Services

Notifyzr is a set of small services behind one Traefik proxy. Each has its own subdomain of DOMAIN.

URL	Service	What it is
api.<domain>	api	Public REST API. API-key auth, rate limited.
app.<domain>	app	Dashboard SPA
app.<domain>/api	app-api	Dashboard auth (JWT) and organization/key management. The only issuer of sessions.
admin.<domain>	admin	Site-admin SPA
admin.<domain>/api	admin-api	Site-admin API — users, organizations, plans, announcements
docs.<domain>	docs	This documentation
mcp.<domain>	mcp-server	MCP server for AI assistants
unsubscribe.<domain>	unsubscribe	Hosted opt-out page
<domain>	landing	Marketing site
—	worker	Consumes every queue: sending, webhooks, automations, fan-out
—	toolbox	Cron-scheduled maintenance (key expiry, quota warnings, message pruning)

Infrastructure

URL	What it is
status.<domain>	Public status page (Gatus). No auth.
grafana.<domain>	Dashboards, pre-provisioned
prometheus.<domain>	Metrics
pgadmin.<domain>	Database admin (development)
mailpit.<domain>	Captured mail (development)
:15672	RabbitMQ management UI
:8090	Traefik dashboard

Datastores

• PostgreSQL — everything persistent. Nearly every table is organization-scoped.
• RabbitMQ — jobs, delayed sends, events, dead letters.
• Redis — API rate limiting, and MCP session state.

Scaling

Services are stateless and can be replicated:

docker compose up --scale api=3 --scale worker=2

Prometheus discovers replicas through Docker DNS, so each new container is scraped automatically with its own instance label — metrics are stored per instance and aggregated at query time, so scaling never double-counts.

Observability

Every PHP service exposes Prometheus metrics at /metrics. That endpoint is blocked at the Traefik edge and scraped container-direct on the Docker network, so it is never reachable from outside.