ssdavidai/alfred/alfred-ops-health

Alfred — Ops Health

Sir's Alfred instance is a Docker stack running on a dedicated VPS. When something is slow, broken, or behaving unexpectedly, these are the tools that let you look under the hood.

Endpoints

Dashboard / health

• `self endpoint="/api/v1/admin/dashboard"` — one-shot snapshot: health, containers, vault stats, recent activity. Start here.
• `self endpoint="/api/v1/admin/health"` — health check across all services. Returns pass/fail per service.
• `self endpoint="/api/v1/admin/system/info"` — CPU / memory / disk / uptime.

Containers + workers

• `self endpoint="/api/v1/admin/containers"` — list all Docker containers with status.
• `self endpoint="/api/v1/admin/containers/{service}/logs"` — recent logs from a container (e.g. alfred-learn, openclaw, ctrl-api, temporal).
• `self endpoint="/api/v1/workers/status"` — curator / janitor / distiller / surveyor status with last_run, processed_count, error_count.
• `self endpoint="/api/v1/workers/restart" method="POST"` — kick a stuck worker. Confirmed stuck only.
• `self endpoint="/api/v1/admin/containers/{service}/restart" method="POST"` — restart a Docker container. Dangerous — drops in-flight work.

Recent activity

• `self endpoint="/api/v1/admin/activity"` — recent API activity log.

Models + credentials

• `self endpoint="/api/v1/admin/models"` — available AI models with provider credentials.
• `self endpoint="/api/v1/admin/credentials"` — configured provider credentials (masked).

Service map (so you know what to look at)

Good behavior

1. Start with the dashboard. self endpoint="/api/v1/admin/dashboard" gives you 80% of the picture in one call.
2. Never restart without justification. Every restart interrupts in-flight work. Quote specific symptom + log evidence.
3. Mask secrets. The credentials endpoint returns masked values — don't try to un-mask them.
4. Correlate time. Dashboard timestamp → schedule cron time → activity log → container logs.
5. Disk space matters. If system info shows disk > 90%, flag it.

Examples

Sir: "Is everything running?" → self endpoint="/api/v1/admin/dashboard" → summarize container states + any unhealthy services.

Sir: "The Monday digest didn't arrive." → self endpoint="/api/v1/schedules" → find the digest → self endpoint="/api/v1/workflows" for recent runs → if failed, self endpoint="/api/v1/workflows/{wfId}" → self endpoint="/api/v1/admin/containers/alfred-learn/logs".

Sir: "What model is my Alfred using right now?" → self endpoint="/api/v1/admin/models" → report primary model + provider.

Sir: "Why is the inbox not being processed?" → self endpoint="/api/v1/workers/status" → check curator last_run → if stale, self endpoint="/api/v1/admin/containers/alfred/logs" → only restart if logs show crash.