Survives ClickHouse outages
When ClickHouse is down or restarting, the collector spools events to disk and replays them when it returns. Maintenance windows do not eat your traffic.
Operations
Boring
on purpose.
Analytics infrastructure should be the least interesting thing you operate. The stack is designed so the failure modes are known, the recovery is automatic, and the upgrades are config swaps — with our engineers in the loop alongside your DevOps.
Runtime behavior
Designed so nothing is lost.
Report cache
The API caches computed reports, so dashboards stay fast under load and ClickHouse is not re-asked the same question by every open tab.
Alerts every minute
Alert rules are evaluated every minute from the materialized rollups, with cooldowns and a weekly digest — signals from aggregates, not raw-event scans.
Deploy webhook
POST /api/deployments with site-token auth — every release lands on the timeline, so metric shifts correlate to deploys automatically.
observer-ctl, audit-logged
Key generation, adding a site or user, changing a tier — one CLI, scriptable and CI-friendly, and every action lands in the audit log your team can read.
Control plane that swaps
JSON-file control plane today; a Postgres adapter sits behind the same interface. When you outgrow files, the change is configuration — not a rewrite.
The joint model
We operate it with you.
Self-hosted does not mean alone. After the deploy session, our engineers stay on the same channel as your DevOps team: monitoring is shared, incidents are handled together, and nobody hands anyone a wiki and wishes them luck.
Runbooks at onboarding
Every procedure used during deployment is handed over as a runbook — written during the session, for your environment, not a generic doc page.
Upgrades are config swaps
Version upgrades do not touch the schema you run — TTLs and rollups are in place. An upgrade is a config swap plus a re-run of the privacy suite, not a migration weekend.
Backups to object storage
Backups and share snapshots land in your object storage (typically €2–5 a month at this scale) — inside your cloud, under your retention rules.
Operate less.
Trust more.
Automatic spool-and-replay, minute-grained alerts, audit-logged admin, config-swap upgrades — and engineers who ran your deploy on the same channel as your DevOps.