6 changed files with 0 additions and 142 deletions
--- a/Netcup-memory-pressure-7.9G-in-swap-1.7G-free.md
+++ b/Netcup-memory-pressure-7.9G-in-swap-1.7G-free.md
@ -1,23 +0,0 @@
---
-id: TASK-LOW.1
-title: 'Netcup memory pressure: 7.9G in swap, 1.7G free'
-status: To Do
-assignee: []
-created_date: '2026-04-16 23:18'
-labels: []
-dependencies: []
-parent_task_id: TASK-LOW
---
-
-## Description
-
-<!-- SECTION:DESCRIPTION:BEGIN -->
-Snapshot 2026-04-17 01:13: 45G/62G used, 1.7G free, 7.9G in swap. Page-swapping regularly. Not urgent but warrants a pass: (a) audit which containers have mem caps > working-set (too generous) vs containers with no cap at all (already fixed by enforce script patch tonight), (b) restart long-running JVM/node containers that leaked, (c) consider killing 'nice-to-have' services if starved. Top mem consumers last checked: mailcow stack, p2pwiki-elasticsearch (3G cap), various twenty-* stacks, gitea (633M / 1G cap = 63%).
-<!-- SECTION:DESCRIPTION:END -->
-
-## Acceptance Criteria
-<!-- AC:BEGIN -->
- [ ] #1 Free mem > 4G under normal load
- [ ] #2 Swap usage < 2G under normal load
- [ ] #3 Identified + documented over-allocated containers
-<!-- AC:END -->
--- a/Deploy-enforce-container-limits.sh-from-dev-ops-repo-replace-unversioned-opt-scripts.md
+++ b/Deploy-enforce-container-limits.sh-from-dev-ops-repo-replace-unversioned-opt-scripts.md
@ -1,24 +0,0 @@
---
-id: TASK-LOW.2
-title: >-
-  Deploy enforce-container-limits.sh from dev-ops repo (replace unversioned
-  /opt/scripts/)
-status: To Do
-assignee: []
-created_date: '2026-04-16 23:18'
-labels: []
-dependencies: []
-parent_task_id: TASK-LOW
---
-
-## Description
-
-<!-- SECTION:DESCRIPTION:BEGIN -->
-Script is now in dev-ops at netcup/scripts/enforce-container-limits.sh (commit dev-ops/73acc6e → main/9b25487). /opt/scripts/enforce-container-limits.sh on Netcup is still a manual copy not tied to git. Consider: (a) symlink /opt/scripts/ → /opt/dev-ops/netcup/scripts/ so git pulls update the script, or (b) add a deploy hook that copies on commit. Option (a) is simpler but exposes directory structure; option (b) is safer.
-<!-- SECTION:DESCRIPTION:END -->
-
-## Acceptance Criteria
-<!-- AC:BEGIN -->
- [ ] #1 Production /opt/scripts/enforce-container-limits.sh tracks dev-ops
- [ ] #2 Script edit in dev-ops flows to Netcup without manual scp
-<!-- AC:END -->
--- a/Sablier-scale-to-zero-for-encryptid-original-TASK-MEDIUM.7-scope.md
+++ b/Sablier-scale-to-zero-for-encryptid-original-TASK-MEDIUM.7-scope.md
@ -1,22 +0,0 @@
---
-id: TASK-LOW.3
-title: Sablier scale-to-zero for encryptid (original TASK-MEDIUM.7 scope)
-status: To Do
-assignee: []
-created_date: '2026-04-16 23:18'
-labels: []
-dependencies: []
-parent_task_id: TASK-LOW
---
-
-## Description
-
-<!-- SECTION:DESCRIPTION:BEGIN -->
-Retargeted tonight to sidecars (TASK-MEDIUM.7 Done). Original idea was to put encryptid + encryptid-db behind Sablier for 256MB RAM savings when auth is idle. Tradeoff: cold-start latency (few seconds) on first login after idle — user-facing annoyance. Probably not worth it for auth, but documenting for future consideration. If pursued: add Sablier labels to encryptid services, configure Traefik dynamic config to route auth.rspace.online / auth.ridentity.online / encryptid.jeffemmett.com through Sablier middleware (see dev-ops/netcup/traefik/config/sablier-voice.yml for the pattern).
-<!-- SECTION:DESCRIPTION:END -->
-
-## Acceptance Criteria
-<!-- AC:BEGIN -->
- [ ] #1 Decision: pursue or permanently close
- [ ] #2 If pursued: labels + Traefik route + verified cold-start acceptable
-<!-- AC:END -->
--- a/Roll-out-canvas-with-widgets-UX-pattern-to-remaining-rApps.md
+++ b/Roll-out-canvas-with-widgets-UX-pattern-to-remaining-rApps.md
@ -1,24 +0,0 @@
---
-id: TASK-MEDIUM.10
-title: Roll out canvas-with-widgets UX pattern to remaining rApps
-status: To Do
-assignee: []
-created_date: '2026-04-16 23:17'
-labels: []
-dependencies: []
-parent_task_id: TASK-MEDIUM
---
-
-## Description
-
-<!-- SECTION:DESCRIPTION:BEGIN -->
-Prototype landed tonight at demo.rspace.online/rtasks/canvas: folk-app-canvas + folk-widget + 3 rTasks widgets (Board/Backlog/Activity). Each rApp's root view becomes an integrated canvas of togglable widgets instead of siloed tab pages. 24 single-view rApps are candidates — feature inventory already done (see conversation log). Pending user review of rTasks prototype before rolling out. Planned tab groupings per rApp: rMeets, rcal, rmaps, rinbox, rtrips, rtime, rfiles, rdocs, rnotes, rfeeds, rchoices, rvote, rbnb, rvnb, rbooks, rdata, rphotos, rforum. Skip: rcred, rgov, rpast, rsplat, rtube, rchats (genuinely single-purpose).
-<!-- SECTION:DESCRIPTION:END -->
-
-## Acceptance Criteria
-<!-- AC:BEGIN -->
- [ ] #1 User approves rTasks prototype after visual review
- [ ] #2 Widget registry pattern documented for new rApp authors
- [ ] #3 Each approved rApp has a /canvas route alongside existing root
- [ ] #4 Mobile fallback (stacked cards) tested on real device
-<!-- AC:END -->
--- a/Create-on-demand-sidecar-containers-kicad-freecad-blender-scribus-open-notebook.md
+++ b/Create-on-demand-sidecar-containers-kicad-freecad-blender-scribus-open-notebook.md
@ -1,26 +0,0 @@
---
-id: TASK-MEDIUM.8
-title: >-
-  Create on-demand sidecar containers
-  (kicad/freecad/blender/scribus/open-notebook)
-status: To Do
-assignee: []
-created_date: '2026-04-16 23:17'
-labels: []
-dependencies: []
-parent_task_id: TASK-MEDIUM
---
-
-## Description
-
-<!-- SECTION:DESCRIPTION:BEGIN -->
-The 5 sidecar containers defined in /opt/rspace-online/docker-compose.yml under profiles:[sidecar] don't exist on Netcup. Sablier can't wake what doesn't exist. Run `cd /opt/rspace-online && docker compose --profile sidecar create` when server load is low — this triggers heavy Docker builds (KiCad/FreeCAD/Blender pull hundreds of MB + compile). Wait for load avg < 8 and free mem > 4GB before running.
-<!-- SECTION:DESCRIPTION:END -->
-
-## Acceptance Criteria
-<!-- AC:BEGIN -->
- [ ] #1 All 5 sidecar images built on Netcup
- [ ] #2 Containers in 'created' state (not started)
- [ ] #3 Sablier can wake each one via /api/strategies/blocking
- [ ] #4 ensureSidecar(name) from rspace server triggers actual container start
-<!-- AC:END -->
--- a/Wire-ollama-into-rspace-sidecar-lifecycle.md
+++ b/Wire-ollama-into-rspace-sidecar-lifecycle.md
@ -1,23 +0,0 @@
---
-id: TASK-MEDIUM.9
-title: Wire ollama into rspace sidecar lifecycle
-status: To Do
-assignee: []
-created_date: '2026-04-16 23:17'
-labels: []
-dependencies: []
-parent_task_id: TASK-MEDIUM
---
-
-## Description
-
-<!-- SECTION:DESCRIPTION:BEGIN -->
-server/sidecar-manager.ts lists ollama in SIDECARS but there's no ollama service in /opt/rspace-online/docker-compose.yml. ensureSidecar('ollama') calls from server/index.ts:2853 silently no-op. Either: (a) add an ollama service to the compose under profiles:[sidecar] so Sablier can wake it, or (b) drop ollama from sidecar-manager and adjust callers. Memory doc (2026-03-31 entry) suggests ollama was intended as a sidecar — option (a) is more likely correct.
-<!-- SECTION:DESCRIPTION:END -->
-
-## Acceptance Criteria
-<!-- AC:BEGIN -->
- [ ] #1 ollama container exists on rspace-internal network at host 'ollama' port 11434
- [ ] #2 Sablier can wake it via blocking API
- [ ] #3 fetch('http://ollama:11434/') from rspace returns 200 after ensureSidecar('ollama')
-<!-- AC:END -->