💚 Service Health Monitoring

Internal monitoring of every Synalux dependency (database, OAuth providers, LiveKit, Inworld TTS, Anthropic, Gemini, OpenRouter, Stripe). Failures email the admin team and surface a status banner to affected users.

🩺 What’s Monitored

• Database — Postgres / Supabase reachability, replication lag, RLS policy presence.
• OAuth providers — Google / Microsoft / Telegram / Meta token-refresh path health.
• LiveKit SFU — TURN reachability, room creation success rate.
• TTS — Inworld TTS-2 latency + error rate; Azure Neural fallback availability.
• AI — Anthropic, Gemini, OpenRouter latency + 5xx rate; trips fallback chain when degraded.
• Stripe — checkout + webhook ingress.
• Storage — Supabase Storage object writes.
• Mail / SMS / chat providers — incoming webhook acceptance rate.

🚨 Alert Path

• Email to admin distribution list when a dependency drops below SLO.
• In-app banner to affected users when their experience is degraded — e.g. “Voice cloning is temporarily unavailable; standard voices still work.”
• Status page at synalux.ai/status (planned) for public visibility.

🛠️ Critical Bug History

• Supabase RLS-disabled critical alert — caught when a migration accidentally dropped RLS policies on the patients table. Auto-detected within 60 seconds; admins paged; rolled back same hour.

🏗️ Architecture

GET  /api/v1/cron/services-health            Aggregate health snapshot (cron-driven)
GET  /api/v1/cron/tts-health                TTS provider latency + availability
GET  /api/v1/cron/chain-health-nightly      Nightly deep probe of all dependencies
GET  /api/v1/integrations/chain-health      Integration health dashboard

Probes run every 60 seconds via Vercel Cron; results written to service_health_checks with TTL retention.

💳 Plans

Always-on for every workspace. Admin-tier sees the full dashboard; users see degraded-feature banners only.