSpeak's
+                       proportional-distribution approach because Edge
+                       emits real per-word offsets during synthesis.
+* GET  /voices       — voice catalog Edge knows about. Filtered to
+                       Hebrew + Greek by default; ?language=all returns
+                       everything Edge supports.
+* GET  /health       — fast readiness check.
+
+Pairs with fc-biblical-tts (eSpeak Ancient Greek + Hebrew). The biblical
+engine handles unpointed Hebrew + Erasmian Greek; this engine handles
+narrative Modern Hebrew + Modern Greek for translations the operator
+might be reading alongside the original.
+"""
+from __future__ import annotations
+
+import io
+import logging
+from typing import Optional
+
+import edge_tts
+from fastapi import FastAPI, HTTPException
+from fastapi.responses import JSONResponse, Response
+from pydantic import BaseModel
+
+LOG = logging.getLogger("modern_tts")
+logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
+
+app = FastAPI(title="FlowerCore modern-tts", version="1.0.0")
+
+# Default voices by short code so AiStation can pick a sensible default
+# when the operator hasn't explicitly asked for one. Edge has multiple
+# voices per locale — these are the calmest male+female narrators.
+DEFAULT_VOICES = {
+    "he":    "he-IL-AvriNeural",
+    "he-IL": "he-IL-AvriNeural",
+    "el":    "el-GR-NestorasNeural",
+    "el-GR": "el-GR-NestorasNeural",
+    "en":    "en-US-AriaNeural",
+}
+
+
+class TtsRequest(BaseModel):
+    text: str
+    voice: Optional[str] = None
+    language: Optional[str] = None
+    rate: str = "+0%"     # Edge accepts +20%, -10%, etc.
+    volume: str = "+0%"
+    pitch: str = "+0Hz"
+
+
+def _resolve_voice(req: TtsRequest) -> str:
+    if req.voice:
+        return req.voice.strip()
+    if req.language and req.language in DEFAULT_VOICES:
+        return DEFAULT_VOICES[req.language]
+    return DEFAULT_VOICES["he"]
+
+
+@app.get("/health")
+def health():
+    return {"status": "ok"}
+
+
+@app.get("/voices")
+async def voices(language: str = "default"):
+    catalog = await edge_tts.list_voices()
+    if language == "all":
+        return {"voices": catalog}
+
+    # Default response: filter to languages relevant to the FlowerCore
+    # biblical workflow (Hebrew + Greek) so the AiStation voice picker
+    # isn't overwhelmed by 400+ Edge voices.
+    keep = ("he-", "el-")
+    filtered = [v for v in catalog if any(v.get("ShortName", "").startswith(k) for k in keep)]
+    return {"voices": filtered}
+
+
+async def _synth_with_subtitles(req: TtsRequest):
+    voice = _resolve_voice(req)
+    LOG.info("edge-tts synth voice=%s len=%d", voice, len(req.text))
+    communicate = edge_tts.Communicate(
+        req.text,
+        voice=voice,
+        rate=req.rate,
+        volume=req.volume,
+        pitch=req.pitch,
+    )
+    audio_buf = io.BytesIO()
+    word_events: list[dict] = []
+    async for chunk in communicate.stream():
+        if chunk["type"] == "audio":
+            audio_buf.write(chunk["data"])
+        elif chunk["type"] == "WordBoundary":
+            word_events.append({
+                "text": chunk.get("text") or "",
+                "offset": chunk.get("offset", 0),       # 100-ns ticks
+                "duration": chunk.get("duration", 0),   # 100-ns ticks
+            })
+    return voice, audio_buf.getvalue(), word_events
+
+
+def _to_ms(ticks_100ns: int) -> int:
+    # Edge emits offsets in 100-nanosecond ticks (.NET TimeSpan style).
+    return int(round(ticks_100ns / 10_000))
+
+
+@app.post("/tts")
+async def tts(req: TtsRequest):
+    if not req.text.strip():
+        raise HTTPException(status_code=400, detail="text is required")
+    try:
+        voice, audio_bytes, _ = await _synth_with_subtitles(req)
+    except edge_tts.exceptions.NoAudioReceived:
+        raise HTTPException(status_code=502, detail="edge-tts returned no audio for the supplied voice/text.")
+    except Exception as ex:
+        raise HTTPException(status_code=502, detail=f"edge-tts failure: {ex}")
+    if not audio_bytes:
+        raise HTTPException(status_code=502, detail="edge-tts returned an empty audio stream.")
+    return Response(content=audio_bytes, media_type="audio/mpeg",
+                    headers={"X-FlowerCore-Voice": voice})
+
+
+def _estimate_duration_ms_from_mp3(audio_bytes: bytes) -> int:
+    """Best-effort duration estimate from raw MP3 bytes by walking frame
+    headers. Edge always returns CBR ~24kbps mono so we can infer total ms
+    from frame count. If parsing fails, return 0 and let the caller fall
+    through to a per-character heuristic."""
+    if not audio_bytes:
+        return 0
+    # MP3 sample rates by version+layer (MPEG1 layer3 / MPEG2 layer3 / MPEG2.5 layer3).
+    # We just walk frame headers and count frames; each frame is 1152 samples.
+    sample_rates_v1 = [44100, 48000, 32000, 0]
+    sample_rates_v2 = [22050, 24000, 16000, 0]
+    sample_rates_v25 = [11025, 12000, 8000, 0]
+    bitrates_v1_l3 = [0,32000,40000,48000,56000,64000,80000,96000,112000,128000,160000,192000,224000,256000,320000,0]
+    bitrates_v2_l3 = [0,8000,16000,24000,32000,40000,48000,56000,64000,80000,96000,112000,128000,144000,160000,0]
+
+    pos = 0
+    total_samples = 0
+    sample_rate = 0
+    while pos + 4 <= len(audio_bytes):
+        b0, b1, b2, b3 = audio_bytes[pos], audio_bytes[pos+1], audio_bytes[pos+2], audio_bytes[pos+3]
+        if b0 != 0xFF or (b1 & 0xE0) != 0xE0:
+            pos += 1
+            continue
+        version_bits = (b1 >> 3) & 0x03
+        layer_bits = (b1 >> 1) & 0x03
+        if layer_bits != 0x01:  # layer 3 only
+            pos += 1
+            continue
+        bitrate_index = (b2 >> 4) & 0x0F
+        sample_rate_index = (b2 >> 2) & 0x03
+        padding = (b2 >> 1) & 0x01
+        if version_bits == 0x03:       # MPEG1
+            sample_rate = sample_rates_v1[sample_rate_index]
+            bitrate = bitrates_v1_l3[bitrate_index]
+            samples_per_frame = 1152
+        elif version_bits == 0x02:     # MPEG2
+            sample_rate = sample_rates_v2[sample_rate_index]
+            bitrate = bitrates_v2_l3[bitrate_index]
+            samples_per_frame = 576
+        elif version_bits == 0x00:     # MPEG2.5
+            sample_rate = sample_rates_v25[sample_rate_index]
+            bitrate = bitrates_v2_l3[bitrate_index]
+            samples_per_frame = 576
+        else:
+            pos += 1
+            continue
+        if not (sample_rate and bitrate):
+            pos += 1
+            continue
+        frame_length = int((samples_per_frame * bitrate / 8) / sample_rate) + padding
+        if frame_length <= 0:
+            pos += 1
+            continue
+        total_samples += samples_per_frame
+        pos += frame_length
+
+    if sample_rate <= 0:
+        return 0
+    return int(round(total_samples * 1000 / sample_rate))
+
+
+@app.post("/timings")
+async def timings(req: TtsRequest):
+    if not req.text.strip():
+        raise HTTPException(status_code=400, detail="text is required")
+    try:
+        voice, audio_bytes, events = await _synth_with_subtitles(req)
+    except Exception as ex:
+        raise HTTPException(status_code=502, detail=f"edge-tts failure: {ex}")
+
+    words: list[dict] = []
+    for event in events:
+        start = _to_ms(event["offset"])
+        end = start + _to_ms(event["duration"])
+        words.append({"text": event.get("text", ""), "startMs": start, "endMs": end})
+
+    # Edge sometimes omits WordBoundary events for non-English voices
+    # (notably he-IL-* and el-GR-*). Fall back to proportional distribution
+    # over the input text — same approach the eSpeak biblical-tts uses.
+    if not words and req.text.strip():
+        total_ms = _estimate_duration_ms_from_mp3(audio_bytes)
+        if total_ms <= 0:
+            # Last-resort fallback: ~600ms per word at average speaking rate.
+            total_ms = max(1, len(req.text.split())) * 600
+        tokens = req.text.split()
+        if tokens:
+            char_total = sum(max(1, len(w)) for w in tokens)
+            cursor = 0
+            for token in tokens:
+                share = int(round(total_ms * max(1, len(token)) / char_total))
+                start = cursor
+                end = start + share
+                words.append({"text": token, "startMs": start, "endMs": end})
+                cursor = end
+            words[-1]["endMs"] = total_ms
+
+    duration_ms = words[-1]["endMs"] if words else 0
+    return JSONResponse({
+        "text": req.text,
+        "voice": voice,
+        "words": words,
+        "durationMs": duration_ms,
+        "audioBytes": len(audio_bytes),
+    })

apps/fc-ttsreader/modern-tts/requirements.txt

@@ -0,0 +1,3 @@
+fastapi==0.115.6
+uvicorn==0.34.0
+edge-tts==7.2.8

apps/fc-ttsreader/speech-align/Dockerfile

@@ -0,0 +1,47 @@
+# FlowerCore speech-align — wraps SYSTRAN/faster-whisper with /align +
+# /transcribe endpoints used by FlowerCore.TtsReader. CPU-only image; the
+# default int8 compute type runs base.en at ~real-time on a single core.
+#
+# Build: podman build -t localhost/fc-speech-align:<ver> .
+# Run:   podman run --rm -p 9200:9200 -v fc-speech-align-models:/models localhost/fc-speech-align:<ver>
+
+FROM python:3.12-slim AS base
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PIP_DISABLE_PIP_VERSION_CHECK=1 \
+    PIP_NO_CACHE_DIR=1 \
+    WHISPER_MODEL=Systran/faster-whisper-base.en \
+    WHISPER_CACHE_DIR=/models \
+    WHISPER_DEVICE=cpu \
+    WHISPER_COMPUTE_TYPE=int8 \
+    DEFAULT_LANGUAGE=en \
+    MAX_AUDIO_BYTES=52428800
+
+# faster-whisper depends on libsndfile1 + libgomp1 (OpenMP runtime). ffmpeg is
+# pulled in for non-WAV inputs (transcribe accepts any container).
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+        libsndfile1 \
+        libgomp1 \
+        ffmpeg \
+        ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+COPY requirements.txt /app/
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY app.py /app/
+
+# Run as a non-root user to satisfy K8s securityContext.runAsNonRoot.
+RUN useradd --create-home --shell /usr/sbin/nologin --uid 1654 align \
+    && mkdir -p /models \
+    && chown -R 1654:1654 /models
+USER 1654
+
+EXPOSE 9200
+HEALTHCHECK --interval=30s --timeout=5s --start-period=120s --retries=3 \
+    CMD python -c "import urllib.request,sys; urllib.request.urlopen('http://127.0.0.1:9200/health',timeout=3); sys.exit(0)" || exit 1
+
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "9200", "--workers", "1"]

apps/fc-ttsreader/speech-align/app.py

@@ -0,0 +1,181 @@
+"""FlowerCore speech-align service.
+
+Wraps SYSTRAN/faster-whisper (https://github.com/SYSTRAN/faster-whisper) in a
+small FastAPI app exposing two endpoints:
+
+* POST /align       — fc-align contract used by FlowerCore.Shared.Speech's
+                       FasterWhisperAlignmentClient on master. Multipart form
+                       (`audio`, `language`) returns
+                       `{text, words: [{word, startSeconds, endSeconds, confidence}],
+                         durationMs, language}`.
+* POST /transcribe  — audio-file-in transcription used by the new TtsReader
+                       audio-import feature. Multipart form (`audio`, optional
+                       `language`) returns `{text, language, durationMs,
+                       segments: [{startSeconds, endSeconds, text}]}` so the
+                       UI can preview the transcript before piping it into
+                       Quick Read or saving as a project.
+
+Both endpoints share the same WhisperModel instance (loaded once at startup).
+Model is pinned by the WHISPER_MODEL env var (defaults to base.en) and cached
+under WHISPER_CACHE_DIR (defaults to /models, backed by a PVC in K8s).
+
+Health: GET /health → {status: ok, model, device, computeType}.
+"""
+from __future__ import annotations
+
+import io
+import logging
+import os
+import time
+from contextlib import asynccontextmanager
+from typing import Optional
+
+from fastapi import FastAPI, File, Form, HTTPException, UploadFile
+from fastapi.responses import JSONResponse
+from faster_whisper import WhisperModel
+
+LOG = logging.getLogger("speech_align")
+logging.basicConfig(
+    level=os.environ.get("LOG_LEVEL", "INFO"),
+    format="%(asctime)s %(levelname)s %(name)s %(message)s",
+)
+
+MODEL_NAME = os.environ.get("WHISPER_MODEL", "Systran/faster-whisper-base.en")
+DEVICE = os.environ.get("WHISPER_DEVICE", "cpu")
+COMPUTE_TYPE = os.environ.get("WHISPER_COMPUTE_TYPE", "int8")
+CACHE_DIR = os.environ.get("WHISPER_CACHE_DIR", "/models")
+MAX_BYTES = int(os.environ.get("MAX_AUDIO_BYTES", str(50 * 1024 * 1024)))  # 50 MB
+DEFAULT_LANGUAGE = os.environ.get("DEFAULT_LANGUAGE", "en")
+
+_state: dict[str, object] = {}
+
+
+@asynccontextmanager
+async def lifespan(_app: FastAPI):
+    LOG.info("Loading faster-whisper model %s (device=%s compute=%s cache=%s)", MODEL_NAME, DEVICE, COMPUTE_TYPE, CACHE_DIR)
+    started = time.time()
+    model = WhisperModel(MODEL_NAME, device=DEVICE, compute_type=COMPUTE_TYPE, download_root=CACHE_DIR)
+    _state["model"] = model
+    LOG.info("Model loaded in %.2fs", time.time() - started)
+    yield
+    _state.clear()
+
+
+app = FastAPI(title="FlowerCore speech-align", version="1.0.0", lifespan=lifespan)
+
+
+def _get_model() -> WhisperModel:
+    model = _state.get("model")
+    if model is None:
+        raise HTTPException(status_code=503, detail="Model not loaded yet")
+    return model  # type: ignore[return-value]
+
+
+async def _read_upload(upload: UploadFile) -> bytes:
+    payload = await upload.read()
+    if not payload:
+        raise HTTPException(status_code=400, detail="audio is empty")
+    if len(payload) > MAX_BYTES:
+        raise HTTPException(
+            status_code=413,
+            detail=f"audio exceeds {MAX_BYTES} byte limit ({len(payload)} bytes received)",
+        )
+    return payload
+
+
+def _normalize_language(value: Optional[str]) -> Optional[str]:
+    if not value or not value.strip():
+        return DEFAULT_LANGUAGE
+    return value.strip().lower()
+
+
+def _transcribe_bytes(audio_bytes: bytes, language: Optional[str], word_timestamps: bool):
+    model = _get_model()
+    started = time.time()
+    segments_iter, info = model.transcribe(
+        io.BytesIO(audio_bytes),
+        language=language,
+        word_timestamps=word_timestamps,
+        beam_size=1,
+        vad_filter=True,
+    )
+    segments = list(segments_iter)
+    elapsed_ms = int((time.time() - started) * 1000)
+    return segments, info, elapsed_ms
+
+
+@app.get("/health")
+def health():
+    return {
+        "status": "ok" if _state.get("model") is not None else "loading",
+        "model": MODEL_NAME,
+        "device": DEVICE,
+        "computeType": COMPUTE_TYPE,
+        "defaultLanguage": DEFAULT_LANGUAGE,
+        "maxBytes": MAX_BYTES,
+    }
+
+
+@app.post("/align")
+async def align(audio: UploadFile = File(...), language: str = Form(DEFAULT_LANGUAGE)):
+    """fc-align contract — used by FlowerCore.Shared.Speech.FasterWhisperAlignmentClient."""
+    payload = await _read_upload(audio)
+    lang = _normalize_language(language)
+    segments, info, elapsed_ms = _transcribe_bytes(payload, lang, word_timestamps=True)
+
+    text_parts: list[str] = []
+    words: list[dict] = []
+    for segment in segments:
+        text_parts.append(segment.text.strip())
+        for word in (segment.words or []):
+            # Field names MUST match the FlowerCore.Shared.Speech contract:
+            # `text` / `startMs` / `endMs`. The deployed FasterWhisperAlignmentClient
+            # ignores any other names — see Common's
+            # FasterWhisperAlignmentResponse / FasterWhisperWord.
+            words.append({
+                "text": word.word.strip(),
+                "startMs": int((word.start or 0.0) * 1000),
+                "endMs": int((word.end or 0.0) * 1000),
+                # Confidence is informational and ignored by the C# client today,
+                # but kept on the wire for future scoring + fc-align operators
+                # that want to surface low-confidence words.
+                "confidence": float(getattr(word, "probability", 0.0) or 0.0),
+            })
+
+    duration_ms = int((info.duration or 0.0) * 1000)
+    return JSONResponse({
+        "text": " ".join(p for p in text_parts if p).strip(),
+        "words": words,
+        "durationMs": duration_ms,
+        "language": info.language or lang,
+        "elapsedMs": elapsed_ms,
+    })
+
+
+@app.post("/transcribe")
+async def transcribe(audio: UploadFile = File(...), language: Optional[str] = Form(None)):
+    """Audio-in transcription contract — used by the new TtsReader audio-import feature.
+
+    Returns full segments (no per-word timestamps) so the UI can preview the
+    transcript before piping it into Quick Read or saving as a project.
+    """
+    payload = await _read_upload(audio)
+    lang = _normalize_language(language)
+    segments, info, elapsed_ms = _transcribe_bytes(payload, lang, word_timestamps=False)
+
+    out_segments = [
+        {
+            "startSeconds": float(segment.start or 0.0),
+            "endSeconds": float(segment.end or 0.0),
+            "text": segment.text.strip(),
+        }
+        for segment in segments
+    ]
+
+    return JSONResponse({
+        "text": " ".join(s["text"] for s in out_segments if s["text"]).strip(),
+        "segments": out_segments,
+        "language": info.language or lang,
+        "durationMs": int((info.duration or 0.0) * 1000),
+        "elapsedMs": elapsed_ms,
+    })

apps/fc-ttsreader/speech-align/requirements.txt

@@ -0,0 +1,8 @@
+faster-whisper==1.0.3
+fastapi==0.115.0
+uvicorn[standard]==0.30.6
+python-multipart==0.0.10
+# faster-whisper 1.0.3's utils module imports requests but doesn't pin it as a
+# transitive dep — pin explicitly so the image isn't relying on whatever
+# happens to be in the base image.
+requests==2.32.3

apps/fc-updater/README.md

@@ -0,0 +1,47 @@
+# fc-updater — Update Center GitOps adoption
+
+**Status:** adopted into `bluejay-infra` on 2026-05-06. The live ArgoCD
+Application is `infra-fc-updater`, generated by the `bluejay-infra`
+ApplicationSet with automated sync, `prune: true`, and `selfHeal: true`.
+
+## Managed manifest set
+
+`apps/fc-updater/fc-updater.yaml` manages:
+
+- `Namespace/fc-updater`
+- `PersistentVolumeClaim/updatecenter-data`
+- `Deployment/updatecenter-web`
+- `Service/updatecenter-web`
+- `Certificate/updatecenter-web-tls`
+- `Certificate/updatecenter-web-internal-tls`
+- `IngressRoute/updatecenter-web`
+- `IngressRoute/updatecenter-web-internal`
+- `IngressRoute/updatecenter-web-public`
+
+The Deployment intentionally sets `revisionHistoryLimit: 3` and
+`strategy.type: Recreate`. The service is singleton + SQLite/local bundle
+storage on `PersistentVolumeClaim/updatecenter-data`, pinned to
+`rke2-server`.
+
+## Runtime dependencies intentionally not stored here
+
+These live Secrets are pre-existing runtime material and are not committed to
+Git:
+
+- `updater-bootstrap-auth`
+- `updater-signing`
+- `updater-webhooks`
+- `cf-origin-flowercore-io`
+
+Rotate the Cloudflare Origin Certificate through
+`FlowerCore.Notes/docs/standards/code-signing-rotation-runbook.md`; the
+shared origin cert must exist in every namespace that serves a
+`*.flowercore.io` public IngressRoute.
+
+## Verification
+
+```powershell
+kubectl.exe --kubeconfig C:\Users\AndrewStoltz\.kube\rke2.yaml -n argocd get application infra-fc-updater
+kubectl.exe --kubeconfig C:\Users\AndrewStoltz\.kube\rke2.yaml -n fc-updater get deploy,svc,ingressroute,certificate,pvc
+curl.exe -sk https://update.flowercore.io/api/v1/manifests/_schema
+```

apps/fc-updater/fc-updater.yaml

@@ -0,0 +1,269 @@
+# FlowerCore Update Center
+# GitOps adoption of the live fc-updater namespace after PUB-1/PUB-3.
+# Runtime credentials remain in existing K8s Secrets; do not store them here.
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: fc-updater
+  labels:
+    app.kubernetes.io/part-of: flowercore
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: updatecenter-data
+  namespace: fc-updater
+  labels:
+    app.kubernetes.io/name: updatecenter-web
+    app.kubernetes.io/part-of: flowercore
+spec:
+  accessModes:
+    - ReadWriteOnce
+  storageClassName: longhorn
+  volumeMode: Filesystem
+  resources:
+    requests:
+      # Sized for fleet bundle storage (LocalFsBundleStore.MaxTotalBytes
+      # soft cap at 25 GiB per project_uc_remaining_4_apps_signed_2026_05_06).
+      # Mike Bundle alone is ~5.1 GiB; cluster live capacity is already
+      # 20 GiB after a manual expand. PVCs cannot shrink, so git must track
+      # at least the live size to avoid the OutOfSync loop.
+      storage: 25Gi
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: updatecenter-web
+  namespace: fc-updater
+  labels:
+    app: updatecenter-web
+    app.kubernetes.io/name: updatecenter-web
+    app.kubernetes.io/part-of: flowercore
+spec:
+  replicas: 1
+  revisionHistoryLimit: 3
+  strategy:
+    # SQLite + local bundle storage live on a single RWO PVC. Recreate avoids
+    # two pods overlapping the same write path during future image bumps.
+    type: Recreate
+  selector:
+    matchLabels:
+      app: updatecenter-web
+  template:
+    metadata:
+      labels:
+        app: updatecenter-web
+    spec:
+      nodeName: rke2-server
+      containers:
+        - name: web
+          image: localhost/fc-updater-web:v20260508-pub3-deepening-2bdf108
+          imagePullPolicy: Never
+          ports:
+            - containerPort: 8080
+              name: http
+          env:
+            - name: ASPNETCORE_URLS
+              value: http://+:8080
+            - name: FlowerCore__Updater__Database__Provider
+              value: sqlite
+            - name: FlowerCore__Updater__Database__ConnectionString
+              value: Data Source=/data/updatecenter.db
+            - name: FlowerCore__Updater__BundleStorage__LocalFs__RootDirectory
+              value: /data/bundles
+            - name: FlowerCore__Updater__PublicShares__RequirePublicVisibilityOnPublicHosts
+              value: "true"
+            - name: FlowerCore__Updater__PublicShares__Links__0__Code
+              value: 8f3c2a9e7d41
+            - name: FlowerCore__Updater__PublicShares__Links__0__AppId
+              value: flowercore.faith-ai-mike
+            - name: FlowerCore__Updater__PublicShares__Links__0__Channel
+              value: stable
+            - name: FlowerCore__Updater__PublicShares__Links__0__RuntimeId
+              value: win-x64
+            - name: FlowerCore__Updater__PublicShares__Links__0__DisplayName
+              value: Faith AI Mike Edition
+            - name: FlowerCore__Updater__PublicShares__Links__0__Headline
+              value: Faith AI Mike Edition
+            - name: FlowerCore__Updater__PublicShares__Links__0__Description
+              value: Private release link for Mike's Faith AI bundle.
+            - name: FlowerCore__Updater__Auth__Bootstrap__Enabled
+              value: "true"
+            - name: FlowerCore__Updater__Auth__Bootstrap__Username
+              valueFrom:
+                secretKeyRef:
+                  name: updater-bootstrap-auth
+                  key: username
+            - name: FlowerCore__Updater__Auth__Bootstrap__Password
+              valueFrom:
+                secretKeyRef:
+                  name: updater-bootstrap-auth
+                  key: password
+            - name: FlowerCore__Updater__Auth__Bootstrap__SigningKey
+              valueFrom:
+                secretKeyRef:
+                  name: updater-bootstrap-auth
+                  key: signing-key
+            - name: FlowerCore__Updater__Signing__AutoSignOnPublish
+              value: "true"
+            - name: FlowerCore__Updater__Signing__RequireSignatureOnPublish
+              value: "true"
+            - name: FlowerCore__Updater__Signing__PfxBase64
+              valueFrom:
+                secretKeyRef:
+                  name: updater-signing
+                  key: pfx-base64
+            - name: FlowerCore__Updater__Signing__PfxPassword
+              valueFrom:
+                secretKeyRef:
+                  name: updater-signing
+                  key: pfx-password
+            - name: FlowerCore__Updater__Signing__OpItemReference
+              value: op://FlowerCore/step-ca-codesign
+            - name: FlowerCore__Updater__Signing__TrustAnchorPath
+              value: /etc/flowercore-updater/signing/root-ca.pem
+            - name: FlowerCore__Updater__GitHub__Token
+              valueFrom:
+                secretKeyRef:
+                  name: updater-webhooks
+                  key: github-token
+            - name: FlowerCore__Updater__GitHub__WebhookSecret
+              valueFrom:
+                secretKeyRef:
+                  name: updater-webhooks
+                  key: github-webhook-secret
+            - name: FlowerCore__Updater__Gitea__Token
+              valueFrom:
+                secretKeyRef:
+                  name: updater-webhooks
+                  key: gitea-token
+            - name: FlowerCore__Updater__Gitea__WebhookSecret
+              valueFrom:
+                secretKeyRef:
+                  name: updater-webhooks
+                  key: gitea-webhook-secret
+          readinessProbe:
+            tcpSocket:
+              port: http
+            initialDelaySeconds: 10
+            periodSeconds: 15
+          livenessProbe:
+            tcpSocket:
+              port: http
+            initialDelaySeconds: 30
+            periodSeconds: 30
+          volumeMounts:
+            - name: data
+              mountPath: /data
+            - name: signing
+              mountPath: /etc/flowercore-updater/signing
+              readOnly: true
+      volumes:
+        - name: data
+          persistentVolumeClaim:
+            claimName: updatecenter-data
+        - name: signing
+          secret:
+            secretName: updater-signing
+            items:
+              - key: root-ca.pem
+                path: root-ca.pem
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: updatecenter-web
+  namespace: fc-updater
+  labels:
+    app: updatecenter-web
+    app.kubernetes.io/name: updatecenter-web
+    app.kubernetes.io/part-of: flowercore
+spec:
+  type: ClusterIP
+  selector:
+    app: updatecenter-web
+  ports:
+    - name: http
+      port: 8080
+      targetPort: http
+---
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: updatecenter-web-tls
+  namespace: fc-updater
+spec:
+  secretName: updatecenter-web-tls
+  issuerRef:
+    name: step-ca-acme
+    kind: ClusterIssuer
+  dnsNames:
+    - updatecenter.iamworkin.lan
+    - updates.iamworkin.lan
+---
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: updatecenter-web-internal-tls
+  namespace: fc-updater
+spec:
+  secretName: updatecenter-web-internal-tls
+  issuerRef:
+    name: step-ca-acme
+    kind: ClusterIssuer
+  dnsNames:
+    - updatecenter-internal.iamworkin.lan
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: updatecenter-web
+  namespace: fc-updater
+spec:
+  entryPoints:
+    - web
+    - websecure
+  routes:
+    - match: (Host(`updatecenter.iamworkin.lan`) || Host(`updates.iamworkin.lan`)) && (Method(`GET`) || Method(`HEAD`) || Method(`POST`) || Method(`OPTIONS`))
+      kind: Rule
+      services:
+        - name: updatecenter-web
+          port: 8080
+  tls:
+    secretName: updatecenter-web-tls
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: updatecenter-web-internal
+  namespace: fc-updater
+spec:
+  entryPoints:
+    - web
+    - websecure
+  routes:
+    - match: Host(`updatecenter-internal.iamworkin.lan`)
+      kind: Rule
+      services:
+        - name: updatecenter-web
+          port: 8080
+  tls:
+    secretName: updatecenter-web-internal-tls
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: updatecenter-web-public
+  namespace: fc-updater
+spec:
+  entryPoints:
+    - websecure
+  routes:
+    - match: (Host(`update.flowercore.io`) || Host(`updates.flowercore.io`)) && (Method(`GET`) || Method(`HEAD`) || Method(`POST`) || Method(`OPTIONS`))
+      kind: Rule
+      services:
+        - name: updatecenter-web
+          port: 8080
+  tls:
+    secretName: cf-origin-flowercore-io

apps/fc-updater/kustomization.yaml

@@ -0,0 +1,7 @@
+# ArgoCD's bluejay-infra ApplicationSet uses a directory generator and does
+# not require kustomization.yaml. Keep this anyway as the manifest inventory
+# and for local `kubectl kustomize apps/fc-updater` previews.
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - fc-updater.yaml

apps/flowercore/flowercore.yaml

@@ -1,5 +1,10 @@
-# FlowerCore Tenant — flowercore.io (main brand)
+# FlowerCore Tenant — retired flowercore.io placeholder.
-# Public-facing placeholder landing page served by nginx
+#
+# Public flowercore.io/www.flowercore.io routing is now owned by
+# apps/fc-landing/fc-landing.yaml. This tenant placeholder remains available
+# only as an in-cluster service; do not create a duplicate public
+# IngressRoute here because it competes with fc-landing and requires a
+# namespace-local cf-origin-flowercore-io Secret.
 # ArgoCD managed - BlueJay Lab
 ---
 apiVersion: v1
@@ -10,12 +15,6 @@ metadata:
     app.kubernetes.io/part-of: bluejay-infra
     flowercore.io/tenant: flowercore
 ---
-# NOTE: The existing cf-origin-flowercore-io secret (covering *.flowercore.io)
-# must be copied into this namespace. It already exists in other namespaces.
-# Copy with: kubectl get secret cf-origin-flowercore-io -n fc-system -o yaml \
-#   | sed 's/namespace: .*/namespace: tenant-flowercore/' \
-#   | kubectl apply -f -
----
 # Landing page HTML
 apiVersion: v1
 kind: ConfigMap
@@ -311,22 +310,3 @@ spec:
     - port: 80
       targetPort: 80
       name: http
----
-# Traefik IngressRoute — public via Cloudflare
-# Uses existing cf-origin-flowercore-io cert (must be copied to this namespace)
-apiVersion: traefik.io/v1alpha1
-kind: IngressRoute
-metadata:
-  name: flowercore-web
-  namespace: tenant-flowercore
-spec:
-  entryPoints:
-    - websecure
-  routes:
-    - match: Host(`flowercore.io`) || Host(`www.flowercore.io`)
-      kind: Rule
-      services:
-        - name: flowercore-web
-          port: 80
-  tls:
-    secretName: cf-origin-flowercore-io

apps/intranet/intranet.yaml

@@ -3,6 +3,28 @@ kind: Namespace
 metadata:
   name: intranet
 ---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: intranet-vector-store
+  namespace: intranet
+spec:
+  accessModes:
+    - ReadWriteOnce
+  storageClassName: longhorn
+  resources:
+    requests:
+      storage: 1Gi
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: intranet-config
+  namespace: intranet
+data:
+  KnowledgeApiKey: ""
+  TrustedHeaderSharedSecret: ""
+---
 apiVersion: apps/v1
 kind: Deployment
 metadata:
@@ -12,6 +34,8 @@ metadata:
     app: intranet-web
 spec:
   replicas: 1
+  strategy:
+    type: Recreate
   selector:
     matchLabels:
       app: intranet-web
@@ -22,7 +46,7 @@ spec:
     spec:
       containers:
         - name: intranet-web
-          image: localhost/fc-intranet-web:latest
+          image: localhost/fc-intranet-web:v20260508-brochure-w1
           imagePullPolicy: Never
           ports:
             - containerPort: 5300
@@ -32,25 +56,58 @@ spec:
               value: Production
             - name: ASPNETCORE_URLS
               value: "http://+:5300"
+            # Bulk corpus indexing on edge1 Pi 5 takes ~6s/chunk × 5665 chunks
+            # ≈ 9 hours. BLUEJAY-WS GPU (R9700, 32GB VRAM) does the same work
+            # in minutes. Memory: feedback_pi5_nomic_embed_slow.
+            - name: IntranetSearch__OllamaBaseUrl
+              value: "http://10.0.56.20:11434"
+            # Sprint E Phase 2α — JSON-file-backed PageReadingOverride persistence
+            # on the writable PVC at /data. Without this env var the
+            # intranet falls back to the in-memory store (loses state on
+            # pod restart). Master's PageReadingOverrideOptions binds
+            # PageReadingOverrides:FilePath.
+            - name: PageReadingOverrides__FilePath
+              value: "/data/page-reading-overrides.json"
+            - name: KnowledgeFleetSearch__BaseUrl
+              value: "https://knowledge.iamworkin.lan"
+            - name: KnowledgeFleetSearch__ApiKey
+              valueFrom:
+                configMapKeyRef:
+                  name: intranet-config
+                  key: KnowledgeApiKey
+                  optional: true
+            - name: TrustedHeaderAuthentication__SharedSecret
+              valueFrom:
+                configMapKeyRef:
+                  name: intranet-config
+                  key: TrustedHeaderSharedSecret
+                  optional: true
           resources:
             requests:
-              memory: "128Mi"
+              memory: "256Mi"
               cpu: "100m"
             limits:
-              memory: "512Mi"
+              memory: "1Gi"
-              cpu: "500m"
+              cpu: "1000m"
           livenessProbe:
             httpGet:
               path: /health
               port: 5300
-            initialDelaySeconds: 10
+            initialDelaySeconds: 30
             periodSeconds: 30
           readinessProbe:
             httpGet:
               path: /health
               port: 5300
-            initialDelaySeconds: 5
+            initialDelaySeconds: 10
             periodSeconds: 10
+          volumeMounts:
+            - name: vector-store
+              mountPath: /data
+      volumes:
+        - name: vector-store
+          persistentVolumeClaim:
+            claimName: intranet-vector-store
 ---
 apiVersion: v1
 kind: Service

apps/knowledge/README.md

@@ -0,0 +1,165 @@
+# knowledge — FlowerCore.Knowledge.Web (Phase 2.4 K8s deploy)
+
+**Status:** **LIVE 2026-04-27** at `https://knowledge.iamworkin.lan` —
+Phase 2.4 closed. Pod running, certificate issued (step-ca-acme), PVC
+bound (Longhorn 20Gi RWO), ArgoCD `infra-knowledge` synced. `/healthz`
+returns 200, `/api/v1/editions` returns `[]` (initial-deploy state — no
+*.db files in the PVC yet; Phase 2.5+ admin UI handles bulk
+population). Phase 1 of the Agent Zero MCP rollout keeps `/healthz`
+anonymous and gates `/mcp` behind `Authorization: Bearer <token>` built
+from the 1Password item `FlowerCore Knowledge MCP Tokens`.
+
+- Plan: [`../../../FlowerCore.Notes/docs/ai-agents/flowercore-knowledge-service-plan.md`](../../../FlowerCore.Notes/docs/ai-agents/flowercore-knowledge-service-plan.md)
+- Sprint: [`../../../FlowerCore.Notes/docs/ai-station/sprint-e-xxl-plan.md`](../../../FlowerCore.Notes/docs/ai-station/sprint-e-xxl-plan.md) (Track B)
+- Repo: `D:\git\FlowerCore\FlowerCore.Knowledge\` (private GitHub repo,
+  bootstrapped Sprint D batch 35)
+
+`FlowerCore.Knowledge.Web` is the fleet-wide vector-indexing & RAG hub —
+a REST + MCP service that scans `*.db` files under
+`/data/vector-stores` and exposes per-edition reachability + corpus
+search to the rest of the FC ecosystem (Agent Zero, Chat.Web persona
+memory, AiStation embeddings explorer, TtsReader chapter context, BMO
+bot, Pi nodes via `fc-index sync`).
+
+Phase 1 MCP routing is explicit:
+
+- in-cluster Agent Zero → `http://knowledge-web.knowledge.svc/mcp`
+- workstation Agent Zero → `https://knowledge.iamworkin.lan/mcp`
+- probe URL for both lanes → `/healthz`
+
+## Deployment order (do NOT skip / reorder)
+
+### 1. FlowerCore.DNS public A record — knowledge.iamworkin.lan -> 10.0.56.200
+
+Required BEFORE the Certificate resource is created, or cert-manager
+HTTP-01 silently backs off ~2h. Memory: `feedback_pfsense_dns_required_for_acme`.
+
+The canonical path is FlowerCore.DNS:
+
+```bash
+curl -sk https://dns.iamworkin.lan/api/v1/servers
+# Find the pfSense serverId, then create the record using the host label only.
+
+curl -sk -X POST https://dns.iamworkin.lan/api/v1/servers/<serverId>/zones/iamworkin.lan/records \
+  -H "Content-Type: application/json" \
+  -d '{"name":"knowledge","type":"A","data":"10.0.56.200","ttl":300}'
+```
+
+If FlowerCore.DNS provider writes are failing 502 with "pfSense
+diag_command.php response did not contain a `<pre>` block" (status as of
+Sprint E Track B authoring 2026-04-27), add the override manually via
+the pfSense web UI:
+
+1. Log in to `https://10.0.56.1` as admin
+2. Services → DNS Resolver → General Settings → Host Overrides
+3. Add: Host=`knowledge`, Domain=`iamworkin.lan`, IP Address=`10.0.56.200`
+4. Save + Apply Changes
+
+Verify resolution from anywhere on LAN:
+
+```bash
+nslookup knowledge.iamworkin.lan 10.0.56.1
+# Expect: 10.0.56.200
+```
+
+Or against FlowerCore.DNS once the provider is fixed:
+
+```bash
+curl -sk "https://dns.iamworkin.lan/api/v1/zones/iamworkin.lan/resolve-preflight?hostname=knowledge.iamworkin.lan"
+# Expect: "resolvable": true
+```
+
+### 2. Build + import the image to ALL RKE2 nodes
+
+Pods may schedule on any RKE2 worker (server, agent1, agent2). The
+Longhorn PVC accepts mounts from any node, so the image must be
+imported to all three. Memory:
+`feedback_rke2_image_import_targets_all_nodes` +
+`feedback_rke2_localhost_imagepullpolicy`.
+
+```bash
+# From BLUEJAY-WS, in D:\git\FlowerCore\FlowerCore.Knowledge
+TAG="v$(date +%Y%m%d%H%M)"
+dotnet.exe publish -c Release -o deploy/app \
+  src/FlowerCore.Knowledge.Web/FlowerCore.Knowledge.Web.csproj
+podman build -t localhost/fc-knowledge-web:$TAG -f deploy/Dockerfile.deploy deploy
+podman save localhost/fc-knowledge-web:$TAG -o /tmp/fc-knowledge-web.tar
+
+# Import to all three RKE2 nodes
+for node in rke2-server rke2-agent1 rke2-agent2; do
+  scp /tmp/fc-knowledge-web.tar $node:/tmp/
+  ssh $node "sudo /var/lib/rancher/rke2/bin/ctr -a /run/k3s/containerd/containerd.sock -n k8s.io images import /tmp/fc-knowledge-web.tar"
+done
+```
+
+The repo's `scripts/deploy-knowledge.sh` automates this loop.
+
+### 3. Bump the image tag + push
+
+Edit `knowledge.yaml`, replace `localhost/fc-knowledge-web:v202604272200`
+with the tag from step 2, then:
+
+```bash
+cd D:/git/FlowerCore/bluejay-infra
+python scripts/check-pfsense-dns.py     # confirms the DNS preflight
+git add apps/knowledge/
+git commit -m "feat(knowledge): deploy Phase 2.4 K8s manifest"
+git push
+```
+
+ArgoCD picks up within ~3 minutes and creates `infra-knowledge`.
+
+### 4. Verify
+
+```bash
+fcadmin_ssh noc1 '
+  kubectl -n argocd get application infra-knowledge
+  kubectl -n knowledge get certificate,pod,pvc
+  curl -sk -m 8 -o /dev/null -w "HTTP %{http_code}\n" https://knowledge.iamworkin.lan/healthz
+  curl -sk -m 8 https://knowledge.iamworkin.lan/api/v1/editions | jq
+'
+```
+
+Expect: Certificate `Ready: True` within ~60s, `/healthz` HTTP 200,
+`/api/v1/editions` returns an empty array (no DBs in the PVC yet) on
+first deploy.
+
+## Initial-deploy state and Phase 2.5 follow-up
+
+The Longhorn PVC is empty on first deploy. Knowledge.Web's filesystem
+catalog will report zero editions until vector-store `*.db` files are
+pushed into `/data/vector-stores`. Initial population is a follow-up
+step (Phase 2.5+, Blazor admin UI's "Rebuild" button); for the first
+deploy the goal is just to prove the pod boots, `/healthz` returns 200,
+and the Traefik IngressRoute serves the Scalar UI.
+
+To copy an existing local DB into the PVC (one-time, manual until
+Phase 2.5 admin UI lands):
+
+```bash
+fcadmin_ssh noc1 '
+  POD=$(kubectl -n knowledge get pod -l app=knowledge-web -o jsonpath="{.items[0].metadata.name}")
+  kubectl -n knowledge cp /var/lib/flowercore/vector-stores/bluejay-ai.db $POD:/data/vector-stores/bluejay-ai.db
+'
+```
+
+## Probes + middleware notes
+
+- `/healthz` is mapped by `Controllers/HealthController.cs` (controller-based
+  attribute route). Cheap — no DB, no dependencies.
+- Liveness uses `tcpSocket` as a defensive fallback in case future
+  middleware accidentally gates `/healthz` behind auth (memory:
+  `feedback_k8s_probes_behind_auth_middleware`).
+- `/openapi/v1.json` and `/scalar/v1` are wired by `UseFlowerCoreApi`.
+  Per memory `feedback_k8s_probes_must_not_hit_openapi`, probes must NOT
+  point at OpenAPI documents — the `MapOpenApi` call can be slow during
+  cold startup.
+
+## Resource sizing
+
+- 256Mi memory request / 1Gi limit.
+- 100m CPU request / 1000m limit.
+- 20Gi Longhorn PVC initial — sufficient for the bluejay-ai 1.94Gi DB +
+  fleet-pi-edge 352Mi + fleet-bmo-bot 141Mi + headroom. Resize via
+  `kubectl -n knowledge edit pvc knowledge-vector-store` if growing
+  past 15Gi.

apps/knowledge/knowledge.yaml

@@ -0,0 +1,266 @@
+# FlowerCore.Knowledge.Web — fleet vector indexing & RAG hub.
+#
+# Phase 2.4 of the Knowledge service plan. REST + MCP service that scans
+# *.db files under /data/vector-stores and exposes:
+#   - REST: /api/v1/editions, /api/v1/corpus/search, /healthz
+#   - MCP:  list_editions, describe_edition, corpus_search
+#   - Static OpenAPI/Scalar via UseFlowerCoreApi
+#
+# Architecture:
+#   Plan:    FlowerCore.Notes/docs/ai-agents/flowercore-knowledge-service-plan.md
+#   Sprint:  FlowerCore.Notes/docs/ai-station/sprint-e-xxl-plan.md (Track B)
+#   Repo:    D:\git\FlowerCore\FlowerCore.Knowledge\
+#   Shared:  FlowerCore.Common -> FlowerCore.Shared.Indexing (chunkers, vector
+#            stores, edition profiles, ICorpusSearchService facade)
+#
+# Deployment order (see apps/knowledge/README.md and the bluejay-infra/README.md
+# top-level checklist):
+#   1. FlowerCore.DNS public A record knowledge.iamworkin.lan -> 10.0.56.200
+#      MUST exist BEFORE the Certificate is created, or cert-manager HTTP-01
+#      backs off ~2h. Memory: feedback_pfsense_dns_required_for_acme.
+#   2. Build + import the image to ALL RKE2 nodes (server + both agents) since
+#      the Pod uses a Longhorn PVC and may schedule anywhere.
+#      Memory: feedback_rke2_localhost_imagepullpolicy.
+#   3. Bump the image tag in this file, git push.
+#   4. ArgoCD ApplicationSet picks up within ~3 minutes and creates
+#      infra-knowledge.
+#
+# Initial-deploy state:
+#   The Longhorn PVC is empty on first deploy. Knowledge.Web's filesystem
+#   catalog will report zero editions until vector-store *.db files are
+#   pushed into /data/vector-stores. Initial population is a follow-up step
+#   (Phase 2.5+, Blazor admin UI's "Rebuild" button); for the first deploy
+#   the goal is just to prove the pod boots, /healthz returns 200, and the
+#   Traefik IngressRoute serves the Scalar UI.
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: knowledge
+  labels:
+    app.kubernetes.io/part-of: bluejay-infra
+---
+# MCP bearer token for the read-only Agent Zero Phase 1 lane. The 1Password
+# item currently stores the raw token in its concealed PASSWORD field, which
+# the operator syncs into the namespaced Secret key `password`.
+apiVersion: onepassword.com/v1
+kind: OnePasswordItem
+metadata:
+  name: knowledge-mcp-tokens
+  namespace: knowledge
+spec:
+  itemPath: "vaults/IAmWorkin/items/FlowerCore Knowledge MCP Tokens"
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: knowledge-vector-store
+  namespace: knowledge
+spec:
+  accessModes:
+    - ReadWriteOnce
+  storageClassName: longhorn
+  resources:
+    requests:
+      storage: 20Gi
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: knowledge-web
+  namespace: knowledge
+  labels:
+    app: knowledge-web
+    app.kubernetes.io/name: knowledge-web
+    app.kubernetes.io/part-of: bluejay-infra
+spec:
+  replicas: 1
+  revisionHistoryLimit: 3
+  # RWO Longhorn PVC blocks rolling updates (multi-attach error). Recreate
+  # is the canonical pattern (memory: feedback_rwo_pvc_blocks_rolling).
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app: knowledge-web
+  template:
+    metadata:
+      labels:
+        app: knowledge-web
+        app.kubernetes.io/name: knowledge-web
+        app.kubernetes.io/part-of: bluejay-infra
+      annotations:
+        prometheus.io/scrape: "true"
+        prometheus.io/port: "8080"
+        prometheus.io/path: "/metrics"
+    spec:
+      securityContext:
+        runAsNonRoot: true
+        fsGroup: 1654
+        fsGroupChangePolicy: OnRootMismatch
+      containers:
+        - name: web
+          # Placeholder tag — bump to the image you built + imported to ALL
+          # RKE2 nodes via scripts/deploy-knowledge.sh before applying.
+          image: localhost/fc-knowledge-web:v20260429232635
+          imagePullPolicy: Never
+          command:
+            - /bin/sh
+            - -c
+          args:
+            - |
+              if [ -n "${KNOWLEDGE_MCP_BEARER_TOKEN:-}" ]; then
+                export FlowerCore__Mcp__ApiKey__Key="Bearer ${KNOWLEDGE_MCP_BEARER_TOKEN}"
+              fi
+              exec dotnet FlowerCore.Knowledge.Web.dll
+          ports:
+            - containerPort: 8080
+              name: http
+          env:
+            - name: ASPNETCORE_URLS
+              value: "http://+:8080"
+            - name: ASPNETCORE_ENVIRONMENT
+              value: "Production"
+            - name: DOTNET_SYSTEM_GLOBALIZATION_INVARIANT
+              value: "false"
+            # Vector-store directory + embedding model + edition profile dir.
+            # Profile JSON is baked into the image at /home/app/editions via the
+            # csproj Content-link from FlowerCore.Common/editions/.
+            - name: Knowledge__VectorStoresDirectory
+              value: "/data/vector-stores"
+            - name: Knowledge__EmbeddingModel
+              value: "nomic-embed-text"
+            - name: Knowledge__DefaultLimit
+              value: "5"
+            - name: Knowledge__MaxLimit
+              value: "50"
+            - name: FlowerCore__Editions__ProfileDirectory
+              value: "/home/app/editions"
+            # Embed via edge1 Pi 5 + AI HAT+ (10.0.57.17:11434). Cluster
+            # services do not depend on BLUEJAY-WS (private dev hardware) per
+            # bluejay-infra@0f9d56e. Query-time embedding is fast enough on
+            # edge1 (~ms per query); bulk index rebuilds (Phase 2.5+) will
+            # need a separate ingestion lane that can opt into the
+            # workstation GPU when present.
+            - name: FlowerCore__Ollama__BaseUrl
+              value: "http://10.0.57.17:11434"
+            - name: FlowerCore__Mcp__ApiKey__Key
+              valueFrom:
+                secretKeyRef:
+                  name: knowledge-mcp-tokens
+                  key: password
+            - name: FlowerCore__Mcp__ApiKey__HeaderName
+              value: "Authorization"
+            - name: KNOWLEDGE_MCP_BEARER_TOKEN
+              valueFrom:
+                secretKeyRef:
+                  name: knowledge-mcp-tokens
+                  key: password
+          resources:
+            requests:
+              cpu: 100m
+              memory: 256Mi
+            limits:
+              cpu: 1000m
+              memory: 1Gi
+          # /healthz is mapped by HealthController (controller-based route).
+          # tcpSocket liveness is the defensive fallback in case middleware
+          # later gates /healthz behind auth (memory:
+          # feedback_k8s_probes_behind_auth_middleware).
+          startupProbe:
+            httpGet:
+              path: /healthz
+              port: 8080
+            initialDelaySeconds: 5
+            periodSeconds: 5
+            failureThreshold: 30
+          readinessProbe:
+            httpGet:
+              path: /healthz
+              port: 8080
+            periodSeconds: 10
+            failureThreshold: 3
+          livenessProbe:
+            tcpSocket:
+              port: 8080
+            initialDelaySeconds: 30
+            periodSeconds: 30
+            failureThreshold: 3
+          securityContext:
+            runAsNonRoot: true
+            runAsUser: 1654
+            runAsGroup: 1654
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            capabilities:
+              drop:
+                - ALL
+          volumeMounts:
+            - name: vector-store
+              mountPath: /data/vector-stores
+            - name: tmp
+              mountPath: /tmp
+            - name: logs
+              mountPath: /home/app/logs
+      volumes:
+        - name: vector-store
+          persistentVolumeClaim:
+            claimName: knowledge-vector-store
+        - name: tmp
+          emptyDir: {}
+        - name: logs
+          emptyDir: {}
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: knowledge-web
+  namespace: knowledge
+  labels:
+    app: knowledge-web
+    app.kubernetes.io/name: knowledge-web
+    app.kubernetes.io/part-of: bluejay-infra
+spec:
+  type: ClusterIP
+  selector:
+    app: knowledge-web
+  ports:
+    - name: http
+      port: 80
+      targetPort: 8080
+---
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: knowledge-tls
+  namespace: knowledge
+spec:
+  secretName: knowledge-tls
+  issuerRef:
+    name: step-ca-acme
+    kind: ClusterIssuer
+  dnsNames:
+    - knowledge.iamworkin.lan
+  # step-ca ACME caps lifetime at 30d; requesting 90d silently capped
+  # made renewBefore=cert-lifetime → perpetual renewal loop (10888+ CRs
+  # in 18h on 2026-05-07). Match working 720h/240h pattern from other
+  # FC services.
+  duration: 720h     # 30d (step-ca cap)
+  renewBefore: 240h  # 10d
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: knowledge
+  namespace: knowledge
+spec:
+  entryPoints:
+    - websecure
+  routes:
+    - match: Host(`knowledge.iamworkin.lan`)
+      kind: Rule
+      services:
+        - name: knowledge-web
+          port: 80
+  tls:
+    secretName: knowledge-tls

apps/knowledge/kustomization.yaml

@@ -0,0 +1,7 @@
+# ArgoCD's bluejay-infra ApplicationSet uses a directory generator and does
+# not require kustomization.yaml. Mirrors the fc-distribution shape so
+# `kubectl kustomize` previews work from a working copy.
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - knowledge.yaml

apps/kubevirt-vms/ci1.yaml

@@ -0,0 +1,487 @@
+# =============================================================================
+# ci1 — Windows Server 2025 KubeVirt VM (GitHub Actions Self-Hosted Runner)
+# =============================================================================
+# Purpose: dedicated CI runner for FlowerCore.Updater Sandbox E2E nightly +
+# future fleet WPF AAT lanes. Replaces the never-registered
+# `bluejay-ws-sandbox-1` runner placeholder. Andrew explicitly does NOT want
+# BLUEJAY-WS registered as a runner (workstation has personal/operator state).
+#
+# Storage layout (2026-05-08):
+#   * ISO is now sourced from Synology NFS (Path B) — see
+#     win2025-iso-nfs-pv.yaml. The Longhorn Filesystem PVC
+#     `windows-server-2025-iso` below is RETAINED but UNUSED so the prior
+#     CDI upload state is preserved as a fallback (and so ArgoCD doesn't
+#     prune it on this commit). It can be deleted in a follow-up commit
+#     after the NFS path is proven on a successful Windows install.
+#
+# Status (2026-05-08): LIVE — Phase 1 prereqs satisfied:
+#   * Multus CNI v4.2.2 thick-plugin DaemonSet running on all 3 RKE2 nodes
+#     (apps/multus/multus.yaml; ApplicationSet `infra-multus` Synced/Healthy)
+#   * CDI v1.65.0 operator + CR Deployed (apps/cdi/; ApplicationSet
+#     `infra-cdi` Synced/Healthy; uploadproxy reachable via kubectl port-forward)
+#   * Windows Server 2025 ISO uploaded via CDI virtctl image-upload to
+#     PVC windows-server-2025-iso (7.7 GiB → 10Gi PVC, Bound, Upload Complete)
+#   * Local Administrator password generated, stored in 1Password vault
+#     IAmWorkin (qaphopopkryhbg353ukzhhuqoq) item id h3ix4mgfk65gmkcmvh6ly3d3hu
+#   * NetworkAttachmentDefinition prod-vlan57 registered (apps/kubevirt-vms/
+#     prod-vlan57-nad.yaml). VM still uses pod-network masquerade until Phase 1.5
+#     host bridge work lands (Puppet br-prod + enp86s0.57); switching is a
+#     one-line YAML edit + git push.
+#
+# See docs/infrastructure/windows-server-build-runner-plan.md "Phase 1 readiness gate".
+#
+# Network choice in this draft: **pod-network fallback** (Calico default).
+# Outbound-only is fine for the Updater Sandbox E2E runner workload (the runner
+# polls GitHub Actions over HTTPS; no inbound listener needed). Switch to a
+# Multus PROD VLAN NetworkAttachmentDefinition once Multus is installed and the
+# operator wants L2 access from `ci1` to other PROD VLAN services.
+#
+# Sizing: 8 vCPU / 16 GB RAM / 200 GB disk on Longhorn (default storageClass).
+# Capacity check 2026-05-08: each RKE2 node has 16 vCPU / ~64Gi allocatable;
+# 8 vCPU is ~17% of one node's allocatable, fits comfortably.
+#
+# Apply (after operator approval + ISO loaded):
+#   kubectl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml apply -f apps/kubevirt-vms/ci1.yaml
+#
+# Connect to console for Windows install:
+#   virtctl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml vnc ci1 -n kubevirt-vms
+#   (Or via Guacamole once a connection profile is added.)
+# =============================================================================
+
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: kubevirt-vms
+  labels:
+    app.kubernetes.io/part-of: kubevirt-stack
+    pod-security.kubernetes.io/enforce: privileged
+
+---
+# ISO PVC — populated via CDI virtctl image-upload (CDI is now installed).
+#
+# **Volume mode (2026-05-08 status):** Filesystem-mode PVC. A migration to
+# `volumeMode: Block` via DataVolume was attempted to address an OVMF SATA
+# CDROM read timeout, but CDI v1.65.0's upload-target pod runs as uid 107
+# with `capabilities.drop: [ALL]` and cannot open the underlying block
+# device (`blockdev: cannot open /dev/cdi-block-volume: Permission denied`).
+# Reverted to Filesystem PVC pending one of:
+#   - CDI deployment override granting CAP_SYS_RAWIO to upload pod
+#   - Pre-populated PVC via privileged init pod that dd's the ISO directly
+#   - Migration to a different storage class that exposes block devices
+#     differently (e.g. iSCSI, where Longhorn's CSI mount path may behave
+#     differently)
+#
+# Population workflow (this PVC, Filesystem mode):
+#   1. virtctl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml image-upload pvc \
+#        windows-server-2025-iso -n kubevirt-vms \
+#        --image-path "$env:USERPROFILE\Downloads\en-us_windows_server_2025_updated_march_2026_x64_dvd_8e06425a.iso" \
+#        --size 10Gi --storage-class longhorn --access-mode ReadWriteOnce \
+#        --uploadproxy-url https://localhost:8443 --insecure
+#   (--uploadproxy-url uses port-forward in practice: `kubectl port-forward
+#   -n cdi service/cdi-uploadproxy 8443:443 &` first.)
+#
+# **Open boot issue:** even with the ISO at bootOrder:1, OVMF console showed:
+#   BdsDxe: starting Boot0001 "UEFI QEMU DVD-ROM QM00001 " from ... Sata(...)
+#   BdsDxe: failed to start Boot0001 ... Time out
+# Diagnosis confirmed PVC content IS a valid bootable ISO9660 image — the
+# timeout is in OVMF reading from the SATA-CDROM-backed-by-filesystem-PVC.
+# Block mode would likely fix it; see CDI permission issue above.
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: windows-server-2025-iso
+  namespace: kubevirt-vms
+  labels:
+    app: ci-runner
+    flowercore.io/managed-by: bluejay-infra
+spec:
+  accessModes:
+    - ReadWriteOnce          # Bump to ReadOnlyMany after population for multi-VM use
+  resources:
+    requests:
+      storage: 10Gi          # Server 2025 ISO is 7.7GB; 10Gi for headroom
+  storageClassName: longhorn
+
+---
+# Root disk PVC — empty 200Gi volume that Windows installs into.
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: ci1-rootdisk
+  namespace: kubevirt-vms
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 200Gi
+  storageClassName: longhorn
+
+---
+# Sysprep ConfigMap — autounattend.xml for hands-off Windows install.
+# Sets local Administrator password (REPLACE the placeholder), enables RDP,
+# enables WinRM, sets hostname, and configures static-ish networking via DHCP.
+# The ISO + VirtIO drivers handle the rest.
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: ci1-autounattend
+  namespace: kubevirt-vms
+data:
+  autounattend.xml: |
+    <?xml version="1.0" encoding="utf-8"?>
+    <unattend xmlns="urn:schemas-microsoft-com:unattend">
+
+      <!-- Pass 1: WindowsPE — Disk setup and VirtIO driver injection -->
+      <settings pass="windowsPE">
+        <component name="Microsoft-Windows-International-Core-WinPE"
+                   processorArchitecture="amd64"
+                   publicKeyToken="31bf3856ad364e35"
+                   language="neutral" versionScope="nonSxS">
+          <SetupUILanguage>
+            <UILanguage>en-US</UILanguage>
+          </SetupUILanguage>
+          <InputLocale>en-US</InputLocale>
+          <SystemLocale>en-US</SystemLocale>
+          <UILanguage>en-US</UILanguage>
+          <UserLocale>en-US</UserLocale>
+        </component>
+
+        <component name="Microsoft-Windows-PnpCustomizationsWinPE"
+                   processorArchitecture="amd64"
+                   publicKeyToken="31bf3856ad364e35"
+                   language="neutral" versionScope="nonSxS">
+          <DriverPaths>
+            <PathAndCredentials wcm:action="add" wcm:keyValue="1">
+              <Path>E:\amd64\2k25</Path>
+            </PathAndCredentials>
+          </DriverPaths>
+        </component>
+
+        <component name="Microsoft-Windows-Setup"
+                   processorArchitecture="amd64"
+                   publicKeyToken="31bf3856ad364e35"
+                   language="neutral" versionScope="nonSxS">
+          <DiskConfiguration>
+            <Disk wcm:action="add">
+              <DiskID>0</DiskID>
+              <WillWipeDisk>true</WillWipeDisk>
+              <CreatePartitions>
+                <CreatePartition wcm:action="add">
+                  <Order>1</Order>
+                  <Size>260</Size>
+                  <Type>EFI</Type>
+                </CreatePartition>
+                <CreatePartition wcm:action="add">
+                  <Order>2</Order>
+                  <Size>128</Size>
+                  <Type>MSR</Type>
+                </CreatePartition>
+                <CreatePartition wcm:action="add">
+                  <Order>3</Order>
+                  <Extend>true</Extend>
+                  <Type>Primary</Type>
+                </CreatePartition>
+              </CreatePartitions>
+              <ModifyPartitions>
+                <ModifyPartition wcm:action="add">
+                  <Order>1</Order>
+                  <PartitionID>1</PartitionID>
+                  <Format>FAT32</Format>
+                  <Label>EFI</Label>
+                </ModifyPartition>
+                <ModifyPartition wcm:action="add">
+                  <Order>2</Order>
+                  <PartitionID>2</PartitionID>
+                </ModifyPartition>
+                <ModifyPartition wcm:action="add">
+                  <Order>3</Order>
+                  <PartitionID>3</PartitionID>
+                  <Format>NTFS</Format>
+                  <Label>Windows</Label>
+                </ModifyPartition>
+              </ModifyPartitions>
+            </Disk>
+          </DiskConfiguration>
+
+          <ImageInstall>
+            <OSImage>
+              <InstallTo>
+                <DiskID>0</DiskID>
+                <PartitionID>3</PartitionID>
+              </InstallTo>
+              <!-- Index 2 = Standard Desktop Experience. Use 4 for Datacenter Desktop. -->
+              <InstallFrom>
+                <MetaData wcm:action="add">
+                  <Key>/IMAGE/INDEX</Key>
+                  <Value>2</Value>
+                </MetaData>
+              </InstallFrom>
+            </OSImage>
+          </ImageInstall>
+
+          <UserData>
+            <AcceptEula>true</AcceptEula>
+            <FullName>FlowerCore CI Runner</FullName>
+            <Organization>FlowerCore</Organization>
+            <!-- Eval install — no product key needed for 180-day evaluation -->
+          </UserData>
+        </component>
+      </settings>
+
+      <!-- Pass 4: Specialize — Hostname, RDP, WinRM -->
+      <settings pass="specialize">
+        <component name="Microsoft-Windows-Shell-Setup"
+                   processorArchitecture="amd64"
+                   publicKeyToken="31bf3856ad364e35"
+                   language="neutral" versionScope="nonSxS">
+          <ComputerName>CI1</ComputerName>
+          <TimeZone>Central Standard Time</TimeZone>
+        </component>
+
+        <component name="Microsoft-Windows-TerminalServices-LocalSessionManager"
+                   processorArchitecture="amd64"
+                   publicKeyToken="31bf3856ad364e35"
+                   language="neutral" versionScope="nonSxS">
+          <fDenyTSConnections>false</fDenyTSConnections>
+        </component>
+      </settings>
+
+      <!-- Pass 7: OOBE — Admin account, RDP firewall, WinRM -->
+      <settings pass="oobeSystem">
+        <component name="Microsoft-Windows-Shell-Setup"
+                   processorArchitecture="amd64"
+                   publicKeyToken="31bf3856ad364e35"
+                   language="neutral" versionScope="nonSxS">
+          <OOBE>
+            <HideEULAPage>true</HideEULAPage>
+            <HideLocalAccountScreen>true</HideLocalAccountScreen>
+            <HideOEMRegistrationScreen>true</HideOEMRegistrationScreen>
+            <HideOnlineAccountScreens>true</HideOnlineAccountScreens>
+            <HideWirelessSetupInOOBE>true</HideWirelessSetupInOOBE>
+            <ProtectYourPC>3</ProtectYourPC>
+          </OOBE>
+          <UserAccounts>
+            <AdministratorPassword>
+              <!-- Real password is in 1Password — vault qaphopopkryhbg353ukzhhuqoq,
+                   item id h3ix4mgfk65gmkcmvh6ly3d3hu, title:
+                   "ci1 Administrator (Windows Server 2025 KubeVirt VM)".
+                   Field "autounattend AdministratorPassword Value (UTF-16-LE base64)"
+                   matches the Value below.
+                   To rotate: regenerate, recompute base64
+                     $combined = $pw + "AdministratorPassword"
+                     [Convert]::ToBase64String([Text.Encoding]::Unicode.GetBytes($combined))
+                   then update both 1P item AND this Value field, recreate VM. -->
+              <Value>bAA3AGsANABOAHcAcgBMAG4AeQBTAHUAYgBBAHQAaQBzAFUAcAB6AEMAWQAhADkAYQBCAEEAZABtAGkAbgBpAHMAdAByAGEAdABvAHIAUABhAHMAcwB3AG8AcgBkAA==</Value>
+              <PlainText>false</PlainText>
+            </AdministratorPassword>
+          </UserAccounts>
+          <FirstLogonCommands>
+            <SynchronousCommand wcm:action="add">
+              <Order>1</Order>
+              <CommandLine>powershell.exe -ExecutionPolicy Bypass -Command "Set-NetFirewallRule -DisplayGroup 'Remote Desktop' -Enabled True"</CommandLine>
+              <Description>Enable RDP firewall rule</Description>
+            </SynchronousCommand>
+            <SynchronousCommand wcm:action="add">
+              <Order>2</Order>
+              <CommandLine>powershell.exe -ExecutionPolicy Bypass -Command "Enable-PSRemoting -Force; Set-Item WSMan:\localhost\Service\Auth\Basic $true; Set-Item WSMan:\localhost\Service\AllowUnencrypted $true"</CommandLine>
+              <Description>Enable WinRM (Phase 2 will pivot to HTTPS via step-ca cert)</Description>
+            </SynchronousCommand>
+            <SynchronousCommand wcm:action="add">
+              <Order>3</Order>
+              <CommandLine>cmd.exe /c reg add "HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\System" /v EnableLUA /t REG_DWORD /d 0 /f</CommandLine>
+              <Description>Disable UAC (Phase 2 Puppet will re-evaluate)</Description>
+            </SynchronousCommand>
+          </FirstLogonCommands>
+        </component>
+      </settings>
+    </unattend>
+
+---
+# VirtualMachine — Windows Server 2025 CI runner.
+apiVersion: kubevirt.io/v1
+kind: VirtualMachine
+metadata:
+  name: ci1
+  namespace: kubevirt-vms
+  labels:
+    app: ci-runner
+    role: github-actions-runner
+    flowercore.io/managed-by: bluejay-infra
+spec:
+  # `running: true` is deprecated in favor of `runStrategy`. They are mutually
+  # exclusive — KubeVirt's validating webhook rejects any VM that sets both:
+  #   admission webhook "virtualmachine-validator.kubevirt.io" denied the request:
+  #   Running and RunStrategy are mutually exclusive.
+  # `Always` keeps a VMI running and restarts it if it crashes/exits — same
+  # semantics as the old `running: true`.
+  #
+  # **2026-05-08 status: VM cannot start due to a stale QEMU flock on the
+  # rootdisk PVC** (qemu reports `Failed to get "write" lock` on
+  # `/var/run/kubevirt-private/vmi-disks/rootdisk/disk.img`). The flock was
+  # left by a previous QEMU process during a force-deleted launcher pod
+  # cycle. Recovery requires either (a) a Longhorn engine restart on
+  # rke2-agent2, (b) a Longhorn volume detach via the longhorn-manager API
+  # (kubectl patch on `volume.longhorn.io/<pvc-name>` does not work — the
+  # spec.nodeID is reconciled back), or (c) a node reboot of rke2-agent2.
+  #
+  # **Confirmed working:** the bootOrder swap (windows-iso=1, rootdisk=2)
+  # and the runStrategy migration (above). The ISO PVC was successfully
+  # repopulated via virtctl image-upload pvc on the Filesystem-mode PVC.
+  #
+  # **Open: SATA CDROM read timeout** — even with bootOrder=1, OVMF reported
+  # `BdsDxe: failed to start Boot0001 ... Time out` reading the SATA CDROM
+  # backed by the Filesystem-mode PVC. A switch to Block-mode DataVolume
+  # was attempted but blocked by a CDI v1.65.0 upload-pod permission issue
+  # (capability drop prevents writing to the underlying block device).
+  # See header docstring on the ISO PVC.
+  runStrategy: Always   # LIVE — ISO uploaded 2026-05-08, password in 1P
+  template:
+    metadata:
+      labels:
+        app: ci-runner
+        role: github-actions-runner
+        kubevirt.io/vm: ci1
+    spec:
+      domain:
+        cpu:
+          cores: 8
+          sockets: 1
+          threads: 1
+        memory:
+          guest: 16Gi
+        resources:
+          requests:
+            memory: 16Gi
+          limits:
+            memory: 16Gi
+        clock:
+          utc: {}
+          timer:
+            hpet:
+              present: false
+            pit:
+              tickPolicy: delay
+            rtc:
+              tickPolicy: catchup
+            hyperv: {}
+        features:
+          acpi: {}
+          apic: {}
+          hyperv:
+            relaxed: {}
+            vapic: {}
+            spinlocks:
+              spinlocks: 8191
+          smm: {}
+        firmware:
+          bootloader:
+            efi:
+              secureBoot: true
+        devices:
+          tpm: {}             # Non-persistent vTPM — sufficient for runner; no BitLocker
+          disks:
+            # bootOrder: ISO must be 1 for first-boot install (the rootdisk has no
+            # EFI bootloader yet). After Windows installs, it writes its own UEFI
+            # Boot#### entries pointing at the rootdisk's EFI partition; UEFI then
+            # boots from rootdisk going forward and the ISO at bootOrder:2 acts as
+            # a fallback for re-install scenarios.
+            #
+            # Original (broken) order had rootdisk=1, windows-iso=2 — UEFI tried
+            # the empty virtio disk first, got nothing, fell back to the SATA
+            # CDROM at Boot0001 with a short timeout, and timed out before the
+            # CDROM enumerated. Console showed:
+            #   BdsDxe: failed to start Boot0001 ... Time out
+            #   BdsDxe: No bootable option or device was found.
+            # Confirmed via debug pod: PVC content IS a real bootable ISO9660
+            # (file: "ISO 9660 CD-ROM filesystem data ... (bootable)"), so the
+            # only bug was boot priority.
+            # 2026-05-08 PM: cdrom bus is SCSI (virtio-scsi controller). Bus
+            # choice is no longer load-bearing since the ISO is delivered via
+            # containerDisk (see volumes block below) — both SATA and SCSI
+            # work fine when the cdrom backing isn't a slow PVC. SCSI is kept
+            # because it's the modern bus and matches the standard FC
+            # KubeVirt VM template.
+            - name: windows-iso
+              bootOrder: 1
+              cdrom:
+                bus: scsi
+            - name: rootdisk
+              bootOrder: 2
+              disk:
+                bus: virtio
+            - name: virtio-drivers
+              cdrom:
+                bus: sata
+            - name: sysprep
+              cdrom:
+                bus: sata
+          interfaces:
+            # Pod-network fallback for Phase 1. To switch to PROD VLAN once Multus
+            # + the prod-vlan57 NAD exist, replace this block with:
+            #   - name: prod-net
+            #     bridge: {}
+            #     model: virtio
+            # and update the networks: stanza to use multus.networkName: kubevirt-vms/prod-vlan57
+            - name: default
+              masquerade: {}
+              model: virtio
+        machine:
+          type: q35
+      networks:
+        - name: default
+          pod: {}
+      volumes:
+        - name: rootdisk
+          persistentVolumeClaim:
+            claimName: ci1-rootdisk
+        - name: windows-iso
+          # 2026-05-08 PM (Path C, CONTAINERDISK): the ISO is now packaged as
+          # a KubeVirt containerDisk OCI image baked from
+          # `FROM scratch ; ADD --chown=107:107 disk.img /disk/disk.img`.
+          # The qemu user (uid 107) reads the ISO directly from a tmpfs view
+          # of the OCI layer, bypassing both:
+          #   - Synology NFS export ACL (Path B failed: uid 107 denied at
+          #     directory level even with mode 0777, see memory
+          #     feedback_synology_iso_export_root_only_uid_107_denied)
+          #   - OVMF cdrom read-window timeout (Path A and Path B's SCSI
+          #     retry both hit `BdsDxe: failed to start Boot0001 ... Time out`
+          #     when the cdrom was backed by a PVC the storage controller
+          #     couldn't satisfy reads from fast enough).
+          #
+          # Image build (one-time, per ISO version):
+          #   1. Copy ISO to disk.img, write Dockerfile
+          #   2. podman build --tag localhost/win-server-2025:1.0 .  (on noc1)
+          #   3. podman save -o win-server-2025-1.0.tar localhost/win-server-2025:1.0
+          #   4. SCP tar to all 3 RKE2 nodes (rke2-server, rke2-agent1, rke2-agent2)
+          #   5. sudo /var/lib/rancher/rke2/bin/ctr -a /run/k3s/containerd/containerd.sock \
+          #        -n k8s.io images import /tmp/win-server-2025-1.0.tar
+          # Standard FC pattern per `feedback_rke2_localhost_imagepullpolicy`.
+          #
+          # When a new Windows ISO version ships, bump the tag (1.1, 1.2, ...),
+          # rebuild + redistribute, and update the image: line below in a new
+          # commit. KubeVirt picks up the new image via a VM restart.
+          #
+          # The legacy NFS PVC + PV (apps/kubevirt-vms/win2025-iso-nfs-pv.yaml)
+          # and CDI Longhorn PVC (`windows-server-2025-iso`) are RETAINED for
+          # this commit so the prior states are recoverable. Once the
+          # containerDisk path proves on a successful Windows install, both
+          # legacy artifacts can be pruned in a follow-up commit.
+          containerDisk:
+            image: localhost/win-server-2025:1.0
+            imagePullPolicy: Never
+        - name: virtio-drivers
+          containerDisk:
+            # Pinned to v1.8.2 (latest stable as of 2026-05-08).
+            # The :latest tag uses Docker manifest v1 schema which containerd
+            # 2.1 (RKE2 v1.34.5) refuses to pull with:
+            #   "media type application/vnd.docker.distribution.manifest.v1+prettyjws
+            #    is no longer supported since containerd v2.1"
+            # v1.8.2 is rebuilt with manifest v2/OCI and works on containerd 2.1.
+            # Bump available: https://quay.io/repository/kubevirt/virtio-container-disk?tab=tags
+            image: quay.io/kubevirt/virtio-container-disk:v1.8.2
+        - name: sysprep
+          sysprep:
+            configMap:
+              name: ci1-autounattend
+      terminationGracePeriodSeconds: 3600

apps/kubevirt-vms/prod-vlan57-nad.yaml

@@ -0,0 +1,69 @@
+# =============================================================================
+# NetworkAttachmentDefinition — PROD VLAN 57 bridge
+# =============================================================================
+# Purpose: makes KubeVirt VMs reachable on the PROD VLAN (10.0.57.0/24)
+# alongside the existing pod network. Required for ci1 to bridge onto PROD
+# (e.g. to provision/scrape edge1, edge2, kiosks, Pis on the same L2 segment).
+#
+# **DEPLOY GATE — Phase 1.5 host work required first**:
+#   On every RKE2 node (rke2-server, rke2-agent1, rke2-agent2):
+#     1. Switch port (UniFi USL16LP) trunks VLAN 57 to the node — usually
+#        already true since BLUEJAY-WS reaches 10.0.57.x services. Verify
+#        with `ip link show enp86s0.57` after configuring sub-interface, OR
+#        `tcpdump -ni enp86s0 vlan 57` and ping a known PROD host.
+#     2. Linux bridge `br-prod` enslaving `enp86s0.57` (VLAN sub-interface).
+#        NetworkManager profile examples in the runbook below.
+#     3. Verify Multus DaemonSet `kube-multus-ds` is Ready on all nodes.
+#
+# Without those, applying this NAD has no effect except to register the CRD.
+# A VM that requests this NAD with no bridge present will fail with:
+#   `error adding pod kubevirt-vms_ci1 to CNI network "prod-vlan57": failed to
+#    plumb VLAN: open /sys/class/net/br-prod/master: no such file or directory`
+#
+# Configuration notes:
+#   - cniVersion 0.3.1 to match Multus daemon-config.json
+#   - mtu 1500 (matches enp86s0 default; bump if jumbo frames configured)
+#   - bridge name `br-prod` is convention; if Puppet picks a different name
+#     (e.g. `br57`, `br-vlan57`), edit BOTH this NAD and the ci1.yaml
+#     interface block. Keep them in sync.
+#   - vlan: 0 because the host bridge already strips VLAN tag (br-prod sits
+#     on top of `enp86s0.57`). If we instead used a VLAN-aware bridge with
+#     trunk port, set vlan: 57 here. Current convention is VLAN-stripped at
+#     the sub-interface, so the bridge passes untagged frames.
+#
+# Apply:
+#   kubectl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml apply -f apps/kubevirt-vms/prod-vlan57-nad.yaml
+#
+# Then update ci1.yaml networks: stanza to:
+#   - name: prod-net
+#     multus:
+#       networkName: kubevirt-vms/prod-vlan57
+# and the interface block from `masquerade` to `bridge`.
+# =============================================================================
+
+---
+# Namespace must exist already (created by ci1.yaml's first document).
+# This file imports a NAD into that same namespace.
+apiVersion: k8s.cni.cncf.io/v1
+kind: NetworkAttachmentDefinition
+metadata:
+  name: prod-vlan57
+  namespace: kubevirt-vms
+  annotations:
+    bluejay.iamworkin.lan/host-bridge: "br-prod (enslaves enp86s0.57)"
+    bluejay.iamworkin.lan/cidr: "10.0.57.0/24"
+    bluejay.iamworkin.lan/gateway: "10.0.57.1"
+    bluejay.iamworkin.lan/dns: "10.0.56.1 (pfSense Unbound)"
+spec:
+  config: |
+    {
+      "cniVersion": "0.3.1",
+      "name": "prod-vlan57",
+      "type": "bridge",
+      "bridge": "br-prod",
+      "ipam": {},
+      "mtu": 1500,
+      "vlan": 0,
+      "promiscMode": true,
+      "preserveDefaultVlan": false
+    }

apps/kubevirt-vms/win2025-iso-nfs-pv.yaml

@@ -0,0 +1,99 @@
+# =============================================================================
+# Windows Server 2025 ISO — Static NFS PV (Path B for SATA-CDROM timeout)
+# =============================================================================
+# Purpose: Mount the ISO from Synology NAS via NFS instead of from a Longhorn-
+# backed Filesystem PVC.
+#
+# Why: SATA-CDROM emulation reading from a Longhorn-backed Filesystem PVC is
+# too slow for OVMF's boot read window — the DVD-ROM enumeration times out
+# before the bootloader can be read. Symptom on the serial console:
+#   BdsDxe: failed to start Boot0001 "UEFI QEMU DVD-ROM QM00001 " from ...
+#   BdsDxe: failed to start Boot0001 ... Time out
+#   BdsDxe: No bootable option or device was found
+# Diagnosis confirmed the ISO content is a perfectly valid bootable ISO9660
+# image — the bug is in the timing path between OVMF and Longhorn-backed
+# storage, not in the ISO itself.
+#
+# Block-mode PVC was tried (`volumeMode: Block` via DataVolume) and would
+# likely fix the timing, but CDI v1.65.0's upload-target pod cannot open the
+# block device due to runAsUser:107 + capabilities.drop:[ALL] and we got:
+#   blockdev: cannot open /dev/cdi-block-volume: Permission denied
+#
+# NFS-mounted ISO bypasses both issues: no Longhorn slowness, no CDI upload
+# pod permission concerns. The ISO is read directly from the NAS over a
+# native NFSv4.1 mount that QEMU's SATA emulator can read at full LAN speed.
+#
+# Layout on Synology:
+#   /volume1/ISOs/                                              (existing export, RKE2 ACL)
+#     en-us_windows_server_2025_updated_march_2026_x64_dvd_8e06425a.iso
+#     win2025-iso-disk/                                         (new subdir, 2026-05-08)
+#       disk.img -> hardlink to ../en-us_windows_server_2025_..._8e06425a.iso
+#
+# KubeVirt's launcher pod expects a PVC mounted at
+# /var/run/kubevirt-private/vmi-disks/<diskName>/disk.img — by mounting the
+# `win2025-iso-disk/` subdir as the NFS PV root, `disk.img` lives at the PV's
+# root and KubeVirt's CDROM emulator finds it without any path manipulation.
+#
+# A symlink would NOT work for sub-path NFS mounts (the relative target
+# `../...iso` falls outside the sub-mount root). A hardlink works because it
+# references the same inode regardless of mount point.
+#
+# Memory references:
+#   - feedback_synology_nfs_volume1_kubernetes_export_scoped (Synology export
+#     scoping pattern — but /volume1/ISOs export, unlike /volume1/kubernetes,
+#     does support sub-path mounts because Synology NFS is configured with
+#     pseudo-fs in NFSv4.1)
+#   - feedback_kubevirt_iso_first_install_bootorder_and_runstrategy (boot
+#     order / runStrategy gotchas, separate from the storage timing issue)
+#
+# Validation (2026-05-08, from rke2-server / rke2-agent1 / rke2-agent2):
+#   mount -t nfs -o nfsvers=4.1,ro 10.0.58.3:/volume1/ISOs/win2025-iso-disk /tmp/m
+#   file /tmp/m/disk.img
+#     -> ISO 9660 CD-ROM filesystem data 'SSS_X64FRE_EN-US_DV9' (bootable)
+# All 3 RKE2 nodes can mount and read.
+# =============================================================================
+
+apiVersion: v1
+kind: PersistentVolume
+metadata:
+  name: windows-server-2025-iso-nfs
+  labels:
+    flowercore.io/iso: windows-server-2025
+    flowercore.io/managed-by: bluejay-infra
+spec:
+  capacity:
+    storage: 8Gi
+  accessModes:
+    - ReadOnlyMany
+  volumeMode: Filesystem
+  persistentVolumeReclaimPolicy: Retain
+  storageClassName: ""              # static, no provisioner
+  mountOptions:
+    - nfsvers=4.1
+    - ro
+    - hard
+    - timeo=600
+    - retrans=3
+  nfs:
+    server: 10.0.58.3               # BlueJayNAS Synology DS1621+ on HOME VLAN 58
+    path: /volume1/ISOs/win2025-iso-disk
+    readOnly: true
+
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: windows-server-2025-iso-nfs
+  namespace: kubevirt-vms
+  labels:
+    app: ci-runner
+    flowercore.io/managed-by: bluejay-infra
+spec:
+  accessModes:
+    - ReadOnlyMany
+  volumeMode: Filesystem
+  resources:
+    requests:
+      storage: 8Gi
+  storageClassName: ""
+  volumeName: windows-server-2025-iso-nfs

apps/monitoring/fc-updatecenter-dashboard.grafana.txt

@@ -0,0 +1,762 @@
+{
+  "annotations": {
+    "list": []
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 1,
+  "id": null,
+  "links": [
+    {
+      "icon": "external link",
+      "includeVars": false,
+      "keepTime": false,
+      "targetBlank": true,
+      "title": "Open Service",
+      "type": "link",
+      "url": "https://updatecenter.iamworkin.lan/"
+    }
+  ],
+  "panels": [
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "mappings": [
+            {
+              "options": {
+                "0": {
+                  "color": "#f87171",
+                  "index": 1,
+                  "text": "DOWN"
+                },
+                "1": {
+                  "color": "#4ade80",
+                  "index": 0,
+                  "text": "UP"
+                }
+              },
+              "type": "value"
+            }
+          ],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "#f87171",
+                "value": null
+              },
+              {
+                "color": "#4ade80",
+                "value": 1
+              }
+            ]
+          }
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 8,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "options": {
+        "colorMode": "background",
+        "graphMode": "none",
+        "justifyMode": "center",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "textMode": "value_and_name"
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "probe_success{job=\"probe-traefik-services\",instance=\"updatecenter.iamworkin.lan\"}",
+          "refId": "A",
+          "legendFormat": "Availability"
+        }
+      ],
+      "title": "Service Availability",
+      "transparent": true,
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "decimals": 2,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "#f87171",
+                "value": null
+              },
+              {
+                "color": "#fbbf24",
+                "value": 95
+              },
+              {
+                "color": "#FFB300",
+                "value": 99
+              },
+              {
+                "color": "#4ade80",
+                "value": 99.9
+              }
+            ]
+          },
+          "unit": "percent"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 8,
+        "x": 8,
+        "y": 0
+      },
+      "id": 2,
+      "options": {
+        "colorMode": "background_solid",
+        "graphMode": "area",
+        "justifyMode": "center",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "textMode": "value_and_name"
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "avg_over_time(probe_success{job=\"probe-traefik-services\",instance=\"updatecenter.iamworkin.lan\"}[24h]) * 100",
+          "refId": "A",
+          "legendFormat": "24h Uptime"
+        }
+      ],
+      "title": "24-Hour Uptime",
+      "transparent": true,
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "max": 30,
+          "min": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "#f87171",
+                "value": null
+              },
+              {
+                "color": "#fbbf24",
+                "value": 2
+              },
+              {
+                "color": "#4ade80",
+                "value": 7
+              }
+            ]
+          },
+          "unit": "d"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 8,
+        "x": 16,
+        "y": 0
+      },
+      "id": 3,
+      "options": {
+        "minVizHeight": 75,
+        "minVizWidth": 75,
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showThresholdLabels": false,
+        "showThresholdMarkers": true
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "(probe_ssl_earliest_cert_expiry{job=\"probe-traefik-services\",instance=\"updatecenter.iamworkin.lan\"} - time()) / 86400",
+          "refId": "A",
+          "legendFormat": "Days Remaining"
+        }
+      ],
+      "title": "Cert Expiry (Days)",
+      "transparent": true,
+      "type": "gauge"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "Response Time (seconds)",
+            "drawStyle": "line",
+            "fillOpacity": 12,
+            "gradientMode": "scheme",
+            "lineInterpolation": "smooth",
+            "lineWidth": 2,
+            "pointSize": 4,
+            "showPoints": "never",
+            "spanNulls": true,
+            "thresholdsStyle": {
+              "mode": "dashed"
+            }
+          },
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "#4ade80",
+                "value": null
+              },
+              {
+                "color": "#fbbf24",
+                "value": 2
+              },
+              {
+                "color": "#f87171",
+                "value": 5
+              }
+            ]
+          },
+          "unit": "s"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 14,
+        "x": 0,
+        "y": 4
+      },
+      "id": 4,
+      "options": {
+        "legend": {
+          "calcs": [
+            "lastNotNull",
+            "mean",
+            "max"
+          ],
+          "displayMode": "table",
+          "placement": "right"
+        },
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "probe_duration_seconds{job=\"probe-traefik-services\",instance=\"updatecenter.iamworkin.lan\"}",
+          "refId": "A",
+          "legendFormat": "Probe Duration"
+        }
+      ],
+      "timeFrom": "1h",
+      "title": "Response Time (1h Trend)",
+      "transparent": true,
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 10,
+        "x": 14,
+        "y": 4
+      },
+      "id": 5,
+      "options": {
+        "alertInstanceLabelFilter": "{instance=\"updatecenter.iamworkin.lan\"}",
+        "alertName": "",
+        "dashboardAlerts": false,
+        "groupBy": [],
+        "groupMode": "default",
+        "maxItems": 10,
+        "sortOrder": 1,
+        "stateFilter": {
+          "error": true,
+          "firing": true,
+          "noData": true,
+          "normal": false,
+          "pending": true
+        },
+        "viewMode": "list"
+      },
+      "title": "Active Alerts",
+      "type": "alertlist"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 12
+      },
+      "id": 20,
+      "title": "OTEL Counters — Track 1D",
+      "type": "row"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "lineWidth": 1,
+            "fillOpacity": 10
+          },
+          "unit": "reqps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 13
+      },
+      "id": 21,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "calcs": ["mean", "lastNotNull"]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "sum by (status) (rate(updatecenter_manifest_requests_total[5m]))",
+          "refId": "A",
+          "legendFormat": "status={{status}}"
+        }
+      ],
+      "title": "Manifest Requests rate by status (5m)",
+      "transparent": true,
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "lineWidth": 1,
+            "fillOpacity": 10
+          },
+          "unit": "Bps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 13
+      },
+      "id": 22,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "calcs": ["mean", "lastNotNull"]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "sum by (slug) (rate(updatecenter_bundle_download_bytes_total[5m]))",
+          "refId": "A",
+          "legendFormat": "{{slug}}"
+        }
+      ],
+      "title": "Bundle Download Throughput by slug (5m)",
+      "transparent": true,
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "lineWidth": 1,
+            "fillOpacity": 10
+          },
+          "unit": "reqps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 21
+      },
+      "id": 23,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "calcs": ["mean", "lastNotNull"]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "sum by (status) (rate(updatecenter_checkins_total[5m]))",
+          "refId": "A",
+          "legendFormat": "status={{status}}"
+        }
+      ],
+      "title": "Agent Check-in Rate by status (5m)",
+      "transparent": true,
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "thresholds"
+          },
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "#4ade80", "value": null },
+              { "color": "#f87171", "value": 1 }
+            ]
+          },
+          "unit": "none",
+          "decimals": 2
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 6,
+        "x": 12,
+        "y": 21
+      },
+      "id": 24,
+      "options": {
+        "colorMode": "background",
+        "graphMode": "area",
+        "justifyMode": "center",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": ["sum"],
+          "fields": "",
+          "values": false
+        },
+        "textMode": "value_and_name"
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "increase(updatecenter_signature_verify_failures_total[1h])",
+          "refId": "A",
+          "legendFormat": "Sig Verify Failures (1h)"
+        }
+      ],
+      "title": "Signature Verify Failures (1h)",
+      "transparent": true,
+      "type": "stat"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "lineWidth": 1,
+            "fillOpacity": 10
+          },
+          "unit": "reqps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 6,
+        "x": 18,
+        "y": 21
+      },
+      "id": 25,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "calcs": ["mean", "lastNotNull"]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "sum by (slug, channel) (rate(updatecenter_release_publishes_total[5m]))",
+          "refId": "A",
+          "legendFormat": "{{slug}}/{{channel}}"
+        }
+      ],
+      "title": "Release Publishes rate by slug/channel (5m)",
+      "transparent": true,
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "lineWidth": 1,
+            "fillOpacity": 10
+          },
+          "unit": "reqps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 29
+      },
+      "id": 26,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "calcs": ["mean", "lastNotNull"]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "sum by (kind, status) (rate(updatecenter_bundle_downloads_total[5m]))",
+          "refId": "A",
+          "legendFormat": "{{kind}} / {{status}}"
+        }
+      ],
+      "title": "Bundle Download Requests by kind/status (5m)",
+      "transparent": true,
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "fffjikve8llhce"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "lineWidth": 2,
+            "fillOpacity": 20
+          },
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "#4ade80", "value": null },
+              { "color": "#f87171", "value": 0.01 }
+            ]
+          },
+          "unit": "reqps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 29
+      },
+      "id": 27,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "calcs": ["mean", "lastNotNull"]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "desc"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "fffjikve8llhce"
+          },
+          "expr": "rate(updatecenter_signature_verify_failures_total[5m])",
+          "refId": "A",
+          "legendFormat": "Sig verify failures/s"
+        }
+      ],
+      "title": "Signature Verify Failure Rate (5m) — Critical if >0",
+      "transparent": true,
+      "type": "timeseries"
+    }
+  ],
+  "refresh": "30s",
+  "schemaVersion": 39,
+  "style": "dark",
+  "tags": [
+    "blue-jay",
+    "flowercore",
+    "synthetic",
+    "updatecenter",
+    "otel"
+  ],
+  "templating": {
+    "list": []
+  },
+  "time": {
+    "from": "now-24h",
+    "to": "now"
+  },
+  "timezone": "browser",
+  "title": "FlowerCore.UpdateCenter Dashboard",
+  "uid": "fc-updatecenter",
+  "version": 2
+}

apps/monitoring/flowercore-remotedesktop-grafana-dashboard.grafana.txt

@@ -0,0 +1,226 @@
+{
+  "annotations": {
+    "list": []
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 0,
+  "id": null,
+  "links": [],
+  "panels": [
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "single"
+        }
+      },
+      "targets": [
+        {
+          "editorMode": "code",
+          "expr": "sum by (event) (increase(fc_desktop_session_events_total[$__rate_interval]))",
+          "legendFormat": "{{event}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "RemoteDesktop Session Events",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 0
+      },
+      "id": 2,
+      "options": {
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "showUnfilled": true
+      },
+      "targets": [
+        {
+          "editorMode": "code",
+          "expr": "sum by (template, event) (increase(fc_desktop_session_events_total[24h]))",
+          "legendFormat": "{{template}} {{event}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "24h Session Events By Template",
+      "type": "bargauge"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 8
+      },
+      "id": 3,
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "single"
+        }
+      },
+      "targets": [
+        {
+          "editorMode": "code",
+          "expr": "fc_desktop_pool_ready",
+          "legendFormat": "{{template}} ready",
+          "range": true,
+          "refId": "A"
+        },
+        {
+          "editorMode": "code",
+          "expr": "fc_desktop_pool_desired",
+          "legendFormat": "{{template}} desired",
+          "range": true,
+          "refId": "B"
+        }
+      ],
+      "title": "Warm Pool Ready vs Desired",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "orange",
+                "value": 1
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 8
+      },
+      "id": 4,
+      "options": {
+        "colorMode": "value",
+        "graphMode": "none",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "textMode": "auto"
+      },
+      "targets": [
+        {
+          "editorMode": "code",
+          "expr": "sum(increase(fc_desktop_session_events_total{event=\"connect\",browser_datasource=\"json\"}[24h])) - sum(increase(fc_desktop_session_events_total{event=\"disconnect\"}[24h]))",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "24h Connect Minus Disconnect",
+      "type": "stat"
+    }
+  ],
+  "refresh": "30s",
+  "schemaVersion": 39,
+  "style": "dark",
+  "tags": [
+    "flowercore",
+    "remotedesktop",
+    "guacamole"
+  ],
+  "templating": {
+    "list": []
+  },
+  "time": {
+    "from": "now-24h",
+    "to": "now"
+  },
+  "timezone": "browser",
+  "title": "FlowerCore RemoteDesktop",
+  "uid": "flowercore-remotedesktop",
+  "version": 1
+}

apps/monitoring/grafana-dashboard-remotedesktop.yaml

@@ -0,0 +1,249 @@
+# Grafana dashboard ConfigMap for FlowerCore.RemoteDesktop.
+#
+# Inlines the JSON from flowercore-remotedesktop-grafana-dashboard.json.
+# Kept as a standalone file (not inlined in noc-monitoring.yaml) so the
+# CRLF-dirty state of noc-monitoring.yaml doesn't have to be normalized
+# in the same pass. To actually load the dashboard, the Grafana Deployment
+# in noc-monitoring.yaml needs a matching 'volumes:' entry:
+#
+#   - name: dashboard-remotedesktop
+#     configMap:
+#       name: grafana-dashboard-remotedesktop
+#
+# ArgoCD will sync this ConfigMap automatically through the bluejay-infra
+# ApplicationSet (infra-monitoring App). The dashboard just won't load
+# until the Grafana Deployment mount is wired.
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: grafana-dashboard-remotedesktop
+  namespace: monitoring
+data:
+  remotedesktop.json: |
+    {
+      "annotations": {
+        "list": []
+      },
+      "editable": true,
+      "fiscalYearStartMonth": 0,
+      "graphTooltip": 0,
+      "id": null,
+      "links": [],
+      "panels": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "unit": "short"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 0
+          },
+          "id": 1,
+          "options": {
+            "legend": {
+              "displayMode": "table",
+              "placement": "bottom"
+            },
+            "tooltip": {
+              "mode": "single"
+            }
+          },
+          "targets": [
+            {
+              "editorMode": "code",
+              "expr": "sum by (event) (increase(fc_desktop_session_events_total[$__rate_interval]))",
+              "legendFormat": "{{event}}",
+              "range": true,
+              "refId": "A"
+            }
+          ],
+          "title": "RemoteDesktop Session Events",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "unit": "short"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 0
+          },
+          "id": 2,
+          "options": {
+            "orientation": "auto",
+            "reduceOptions": {
+              "calcs": [
+                "lastNotNull"
+              ],
+              "fields": "",
+              "values": false
+            },
+            "showUnfilled": true
+          },
+          "targets": [
+            {
+              "editorMode": "code",
+              "expr": "sum by (template, event) (increase(fc_desktop_session_events_total[24h]))",
+              "legendFormat": "{{template}} {{event}}",
+              "range": true,
+              "refId": "A"
+            }
+          ],
+          "title": "24h Session Events By Template",
+          "type": "bargauge"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "unit": "short"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 8
+          },
+          "id": 3,
+          "options": {
+            "legend": {
+              "displayMode": "table",
+              "placement": "bottom"
+            },
+            "tooltip": {
+              "mode": "single"
+            }
+          },
+          "targets": [
+            {
+              "editorMode": "code",
+              "expr": "fc_desktop_pool_ready",
+              "legendFormat": "{{template}} ready",
+              "range": true,
+              "refId": "A"
+            },
+            {
+              "editorMode": "code",
+              "expr": "fc_desktop_pool_desired",
+              "legendFormat": "{{template}} desired",
+              "range": true,
+              "refId": "B"
+            }
+          ],
+          "title": "Warm Pool Ready vs Desired",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": null
+                  },
+                  {
+                    "color": "orange",
+                    "value": 1
+                  }
+                ]
+              },
+              "unit": "short"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 8
+          },
+          "id": 4,
+          "options": {
+            "colorMode": "value",
+            "graphMode": "none",
+            "justifyMode": "auto",
+            "orientation": "auto",
+            "reduceOptions": {
+              "calcs": [
+                "lastNotNull"
+              ],
+              "fields": "",
+              "values": false
+            },
+            "textMode": "auto"
+          },
+          "targets": [
+            {
+              "editorMode": "code",
+              "expr": "sum(increase(fc_desktop_session_events_total{event=\"connect\",browser_datasource=\"json\"}[24h])) - sum(increase(fc_desktop_session_events_total{event=\"disconnect\"}[24h]))",
+              "range": true,
+              "refId": "A"
+            }
+          ],
+          "title": "24h Connect Minus Disconnect",
+          "type": "stat"
+        }
+      ],
+      "refresh": "30s",
+      "schemaVersion": 39,
+      "style": "dark",
+      "tags": [
+        "flowercore",
+        "remotedesktop",
+        "guacamole"
+      ],
+      "templating": {
+        "list": []
+      },
+      "time": {
+        "from": "now-24h",
+        "to": "now"
+      },
+      "timezone": "browser",
+      "title": "FlowerCore RemoteDesktop",
+      "uid": "flowercore-remotedesktop",
+      "version": 1
+    }

apps/monitoring/noc-monitoring.yaml

@@ -104,21 +104,27 @@ data:
           - target_label: __address__
             replacement: snmp-exporter.monitoring.svc:9116
 
-      # UniFi Cloud Key SNMP
+      # UniFi Cloud Key SNMP — DISABLED 2026-04-26
-      - job_name: "snmp-cloudkey"
+      # The Cloud Key Gen2+ runs unifi-core (controller) only — not a network
-        static_configs:
+      # device — and does NOT run an SNMP agent on UDP/161. Scrapes were
-          - targets: ["10.0.56.3"]
+      # silently failing with "connection refused" from 10.42.x.x:161 every
-        metrics_path: /snmp
+      # 30s, polluting up{} = 0 and lastError on the Targets page. Hardware
-        params:
+      # health (CPU/mem/disk) for the Cloud Key host should come from
-          module: [if_mib]
+      # node_exporter via SSH — not SNMP.
-          auth: [bluejay_v2]
+      # - job_name: "snmp-cloudkey"
-        relabel_configs:
+      #   static_configs:
-          - source_labels: [__address__]
+      #     - targets: ["10.0.56.3"]
-            target_label: __param_target
+      #   metrics_path: /snmp
-          - source_labels: [__param_target]
+      #   params:
-            target_label: instance
+      #     module: [if_mib]
-          - target_label: __address__
+      #     auth: [bluejay_v2]
-            replacement: snmp-exporter.monitoring.svc:9116
+      #   relabel_configs:
+      #     - source_labels: [__address__]
+      #       target_label: __param_target
+      #     - source_labels: [__param_target]
+      #       target_label: instance
+      #     - target_label: __address__
+      #       replacement: snmp-exporter.monitoring.svc:9116
 
       # UniFi Switch SNMP
       - job_name: "snmp-switch"
@@ -278,6 +284,38 @@ data:
           - target_label: __address__
             replacement: blackbox-exporter.monitoring.svc:9115
 
+      # FlowerCore.RemoteDesktop web health (public cluster VIP)
+      # Module is https_internal — desktop.iamworkin.lan uses a step-ca leaf
+      # cert; blackbox does NOT trust step-ca root, so http_2xx fails with
+      # x509 unknown authority and probe_success=0 even when /health 200s.
+      - job_name: "probe-remotedesktop"
+        metrics_path: /probe
+        params:
+          module: [https_internal]
+        scrape_interval: 30s
+        static_configs:
+          - targets: ["https://desktop.iamworkin.lan/health"]
+            labels:
+              instance: "https://desktop.iamworkin.lan/health"
+              service: "remotedesktop-web"
+        relabel_configs:
+          - source_labels: [__address__]
+            target_label: __param_target
+          - target_label: __address__
+            replacement: blackbox-exporter.monitoring.svc:9115
+
+      # FlowerCore.RemoteDesktop /metrics (direct scrape for counters)
+      - job_name: "fc-remotedesktop"
+        metrics_path: /metrics
+        scheme: https
+        scrape_interval: 30s
+        tls_config:
+          insecure_skip_verify: true
+        static_configs:
+          - targets: ["desktop.iamworkin.lan"]
+            labels:
+              service: "remotedesktop-web"
+
       # CUPS web UI health (port 631)
       - job_name: "probe-cups"
         metrics_path: /probe
@@ -301,26 +339,12 @@ data:
       # AI Stack Health Probes (Blackbox Exporter)
       # =============================================================================
 
-      # Ollama API — workstation (LOCAL Agent Zero)
+      # NOTE: probe-ollama-local and probe-agentzero-local were REMOVED
-      - job_name: "probe-ollama-local"
+      # 2026-04-26. They pointed at 10.0.58.100 (HOME VLAN) which is not
-        metrics_path: /probe
+      # reachable from cluster pods (firewalled). They had been firing as
-        params:
+      # OllamaDown / AgentZeroDown since 2026-04-24. Workstation/AI-laptop
-          module: [http_ollama]
+      # Ollama and Agent Zero should be monitored via host-side Puppet
-        scrape_interval: 30s
+      # (node_exporter on the box) once the AI laptop is running 24/7.
-        static_configs:
-          - targets: ["http://10.0.58.100:11434/api/tags"]
-            labels:
-              instance: "ollama-local"
-              service: "ollama"
-              deployment: "local"
-              gpu: "r9700"
-        relabel_configs:
-          - source_labels: [__address__]
-            target_label: __param_target
-          - source_labels: [__param_target]
-            target_label: instance
-          - target_label: __address__
-            replacement: blackbox-exporter.monitoring.svc:9115
 
       # Ollama API — edge1 Pi 5 (NUC Agent Zero)
       - job_name: "probe-ollama-edge1"
@@ -343,34 +367,18 @@ data:
           - target_label: __address__
             replacement: blackbox-exporter.monitoring.svc:9115
 
-      # Agent Zero Web UI — local (K3s)
+      # Agent Zero Web UI — in-cluster (RKE2)
-      - job_name: "probe-agentzero-local"
+      # Target uses short svc form (agent-zero.agent-zero.svc) NOT
-        metrics_path: /probe
+      # cluster.local FQDN — the *.cluster.local form gets rewritten to
-        params:
+      # 10.0.56.200 (Traefik VIP) by the CoreDNS iamworkin.lan template +
-          module: [http_2xx]
+      # ndots:5 search-suffix expansion. Memory: feedback_coredns_ndots_template_collision.
-        scrape_interval: 30s
-        static_configs:
-          - targets: ["http://10.0.58.100:30050/"]
-            labels:
-              instance: "agent-zero-local"
-              service: "agent-zero"
-              deployment: "local"
-        relabel_configs:
-          - source_labels: [__address__]
-            target_label: __param_target
-          - source_labels: [__param_target]
-            target_label: instance
-          - target_label: __address__
-            replacement: blackbox-exporter.monitoring.svc:9115
-
-      # Agent Zero Web UI — NUC (RKE2 via Traefik)
       - job_name: "probe-agentzero-nuc"
         metrics_path: /probe
         params:
           module: [http_2xx]
         scrape_interval: 30s
         static_configs:
-          - targets: ["http://agent-zero.agent-zero.svc.cluster.local/"]
+          - targets: ["http://agent-zero.agent-zero.svc:80/"]
             labels:
               instance: "agent-zero-nuc"
               service: "agent-zero"
@@ -383,6 +391,119 @@ data:
           - target_label: __address__
             replacement: blackbox-exporter.monitoring.svc:9115
 
+      # =============================================================================
+      # K8s Cluster State (kube-state-metrics, cert-manager, traefik)
+      # =============================================================================
+      # Use in-cluster ClusterIP service DNS — NOT NodePorts — so a same-node
+      # NodePort hairpin doesn't break the scrape (hit on rke2-agent1 hosting
+      # both prometheus and traefik on 2026-04-26: 10.0.56.12:30900 timed out
+      # from prometheus while .11/.13 worked). NodePorts at 30900-30902 are
+      # still useful for noc1-Podman-style external scrapers, but in-cluster
+      # we should always use the svc DNS form.
+
+      # kube-state-metrics — exposes K8s object state (pods, deployments, nodes)
+      # Required for KubeContainerRestartingFrequently / KubePodNotReady alerts.
+      - job_name: "kube-state-metrics"
+        scrape_interval: 30s
+        static_configs:
+          - targets: ["kube-state-metrics.kube-system.svc:8080"]
+            labels:
+              cluster: "rke2"
+
+      # cert-manager — exposes certmanager_certificate_ready_status,
+      # certmanager_certificate_expiration_timestamp_seconds, etc. Drives the
+      # CertManagerCertificateNotReady / CertManagerCertificateRenewalFailed
+      # alerts. Memory: project_cert_manager_prometheus_scrape.
+      - job_name: "cert-manager"
+        scrape_interval: 30s
+        static_configs:
+          - targets: ["cert-manager-metrics.cert-manager.svc:9402"]
+            labels:
+              cluster: "rke2"
+
+      # Traefik — request rates, latency, TLS cert metadata, router state.
+      # ClusterIP svc routes to one of the traefik pods; per-pod scrape via
+      # the headless `traefik-metrics` selector would be nicer for failover
+      # visibility but the single-replica scrape is enough for steady-state.
+      - job_name: "traefik"
+        scrape_interval: 15s
+        static_configs:
+          - targets: ["traefik-metrics.traefik-system.svc:9100"]
+            labels:
+              service: "traefik"
+              cluster: "rke2"
+
+      # Longhorn — exposes longhorn_volume_robustness, longhorn_backup_*,
+      # longhorn_node_status_*. Enables LonghornVolumeUnhealthy +
+      # LonghornBackupFailed alerts (no real visibility into Longhorn
+      # health before this — was relying on K8s events which are noisy
+      # transient lifecycle messages, not actionable signals).
+      - job_name: "longhorn"
+        scrape_interval: 30s
+        static_configs:
+          - targets: ["longhorn-backend.longhorn-system.svc:9500"]
+            labels:
+              service: "longhorn"
+              cluster: "rke2"
+
+      # FC web services through Traefik — single probe surface to spot any
+      # iamworkin.lan host returning non-200. Uses https_internal because all
+      # certs are step-ca leaves; blackbox would x509-fail with http_2xx.
+      # Some services need explicit healthcheck paths because root returns
+      # 404 (acme, guac) or 401 (grafana, prometheus). Drop them or point at
+      # the right endpoint — don't lower valid_status_codes globally because
+      # 401 from a healthy pod and 401 from an outage look identical.
+      - job_name: "probe-traefik-services"
+        metrics_path: /probe
+        params:
+          module: [https_internal]
+        scrape_interval: 60s
+        static_configs:
+          - targets:
+              # Root-reachable services (200 or 3xx)
+              - "https://gitea.iamworkin.lan/"
+              - "https://argocd.iamworkin.lan/"
+              - "https://intranet.iamworkin.lan/"
+              - "https://signage.iamworkin.lan/"
+              - "https://kiosk.iamworkin.lan/"
+              - "https://media.iamworkin.lan/"
+              - "https://mysql.iamworkin.lan/"
+              - "https://php.iamworkin.lan/"
+              - "https://zabbix.iamworkin.lan/"
+              - "https://desktop.iamworkin.lan/"
+              - "https://print.iamworkin.lan/"
+              - "https://dns.iamworkin.lan/"
+              - "https://chat.iamworkin.lan/"
+              - "https://dist.iamworkin.lan/"
+              - "https://dms.iamworkin.lan/"
+              - "https://menuboard.iamworkin.lan/"
+              - "https://messageboard.iamworkin.lan/"
+              - "https://presentations.iamworkin.lan/"
+              - "https://retail.iamworkin.lan/"
+              - "https://ttsreader.iamworkin.lan/"
+              # Explicit healthcheck paths
+              - "https://fc-llm-bridge.iamworkin.lan/healthz"
+              - "https://acme.iamworkin.lan/health"
+              # NOTE: services intentionally NOT in this probe surface
+              #   - grafana.iamworkin.lan: every endpoint (incl. /api/health
+              #     and /login) returns 401 behind Traefik basic-auth.
+              #     Health covered by in-cluster monitoring-grafana scrape.
+              #   - prometheus.iamworkin.lan: same auth pattern. Health covered
+              #     by the prometheus self-scrape job.
+              #   - guac.iamworkin.lan: deprecated — Guacamole moved to
+              #     desktop.iamworkin.lan/guacamole/ (memory:
+              #     feedback_traefik_cross_namespace_refs_disabled).
+            labels:
+              probe_type: "traefik-service"
+        relabel_configs:
+          - source_labels: [__address__]
+            target_label: __param_target
+          - source_labels: [__param_target]
+            regex: "https?://([^/:]+).*"
+            target_label: instance
+          - target_label: __address__
+            replacement: blackbox-exporter.monitoring.svc:9115
+
       # =============================================================================
       # Self-monitoring (K8s monitoring namespace)
       # =============================================================================
@@ -521,6 +642,42 @@ data:
               summary: "Print queue backlog on edge2 ({{ $value }} active jobs)"
               description: "CUPS has {{ $value }} active jobs queued. Possible printer jam, USB disconnect, or paper out."
 
+          # Paper roll lifecycle alerts (XL Track I, 2026-04-26).
+          # Source-of-truth gauge: print_paper_remaining_percent (Print.Web OTEL,
+          # hydrated on startup from the active PaperRoll row).
+          # alert_channel=thermal_print routes through irc-notify -> Print.Web
+          # /api/print/alert so the printer announces its own paper-out warning
+          # on its remaining paper. Self-referential humor + operator nudge.
+          - alert: PrintPaperRollLow
+            expr: print_paper_remaining_percent{job="printweb-otel"} < 10 and print_paper_remaining_percent{job="printweb-otel"} > 5
+            for: 5m
+            labels:
+              severity: warning
+              alert_channel: thermal_print
+            annotations:
+              summary: "Print roll low on edge2 ({{ $value | printf \"%.1f\" }}% remaining)"
+              description: "NuPrint 210 paper roll has {{ $value | printf \"%.1f\" }}% remaining. Operator should load a fresh roll soon. Run /api/paper/status for the precise mm + estimated jobs left."
+
+          - alert: PrintPaperRollCritical
+            expr: print_paper_remaining_percent{job="printweb-otel"} <= 5
+            for: 2m
+            labels:
+              severity: critical
+              alert_channel: thermal_print
+            annotations:
+              summary: "Print roll critical on edge2 ({{ $value | printf \"%.1f\" }}% remaining)"
+              description: "NuPrint 210 paper roll at {{ $value | printf \"%.1f\" }}% — load a new roll NOW. The 50ft roll has a ~12% red-stripe zone; once paper passes that, the printer can run dry mid-job."
+
+          - alert: PrintJobDeadLetter
+            expr: increase(print_jobs_dead_letter_total[15m]) > 0
+            for: 1m
+            labels:
+              severity: warning
+              alert_channel: thermal_print
+            annotations:
+              summary: "Print job(s) entered dead-letter on edge2 ({{ $value | printf \"%.0f\" }} in last 15m)"
+              description: "{{ $value | printf \"%.0f\" }} print job(s) exhausted MaxRetries and need operator action. Open /print-log, filter Status=DeadLetter, click 'Retry From Start' after fixing the underlying cause (paper jam, USB disconnect, printer power-cycle)."
+
           - alert: CUPSHighJobRate
             expr: rate(cups_job_total[5m]) * 60 > 30
             for: 5m
@@ -540,6 +697,89 @@ data:
               summary: "Print.Web Ollama runner held for >10m ({{ $labels.model }})"
               description: "Print.Web reports model {{ $labels.model }} with {{ $value | printf \"%.0f\" }}s of keep-alive remaining. Check concurrent requests before the Pi 5 Ollama lane thrashes."
 
+      - name: remote-desktop
+        rules:
+          - alert: RemoteDesktopWebDown
+            expr: probe_success{job="probe-remotedesktop",instance="https://desktop.iamworkin.lan/health"} == 0
+            for: 3m
+            labels:
+              severity: warning
+            annotations:
+              summary: "FlowerCore RemoteDesktop web is down"
+              description: "https://desktop.iamworkin.lan/health probe has failed for 3 minutes. Catalog + session launch surface offline."
+
+          - alert: RemoteDesktopMetricsStale
+            expr: absent(fc_desktop_session_events_total)
+            for: 10m
+            labels:
+              severity: warning
+            annotations:
+              summary: "RemoteDesktop /metrics scrape returning no data"
+              description: "No fc_desktop_session_events_total series for 10 minutes. Either the Prometheus scrape target is misconfigured or the web deployment stopped exporting metrics. Zabbix template carries the same 10m no-data trigger for cross-monitor parity."
+
+          # PUBLISHER QUIRK: fc_desktop_pool_depleted / _deficit emit one
+          # series per template per status (Ready/Warming/BelowDesiredSize/
+          # Disabled), and the historical series for non-current statuses
+          # stay at their last value. So just `_depleted > 0` fires forever
+          # on any template that ever entered a bad state.
+          #
+          # SAFE PATTERN: alert only when the canonical "Ready" status
+          # gauge does NOT report ready=1 for the enabled template. This
+          # is the publisher's own canary — _ready{status="Ready"}==1 is
+          # always the current "everything is fine" signal.
+          - alert: RemoteDesktopPoolDepleted
+            expr: |
+              group by(template) (fc_desktop_pool_ready{enabled="true"})
+              unless on(template) (fc_desktop_pool_ready{enabled="true",status="Ready"} == 1)
+            for: 5m
+            labels:
+              severity: warning
+            annotations:
+              summary: "RemoteDesktop pool depleted ({{ $labels.template }})"
+              description: "Pool for template {{ $labels.template }} has no Ready warm pod for 5 minutes. New launches will cold-start. Check pod-scheduling failures, image pull issues, or exhausted node capacity."
+
+          # Same pattern, but only fires when template explicitly reports
+          # a sustained Warning-level alert state (current-status series).
+          - alert: RemoteDesktopPoolDeficitSustained
+            expr: |
+              fc_desktop_pool_deficit{enabled="true",alert_level="Warning"} > 0
+              unless on(template) (fc_desktop_pool_ready{enabled="true",status="Ready"} == 1)
+            for: 10m
+            labels:
+              severity: info
+            annotations:
+              summary: "RemoteDesktop pool {{ $labels.template }} below desired for 10m"
+              description: "Pool {{ $labels.template }} has a persistent deficit of {{ $value }} warm pods AND no Ready series. Likely image pull, NFS affinity, or claim-init issue."
+
+          - alert: RemoteDesktopSessionChurnSpike
+            expr: sum(rate(fc_desktop_session_events_total{event="launch"}[5m])) * 60 > 20
+            for: 5m
+            labels:
+              severity: info
+            annotations:
+              summary: "RemoteDesktop launch rate high ({{ $value | printf \"%.0f\" }}/min)"
+              description: "Launch events exceed 20/min for 5 minutes. Could be a user-facing feature launch, a pooled template thrashing, or a runaway automation loop."
+
+          - alert: RemoteDesktopRecordingEventsDropped
+            expr: absent_over_time(fc_desktop_session_events_total{event="recording"}[30m]) and on() (sum(fc_desktop_session_events_total{event="launch"}) > 0)
+            for: 15m
+            labels:
+              severity: info
+            annotations:
+              summary: "RemoteDesktop recording events silent for 30m despite active launches"
+              description: "No recording events in 30 minutes while launches are happening. Recording may be silently disabled on all templates (SessionRecordingEnabled=false), the guacd NFS mount may be unhealthy, or the retention sweep isn't emitting events. Not an error by itself — worth checking."
+
+          # Match by job — instance label carries full URL incl. /health,
+          # not just hostname, so a hostname-only match never fires.
+          - alert: RemoteDesktopTlsExpiry
+            expr: probe_ssl_earliest_cert_expiry{job="probe-remotedesktop"} - time() < 2 * 86400
+            for: 6h
+            labels:
+              severity: critical
+            annotations:
+              summary: "desktop.iamworkin.lan TLS cert expires within 2 days"
+              description: "The desktop.iamworkin.lan cert is inside the 2-day renewal window and cert-manager has not renewed. Check cert-manager logs, step-ca reachability, and pfSense DNS overrides per the ACME DNS-01 gate."
+
       - name: pi-fleet
         rules:
           - alert: PiManagerDown
@@ -619,13 +859,16 @@ data:
             annotations:
               summary: "Epson ink CRITICAL: {{ $labels.prtMarkerSuppliesDescription }} at {{ $value }}%"
 
+          # for: 30m absorbs sleep cycles. The EcoTank sleeps after ~5 min
+          # of idle and SNMP times out, so 5m for: would page nightly. A
+          # genuine printer outage (jam, disconnected) lasts well over 30m.
           - alert: EpsonPrinterDown
             expr: up{job="snmp-printer"} == 0
-            for: 5m
+            for: 30m
             labels:
               severity: warning
             annotations:
-              summary: "Epson ET-3750 SNMP unreachable"
+              summary: "Epson ET-3750 SNMP unreachable for >30m (likely actual fault, not sleep)"
 
           - alert: SynologyDiskLow
             expr: hrStorageUsed{job="snmp-nas"} / hrStorageSize{job="snmp-nas"} * 100 > 85
@@ -679,6 +922,174 @@ data:
             annotations:
               summary: "Disk usage high on {{ $labels.instance }} ({{ $value | printf \"%.1f\" }}%)"
 
+      # K8s pod-state alerts. Require kube-state-metrics scrape (added
+      # 2026-04-26 — see scrape_configs above). Would have surfaced the
+      # agent-zero ollama-proxy 172x crash-loop instead of letting it
+      # silently churn for ~3 days.
+      - name: kubernetes-state
+        rules:
+          - alert: KubeContainerRestartingFrequently
+            expr: increase(kube_pod_container_status_restarts_total[1h]) > 5
+            for: 15m
+            labels:
+              severity: warning
+            annotations:
+              summary: "{{ $labels.namespace }}/{{ $labels.pod }} container {{ $labels.container }} restarting >5x/hr"
+              description: "Container {{ $labels.container }} in pod {{ $labels.namespace }}/{{ $labels.pod }} has restarted {{ $value | printf \"%.0f\" }} times in the last hour. Check 'kubectl describe pod' + last-state termination reason."
+
+          - alert: KubeContainerCrashLooping
+            expr: increase(kube_pod_container_status_restarts_total[15m]) > 3
+            for: 5m
+            labels:
+              severity: critical
+              alert_channel: thermal_print
+            annotations:
+              summary: "{{ $labels.namespace }}/{{ $labels.pod }} crashlooping ({{ $value | printf \"%.0f\" }} restarts/15m)"
+              description: "Container {{ $labels.container }} restarted {{ $value | printf \"%.0f\" }} times in 15 minutes — actively crashlooping."
+
+          - alert: KubePodNotReady
+            expr: sum by(namespace, pod) (kube_pod_status_phase{phase=~"Pending|Failed|Unknown"}) > 0
+            for: 15m
+            labels:
+              severity: warning
+            annotations:
+              summary: "{{ $labels.namespace }}/{{ $labels.pod }} not Ready for >15m"
+              description: "Pod is in a non-Running, non-Succeeded phase for over 15 minutes. Common causes: ImagePullBackOff (registry/Nexus down, wrong image tag), pending PVC, scheduling failure (taint/resources)."
+
+          - alert: KubePodImagePullBackOff
+            expr: sum by(namespace, pod) (kube_pod_container_status_waiting_reason{reason=~"ImagePullBackOff|ErrImagePull"}) > 0
+            for: 10m
+            labels:
+              severity: warning
+            annotations:
+              summary: "{{ $labels.namespace }}/{{ $labels.pod }} ImagePullBackOff for >10m"
+              description: "Pod can't pull image. Check the image ref (often a stale tag or unreachable registry) and clean up if it's an orphan."
+
+          - alert: KubeDeploymentReplicasMismatch
+            expr: kube_deployment_spec_replicas != kube_deployment_status_replicas_available
+            for: 15m
+            labels:
+              severity: warning
+            annotations:
+              summary: "Deployment {{ $labels.namespace }}/{{ $labels.deployment }} replica mismatch"
+              description: "Spec wants {{ $labels.spec_replicas }} but only {{ $value }} available. Likely a rollout stuck on probe failure, scheduling, or PVC."
+
+      # Longhorn storage health alerts. Required: longhorn scrape job
+      # (added 2026-04-26 — see scrape_configs above). The K8s events
+      # for "snapshot becomes not ready to use" are transient lifecycle
+      # noise, not actionable — these alerts use the actual Longhorn
+      # gauges that reflect persistent state.
+      - name: longhorn-storage
+        rules:
+          # Volume robustness: 0=unknown, 1=healthy, 2=degraded, 3=faulted.
+          # Detached volumes report 0 — that's normal for unattached PVCs,
+          # so filter to only attached.
+          - alert: LonghornVolumeDegraded
+            expr: longhorn_volume_robustness{robustness="degraded"} == 1
+            for: 15m
+            labels:
+              severity: warning
+            annotations:
+              summary: "Longhorn volume {{ $labels.volume }} degraded for >15m"
+              description: "Volume {{ $labels.volume }} on node {{ $labels.node }} has been degraded (one or more replicas unhealthy) for 15+ minutes. Auto-rebuild may need help — check 'kubectl describe volume.longhorn.io {{ $labels.volume }} -n longhorn-system'."
+
+          - alert: LonghornVolumeFaulted
+            expr: longhorn_volume_robustness{robustness="faulted"} == 1
+            for: 5m
+            labels:
+              severity: critical
+              alert_channel: thermal_print
+            annotations:
+              summary: "Longhorn volume {{ $labels.volume }} FAULTED"
+              description: "Volume {{ $labels.volume }} on node {{ $labels.node }} is faulted — all replicas unavailable. Data inaccessible. Manual intervention required."
+
+          # No backup in 36h indicates the daily-backup recurringJob is
+          # silently failing. Allows for one missed run + slack.
+          - alert: LonghornBackupStale
+            expr: |
+              (time() - max by(volume) (longhorn_backup_state{state="Completed"} * on(backup) group_left() longhorn_backup_actual_size_bytes)) > 36 * 3600
+            for: 1h
+            labels:
+              severity: warning
+            annotations:
+              summary: "Longhorn volume {{ $labels.volume }} has no completed backup in >36h"
+              description: "Daily backup recurringJob (cron 0 2 * * *) appears to have skipped this volume. Check 'kubectl get backups.longhorn.io -n longhorn-system' and the daily-backup CronJob logs."
+
+          - alert: LonghornNodeUnhealthy
+            expr: longhorn_node_status{condition="ready",condition_reason!=""} == 0
+            for: 5m
+            labels:
+              severity: warning
+            annotations:
+              summary: "Longhorn node {{ $labels.node }} not Ready"
+              description: "Node {{ $labels.node }} reports ready=false (reason: {{ $labels.condition_reason }}). Volumes scheduled to this node will be unavailable until it recovers."
+
+      # ============================================================
+      # FC Signage Marquee Performance — Track 3 + 8 (2026-05-06)
+      # Live-mirrored from FlowerCore.Notes/scripts/monitoring/alerts.yml.
+      # Source-of-truth for the live Podman Prometheus on noc1 is the
+      # Notes file; this K8s ConfigMap exists so a future migration to
+      # in-cluster Prometheus inherits the ruleset automatically.
+      # See feedback_monitoring_k8s_target_vs_live_podman.
+      # ============================================================
+      - name: fc-signage-marquee
+        rules:
+          - alert: MarqueeDroppedFramesHigh
+            expr: |
+              (
+                sum by (renderer, phase, node_id) (rate(marquee_dropped_frames_total[5m]))
+                /
+                sum by (renderer, phase, node_id) (rate(marquee_render_latency_ms_count[5m]))
+              ) > 0.05
+              unless on()
+              absent_over_time(marquee_dropped_frames_total[7d])
+            for: 5m
+            labels:
+              severity: warning
+              service: signage
+              alert_channel: irc
+            annotations:
+              summary: "Marquee dropped-frame rate >5% on {{ $labels.renderer }}/{{ $labels.node_id }} ({{ $labels.phase }})"
+              description: "Renderer {{ $labels.renderer }} on {{ $labels.node_id }} drops >5% of frames during {{ $labels.phase }}. Animation visibly stuttery."
+
+          - alert: MarqueeRenderLatencyP99High
+            expr: |
+              histogram_quantile(
+                0.99,
+                sum by (renderer, phase, node_id, le) (rate(marquee_render_latency_ms_bucket[5m]))
+              ) > 16
+              unless on()
+              absent_over_time(marquee_render_latency_ms_bucket[7d])
+            for: 10m
+            labels:
+              severity: warning
+              service: signage
+              alert_channel: irc
+            annotations:
+              summary: "Marquee render latency p99 > 16ms on {{ $labels.renderer }}/{{ $labels.node_id }} ({{ $labels.phase }})"
+              description: "Per-frame render latency p99 has exceeded the Pi-class 16ms budget for 10 minutes."
+
+          - alert: MarqueeAnimationDurationDrift
+            expr: |
+              abs(
+                histogram_quantile(0.5, sum by (renderer, phase, le) (rate(marquee_animation_duration_ms_bucket[15m])))
+                -
+                on (phase) group_left() avg by (phase) (marquee_animation_duration_target_ms)
+              )
+              /
+              on (phase) group_left() avg by (phase) (marquee_animation_duration_target_ms)
+              > 0.10
+              unless on()
+              absent_over_time(marquee_animation_duration_ms_bucket[7d])
+            for: 15m
+            labels:
+              severity: info
+              service: signage
+              alert_channel: irc
+            annotations:
+              summary: "Marquee animation duration drifting > 10% on {{ $labels.renderer }} ({{ $labels.phase }})"
+              description: "Median observed cycle duration deviates from target DurationMs by >10%. Could indicate browser tab throttling, GPU pressure, or phase-advancement bug."
+
 # =============================================================================
 # ConfigMap: Blackbox Exporter Configuration
 # =============================================================================
@@ -710,6 +1121,22 @@ data:
           fail_if_body_not_matches_regexp:
             - '"models"'
           preferred_ip_protocol: ip4
+      # https_internal — for Traefik-fronted services with step-ca leaf
+      # certs. blackbox does not trust the step-ca root CA, so http_2xx
+      # against any *.iamworkin.lan host fails with x509 unknown authority.
+      # Redirects + multiple status codes are accepted because some hosts
+      # 302 to /login or /scalar.
+      https_internal:
+        prober: http
+        timeout: 10s
+        http:
+          valid_http_versions: ["HTTP/1.1", "HTTP/2.0"]
+          valid_status_codes: [200, 301, 302, 303, 307, 308]
+          method: GET
+          follow_redirects: true
+          preferred_ip_protocol: ip4
+          tls_config:
+            insecure_skip_verify: true
 
 # =============================================================================
 # ConfigMap: IRC Notify Script
@@ -3045,6 +3472,172 @@ data:
                 relativeTimeRange: {from: 600, to: 0}
                 datasourceUid: __expr__
                 model: {type: threshold, expression: B, conditions: [{evaluator: {params: [85], type: gt}}], refId: C}
+      - orgId: 1
+        name: RemoteDesktop
+        folder: AI Stack Alerts
+        interval: 1m
+        rules:
+          - uid: remotedesktop-web-down
+            title: RemoteDesktop Web DOWN
+            condition: C
+            for: 3m
+            noDataState: Alerting
+            execErrState: OK
+            annotations:
+              summary: FlowerCore RemoteDesktop /health probe failing
+              description: "https://desktop.iamworkin.lan/health has failed for 3 minutes. Catalog + session launch surface offline."
+              runbook: "1. kubectl -n fc-desktop get pods -l app.kubernetes.io/name=remotedesktop-web 2. kubectl -n fc-desktop logs deploy/remotedesktop-web --tail=50 3. Check Traefik IngressRoute + step-ca cert 4. Rollout restart if pod is stuck"
+            labels:
+              severity: warning
+              service: remotedesktop
+            data:
+              - refId: A
+                relativeTimeRange: {from: 180, to: 0}
+                datasourceUid: prometheus
+                model: {expr: 'probe_success{job="probe-remotedesktop"}', instant: true, refId: A}
+              - refId: B
+                relativeTimeRange: {from: 180, to: 0}
+                datasourceUid: __expr__
+                model: {type: reduce, expression: A, reducer: last, refId: B}
+              - refId: C
+                relativeTimeRange: {from: 180, to: 0}
+                datasourceUid: __expr__
+                model: {type: threshold, expression: B, conditions: [{evaluator: {params: [1], type: lt}}], refId: C}
+
+          - uid: remotedesktop-metrics-stale
+            title: RemoteDesktop metrics stale
+            condition: C
+            for: 10m
+            noDataState: Alerting
+            execErrState: OK
+            annotations:
+              summary: RemoteDesktop /metrics returning no series
+              description: "No fc_desktop_session_events_total series for 10 minutes. Either the Prometheus scrape is misconfigured or the web deployment stopped exporting metrics. Cross-checked by Zabbix template's identical 10m no-data trigger."
+              runbook: "1. curl -sk https://desktop.iamworkin.lan/metrics | head 2. kubectl -n monitoring exec deploy/prometheus -- wget -qO- localhost:9090/api/v1/targets?scrapePool=fc-remotedesktop 3. Check monitoring-netpol egress allows to fc-desktop:8080"
+            labels:
+              severity: warning
+              service: remotedesktop
+            data:
+              - refId: A
+                relativeTimeRange: {from: 600, to: 0}
+                datasourceUid: prometheus
+                model: {expr: 'count(fc_desktop_session_events_total) or vector(0)', instant: true, refId: A}
+              - refId: B
+                relativeTimeRange: {from: 600, to: 0}
+                datasourceUid: __expr__
+                model: {type: reduce, expression: A, reducer: last, refId: B}
+              - refId: C
+                relativeTimeRange: {from: 600, to: 0}
+                datasourceUid: __expr__
+                model: {type: threshold, expression: B, conditions: [{evaluator: {params: [1], type: lt}}], refId: C}
+
+          - uid: remotedesktop-pool-depleted
+            title: RemoteDesktop pool depleted
+            condition: C
+            for: 5m
+            noDataState: OK
+            execErrState: OK
+            annotations:
+              summary: RemoteDesktop warm pool depleted for 5m
+              description: "A RemoteDesktop warm pool has fc_desktop_pool_depleted=1 for 5 minutes. New launches will cold-start. Check pod scheduling, image pull, node capacity."
+              runbook: "1. kubectl -n fc-desktop get pods -l app.kubernetes.io/name=remote-desktop --sort-by=.status.startTime 2. kubectl -n fc-desktop describe desktoppool <name> 3. Verify localhost/fc-desktop:* images imported on all 3 RKE2 nodes"
+            labels:
+              severity: warning
+              service: remotedesktop
+            data:
+              - refId: A
+                relativeTimeRange: {from: 300, to: 0}
+                datasourceUid: prometheus
+                model: {expr: 'max(fc_desktop_pool_depleted)', instant: true, refId: A}
+              - refId: B
+                relativeTimeRange: {from: 300, to: 0}
+                datasourceUid: __expr__
+                model: {type: reduce, expression: A, reducer: last, refId: B}
+              - refId: C
+                relativeTimeRange: {from: 300, to: 0}
+                datasourceUid: __expr__
+                model: {type: threshold, expression: B, conditions: [{evaluator: {params: [0.5], type: gt}}], refId: C}
+
+          - uid: remotedesktop-pool-deficit-sustained
+            title: RemoteDesktop pool below desired
+            condition: C
+            for: 10m
+            noDataState: OK
+            execErrState: OK
+            annotations:
+              summary: RemoteDesktop pool sustained deficit
+              description: "A pool has fc_desktop_pool_deficit>0 for 10 minutes. Operator is reconciling but can't reach desired size — likely image pull, NFS affinity, or claim-init issue."
+              runbook: "1. kubectl -n fc-desktop get pods -l flowercore.io/pool=<pool> 2. kubectl logs -n fc-desktop deploy/remotedesktop-operator 3. Check claim-init hook env on template"
+            labels:
+              severity: info
+              service: remotedesktop
+            data:
+              - refId: A
+                relativeTimeRange: {from: 600, to: 0}
+                datasourceUid: prometheus
+                model: {expr: 'max(fc_desktop_pool_deficit)', instant: true, refId: A}
+              - refId: B
+                relativeTimeRange: {from: 600, to: 0}
+                datasourceUid: __expr__
+                model: {type: reduce, expression: A, reducer: last, refId: B}
+              - refId: C
+                relativeTimeRange: {from: 600, to: 0}
+                datasourceUid: __expr__
+                model: {type: threshold, expression: B, conditions: [{evaluator: {params: [0], type: gt}}], refId: C}
+
+          - uid: remotedesktop-session-churn-spike
+            title: RemoteDesktop launch rate spike
+            condition: C
+            for: 5m
+            noDataState: OK
+            execErrState: OK
+            annotations:
+              summary: RemoteDesktop launch rate exceeds 20/min
+              description: "Launch events >20/min for 5 minutes. Could be a user-facing feature launch, pooled template thrashing, or runaway automation loop."
+              runbook: "1. kubectl -n fc-desktop get pods -l app.kubernetes.io/name=remote-desktop -o wide | wc -l 2. curl -sk https://desktop.iamworkin.lan/api/sessions/active 3. Check operator logs for reconcile loops"
+            labels:
+              severity: info
+              service: remotedesktop
+            data:
+              - refId: A
+                relativeTimeRange: {from: 300, to: 0}
+                datasourceUid: prometheus
+                model: {expr: 'sum(rate(fc_desktop_session_events_total{event="launch"}[5m])) * 60', instant: true, refId: A}
+              - refId: B
+                relativeTimeRange: {from: 300, to: 0}
+                datasourceUid: __expr__
+                model: {type: reduce, expression: A, reducer: last, refId: B}
+              - refId: C
+                relativeTimeRange: {from: 300, to: 0}
+                datasourceUid: __expr__
+                model: {type: threshold, expression: B, conditions: [{evaluator: {params: [20], type: gt}}], refId: C}
+
+          - uid: remotedesktop-tls-expiry
+            title: RemoteDesktop TLS cert expiring
+            condition: C
+            for: 6h
+            noDataState: OK
+            execErrState: OK
+            annotations:
+              summary: desktop.iamworkin.lan cert <2d to expiry
+              description: "The desktop.iamworkin.lan certificate is inside the 2-day renewal window and cert-manager has not renewed. Check cert-manager logs, step-ca reachability, FlowerCore.DNS preflight for dnsNames."
+              runbook: "1. kubectl -n fc-desktop get certificate remotedesktop-web-tls 2. kubectl -n cert-manager logs deploy/cert-manager --tail=50 3. Verify pfSense DNS override for desktop.iamworkin.lan"
+            labels:
+              severity: critical
+              service: remotedesktop
+            data:
+              - refId: A
+                relativeTimeRange: {from: 21600, to: 0}
+                datasourceUid: prometheus
+                model: {expr: '(probe_ssl_earliest_cert_expiry{job="probe-remotedesktop"} - time()) / 86400', instant: true, refId: A}
+              - refId: B
+                relativeTimeRange: {from: 21600, to: 0}
+                datasourceUid: __expr__
+                model: {type: reduce, expression: A, reducer: last, refId: B}
+              - refId: C
+                relativeTimeRange: {from: 21600, to: 0}
+                datasourceUid: __expr__
+                model: {type: threshold, expression: B, conditions: [{evaluator: {params: [2], type: lt}}], refId: C}
 
 # =============================================================================
 # Deployment: Grafana
@@ -3122,6 +3715,9 @@ spec:
             - name: dashboards-infra-overview
               mountPath: /var/lib/grafana/dashboards/infra-overview
               readOnly: true
+            - name: dashboards-remotedesktop
+              mountPath: /var/lib/grafana/dashboards/remotedesktop
+              readOnly: true
             - name: datasource-provisioning
               mountPath: /etc/grafana/provisioning/datasources
               readOnly: true
@@ -3172,6 +3768,9 @@ spec:
         - name: dashboards-infra-overview
           configMap:
             name: grafana-dashboard-infra-overview
+        - name: dashboards-remotedesktop
+          configMap:
+            name: grafana-dashboard-remotedesktop
         - name: datasource-provisioning
           configMap:
             name: grafana-datasource-provisioning
@@ -3733,6 +4332,66 @@ spec:
       ports:
         - port: 80
           protocol: TCP
+    # FlowerCore.RemoteDesktop /metrics scrape via the fc-desktop
+    # ClusterIP Service (remotedesktop-web:8080). Also covers the
+    # Traefik VIP hairpin path since after kube-proxy DNAT, the egress
+    # destination is the backend pod IP on the service port (see
+    # feedback_netpol_dnat_backend_port).
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: fc-desktop
+      ports:
+        - port: 8080
+          protocol: TCP
+    # Traefik backend ports — needed for in-cluster egress to public
+    # iamworkin.lan hostnames that CoreDNS wildcard resolves to the
+    # LoadBalancer VIP. Post-DNAT destination is a Traefik pod on 8080/8443.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+          podSelector:
+            matchLabels:
+              app.kubernetes.io/name: traefik
+      ports:
+        - port: 8080
+          protocol: TCP
+        - port: 8443
+          protocol: TCP
+    # Traefik /metrics endpoint (port 9100) — separate from the data-path
+    # ports above. Required for the in-cluster `traefik` scrape job.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: traefik-system
+      ports:
+        - port: 9100
+          protocol: TCP
+    # kube-state-metrics — required for kubernetes-state alert group.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: kube-system
+      ports:
+        - port: 8080
+          protocol: TCP
+    # cert-manager metrics — required for CertManagerCertificate* alerts.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: cert-manager
+      ports:
+        - port: 9402
+          protocol: TCP
+    # Longhorn manager metrics — required for Longhorn* alerts.
+    - to:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: longhorn-system
+      ports:
+        - port: 9500
+          protocol: TCP
     # IRC (irc-notify → UnrealIRCd in irc namespace via K8s DNS)
     - to:
         - namespaceSelector:

apps/multus/multus.yaml

@@ -0,0 +1,286 @@
+# =============================================================================
+# Multus CNI — Meta-CNI for multi-network attachment to pods/VMs
+# =============================================================================
+# Purpose: enable KubeVirt VMs (and any future workload) to attach additional
+# network interfaces beyond the default Calico-managed pod network. Required
+# for ci1 (Windows Server 2025 KubeVirt VM) to bridge onto PROD VLAN 57.
+#
+# Source: upstream k8snetworkplumbingwg/multus-cni v4.2.2
+#   https://github.com/k8snetworkplumbingwg/multus-cni/blob/v4.2.2/deployments/multus-daemonset-thick.yml
+#
+# Inlined verbatim (with project header + version pin annotation) for
+# reproducibility and air-gap safety. Bumping versions = edit this file +
+# git push. ArgoCD picks up via the bluejay-infra ApplicationSet
+# (apps/* directory generator on main).
+#
+# Why thick plugin (not thin):
+#   - Thick = daemon + thin shim binary; daemon handles NAD watch + CRD reads
+#     centrally so each pod's CNI ADD doesn't hit the K8s API server. Better
+#     for clusters with many NAD-using pods.
+#   - Thin = each CNI ADD process directly contacts K8s API. Simpler but
+#     scales worse and has more failure modes.
+#   - KubeVirt + multi-VM workload pattern fits thick perfectly.
+#
+# Cluster context (verified 2026-05-08):
+#   - RKE2 v1.34.5 on 3 nodes (rke2-server, rke2-agent1, rke2-agent2)
+#   - Calico CNI (Tigera-managed) at /etc/cni/net.d + /opt/cni/bin (default)
+#   - openSUSE Leap 16, kernel 6.12, containerd 2.1.5
+#   - host bridge for PROD VLAN 57 = `br-prod` (PUPPET HOST WORK — see Phase 1.5
+#     in docs/infrastructure/windows-server-build-runner-plan.md)
+#
+# Version pin: snapshot-thick → pinning to v4.2.2 release tag at deploy time
+# would require a private mirror of the image. Upstream `snapshot-thick` tag
+# is updated on every release, so for now we trust upstream + Calico's
+# established pattern. Pin to a specific SHA256 once we mirror to Gitea OCI.
+#
+# Apply (once committed to bluejay-infra main, ApplicationSet auto-syncs):
+#   git add apps/multus/multus.yaml && git commit && git push origin main
+#   # ArgoCD `infra-multus` Application appears within 3 min via ApplicationSet
+#
+# Verify:
+#   kubectl -n kube-system get ds kube-multus-ds
+#   kubectl -n kube-system rollout status ds kube-multus-ds
+#   kubectl get crd network-attachment-definitions.k8s.cni.cncf.io
+# =============================================================================
+
+---
+apiVersion: apiextensions.k8s.io/v1
+kind: CustomResourceDefinition
+metadata:
+  name: network-attachment-definitions.k8s.cni.cncf.io
+  annotations:
+    bluejay.iamworkin.lan/source: "k8snetworkplumbingwg/multus-cni v4.2.2"
+spec:
+  group: k8s.cni.cncf.io
+  scope: Namespaced
+  names:
+    plural: network-attachment-definitions
+    singular: network-attachment-definition
+    kind: NetworkAttachmentDefinition
+    shortNames:
+      - net-attach-def
+  versions:
+    - name: v1
+      served: true
+      storage: true
+      schema:
+        openAPIV3Schema:
+          description: 'NetworkAttachmentDefinition is a CRD schema specified by the Network Plumbing
+            Working Group to express the intent for attaching pods to one or more logical or physical
+            networks. More information available at: https://github.com/k8snetworkplumbingwg/multi-net-spec'
+          type: object
+          properties:
+            apiVersion:
+              type: string
+            kind:
+              type: string
+            metadata:
+              type: object
+            spec:
+              description: 'NetworkAttachmentDefinition spec defines the desired state of a network attachment'
+              type: object
+              properties:
+                config:
+                  description: 'NetworkAttachmentDefinition config is a JSON-formatted CNI configuration'
+                  type: string
+---
+kind: ClusterRole
+apiVersion: rbac.authorization.k8s.io/v1
+metadata:
+  name: multus
+rules:
+  - apiGroups: ["k8s.cni.cncf.io"]
+    resources:
+      - '*'
+    verbs:
+      - '*'
+  - apiGroups:
+      - ""
+    resources:
+      - pods
+      - pods/status
+    verbs:
+      - get
+      - list
+      - update
+      - watch
+  - apiGroups:
+      - ""
+      - events.k8s.io
+    resources:
+      - events
+    verbs:
+      - create
+      - patch
+      - update
+---
+kind: ClusterRoleBinding
+apiVersion: rbac.authorization.k8s.io/v1
+metadata:
+  name: multus
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: multus
+subjects:
+  - kind: ServiceAccount
+    name: multus
+    namespace: kube-system
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: multus
+  namespace: kube-system
+---
+kind: ConfigMap
+apiVersion: v1
+metadata:
+  name: multus-daemon-config
+  namespace: kube-system
+  labels:
+    tier: node
+    app: multus
+data:
+  daemon-config.json: |
+    {
+        "chrootDir": "/hostroot",
+        "cniVersion": "0.3.1",
+        "logLevel": "verbose",
+        "logToStderr": true,
+        "cniConfigDir": "/host/etc/cni/net.d",
+        "multusAutoconfigDir": "/host/etc/cni/net.d",
+        "multusConfigFile": "auto",
+        "socketDir": "/host/run/multus/"
+    }
+---
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: kube-multus-ds
+  namespace: kube-system
+  labels:
+    tier: node
+    app: multus
+    name: multus
+spec:
+  selector:
+    matchLabels:
+      name: multus
+  updateStrategy:
+    type: RollingUpdate
+  template:
+    metadata:
+      labels:
+        tier: node
+        app: multus
+        name: multus
+    spec:
+      hostNetwork: true
+      hostPID: true
+      tolerations:
+        - operator: Exists
+          effect: NoSchedule
+        - operator: Exists
+          effect: NoExecute
+      serviceAccountName: multus
+      containers:
+        - name: kube-multus
+          image: ghcr.io/k8snetworkplumbingwg/multus-cni:snapshot-thick
+          command: [ "/usr/src/multus-cni/bin/multus-daemon" ]
+          resources:
+            requests:
+              cpu: "100m"
+              memory: "50Mi"
+            limits:
+              cpu: "100m"
+              memory: "50Mi"
+          securityContext:
+            privileged: true
+          terminationMessagePolicy: FallbackToLogsOnError
+          volumeMounts:
+            - name: cni
+              mountPath: /host/etc/cni/net.d
+            # multus-daemon expects that cnibin path must be identical between pod and container host.
+            # e.g. if the cni bin is in '/opt/cni/bin' on the container host side, then it should be mount to '/opt/cni/bin' in multus-daemon,
+            # not to any other directory, like '/opt/bin' or '/usr/bin'.
+            - name: cnibin
+              mountPath: /opt/cni/bin
+            - name: host-run
+              mountPath: /host/run
+            - name: host-var-lib-cni-multus
+              mountPath: /var/lib/cni/multus
+            - name: host-var-lib-kubelet
+              mountPath: /var/lib/kubelet
+              mountPropagation: HostToContainer
+            - name: host-run-k8s-cni-cncf-io
+              mountPath: /run/k8s.cni.cncf.io
+            - name: host-run-netns
+              mountPath: /run/netns
+              mountPropagation: HostToContainer
+            - name: multus-daemon-config
+              mountPath: /etc/cni/net.d/multus.d
+              readOnly: true
+            - name: hostroot
+              mountPath: /hostroot
+              mountPropagation: HostToContainer
+            - mountPath: /etc/cni/multus/net.d
+              name: multus-conf-dir
+          env:
+            - name: MULTUS_NODE_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: spec.nodeName
+      initContainers:
+        - name: install-multus-binary
+          image: ghcr.io/k8snetworkplumbingwg/multus-cni:snapshot-thick
+          command:
+            - "sh"
+            - "-c"
+            - "cp /usr/src/multus-cni/bin/multus-shim /host/opt/cni/bin/multus-shim && cp /usr/src/multus-cni/bin/passthru /host/opt/cni/bin/passthru"
+          resources:
+            requests:
+              cpu: "10m"
+              memory: "15Mi"
+          securityContext:
+            privileged: true
+          terminationMessagePolicy: FallbackToLogsOnError
+          volumeMounts:
+            - name: cnibin
+              mountPath: /host/opt/cni/bin
+              mountPropagation: Bidirectional
+      terminationGracePeriodSeconds: 10
+      volumes:
+        - name: cni
+          hostPath:
+            path: /etc/cni/net.d
+        - name: cnibin
+          hostPath:
+            path: /opt/cni/bin
+        - name: hostroot
+          hostPath:
+            path: /
+        - name: multus-daemon-config
+          configMap:
+            name: multus-daemon-config
+            items:
+            - key: daemon-config.json
+              path: daemon-config.json
+        - name: host-run
+          hostPath:
+            path: /run
+        - name: host-var-lib-cni-multus
+          hostPath:
+            path: /var/lib/cni/multus
+        - name: host-var-lib-kubelet
+          hostPath:
+            path: /var/lib/kubelet
+        - name: host-run-k8s-cni-cncf-io
+          hostPath:
+            path: /run/k8s.cni.cncf.io
+        - name: host-run-netns
+          hostPath:
+            path: /run/netns/
+        - name: multus-conf-dir
+          hostPath:
+            path: /etc/cni/multus/net.d

apps/noc-services/noc-services.yaml

@@ -219,6 +219,65 @@ spec:
   tls:
     secretName: cockpit-tls
 ---
+# ============================================================
+# PuppetDB Dashboard - noc1:8080 (HTTP, web UI only)
+# Agent-to-PuppetDB mTLS still uses port 8081 directly via Puppet CA
+# (NOT via this proxy). See docs/infrastructure/cert-recovery-2026-04-28.md
+# ============================================================
+apiVersion: v1
+kind: Service
+metadata:
+  name: puppetdb-external
+  namespace: noc-proxy
+spec:
+  ports:
+    - port: 8080
+      targetPort: 8080
+      name: http
+  clusterIP: None
+---
+apiVersion: v1
+kind: Endpoints
+metadata:
+  name: puppetdb-external
+  namespace: noc-proxy
+subsets:
+  - addresses:
+      - ip: 10.0.56.10
+    ports:
+      - port: 8080
+        name: http
+---
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: puppetdb-tls
+  namespace: noc-proxy
+spec:
+  secretName: puppetdb-tls
+  issuerRef:
+    name: step-ca-acme
+    kind: ClusterIssuer
+  dnsNames:
+    - puppetdb.iamworkin.lan
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: puppetdb
+  namespace: noc-proxy
+spec:
+  entryPoints:
+    - websecure
+  routes:
+    - kind: Rule
+      match: Host(`puppetdb.iamworkin.lan`)
+      services:
+        - name: puppetdb-external
+          port: 8080
+  tls:
+    secretName: puppetdb-tls
+---
 # NetworkPolicy: allow Traefik ingress, allow egress to noc1
 apiVersion: networking.k8s.io/v1
 kind: NetworkPolicy
@@ -242,6 +301,8 @@ spec:
       ports:
         - port: 3000
           protocol: TCP
+        - port: 8080
+          protocol: TCP
         - port: 9090
           protocol: TCP
         - port: 9091

apps/selenium/network-policy.yaml

@@ -0,0 +1,210 @@
+# Selenium Grid NetworkPolicy.
+#
+# Captured into bluejay-infra 2026-05-07 during the regroup audit. This
+# NetworkPolicy was previously applied via `kubectl apply` directly to
+# the cluster with no source-of-truth anywhere — a fresh cluster rebuild
+# would have lost all of it (including the Selenium Grid → Traefik VIP
+# allow rule for AAT runs against `*.iamworkin.lan` services).
+#
+# The Selenium Grid Deployment + Services themselves are still managed
+# outside ArgoCD (deployed via raw kubectl from the original Selenium
+# Grid bring-up). Migrating those into bluejay-infra is a separate lane —
+# this commit only restores GitOps repeatability for the NetworkPolicy.
+#
+# Rules captured from the live cluster's `kubectl get netpol -n selenium
+# selenium-netpol -o yaml` on 2026-05-07. Originally applied 2026-03-15
+# (from `metadata.creationTimestamp` before the field was stripped).
+#
+# Allows:
+#   - Egress: CoreDNS, intra-namespace pod-to-pod (4442/4443/4444/5555),
+#     Traefik VIP for `*.iamworkin.lan` AAT runs, all FC namespaces on
+#     standard FC service ports (5100/5200/5300/5400/8080), pod CIDR
+#     (10.42.0.0/16) + service CIDR (10.43.0.0/16) for the same ports,
+#     LAN gateway range (10.0.56.0/24) for HTTPS, edge2 CUPS print
+#     (10.0.57.16:5200), public internet 80/443 (excluding RFC1918), and
+#     fc-signage:5190 for the signage AAT lane.
+#   - Ingress: Traefik (4444 + 8089 ACME-solver-style), intra-pod,
+#     telephony / gitea / fc-system / fc-signage namespaces on 4444.
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: selenium-netpol
+  namespace: selenium
+  labels:
+    app.kubernetes.io/part-of: selenium
+    app.kubernetes.io/component: isolation
+spec:
+  egress:
+  - ports:
+    - port: 53
+      protocol: UDP
+    - port: 53
+      protocol: TCP
+    to:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: kube-system
+  - ports:
+    - port: 4442
+      protocol: TCP
+    - port: 4443
+      protocol: TCP
+    - port: 4444
+      protocol: TCP
+    - port: 5555
+      protocol: TCP
+    to:
+    - podSelector: {}
+  - ports:
+    - port: 443
+      protocol: TCP
+    - port: 80
+      protocol: TCP
+    to:
+    - ipBlock:
+        cidr: 10.0.56.200/32
+  - ports:
+    - port: 443
+      protocol: TCP
+    - port: 80
+      protocol: TCP
+    - port: 5200
+      protocol: TCP
+    - port: 5300
+      protocol: TCP
+    - port: 5400
+      protocol: TCP
+    - port: 5100
+      protocol: TCP
+    - port: 8080
+      protocol: TCP
+    to:
+    - namespaceSelector: {}
+  - ports:
+    - port: 443
+      protocol: TCP
+    - port: 80
+      protocol: TCP
+    - port: 8443
+      protocol: TCP
+    - port: 8080
+      protocol: TCP
+    - port: 5200
+      protocol: TCP
+    - port: 5300
+      protocol: TCP
+    - port: 5400
+      protocol: TCP
+    - port: 5100
+      protocol: TCP
+    to:
+    - ipBlock:
+        cidr: 10.43.0.0/16
+  - ports:
+    - port: 443
+      protocol: TCP
+    - port: 80
+      protocol: TCP
+    - port: 8443
+      protocol: TCP
+    - port: 8080
+      protocol: TCP
+    - port: 5200
+      protocol: TCP
+    - port: 5300
+      protocol: TCP
+    - port: 5400
+      protocol: TCP
+    - port: 5100
+      protocol: TCP
+    to:
+    - ipBlock:
+        cidr: 10.42.0.0/16
+  - ports:
+    - port: 443
+      protocol: TCP
+    - port: 80
+      protocol: TCP
+    - port: 8443
+      protocol: TCP
+    to:
+    - ipBlock:
+        cidr: 10.0.56.0/24
+  - ports:
+    - port: 5200
+      protocol: TCP
+    to:
+    - ipBlock:
+        cidr: 10.0.57.16/32
+  - ports:
+    - port: 80
+      protocol: TCP
+    - port: 443
+      protocol: TCP
+    to:
+    - ipBlock:
+        cidr: 0.0.0.0/0
+        except:
+        - 172.16.0.0/12
+        - 192.168.0.0/16
+  - ports:
+    - port: 5190
+      protocol: TCP
+    to:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: fc-signage
+  ingress:
+  - from:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: traefik-system
+    ports:
+    - port: 4444
+      protocol: TCP
+    - port: 8089
+      protocol: TCP
+  - from:
+    - podSelector: {}
+    ports:
+    - port: 4442
+      protocol: TCP
+    - port: 4443
+      protocol: TCP
+    - port: 4444
+      protocol: TCP
+    - port: 5555
+      protocol: TCP
+  - from:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: telephony
+    ports:
+    - port: 4444
+      protocol: TCP
+  - from:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: gitea
+    ports:
+    - port: 4444
+      protocol: TCP
+  - from:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: fc-system
+    ports:
+    - port: 4444
+      protocol: TCP
+  - from:
+    - namespaceSelector:
+        matchLabels:
+          kubernetes.io/metadata.name: fc-signage
+    ports:
+    - port: 4444
+      protocol: TCP
+  podSelector: {}
+  policyTypes:
+  - Ingress
+  - Egress
+

apps/telephony/telephony.yaml

@@ -148,7 +148,7 @@ spec:
               topologyKey: kubernetes.io/hostname
       containers:
         - name: telephony-web
-          image: localhost/fc-telephony-web:v202604170153
+          image: localhost/fc-telephony-web:v202604252156
           imagePullPolicy: Never
           securityContext:
             readOnlyRootFilesystem: true

apps/worldbuilder/README.md

@@ -0,0 +1,60 @@
+# FlowerCore.WorldBuilder
+
+ArgoCD-managed manifest for FlowerCore.WorldBuilder.Web — comic / storyboard
+authoring service that drives ComfyUI for panel image generation and
+QuestPDF for letter / A4 export.
+
+Source: `D:\git\FlowerCore\FlowerCore.WorldBuilder` (master)
+
+## Deployment order
+
+1. **DNS preflight** — `worldbuilder.iamworkin.lan -> 10.0.56.200` MUST exist
+   in pfSense Unbound before this manifest is applied, or cert-manager
+   HTTP-01 silently exponential-backs-off ~2h.
+   Memory: `feedback_pfsense_dns_required_for_acme`.
+2. **Image import to ALL RKE2 nodes** — pod can schedule to any of
+   `rke2-server` (10.0.56.11), `rke2-agent1` (10.0.56.12),
+   `rke2-agent2` (10.0.56.13). Build with:
+   ```bash
+   bash deploy/build.sh   # in FlowerCore.WorldBuilder repo
+   podman save localhost/fc-worldbuilder:v<TAG> -o /tmp/fc-worldbuilder-v<TAG>.tar
+   for h in 10.0.56.11 10.0.56.12 10.0.56.13; do
+     scp /tmp/fc-worldbuilder-v<TAG>.tar fcadmin@$h:/tmp/
+     ssh fcadmin@$h \
+       "sudo /var/lib/rancher/rke2/bin/ctr -a /run/k3s/containerd/containerd.sock \
+         -n k8s.io images import /tmp/fc-worldbuilder-v<TAG>.tar"
+   done
+   ```
+   Memory: `feedback_rke2_image_import_per_node_scp`.
+3. **Bump image tag** in `worldbuilder.yaml` and git push.
+   ArgoCD ApplicationSet picks up within ~3 minutes.
+4. **First production render** — open `https://worldbuilder.iamworkin.lan`,
+   create World → Character → Storyboard → ExportJob, confirm artifact
+   downloads. ComfyUI lives on BLUEJAY-WS at `http://10.0.56.20:8188`.
+
+## Health probes
+
+- `startupProbe` + `readinessProbe`: `httpGet /healthz` (registered explicitly
+  in Program.cs — anonymous, no DB or OpenAPI dependency).
+- `livenessProbe`: `tcpSocket` as a cheap fallback.
+  Memory: `feedback_k8s_probes_must_not_hit_openapi`,
+  `feedback_k8s_probes_behind_auth_middleware`.
+
+## Storage
+
+- Longhorn RWO PVC `worldbuilder-data` (5Gi) mounted at `/data`. SQLite DB
+  lives at `/data/worldbuilder.db`, generated images under `/data/gallery/`,
+  PDF/PNG exports under `/data/exports/`.
+- DataProtection keys persist to the same SQLite via
+  `AddFlowerCoreDataProtection<WorldBuilderDbContext>` — explicit migration
+  `20260429133417_Initial` already creates `fc_dp_keys`.
+  Memory: `feedback_dataprotection_keys_persist_to_app_dbcontext`,
+  `feedback_intranet_dataprotection_table_must_have_explicit_migration`.
+
+## Image generation backend
+
+`FlowerCore:WorldBuilder:ImageGeneration:BaseUrl=http://10.0.56.20:8188` —
+ComfyUI runs on BLUEJAY-WS Windows (R9700 / gfx1201 / ROCm 7.2.1). Pod reaches
+the workstation directly across the 10.0.56.0/24 VLAN (no Podman-style host-
+filter issues — K8s pods route via Calico, which is L3-routed across the
+VLAN).

apps/worldbuilder/worldbuilder.yaml

@@ -0,0 +1,213 @@
+# FlowerCore.WorldBuilder — comic / storyboard authoring service.
+#
+# Deployment + Service + PVC + Certificate + IngressRoute. ArgoCD-managed
+# end-to-end. See apps/worldbuilder/README.md for the per-deploy runbook.
+#
+# Image build (BLUEJAY-WS):
+#   bash deploy/build.sh                 # in FlowerCore.WorldBuilder repo
+#   podman save localhost/fc-worldbuilder:v<TAG> -o /tmp/fc-worldbuilder-v<TAG>.tar
+#   for h in 10.0.56.11 10.0.56.12 10.0.56.13; do
+#     scp /tmp/fc-worldbuilder-v<TAG>.tar fcadmin@$h:/tmp/
+#     ssh fcadmin@$h "sudo /var/lib/rancher/rke2/bin/ctr -a /run/k3s/containerd/containerd.sock -n k8s.io images import /tmp/fc-worldbuilder-v<TAG>.tar"
+#   done
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: fc-worldbuilder
+  labels:
+    app.kubernetes.io/part-of: flowercore
+---
+# SQLite DB + generated image gallery + PDF/PNG exports.
+# Longhorn RWO — single replica with `Recreate` rollout strategy keeps it safe.
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: worldbuilder-data
+  namespace: fc-worldbuilder
+spec:
+  accessModes:
+    - ReadWriteOnce
+  storageClassName: longhorn
+  resources:
+    requests:
+      storage: 5Gi
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: worldbuilder-web
+  namespace: fc-worldbuilder
+  labels:
+    app.kubernetes.io/name: worldbuilder-web
+    app.kubernetes.io/part-of: flowercore
+spec:
+  replicas: 1
+  revisionHistoryLimit: 3
+  strategy:
+    # RWO PVC + single replica. Recreate avoids multi-attach overlap.
+    type: Recreate
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: worldbuilder-web
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: worldbuilder-web
+        app.kubernetes.io/part-of: flowercore
+      annotations:
+        prometheus.io/scrape: "true"
+        prometheus.io/port: "8080"
+        prometheus.io/path: "/metrics/prometheus"
+    spec:
+      securityContext:
+        fsGroup: 1654
+        fsGroupChangePolicy: OnRootMismatch
+      containers:
+        - name: web
+          # Bump tag for each rebuild. Initial deploy: v202605062048
+          image: localhost/fc-worldbuilder:v202605062048
+          imagePullPolicy: Never
+          ports:
+            - containerPort: 8080
+              name: http
+          env:
+            - name: ASPNETCORE_URLS
+              value: "http://+:8080"
+            - name: ASPNETCORE_ENVIRONMENT
+              value: "Production"
+            - name: DOTNET_RUNNING_IN_CONTAINER
+              value: "true"
+            - name: DOTNET_SYSTEM_GLOBALIZATION_INVARIANT
+              value: "false"
+            # SQLite path overrides (default appsettings uses relative paths).
+            - name: ConnectionStrings__DefaultConnection
+              value: "Data Source=/data/worldbuilder.db"
+            - name: FlowerCore__Database__Provider
+              value: "Sqlite"
+            - name: FlowerCore__Database__ConnectionStrings__Sqlite
+              value: "Data Source=/data/worldbuilder.db"
+            # Generated image gallery + exports persist on /data.
+            - name: FlowerCore__WorldBuilder__ImageStore__RootPath
+              value: "/data/gallery"
+            - name: FlowerCore__WorldBuilder__Export__RootPath
+              value: "/data/exports"
+            # ComfyUI on BLUEJAY-WS (R9700 / gfx1201 / ROCm 7.2.1).
+            - name: FlowerCore__WorldBuilder__ImageGeneration__BaseUrl
+              value: "http://10.0.56.20:8188"
+            - name: FlowerCore__WorldBuilder__ImageGeneration__ClientMode
+              value: "comfyui"
+          resources:
+            # Cluster CPU-request budget runs hot (99% on all 3 nodes at deploy
+            # time) while actual CPU usage is well below capacity. Idle Blazor
+            # Server + SignalR + a single ComfyUI poller uses ~5m, so 25m is
+            # generous. Re-evaluate if active rendering/export workers ever
+            # push past the limit.
+            requests:
+              cpu: 25m
+              memory: 256Mi
+            limits:
+              cpu: 1000m
+              memory: 768Mi
+          # /healthz is registered explicitly in Program.cs (anonymous, no DB
+          # or OpenAPI dependency). Liveness uses tcpSocket as a cheap fallback
+          # in case future middleware changes accidentally gate /healthz.
+          # Memory: feedback_k8s_probes_must_not_hit_openapi,
+          #         feedback_k8s_probes_behind_auth_middleware.
+          startupProbe:
+            httpGet:
+              path: /healthz
+              port: 8080
+            initialDelaySeconds: 5
+            periodSeconds: 5
+            failureThreshold: 30
+          readinessProbe:
+            httpGet:
+              path: /healthz
+              port: 8080
+            periodSeconds: 10
+            failureThreshold: 3
+          livenessProbe:
+            tcpSocket:
+              port: 8080
+            initialDelaySeconds: 30
+            periodSeconds: 30
+            failureThreshold: 3
+          securityContext:
+            runAsNonRoot: true
+            runAsUser: 1654
+            runAsGroup: 1654
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            capabilities:
+              drop:
+                - ALL
+          volumeMounts:
+            - name: data
+              mountPath: /data
+            - name: tmp
+              mountPath: /tmp
+            - name: logs
+              mountPath: /app/logs
+      volumes:
+        - name: data
+          persistentVolumeClaim:
+            claimName: worldbuilder-data
+        - name: tmp
+          emptyDir: {}
+        - name: logs
+          emptyDir: {}
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: worldbuilder-web
+  namespace: fc-worldbuilder
+  labels:
+    app.kubernetes.io/name: worldbuilder-web
+    app.kubernetes.io/part-of: flowercore
+spec:
+  type: ClusterIP
+  selector:
+    app.kubernetes.io/name: worldbuilder-web
+  ports:
+    - name: http
+      port: 80
+      targetPort: 8080
+---
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: worldbuilder-web-tls
+  namespace: fc-worldbuilder
+spec:
+  secretName: worldbuilder-web-tls
+  issuerRef:
+    name: step-ca-acme
+    kind: ClusterIssuer
+  dnsNames:
+    - worldbuilder.iamworkin.lan
+  # step-ca ACME provisioner caps lifetime at 30d. Requesting 90d
+  # silently capped to 30d, making renewBefore 720h (30d) equal to the
+  # actual cert lifetime — triggered a perpetual renewal loop that
+  # generated 2365+ CertificateRequest objects in 18h. Match the working
+  # 720h/240h pattern used by every other FC service cert.
+  duration: 720h     # 30d (step-ca cap)
+  renewBefore: 240h  # 10d
+---
+apiVersion: traefik.io/v1alpha1
+kind: IngressRoute
+metadata:
+  name: worldbuilder-web
+  namespace: fc-worldbuilder
+spec:
+  entryPoints:
+    - websecure
+  routes:
+    - match: Host(`worldbuilder.iamworkin.lan`)
+      kind: Rule
+      services:
+        - name: worldbuilder-web
+          port: 80
+  tls:
+    secretName: worldbuilder-web-tls

tests/bluejay-infra-lint/BluejayInfraLint.Tests.csproj

@@ -0,0 +1,24 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <TargetFramework>net10.0</TargetFramework>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+    <IsPackable>false</IsPackable>
+    <TreatWarningsAsErrors>true</TreatWarningsAsErrors>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="coverlet.collector" Version="6.0.2">
+      <PrivateAssets>all</PrivateAssets>
+      <IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
+    </PackageReference>
+    <PackageReference Include="FluentAssertions" Version="6.12.1" />
+    <PackageReference Include="Microsoft.NET.Test.Sdk" Version="17.12.0" />
+    <PackageReference Include="xunit" Version="2.9.2" />
+    <PackageReference Include="xunit.runner.visualstudio" Version="2.8.2">
+      <PrivateAssets>all</PrivateAssets>
+      <IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
+    </PackageReference>
+    <PackageReference Include="YamlDotNet" Version="16.2.0" />
+  </ItemGroup>
+</Project>

tests/bluejay-infra-lint/FleetManifestLintTests.cs

@@ -0,0 +1,639 @@
+using FluentAssertions;
+using System.Text.RegularExpressions;
+using Xunit;
+using YamlDotNet.Core;
+using YamlDotNet.RepresentationModel;
+
+namespace BluejayInfraLint.Tests;
+
+[Trait("Category", "Unit")]
+public sealed class FleetManifestLintTests
+{
+    private static readonly ManifestInventory Inventory = ManifestInventory.Load();
+
+    private static readonly HashSet<string> PublicReadOnlyHosts = new(StringComparer.Ordinal)
+    {
+        "dist.flowercore.io",
+        "dns.iamworkin.lan",
+    };
+
+    // Public hosts that allow a tightly bounded write surface in addition to
+    // GET/HEAD. updatecenter.iamworkin.lan accepts POST /api/v1/checkin/{id}
+    // (bootstrap-JWT) so its allowlist is GET||HEAD||POST||OPTIONS — but
+    // PUT/PATCH/DELETE must still 404 at the route. Anything wider than this
+    // set should fail this lint.
+    //
+    // PUB-1 (2026-05-06): update.flowercore.io / updates.flowercore.io were
+    // added for the Cloudflare-proxied public Update Center edge. They use the
+    // same bounded read-write allowlist as the LAN pair.
+    private static readonly HashSet<string> PublicReadWriteAllowlistHosts = new(StringComparer.Ordinal)
+    {
+        "updatecenter.iamworkin.lan",
+        "updates.iamworkin.lan",
+        "update.flowercore.io",
+        "updates.flowercore.io",
+    };
+
+    private static readonly HashSet<string> ApiKeyProtectedDeployments = new(StringComparer.Ordinal)
+    {
+        "messageboard-web",
+        "scoreboard-web",
+        "segmentdisplay-web",
+        "signalcontrol-web",
+    };
+
+    private static readonly HashSet<string> PublicEgressDeployments = new(StringComparer.Ordinal)
+    {
+        "asterisk",
+        "fc-llm-bridge",
+        "mysql-web",
+        "php-web",
+        "ttsreader-align",
+        "ttsreader-kokoro",
+        "ttsreader-modern",
+        "ttsreader-piper",
+    };
+
+    [Fact]
+    public void IngressRoutes_MustKeepServiceReferencesInTheSameNamespace()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.Kind == "IngressRoute")
+            .SelectMany(document =>
+                document.MappingSequence("spec", "routes")
+                    .SelectMany(route =>
+                        route.MappingSequence("services")
+                            .Select(service => new
+                            {
+                                Document = document,
+                                ServiceName = ManifestNodeExtensions.Scalar(service, "name"),
+                                ServiceNamespace = ManifestNodeExtensions.Scalar(service, "namespace"),
+                            })))
+            .Where(entry => !string.IsNullOrWhiteSpace(entry.ServiceNamespace))
+            .Where(entry => !string.Equals(entry.ServiceNamespace, entry.Document.Namespace, StringComparison.Ordinal))
+            .Select(entry =>
+                $"{entry.Document.Descriptor} references Service '{entry.ServiceName}' in namespace '{entry.ServiceNamespace}'.")
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void PublicReadOnlyIngressRoutes_MustExplicitlyAllowOnlyGetAndHead()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.Kind == "IngressRoute")
+            .SelectMany(document =>
+                document.MappingSequence("spec", "routes")
+                    .Select(route => new
+                    {
+                        Document = document,
+                        Match = ManifestNodeExtensions.Scalar(route, "match") ?? string.Empty,
+                    }))
+            .Where(entry => PublicReadOnlyHosts.Any(host => entry.Match.Contains($"Host(`{host}`)", StringComparison.Ordinal)))
+            .Where(entry => !entry.Match.Contains("Method(`GET`)", StringComparison.Ordinal)
+                || !entry.Match.Contains("Method(`HEAD`)", StringComparison.Ordinal))
+            .Select(entry => $"{entry.Document.Descriptor} is missing an explicit GET/HEAD method allowlist.")
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void PublicReadWriteIngressRoutes_MustPinGetHeadPostOptionsAllowlist()
+    {
+        // For hosts in PublicReadWriteAllowlistHosts, the route match MUST
+        // contain Method(`GET`), Method(`HEAD`), Method(`POST`), and
+        // Method(`OPTIONS`) AND MUST NOT contain Method(`PUT`),
+        // Method(`PATCH`), or Method(`DELETE`). This keeps the public
+        // allowlist invariant against regression — see Track A's
+        // updatecenter-web ingressroute hardening.
+        var violations = Inventory.Documents
+            .Where(document => document.Kind == "IngressRoute")
+            .SelectMany(document =>
+                document.MappingSequence("spec", "routes")
+                    .Select(route => new
+                    {
+                        Document = document,
+                        Match = ManifestNodeExtensions.Scalar(route, "match") ?? string.Empty,
+                    }))
+            .Where(entry => PublicReadWriteAllowlistHosts.Any(host => entry.Match.Contains($"Host(`{host}`)", StringComparison.Ordinal)))
+            .SelectMany(entry =>
+            {
+                var localViolations = new List<string>();
+
+                foreach (var required in new[] { "GET", "HEAD", "POST", "OPTIONS" })
+                {
+                    if (!entry.Match.Contains($"Method(`{required}`)", StringComparison.Ordinal))
+                    {
+                        localViolations.Add($"{entry.Document.Descriptor} is missing required Method(`{required}`).");
+                    }
+                }
+
+                foreach (var forbidden in new[] { "PUT", "PATCH", "DELETE" })
+                {
+                    if (entry.Match.Contains($"Method(`{forbidden}`)", StringComparison.Ordinal))
+                    {
+                        localViolations.Add($"{entry.Document.Descriptor} must not include Method(`{forbidden}`) on a public host.");
+                    }
+                }
+
+                return localViolations;
+            })
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void TraefikVipNetworkPolicies_MustAllowPostDnatBackendPorts()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.Kind == "NetworkPolicy")
+            .Where(document => document.AllScalars().Any(value => value.Contains("10.0.56.200", StringComparison.Ordinal)))
+            .SelectMany(document =>
+            {
+                var ports = document.EgressPorts().ToHashSet(StringComparer.Ordinal);
+                var localViolations = new List<string>();
+
+                if (ports.Contains("443") && !ports.Contains("8443"))
+                {
+                    localViolations.Add($"{document.Descriptor} allows Traefik VIP 443 without backend port 8443.");
+                }
+
+                if (ports.Contains("80") && !ports.Contains("8000") && !ports.Contains("8080"))
+                {
+                    localViolations.Add($"{document.Descriptor} allows Traefik VIP 80 without a backend HTTP port (8000/8080).");
+                }
+
+                return localViolations;
+            })
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void ApiKeyProtectedDeployments_MustUseTcpSocketHealthProbes()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.Kind == "Deployment")
+            .Where(document => ApiKeyProtectedDeployments.Contains(document.Name))
+            .SelectMany(document => document.ContainerMappings().SelectMany(container =>
+                ProbeViolations(document, container, "readinessProbe")
+                    .Concat(ProbeViolations(document, container, "livenessProbe"))))
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void StatefulSets_WithVolumeClaimTemplates_MustDeclareFilesystemDefaults()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.Kind == "StatefulSet")
+            .Where(document => document.MappingSequence("spec", "volumeClaimTemplates").Count > 0)
+            .SelectMany(document =>
+            {
+                var localViolations = new List<string>();
+
+                if (string.IsNullOrWhiteSpace(document.Scalar("spec", "podManagementPolicy")))
+                {
+                    localViolations.Add($"{document.Descriptor} is missing spec.podManagementPolicy.");
+                }
+
+                if (string.IsNullOrWhiteSpace(document.Scalar("spec", "revisionHistoryLimit")))
+                {
+                    localViolations.Add($"{document.Descriptor} is missing spec.revisionHistoryLimit.");
+                }
+
+                foreach (var claimTemplate in document.MappingSequence("spec", "volumeClaimTemplates"))
+                {
+                    if (!string.Equals(
+                            ManifestNodeExtensions.Scalar(claimTemplate, "spec", "volumeMode"),
+                            "Filesystem",
+                            StringComparison.Ordinal))
+                    {
+                        var claimName = ManifestNodeExtensions.Scalar(claimTemplate, "metadata", "name") ?? "<unnamed>";
+                        localViolations.Add($"{document.Descriptor} volumeClaimTemplate '{claimName}' is missing volumeMode: Filesystem.");
+                    }
+                }
+
+                return localViolations;
+            })
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void LocallyImportedImages_MustUseLocalhostPrefixAndNeverPullPolicy()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.PodSpec() is not null)
+            .SelectMany(document => document.ContainerSpecs()
+                .Where(container => !string.IsNullOrWhiteSpace(container.Image))
+                .Select(container => new
+                {
+                    Document = document,
+                    Container = container,
+                }))
+            .Where(entry =>
+                (entry.Container.Image.StartsWith("localhost/", StringComparison.Ordinal)
+                    && !string.Equals(entry.Container.ImagePullPolicy, "Never", StringComparison.Ordinal))
+                || (entry.Container.Image.StartsWith("fc-", StringComparison.Ordinal)
+                    && !entry.Container.Image.Contains('/', StringComparison.Ordinal)))
+            .Select(entry =>
+            {
+                if (entry.Container.Image.StartsWith("localhost/", StringComparison.Ordinal))
+                {
+                    return $"{entry.Document.Descriptor} container '{entry.Container.Name}' uses {entry.Container.Image} without imagePullPolicy: Never.";
+                }
+
+                return $"{entry.Document.Descriptor} container '{entry.Container.Name}' uses non-local image '{entry.Container.Image}' for a node-imported FlowerCore workload.";
+            })
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    [Fact]
+    public void PublicEgressDeployments_MustOptOutOfIamworkinLanSearchSuffixes()
+    {
+        var violations = Inventory.Documents
+            .Where(document => document.PodSpec() is not null)
+            .Where(document => PublicEgressDeployments.Contains(document.Name))
+            .SelectMany(document =>
+            {
+                var localViolations = new List<string>();
+                var podSpec = document.PodSpec()!;
+                var dnsPolicy = ManifestNodeExtensions.Scalar(podSpec, "dnsPolicy");
+                var searches = ManifestNodeExtensions.ScalarSequence(podSpec, "dnsConfig", "searches").ToList();
+
+                if (!string.Equals(dnsPolicy, "None", StringComparison.Ordinal))
+                {
+                    localViolations.Add($"{document.Descriptor} is missing dnsPolicy: None.");
+                }
+
+                if (searches.Count == 0)
+                {
+                    localViolations.Add($"{document.Descriptor} is missing dnsConfig.searches.");
+                }
+                else if (searches.Any(search => search.Contains("iamworkin.lan", StringComparison.OrdinalIgnoreCase)))
+                {
+                    localViolations.Add($"{document.Descriptor} still includes iamworkin.lan in dnsConfig.searches.");
+                }
+
+                return localViolations;
+            })
+            .ToList();
+
+        violations.Should().BeEmpty();
+    }
+
+    private static IEnumerable<string> ProbeViolations(
+        ManifestDocument document,
+        YamlMappingNode container,
+        string probeKey)
+    {
+        if (!ManifestNodeExtensions.TryGetMapping(container, probeKey, out var probe)
+            || !ManifestNodeExtensions.TryGetMapping(probe, "httpGet", out var httpGet))
+        {
+            return Array.Empty<string>();
+        }
+
+        var path = ManifestNodeExtensions.Scalar(httpGet, "path");
+        if (!string.Equals(path, "/health", StringComparison.Ordinal))
+        {
+            return Array.Empty<string>();
+        }
+
+        var containerName = ManifestNodeExtensions.Scalar(container, "name") ?? "<unnamed>";
+        return new[]
+        {
+            $"{document.Descriptor} container '{containerName}' still uses {probeKey}.httpGet on /health.",
+        };
+    }
+}
+
+internal sealed class ManifestInventory
+{
+    private ManifestInventory(string workspaceRoot, string bluejayRoot, IReadOnlyList<ManifestDocument> documents)
+    {
+        WorkspaceRoot = workspaceRoot;
+        BluejayRoot = bluejayRoot;
+        Documents = documents;
+    }
+
+    public string WorkspaceRoot { get; }
+
+    public string BluejayRoot { get; }
+
+    public IReadOnlyList<ManifestDocument> Documents { get; }
+
+    public static ManifestInventory Load()
+    {
+        var bluejayRoot = FindBluejayInfraRoot();
+        var workspaceRoot = Directory.GetParent(bluejayRoot)?.FullName
+            ?? throw new DirectoryNotFoundException($"Could not resolve workspace root from '{bluejayRoot}'.");
+
+        var documents = ManifestRoots(workspaceRoot, bluejayRoot)
+            .SelectMany(LoadDocumentsFromRoot)
+            .ToList();
+
+        return new ManifestInventory(workspaceRoot, bluejayRoot, documents);
+    }
+
+    private static string FindBluejayInfraRoot()
+    {
+        var current = new DirectoryInfo(AppContext.BaseDirectory);
+        while (current is not null)
+        {
+            if (Directory.Exists(Path.Combine(current.FullName, "apps"))
+                && File.Exists(Path.Combine(current.FullName, "README.md")))
+            {
+                return current.FullName;
+            }
+
+            current = current.Parent;
+        }
+
+        throw new DirectoryNotFoundException("Could not find the bluejay-infra repository root from the test output directory.");
+    }
+
+    private static IEnumerable<string> ManifestRoots(string workspaceRoot, string bluejayRoot)
+    {
+        var roots = new[]
+        {
+            Path.Combine(bluejayRoot, "apps"),
+            Path.Combine(workspaceRoot, "FlowerCore.Chat", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.DMS", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.DNS", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Intranet.Web", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Kiosk", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Media", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.MenuBoard", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.MessageBoard", "k8s"),
+            // FlowerCore.Notes/k8s/selenium/ is the live Selenium Grid
+            // manifest tree (consumed by deploy-selenium scripts).
+            // FlowerCore.Notes/k8s/guacamole/ + FlowerCore.Notes/k8s/monitoring/
+            // are historical scaffolds that have diverged from the live state
+            // (bluejay-infra/apps/guacamole + bluejay-infra/apps/monitoring are
+            // canonical). Operator review is required before bringing them in
+            // line OR decommissioning them — keep them out of the lint scope
+            // until that decision lands. See xxl-regroup-2026-05-03-followup.md
+            // "Codex 7 §0 stop conditions" + the C7 close-session output.
+            Path.Combine(workspaceRoot, "FlowerCore.Notes", "k8s", "selenium"),
+            Path.Combine(workspaceRoot, "FlowerCore.MySQL", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.PHP", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Presentations", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Print.Web", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.RemoteDesktop", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Scoreboard", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.SegmentDisplay", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.SignalControl", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.TtsReader", "k8s"),
+            Path.Combine(workspaceRoot, "FlowerCore.Updater", "k8s"),
+        };
+
+        return roots.Where(Directory.Exists);
+    }
+
+    private static IEnumerable<ManifestDocument> LoadDocumentsFromRoot(string root)
+    {
+        foreach (var filePath in Directory.EnumerateFiles(root, "*.yaml", SearchOption.AllDirectories))
+        {
+            var fileText = File.ReadAllText(filePath);
+            var segments = SplitManifestDocuments(fileText);
+
+            for (var index = 0; index < segments.Count; index++)
+            {
+                var yaml = new YamlStream();
+                try
+                {
+                    using var reader = new StringReader(segments[index]);
+                    yaml.Load(reader);
+                }
+                catch (YamlException exception)
+                {
+                    _ = exception;
+                    continue;
+                }
+
+                if (yaml.Documents.Count == 0)
+                {
+                    continue;
+                }
+
+                if (yaml.Documents[0].RootNode is YamlMappingNode mapping
+                    && ManifestNodeExtensions.Scalar(mapping, "kind") is not null)
+                {
+                    yield return new ManifestDocument(root, filePath, index, fileText, mapping);
+                }
+            }
+        }
+    }
+
+    private static IReadOnlyList<string> SplitManifestDocuments(string fileText)
+    {
+        var documents = new List<string>();
+        var currentLines = new List<string>();
+        var seenApiVersion = false;
+
+        foreach (var line in Regex.Split(fileText, @"\r?\n"))
+        {
+            if (Regex.IsMatch(line, @"^\s*---\s*$"))
+            {
+                FlushCurrentDocument();
+                continue;
+            }
+
+            if (Regex.IsMatch(line, @"^\s*apiVersion:\s*")
+                && seenApiVersion
+                && currentLines.Any(existing => !string.IsNullOrWhiteSpace(existing)))
+            {
+                FlushCurrentDocument();
+            }
+
+            currentLines.Add(line);
+            if (Regex.IsMatch(line, @"^\s*apiVersion:\s*"))
+            {
+                seenApiVersion = true;
+            }
+        }
+
+        FlushCurrentDocument();
+        return documents;
+
+        void FlushCurrentDocument()
+        {
+            var text = string.Join(Environment.NewLine, currentLines).Trim();
+            if (!string.IsNullOrWhiteSpace(text))
+            {
+                documents.Add(text);
+            }
+
+            currentLines.Clear();
+            seenApiVersion = false;
+        }
+    }
+}
+
+internal sealed record ManifestDocument(
+    string RootPath,
+    string FilePath,
+    int DocumentIndex,
+    string FileText,
+    YamlMappingNode Root)
+{
+    public string Kind => Scalar("kind") ?? string.Empty;
+
+    public string Name => Scalar("metadata", "name") ?? $"document-{DocumentIndex}";
+
+    public string Namespace => Scalar("metadata", "namespace") ?? string.Empty;
+
+    public string RelativePath => Path.GetRelativePath(RootPath, FilePath).Replace('\\', '/');
+
+    public string Descriptor => $"{Kind} {Namespace}/{Name} [{RelativePath}#{DocumentIndex + 1}]";
+
+    public string? Scalar(params string[] path) => ManifestNodeExtensions.Scalar(Root, path);
+
+    public IReadOnlyList<YamlMappingNode> MappingSequence(params string[] path) => ManifestNodeExtensions.MappingSequence(Root, path);
+
+    public IEnumerable<string> AllScalars() => ManifestNodeExtensions.AllScalars(Root);
+
+    public IReadOnlyList<string> EgressPorts()
+    {
+        return MappingSequence("spec", "egress")
+            .SelectMany(egressRule => ManifestNodeExtensions.MappingSequence(egressRule, "ports"))
+            .Select(portMapping => ManifestNodeExtensions.Scalar(portMapping, "port"))
+            .Where(value => !string.IsNullOrWhiteSpace(value))
+            .Cast<string>()
+            .ToList();
+    }
+
+    public YamlMappingNode? PodSpec()
+    {
+        return Kind switch
+        {
+            "Deployment" or "StatefulSet" or "DaemonSet" or "Job" =>
+                ManifestNodeExtensions.Mapping(Root, "spec", "template", "spec"),
+            "CronJob" =>
+                ManifestNodeExtensions.Mapping(Root, "spec", "jobTemplate", "spec", "template", "spec"),
+            _ => null,
+        };
+    }
+
+    public IReadOnlyList<YamlMappingNode> ContainerMappings()
+    {
+        var podSpec = PodSpec();
+        if (podSpec is null)
+        {
+            return Array.Empty<YamlMappingNode>();
+        }
+
+        return ManifestNodeExtensions.MappingSequence(podSpec, "containers")
+            .Concat(ManifestNodeExtensions.MappingSequence(podSpec, "initContainers"))
+            .ToList();
+    }
+
+    public IReadOnlyList<ContainerSpec> ContainerSpecs()
+    {
+        return ContainerMappings()
+            .Select(container => new ContainerSpec(
+                ManifestNodeExtensions.Scalar(container, "name") ?? "<unnamed>",
+                ManifestNodeExtensions.Scalar(container, "image") ?? string.Empty,
+                ManifestNodeExtensions.Scalar(container, "imagePullPolicy") ?? string.Empty))
+            .ToList();
+    }
+}
+
+internal sealed record ContainerSpec(string Name, string Image, string ImagePullPolicy);
+
+internal static class ManifestNodeExtensions
+{
+    public static string? Scalar(this YamlMappingNode mapping, params string[] path)
+    {
+        return TryGetNode(mapping, path, out var node) && node is YamlScalarNode scalar
+            ? scalar.Value
+            : null;
+    }
+
+    public static YamlMappingNode? Mapping(this YamlMappingNode mapping, params string[] path)
+    {
+        return TryGetNode(mapping, path, out var node) ? node as YamlMappingNode : null;
+    }
+
+    public static bool TryGetMapping(this YamlMappingNode mapping, string key, out YamlMappingNode result)
+    {
+        if (TryGetChild(mapping, key, out var child) && child is YamlMappingNode childMapping)
+        {
+            result = childMapping;
+            return true;
+        }
+
+        result = null!;
+        return false;
+    }
+
+    public static IReadOnlyList<YamlMappingNode> MappingSequence(this YamlMappingNode mapping, params string[] path)
+    {
+        return TryGetNode(mapping, path, out var node) && node is YamlSequenceNode sequence
+            ? sequence.Children.OfType<YamlMappingNode>().ToList()
+            : Array.Empty<YamlMappingNode>();
+    }
+
+    public static IReadOnlyList<string> ScalarSequence(this YamlMappingNode mapping, params string[] path)
+    {
+        return TryGetNode(mapping, path, out var node) && node is YamlSequenceNode sequence
+            ? sequence.Children.OfType<YamlScalarNode>()
+                .Select(child => child.Value)
+                .Where(value => !string.IsNullOrWhiteSpace(value))
+                .Cast<string>()
+                .ToList()
+            : Array.Empty<string>();
+    }
+
+    public static IEnumerable<string> AllScalars(YamlNode node)
+    {
+        return node switch
+        {
+            YamlScalarNode scalar when !string.IsNullOrWhiteSpace(scalar.Value) => new[] { scalar.Value! },
+            YamlSequenceNode sequence => sequence.Children.SelectMany(AllScalars),
+            YamlMappingNode mapping => mapping.Children.SelectMany(entry => AllScalars(entry.Key).Concat(AllScalars(entry.Value))),
+            _ => Array.Empty<string>(),
+        };
+    }
+
+    private static bool TryGetNode(YamlMappingNode mapping, IReadOnlyList<string> path, out YamlNode node)
+    {
+        YamlNode current = mapping;
+        foreach (var segment in path)
+        {
+            if (current is not YamlMappingNode currentMapping || !TryGetChild(currentMapping, segment, out current))
+            {
+                node = null!;
+                return false;
+            }
+        }
+
+        node = current;
+        return true;
+    }
+
+    private static bool TryGetChild(YamlMappingNode mapping, string key, out YamlNode value)
+    {
+        foreach (var entry in mapping.Children)
+        {
+            if (entry.Key is YamlScalarNode scalar
+                && string.Equals(scalar.Value, key, StringComparison.Ordinal))
+            {
+                value = entry.Value;
+                return true;
+            }
+        }
+
+        value = null!;
+        return false;
+    }
+}

tests/bluejay-infra-lint/conftest.dev/01_cross_namespace_ingressroute.rego

@@ -0,0 +1,12 @@
+package bluejayinfra.cross_namespace_ingressroute
+
+deny[msg] {
+  input.kind == "IngressRoute"
+  ns := object.get(input.metadata, "namespace", "")
+  route := input.spec.routes[_]
+  service := route.services[_]
+  svc_ns := object.get(service, "namespace", "")
+  svc_ns != ""
+  svc_ns != ns
+  msg := sprintf("IngressRoute %s/%s references Service %s in namespace %s", [ns, input.metadata.name, service.name, svc_ns])
+}

tests/bluejay-infra-lint/conftest.dev/02_public_method_allowlist.rego

@@ -0,0 +1,23 @@
+package bluejayinfra.public_method_allowlist
+
+public_hosts := {"dist.flowercore.io", "dns.iamworkin.lan"}
+
+deny[msg] {
+  input.kind == "IngressRoute"
+  route := input.spec.routes[_]
+  match := object.get(route, "match", "")
+  host := public_hosts[_]
+  contains(match, sprintf("Host(`%s`)", [host]))
+  not contains(match, "Method(`GET`)")
+  msg := sprintf("IngressRoute %s/%s is missing Method(GET) for public read-only host %s", [input.metadata.namespace, input.metadata.name, host])
+}
+
+deny[msg] {
+  input.kind == "IngressRoute"
+  route := input.spec.routes[_]
+  match := object.get(route, "match", "")
+  host := public_hosts[_]
+  contains(match, sprintf("Host(`%s`)", [host]))
+  not contains(match, "Method(`HEAD`)")
+  msg := sprintf("IngressRoute %s/%s is missing Method(HEAD) for public read-only host %s", [input.metadata.namespace, input.metadata.name, host])
+}

tests/bluejay-infra-lint/conftest.dev/03_traefik_vip_backend_ports.rego

@@ -0,0 +1,30 @@
+package bluejayinfra.traefik_vip_backend_ports
+
+has_vip {
+  some i
+  some j
+  input.spec.egress[i].to[j].ipBlock.cidr == "10.0.56.200/32"
+}
+
+has_port(port) {
+  some i
+  some j
+  input.spec.egress[i].ports[j].port == port
+}
+
+deny[msg] {
+  input.kind == "NetworkPolicy"
+  has_vip
+  has_port(443)
+  not has_port(8443)
+  msg := sprintf("NetworkPolicy %s/%s allows 10.0.56.200:443 without backend port 8443", [input.metadata.namespace, input.metadata.name])
+}
+
+deny[msg] {
+  input.kind == "NetworkPolicy"
+  has_vip
+  has_port(80)
+  not has_port(8080)
+  not has_port(8000)
+  msg := sprintf("NetworkPolicy %s/%s allows 10.0.56.200:80 without backend HTTP port 8080 or 8000", [input.metadata.namespace, input.metadata.name])
+}

tests/bluejay-infra-lint/conftest.dev/04_auth_probe_path.rego

@@ -0,0 +1,28 @@
+package bluejayinfra.auth_probe_path
+
+protected_deployments := {
+  "messageboard-web",
+  "scoreboard-web",
+  "segmentdisplay-web",
+  "signalcontrol-web",
+}
+
+deny[msg] {
+  input.kind == "Deployment"
+  protected_deployments[input.metadata.name]
+  container := input.spec.template.spec.containers[_]
+  probe := object.get(container, "readinessProbe", {})
+  http_get := object.get(probe, "httpGet", {})
+  object.get(http_get, "path", "") == "/health"
+  msg := sprintf("Deployment %s/%s must not use readinessProbe.httpGet /health behind API key middleware", [input.metadata.namespace, input.metadata.name])
+}
+
+deny[msg] {
+  input.kind == "Deployment"
+  protected_deployments[input.metadata.name]
+  container := input.spec.template.spec.containers[_]
+  probe := object.get(container, "livenessProbe", {})
+  http_get := object.get(probe, "httpGet", {})
+  object.get(http_get, "path", "") == "/health"
+  msg := sprintf("Deployment %s/%s must not use livenessProbe.httpGet /health behind API key middleware", [input.metadata.namespace, input.metadata.name])
+}

tests/bluejay-infra-lint/conftest.dev/05_statefulset_volumeclaim_defaults.rego

@@ -0,0 +1,23 @@
+package bluejayinfra.statefulset_volumeclaim_defaults
+
+deny[msg] {
+  input.kind == "StatefulSet"
+  count(object.get(input.spec, "volumeClaimTemplates", [])) > 0
+  object.get(input.spec, "podManagementPolicy", "") == ""
+  msg := sprintf("StatefulSet %s/%s is missing spec.podManagementPolicy", [input.metadata.namespace, input.metadata.name])
+}
+
+deny[msg] {
+  input.kind == "StatefulSet"
+  count(object.get(input.spec, "volumeClaimTemplates", [])) > 0
+  object.get(input.spec, "revisionHistoryLimit", 0) == 0
+  msg := sprintf("StatefulSet %s/%s is missing spec.revisionHistoryLimit", [input.metadata.namespace, input.metadata.name])
+}
+
+deny[msg] {
+  input.kind == "StatefulSet"
+  claim := input.spec.volumeClaimTemplates[_]
+  object.get(claim.spec, "volumeMode", "") != "Filesystem"
+  claim_name := object.get(claim.metadata, "name", "<unnamed>")
+  msg := sprintf("StatefulSet %s/%s volumeClaimTemplate %s is missing volumeMode: Filesystem", [input.metadata.namespace, input.metadata.name, claim_name])
+}

tests/bluejay-infra-lint/conftest.dev/06_localhost_image_pull_policy.rego

@@ -0,0 +1,40 @@
+package bluejayinfra.localhost_image_pull_policy
+
+pod_spec(spec) = pod {
+  input.kind == "Deployment"
+  pod := spec.template.spec
+}
+
+pod_spec(spec) = pod {
+  input.kind == "StatefulSet"
+  pod := spec.template.spec
+}
+
+pod_spec(spec) = pod {
+  input.kind == "DaemonSet"
+  pod := spec.template.spec
+}
+
+deny[msg] {
+  pod := pod_spec(input.spec)
+  container := pod.containers[_]
+  startswith(object.get(container, "image", ""), "localhost/")
+  object.get(container, "imagePullPolicy", "") != "Never"
+  msg := sprintf("%s/%s container %s uses a localhost image without imagePullPolicy: Never", [input.metadata.namespace, input.metadata.name, container.name])
+}
+
+deny[msg] {
+  pod := pod_spec(input.spec)
+  container := pod.initContainers[_]
+  startswith(object.get(container, "image", ""), "localhost/")
+  object.get(container, "imagePullPolicy", "") != "Never"
+  msg := sprintf("%s/%s initContainer %s uses a localhost image without imagePullPolicy: Never", [input.metadata.namespace, input.metadata.name, container.name])
+}
+
+deny[msg] {
+  pod := pod_spec(input.spec)
+  container := pod.containers[_]
+  startswith(object.get(container, "image", ""), "fc-")
+  not contains(object.get(container, "image", ""), "/")
+  msg := sprintf("%s/%s container %s uses a non-localhost FlowerCore image reference %s", [input.metadata.namespace, input.metadata.name, container.name, container.image])
+}

tests/bluejay-infra-lint/conftest.dev/07_public_egress_dns_none.rego

@@ -0,0 +1,27 @@
+package bluejayinfra.public_egress_dns_none
+
+public_egress_workloads := {
+  "asterisk",
+  "fc-llm-bridge",
+  "mysql-web",
+  "php-web",
+  "ttsreader-align",
+  "ttsreader-kokoro",
+  "ttsreader-modern",
+  "ttsreader-piper",
+}
+
+deny[msg] {
+  input.kind == "Deployment"
+  public_egress_workloads[input.metadata.name]
+  object.get(input.spec.template.spec, "dnsPolicy", "") != "None"
+  msg := sprintf("Deployment %s/%s must set dnsPolicy: None for public-internet egress", [input.metadata.namespace, input.metadata.name])
+}
+
+deny[msg] {
+  input.kind == "Deployment"
+  public_egress_workloads[input.metadata.name]
+  search := object.get(object.get(input.spec.template.spec, "dnsConfig", {}), "searches", [])[_]
+  contains(lower(search), "iamworkin.lan")
+  msg := sprintf("Deployment %s/%s must not include iamworkin.lan in dnsConfig.searches", [input.metadata.namespace, input.metadata.name])
+}

tests/bluejay-infra-lint/conftest.dev/08_public_readwrite_allowlist.rego

@@ -0,0 +1,40 @@
+package bluejayinfra.public_readwrite_allowlist
+
+# Public hosts that allow a tightly bounded write surface in addition to
+# GET/HEAD. updatecenter.iamworkin.lan accepts POST /api/v1/checkin/{id}
+# (bootstrap-JWT) so its allowlist is GET||HEAD||POST||OPTIONS — but
+# PUT/PATCH/DELETE must still 404 at the route. Any host in this set MUST
+# include all four required methods AND MUST NOT include any forbidden
+# method.
+public_readwrite_hosts := {
+  "updatecenter.iamworkin.lan",
+  "updates.iamworkin.lan",
+  "update.flowercore.io",
+  "updates.flowercore.io",
+}
+
+required_methods := {"GET", "HEAD", "POST", "OPTIONS"}
+
+forbidden_methods := {"PUT", "PATCH", "DELETE"}
+
+deny[msg] {
+  input.kind == "IngressRoute"
+  route := input.spec.routes[_]
+  match := object.get(route, "match", "")
+  host := public_readwrite_hosts[_]
+  contains(match, sprintf("Host(`%s`)", [host]))
+  required := required_methods[_]
+  not contains(match, sprintf("Method(`%s`)", [required]))
+  msg := sprintf("IngressRoute %s/%s is missing required Method(%s) for public read-write host %s", [input.metadata.namespace, input.metadata.name, required, host])
+}
+
+deny[msg] {
+  input.kind == "IngressRoute"
+  route := input.spec.routes[_]
+  match := object.get(route, "match", "")
+  host := public_readwrite_hosts[_]
+  contains(match, sprintf("Host(`%s`)", [host]))
+  forbidden := forbidden_methods[_]
+  contains(match, sprintf("Method(`%s`)", [forbidden]))
+  msg := sprintf("IngressRoute %s/%s must not include Method(%s) on public read-write host %s", [input.metadata.namespace, input.metadata.name, forbidden, host])
+}