# Archived Memory — Jun 10–12, 2026

## Jun 12, 2026 (00:45 UTC) — n8n Restart Broke SocialSync Workflows

- **All 5 services healthy (200 OK)** at heartbeat, but n8n restarted at 00:25:40 UTC
- **3 SocialSync workflows failed to activate** on restart with `TypeError: Cannot read properties of undefined (reading 'execute')`:
  - SocialSync - AI Content Generator (6829c6ff-...)
  - SocialSync - Engagement Auto-Reply (929a4473-...)
  - SocialSync - Google Sheet Queue Processor (028f612f-...)
  - 3 more (Master Orchestrator, Multi-Platform Publisher, Workspace Manager) WILL fail on next retry — all 6 use the same node type
- **Root cause:** `n8n-nodes-base.function v2` (legacy Function node) was **removed in n8n 2.x**. Container is `n8nio/n8n:2.20.7`. Stack: `shouldAssignExecuteMethod` in `utils.ts:88` fails because the node type isn't registered.
- **Fix:** replace Function nodes with Code nodes (`n8n-nodes-base.code`) in all 6 SocialSync workflows. Mechanical but per-function verification needed.
- **Why n8n restarted:** unknown (clean exit code 0, no docker events available, no syslog entry). Not urgent to find — restart was clean. The breakage was latent: 6 workflows were silently depending on a removed node type, waiting for a restart to surface.
- **Logged silently** to /var/log/larry-notifications.log per Larry's "stop pinging" preference
- **Daily log:** `memory/2026-06-12-1.md`

## Jun 11, 2026 (16:14 UTC) — Heartbeat Health Check + Two Bug Fixes

- **All 5 services healthy** (cars:5060, paperclip:3100, n8n:5678, openclaw:18789, dashboard:5000) — all 200 OK
- **Disk 68% used, 65G free** — comfortable
- **n8n-monitor.sh false-positive bug fixed.** Script was checking `/var/lib/docker/volumes/n8n_n8n-data/_data/database.sqlite` (wrong — extra `n8n-` segment that never existed). Real volume is `/var/lib/docker/volumes/n8n_data/_data/`. So the "n8n db missing" alert has been firing every 5 min for ~10 days (since n8n container was created Jun 1) and silently filling the log. Fixed: extracted `N8N_DB` variable, used in both checks. Verified: runs silently now, workflow count check passes.
- **`chattr +i` flag was GONE from dashboard files.** `lsattr` showed `----i---------e-------` was missing on `index.html` and `server.py`. Files had 154 lines of uncommitted changes (my own graph tab wiki-link work from this morning per MEMORY.md). So: changes are mine, but immutability was lost — likely from a Docker volume backup or some other touch that reset xattrs. **Committed** the wiki-link work (e4c1a4e) and **re-applied chattr +i**. Need to investigate what stripped the flag if it happens again.
- **Daily log:** `memory/2026-06-11-3.md` (this turn)

## Jun 11, 2026 (earlier, from MEMORY.md) — Graph Tab Wiki-Links + Memory Search Fix + Service Health

- **Memory search fixed:** index rebuilt with `openclaw memory index --force`, gateway restarted, now using `ollama/nomic-embed-text` (186ms searches)
- **Graph tab wiki-link extraction:** `_extract_wiki_links()` added to server.py — parses `[[wiki-link]]` and `[[wiki-link|display text]]` from body text, creates `kind:"wiki"` edges (amber, 2px). No files use this syntax yet but it's live.
- **Graph tab time slider:** range slider filters nodes by mtime, re-renders D3 in-place
- **SSH to WSL:** working fine (direct Tailscale, <1s connect)
- **All 5 services healthy:** cars(5060), paperclip(3100), n8n(5678), openclaw(18789), dashboard(5000) — all 200 OK
- **Dashboard files:** still `chattr +i` protected after edits

## Jun 11, 2026 — Model Quota Crisis, Fallbacks & Cron Fixes

- **Ollama `minimax-m3:cloud` hit session quota** — "you (BuckalewFinancial) have reached your session usage limit" — root cause of connection drops
- **All 8 user-facing agents got 3-tier fallbacks:** primary=`minimax-m3:cloud`, fallback1=`openrouter/owl-alpha` (via freellmapi), fallback2=`qwen3:8b`
- **All 4 OpenClaw crons switched** from `ollama/minimax-m3:cloud` → `ollama/qwen3:8b`
- **Models tab v2** shipped: health indicators, error counts, agent counts, recommendations, backup status
- **OpenRouter provider has no API key** — `openrouter/owl-alpha` works through `freellmapi` proxy on WSL (`100.109.176.73:3001`)
- **Bug:** `openclaw config patch` REPLACES objects (doesn't deep merge) — always include all fields
- **SSH to WSL timing out** from VPS — need to investigate Tailscale routing
- Vault: 60 files, all pipelines (shared/paperclip/discord) flowing
- System cron `obsidian-memory-dump.sh` (4am UTC) working; OpenClaw cron now on qwen3:8b
- Docs: `memory/2026-06-11-obsidian-dump.md`
- Agent OS built production-ready in a single massive day (Jun 5): 6 agents, feedback loop, frontmatter schema (26/26 compliant), first end-to-end content pipeline
- Dashboard upgraded with 6 widgets + critical threading fix (no more wedging on chat SSE)
- Model routing standardized: specialty agents on `ollama/minimax-m3:cloud` → `ollama/kimi-k2.6:cloud`, OpenRouter fallback only
- Two content pipelines proved: AISetupPros AI article (v2 QA-passed) + dealership homepage v2 (0 blockers)
- Tool sprawl audit done (4 coding agents) — recommendation to cut to 1-2, Larry hasn't decided yet
- **Gap:** 3 days with no daily memory logs (Jun 6–8). Habit needs restarting.
- Full review: `/root/Documents/Obsidian Vault/Reviews/2026-06-08_weekly-review.md`

## Jun 10, 2026 — Memory Backfill (Jun 6/7/8)

- Backfilled `memory/2026-06-06.md`, `2026-06-07.md`, `2026-06-08.md` from git history + shared/ artifacts + Obsidian weekly review
- **Jun 6:** Odysseus VPS install shipped (7 Ollama models, ChromaDB, SearXNG, ntfy, Caddy proxy). Shared-brain sync went live (VPS↔WSL). First real cross-machine agent comm: bill (VPS) asked, scout (WSL) answered via shared/notes.
- **Jun 7:** Quiet Sunday, no activity
- **Jun 8:** Weekly review cron ran, 26/26 docs compliant, 3 friction points flagged (memory gap, WSL mirror, tool sprawl decision pending)
- **Lesson:** When quiet days are recorded in shared/memory/, copy them to workspace/memory/ even if nothing new — otherwise the daily-log habit breaks silently.

- **User request:** "stop pinging me with automated reports, keep conversations between me and the agents when I initiate them"
- **Created:** `/usr/local/bin/notify-larry.sh` — wrapper that defaults to silent logging to `/var/log/larry-notifications.log`, has `--urgent` flag for emergencies
- **Patched 7 proactive scripts** to log to file instead of sending to chat_id 5273258651:
  - ghl_monitor.py, bfs-auto-blog.py, dashboard-integrity-check.sh, florida_events_scraper.py, generate-social-vps.py, n8n-monitor.sh, weekly-faceless-videos.sh
- **OpenClaw agents still respond to user-initiated messages** — only cron pings are silenced
- **Log location:** `/var/log/larry-notifications.log`