`meetings/2026-05-01_183915-morning.md`

53 lines
# Daily Stand-up — Fri 2026-05-01 18:39 PDT

---

### 1. Carryover from last meeting

- ✅ `pd-api` — back online, now 5 restarts (was 4). One additional restart since last check; stable but watch closely.
- ⏳ `am-recrawl` — still stopped, 90 restarts. No fix landed. Enrichment queue coasting on stale data.
- ⏳ `sf-orchestrator` — 7 restarts, undiagnosed. No change.
- ⏳ `george-gmail` — 3 restarts / 6h uptime. No new crash this hour; OAuth still unverified.
- ⏳ Directory platform decision — lawyers + doctors still unresolved. Repos diverging.

---

### 2. Overnight (last 12 hours)

**Crashes / instability:**
- `animals-viewer` — 34 restarts, actively climbing. Debate-driven fix committed 4h ago is not holding. Second crash path likely.
- `visual-factory-orchestrator` — 17 restarts. Prefetch fix committed 19h ago (`agent:false` socket reset) not stabilizing it.
- `gemini-catalog-tagger` — 7 restarts in 3h uptime. Fast new crasher; no root cause named yet.
- `am-recrawl` — stopped, 90 restarts. No log pull, no fix.
- `sf-orchestrator` — 7 restarts, no diagnosis.

**Long-running jobs:** Enrichment agents (full-monty, gap, update-sku, silas-sku) all stable at 0 restarts / 33h uptime. Idle — no crawl input since am-recrawl is down.

**External signals:** No remote commits from collaborators. Fleet otherwise stable — 30+ agents at 0 restarts.

---

### 3. Today's plan + agent assignments

- `Steve` — verify george-gmail OAuth credentials — 3 restarts in 6h is the OAuth signature; every hour without this risks losing email routing — credentials confirmed or rotated
- `debugger-agent` — pull am-recrawl crash logs, identify root cause — downstream enrichment is coasting; no new data ingesting — commit fix or clear diagnosis within 24h
- `debugger-agent` — read animals-viewer logs post-community-fix (34 restarts, still climbing) — patch debates won't surface a memory leak or dep timeout — second crash path named or ruled out
- `debugger-agent` — audit visual-factory-orchestrator logs; 17 restarts says prefetch fix didn't close it — root cause identified or follow-up patch committed
- `pm2-crash-watcher` — deep-triage gemini-catalog-tagger (7 restarts / 3h uptime) — fastest new crasher, no diagnosis yet — root cause named
- `Steve` — write the lawyers + doctors platform decision — two active repos, no architecture decision; pre-scale window closing — decision in writing today

---

### 4. Blockers and decisions Steve owes

- **george-gmail OAuth** — 2-minute credential check; blocked on Steve. Every restart risks missed mail.
- **am-recrawl log pull** — no one can fix it without reading the logs first; enrichment agents are idle.
- **animals-viewer second crash path** — patch debates are hitting a wall; log read is prerequisite before another fix attempt.
- **Directory platform decision** — `lawyer-directory-builder` + `professional-directory` are diverging daily; no decision = accumulating rework.

---

### 5. One thing to think about

`animals-viewer` has 34 restarts and is still climbing right now — the debate-driven community fix committed 4h ago is clearly insufficient. Before triggering another round of code patches, a single log read may reveal a completely different failure mode (memory ceiling, race condition, external dep timeout) that no amount of community-logic debate will surface. The next action should be read-then-fix, not fix-then-observe.