Keep Forgejo prune inventory outside replaceable checkouts

Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a06ecb-456a-71c2-b41e-0755d336e883

activity-definitions/weekly-forgejo-package-prune.md

@@ -17,7 +17,7 @@ context_sources:
     required: true
     params:
       prune_script: /opt/railiance-platform/tools/cmd/forgejo-package-prune
-      live_images_file: /opt/railiance-platform/docs/evidence/live-images-all.txt
+      live_images_file: /var/lib/railiance-platform/live-images/all.txt
       apply: true
       max_versions: 3
       evidence_sinks:

docs/runbook.md

@@ -602,7 +602,7 @@ Refresh protection list after cluster image rollouts:
 # From workstation with both contexts, or merge scp'd exports on railiance01:
 ./scripts/refresh_live_images.sh
 # railiance01 worker hostPath target:
-OUT=~/railiance-platform/docs/evidence/live-images-all.txt \
+OUT=~/.local/state/railiance-platform/live-images/all.txt \
   EXTRA_LIVE_FILES=/path/to/coulombcore-export.txt \
   ./scripts/refresh_live_images.sh
 ```
@@ -614,7 +614,7 @@ cd ~/railiance-platform
 export VAULT_ADDR=https://bao.coulomb.social
 # OIDC or platform token — never paste PAT into chat
 export FORGEJO_TOKEN=$(bao kv get -field=API_TOKEN platform/workloads/forgejo/forgejo-admin)
-./tools/cmd/forgejo-package-prune --apply --live-images-file docs/evidence/live-images-all.txt
+./tools/cmd/forgejo-package-prune --apply --live-images-file ~/.local/state/railiance-platform/live-images/all.txt
 ```
 
 `weekly-sbom-staleness` is emergency-paused in source and production. Its
@@ -1025,3 +1025,11 @@ uv run alembic upgrade head
 uv run python src/activity_core/seed.py
 # Re-register search attributes (see Dev environment step 4)
 ```
+
+
+Inventory durability (RPF-WP-0028): the worker mounts the host inventory directory
+read-only at `/var/lib/railiance-platform/live-images`; the weekly definition uses
+`all.txt` there. `make refresh-live-images` uses the platform publisher and keeps
+prior cluster tags on export failure. Run the host hook
+`~/.local/bin/railiance-live-images-refresh` after each cluster image rollout.
+The helper installation is versioned outside both repository checkouts.

k8s/railiance/20-runtime.yaml

@@ -48,6 +48,46 @@ metadata:
     app.kubernetes.io/name: activity-core
     app.kubernetes.io/part-of: activity-core
 data:
+  weekly-forgejo-package-prune.md: |
+    ---
+    id: weekly-forgejo-package-prune
+    name: Weekly Forgejo Package Prune
+    enabled: true
+    owner: custodian-agent
+    governance: custodian
+    status: active
+    trigger:
+      type: cron
+      cron_expression: "30 3 * * 0"
+      timezone: UTC
+      misfire_policy: skip
+    context_sources:
+      - type: shell
+        query: forgejo_package_prune
+        operation: forgejo_package_prune
+        required: true
+        params:
+          prune_script: /opt/railiance-platform/tools/cmd/forgejo-package-prune
+          live_images_file: /var/lib/railiance-platform/live-images/all.txt
+          apply: true
+          max_versions: 3
+          evidence_sinks:
+            - type: state-hub-progress
+              event_type: forgejo_package_prune
+              author: activity-core
+        bind_to: context.prune
+    ---
+
+    # Weekly Forgejo Package Prune
+
+    Runs every Sunday at 03:30 UTC (after the 02:15 `forgejo-backup` cron). Invokes
+    `railiance-platform/tools/cmd/forgejo-package-prune` to retain the newest **3**
+    versions per `coulomb` package (OCI, PyPI, npm, generic). Production image tags
+    (live cluster + Helm values) are protected.
+
+    **Enabled 2026-07-21** (`ACTIVITY-WP-0020` T05/T06): first apply deleted 38
+    stale package versions; worker has `FORGEJO_TOKEN` + host-mounted
+    `/opt/railiance-platform` prune tools.
   glas-profile-pilot.md: |
     ---
     id: glas-profile-pilot
@@ -1361,6 +1401,9 @@ spec:
             - name: railiance-platform
               mountPath: /opt/railiance-platform
               readOnly: true
+            - name: live-image-inventory
+              mountPath: /var/lib/railiance-platform/live-images
+              readOnly: true
             - name: kubeconfigs
               mountPath: /kube
               readOnly: true
@@ -1385,6 +1428,10 @@ spec:
           hostPath:
             path: /home/tegwick/railiance-platform
             type: Directory
+        - name: live-image-inventory
+          hostPath:
+            path: /home/tegwick/.local/state/railiance-platform/live-images
+            type: Directory
         - name: kubeconfigs
           hostPath:
             path: /home/tegwick/.kube

scripts/refresh_live_images.sh

@@ -1,86 +1,8 @@
 #!/usr/bin/env bash
-# ACTIVITY-WP-0023-T04: refresh multi-cluster live Forgejo image protection list.
-#
-# Exports container images matching forgejo.coulomb.social from one or more
-# kubectl contexts and merges them into a single non-secret list used by
-# weekly-forgejo-package-prune (live_images_file).
-#
-# Usage:
-#   ./scripts/refresh_live_images.sh
-#   OUT=/path/to/live-images-all.txt CONTEXTS="default hosteurope" ./scripts/refresh_live_images.sh
-#
-# On railiance01 (worker hostPath target):
-#   OUT=~/railiance-platform/docs/evidence/live-images-all.txt ./scripts/refresh_live_images.sh
+# Platform-owned, additive multi-cluster protection refresh after image rollouts.
 set -euo pipefail
-
-OUT="${OUT:-${HOME}/railiance-platform/docs/evidence/live-images-all.txt}"
-# Space-separated kubeconfig contexts (empty = current default context only)
-CONTEXTS="${CONTEXTS:-}"
-PATTERN="${FORGEJO_IMAGE_PATTERN:-forgejo.coulomb.social}"
-
-tmpdir=$(mktemp -d)
-trap 'rm -rf "$tmpdir"' EXIT
-
-export_one() {
-  local ctx="$1"
-  local dest="$2"
-  local args=()
-  if [[ -n "$ctx" ]]; then
-    args=(--context "$ctx")
-  fi
-  if ! kubectl "${args[@]}" get pods -A -o jsonpath='{range .items[*]}{range .spec.containers[*]}{.image}{"\n"}{end}{end}' \
-    2>/dev/null | grep -E "$PATTERN" | sort -u >"$dest"; then
-    # also try hosteurope-style if default failed
-    kubectl "${args[@]}" get pods -A -o json 2>/dev/null \
-      | python3 -c "
-import json,sys,re
-pat=re.compile(r'${PATTERN}')
-try:
-  d=json.load(sys.stdin)
-except Exception:
-  sys.exit(0)
-for it in d.get('items',[]):
-  for c in (it.get('spec') or {}).get('containers') or []:
-    img=c.get('image') or ''
-    if pat.search(img):
-      print(img)
-" | sort -u >"$dest" || true
-  fi
-  # hostPath / crictl fallback not required; empty file is ok for this context
-  local n
-  n=$(wc -l <"$dest" | tr -d ' ')
-  echo "context=${ctx:-current}: ${n} forgejo images" >&2
-}
-
-if [[ -z "$CONTEXTS" ]]; then
-  export_one "" "$tmpdir/a.txt"
-  cat "$tmpdir/a.txt" >"$tmpdir/all.txt"
-else
-  : >"$tmpdir/all.txt"
-  i=0
-  for ctx in $CONTEXTS; do
-    export_one "$ctx" "$tmpdir/c$i.txt"
-    cat "$tmpdir/c$i.txt" >>"$tmpdir/all.txt"
-    i=$((i + 1))
-  done
-fi
-
-# Also accept extra files to merge (e.g. coulombcore export scp'd earlier)
-if [[ -n "${EXTRA_LIVE_FILES:-}" ]]; then
-  for f in $EXTRA_LIVE_FILES; do
-    if [[ -f "$f" ]]; then
-      cat "$f" >>"$tmpdir/all.txt"
-      echo "merged extra $f" >&2
-    fi
-  done
-fi
-
-sort -u "$tmpdir/all.txt" | grep -v '^$' >"$tmpdir/merged.txt" || true
-mkdir -p "$(dirname "$OUT")"
-cp "$tmpdir/merged.txt" "$OUT"
-count=$(wc -l <"$OUT" | tr -d ' ')
-echo "wrote $OUT ($count unique images)" >&2
-if [[ "$count" -eq 0 ]]; then
-  echo "WARNING: empty live-images list — prune apply will refuse (ACTIVITY-WP-0023-T03)" >&2
-  exit 2
+PLATFORM_INVENTORY_ROOT="${PLATFORM_INVENTORY_ROOT:-${HOME}/.local/lib/railiance-platform-inventory/current}"
+if [[ ! -d "$PLATFORM_INVENTORY_ROOT" ]]; then
+  PLATFORM_INVENTORY_ROOT="${HOME}/railiance-platform"
 fi
+exec bash "$PLATFORM_INVENTORY_ROOT/tools/cmd/refresh-live-images"