System health & Wingman
One status wall over every probed dependency across the family, plus Wingman’s daily scan — it self-heals the low-level, reports the high-level, and routes alerts where you actually read them.

One status wall + Wingman’s daily scan.

Demo dataAuto-refresh · 45s

Services & status 12 of 14 nominallist grows at implementation
Probed dependencies, primary ↔ fallback labelled. Status · latency · capacity per service.

No endpoints or tokens are ever read here. The wall shows probe results only; connection strings and keys never render on any admin surface.

Postgres · NeonPrimary

lat 12ms·cap 31% pool

Redis · UpstashPrimary

lat 3ms·cap 18% mem

GPU Box 1 · AI media genMojoHost

lat 210ms·cap queue 84%

GPU Box 2 · AI ConverseMojoHost

lat 96ms·cap queue 22%

CDN / object storageMojoHost

lat 28ms·buckets media · vault · watermark · exports· 46% quota

LLM · AnthropicPrimary

lat 640ms· healthy

LLM · OpenRouterFallback

lat 820ms· standby

Auth · Clerk

lat 91ms· SSO issuer OK

Payment rails

lat 130ms· live mode

Background workers

depth 42 jobs· draining

Telegram bot service

lat 76ms· webhook OK

Media pipeline

staged 123,951 posts

Search indexFallback

reindex stalled 2h

Email delivery

lat 340ms· 99.2% 24h

API usage & anomalies Demo data
Per integration: yesterday (accent) against the 7-day baseline (grey). Shape detection flags what a threshold would miss.

Clerk authsessions verified

+7%

Payment railscharges + webhooks

−5%

Notificationspush + email sends

×4.3

LLM tokensorchestration total

+11%

Media presignersigned URL issues

−3%

⚠

1 anomaly — notifications ×4.3 baseline · Wingman reported it below.

Wingman daily scan Ran today 04:00 Demo data
Every day Wingman scans engines + platforms for bugs, errors and fixes. Low-level issues it heals itself and logs; high-level ones it reports — routed to humans and AI. It also picks up from the admins, Slack and Asana.

Self-healed 14
Low-level, fixed autonomously — each one logged like a commit. Wingman contract (2026-08-09): this self-heal leg is NEW scope, gated on the founder’s dispatcher confirm-mode (“Wingman may act autonomously” is a founder gate). The REPORTED leg rides the existing /api/errors intake (live since July, posts to #_error_alerts) — its count deep-links to a Wingman trace that exists today.

wm-4f21crestarted stalled search reindex worker after 2 idle heartbeatsinfra

wm-4f1becleared 3 stale feature-flag cache keys drifted from DBfolo

wm-4f1a2retried 11 failed webhook deliveries · all succeeded second passreceipt

wm-4f188rotated an expiring presigner cert 6d before cut-offmedia

wm-4f150pruned 9 orphaned background jobs older than the queue TTLworkers

+ 9 more today · full list in Wingman’s own admin

Reported 3
High-level — Wingman diagnoses, never decides. Routed to a human or an AI lane for confirmation.

⚠

Notification fan-out loop4.3× baseline · matches the 16M-sends shape · confidence 88%

→ Human

⚠

GPU queue saturation trend84% and rising 3 days straight · capacity decision needed

→ Human

◆

Flaky Folo finance redirectfirst-hit bounce to another route, self-corrects on reload · repro attached

→ AI lane

Traces in Wingman’s own admin ↗

Alerts & billing 2 need attention
API outages, billing issues and low credit on paid services — the things that fail quietly until they fail loudly. Billing + outage alerts also post to Wingman and Slack, and land in our own admin notifications — routing below.

Service credit

Anthropic API

72% · ~19d left

OpenRouter

64% · ~15d left

SMS / OTP provider

18% · ~4d left

Image moderation

6% · ~1d left

Email delivery

55% · ~12d left

Open alerts

▲

Image moderation credit criticalHard-fail closed when exhausted — uploads queue, nothing slips through

Critical

▲

SMS credit lowSign-in OTPs at risk in ~4 days at current burn

Warning

✓

No API outages in 7 daysLast incident: payment-rail webhook delay, resolved 6d ago

Clear

Notification routing Per severity
Where each severity goes. TG notifier is the founder’s always-on channel; Slack for the team; the admin bell is the record.

Awaiting source

GPU capacity history · Box 1 (media gen) + Box 2 (Converse)

Awaiting source

Paging / on-call escalation

Awaiting source

System health & WingmaniOne status wall over every probed dependency across the family, plus Wingman’s daily scan — it self-heals the low-level, reports the high-level, and routes alerts where you actually read them.

Services & status 12 of 14 nominallist grows at implementationiProbed dependencies, primary ↔ fallback labelled. Status · latency · capacity per service.No endpoints or tokens are ever read here. The wall shows probe results only; connection strings and keys never render on any admin surface.

API usage & anomalies Demo dataiPer integration: yesterday (accent) against the 7-day baseline (grey). Shape detection flags what a threshold would miss.

Wingman daily scan Ran today 04:00 Demo dataiEvery day Wingman scans engines + platforms for bugs, errors and fixes. Low-level issues it heals itself and logs; high-level ones it reports — routed to humans and AI. It also picks up from the admins, Slack and Asana.

Reported 3iHigh-level — Wingman diagnoses, never decides. Routed to a human or an AI lane for confirmation.

Alerts & billing 2 need attentioniAPI outages, billing issues and low credit on paid services — the things that fail quietly until they fail loudly. Billing + outage alerts also post to Wingman and Slack, and land in our own admin notifications — routing below.

Service credit

Open alerts

Notification routing Per severityiWhere each severity goes. TG notifier is the founder’s always-on channel; Slack for the team; the admin bell is the record.

System health & Wingman
One status wall over every probed dependency across the family, plus Wingman’s daily scan — it self-heals the low-level, reports the high-level, and routes alerts where you actually read them.

Services & status 12 of 14 nominallist grows at implementation
Probed dependencies, primary ↔ fallback labelled. Status · latency · capacity per service.

No endpoints or tokens are ever read here. The wall shows probe results only; connection strings and keys never render on any admin surface.

API usage & anomalies Demo data
Per integration: yesterday (accent) against the 7-day baseline (grey). Shape detection flags what a threshold would miss.

Wingman daily scan Ran today 04:00 Demo data
Every day Wingman scans engines + platforms for bugs, errors and fixes. Low-level issues it heals itself and logs; high-level ones it reports — routed to humans and AI. It also picks up from the admins, Slack and Asana.

Reported 3
High-level — Wingman diagnoses, never decides. Routed to a human or an AI lane for confirmation.

Alerts & billing 2 need attention
API outages, billing issues and low credit on paid services — the things that fail quietly until they fail loudly. Billing + outage alerts also post to Wingman and Slack, and land in our own admin notifications — routing below.

Notification routing Per severity
Where each severity goes. TG notifier is the founder’s always-on channel; Slack for the team; the admin bell is the record.