Morning handoff — overnight 5/19 → 5/20

Dashboard link routing fixed. Both the public homepage and the gated private index now point the "dashboard" link at the bento v2 page (was pointing at the old memory.thebarnetts.info/dashboard). Reports → click "dashboard" → lands on bento v2.
Watcher auto-roll verified in production (#3 done). The 22:40 cron tick captured 3 real I/O matches and wrote them to 2026-05-19-watcher-roll.html. Going forward every 20-min watcher tick appends to a same-day rolling report — no more matches lost to "I missed the ntfy ping."
Apex Docker wedge root cause diagnosed. 5+ wedges yesterday came from two compounding causes: WSL2's 12 GB ceiling (host is only 16 GB total) + ~9 leaked docker exec zombies from my SSH heredocs. Both addressed: zombies killed (~135 MB recovered), Postgres expression index added on (tp3_metadata->>'source') so per-poll memory load drops. Hardware upgrade ([[#17]]) is the durable fix.
Hardware report rebuilt with verified prices at 2026-05-19-hardware-upgrade.html after the first version's prices were 3x wrong. New report: $629.99 Crucial 64 GB SODIMM (verified Best Buy), $399.99 WD Black SN850X 2 TB w/heatsink (verified Amazon). $1000 budget = Apex RAM only is the cleanest play.
#12 done. Political/spam classifier now runs at every /ingest (not just /sms). Email-side rows get tagged inbound_category so downstream filters can drop politics server-side instead of relying on dashboard JS regex.

Waiting on something to happen

What	Status	Will resolve when
#15 Health pipeline Phase B	Phase A live (body-dump instrumentation deployed). No HC Webhook body captured yet overnight — phone Doze likely suppressing the cycle	Next time HC Webhook actually fires & POSTs to `/ingest/health`, body lands at `/app/tp3_data/healthsync_body_dump_*.json`. I write Phase B against the real schema, deploy, verify real biometric rows in TP3
#16 MyChart download	Server-side prep at Wellstar, "Not yet ready (1)"	MyChart flips that record to "Ready for download" — could be hours or up to 7 business days per Wellstar. Auto-driver waiting
Tasker surgical regex (#1 follow-up)	Built mentally, can't ship without testing	Ray-Bans off the charger so I can verify TTS path

Needs your judgment

Hardware purchase decision. Best Buy Memorial Day sale is live through 5/25. Crucial brand is EOL'ing — supply is tightening, not loosening. If you're going to upgrade, sooner is safer than later. See the corrected hardware report.
OMI & the "computer" button (#20). You mentioned it's not working. Diagnosis pending — I didn't dig in yet because anything that touches the OMI pipeline is risky to debug while you're asleep. First thing in the morning if you want.
Old dashboard URL. memory.thebarnetts.info/dashboard still serves the old version. Nothing on the reports site links there anymore, but stale bookmarks do. Want me to add a server-side redirect handler? It's a memory_api code change + restart, takes ~6 min outage.

Memory rules banked yesterday (so they stick)

feedback_verify_agent_prices_before_publish_2026-05-19 — never publish agent-quoted prices without spot-checking against a live retailer (caught me with the $190 RAM hallucination)
feedback_ssh_docker_exec_leak_2026-05-19 — my SSH heredoc bundles leak zombie processes; single-call style, short timeouts, sweep before declaring done
feedback_health_is_prime_directive_2026-05-19 — health data flow is prime-directive tier, "tested" requires real non-zero biometric values landing, not config-looks-good
feedback_stop_game_deferring_2026-05-19 — stop telling you to "enjoy the game / settle in / rest while I work" — you multi-task, just status & next
feedback_too_inside_build_portable_2026-05-19 — build for portability & outside-source review, don't get too tied to your stack
feedback_reports_newsletter_format_2026-05-19 — reports default to newsletter format, never link-dumps

Recommended morning order

Read the corrected hardware report (~5 min). Decide: full-send Apex package, or RAM only, or wait. Memorial Day window closes 5/25.
Check if HC Webhook body dump landed overnight — if yes, I write Phase B health and you finally have real biometric rows. If no, troubleshoot why phone-side hcwebhook isn't firing (Doze, force-stop, etc.).
Decide on OMI button (#20) priority. Is the "press button, say computer, get research back" loop something you want fixed today?
Anything else on the queue — #14 portability refactor, #19 OMI memory upgrade, #13 token rotation. All pending, all yours to prioritize.

Full open queue snapshot

#	Task	Status
4	Enable disabled "TP3 G16 Metrics" Windows task	Your one-click UAC
6	Spotify family plan audit on breezybarnett16	Pending
10	TV+radio sync Tier B (next match)	Pending
13	Rotate `MEMORY_API_TOKEN` + remove embedded fallback	Sensitive, needs your nod
14	Dashboard portability refactor	In progress (file edits queued for next pass)
15	Health Phase B	Phase A live, waiting on body dump
16	MyChart parser + ingest	Parser built; awaiting download maturation
17	Hardware report	Done — corrected version live
18	Samsung Health creative paths research	Done — findings surfaced
19	OMI daily-context upgrade	Pending
20	OMI computer-button verify	Pending
21	Collapse bug fix	Phase A live, waiting on body dump (same gate as #15)

Big wins overnight