Skip to content
insightd

Introduction

What is insightd?

Section titled “What is insightd?”

insightd is a self-hosted server awareness tool designed for homelabbers. It fills the gap between reactive alerting tools (like Uptime Kuma) and heavyweight monitoring stacks (like Prometheus + Grafana).

With insightd you get:

  • Multi-runtime support — monitor Docker, Kubernetes / k3s, and Proxmox VE hypervisors from the same hub
  • Container monitoring across multiple hosts — status, CPU, RAM, restarts, network I/O, health checks
  • Host system metrics — CPU, memory, load average, uptime, disk usage, GPU, temperature
  • HTTP endpoint monitoring — uptime, response time, alerting
  • Smart insights engine — capacity-based health scoring, predictions, trends, and a correlation-based diagnosis engine that explains why a container is unhealthy
  • Explainable alerts — every alert stores why it fired (value + threshold), with email and webhook delivery (Slack, Discord, Telegram, ntfy)
  • Metric personalities — baseline-aware human-friendly moods on every metric (e.g. ”😌 Normal”, ”🔥 Way above normal”)
  • Health score breakdown — click to see per-host factor analysis explaining your system health
  • Container actions — start, stop, restart, and remove containers from the UI (Docker mode)
  • Kubernetes-aware — cluster Warning events, node conditions as alerts, container resource-limit saturation alerts, PV / PVC inventory, and Ingress auto-discovery into the HTTP endpoint monitor
  • Weekly digest emails summarizing what happened
  • Modern web dashboard — setup wizard configures everything from the UI, no .env file required

Who is it for?

Section titled “Who is it for?”
  • Homelabbers running Docker containers across one or more servers
  • Self-hosters who want to know when things break without setting up Prometheus
  • Small teams who need basic monitoring without SaaS costs

Architecture

Section titled “Architecture”

One hub, one Mosquitto broker, and one agent per monitored host. To monitor ten hosts instead of one, you run nine more agents — nothing else changes.

[agent per host] ─┐
[agent per host] ─┼─► MQTT (Mosquitto) ─► [hub] ─► SQLite ─► Web UI + Email + Webhooks
[agent per host] ─┘

On Kubernetes / k3s the agent runs as a DaemonSet — one pod per node, reporting to the same hub and Mosquitto broker. Each pod’s containers appear in the dashboard as {namespace}/{pod-name}/{container-name}. See the Kubernetes guide.

For Proxmox VE hypervisors, an agent talks to PVE’s REST API from any guest VM and surfaces the LXC/QEMU guests as containers — plus ZFS pools, storage saturation, cluster quorum, and per-guest backup state. No install on the hypervisor. See the Proxmox VE guide.

Everything that isn’t installation — email, alerts, webhooks, AI diagnosis, retention, the status page — is configured from the hub’s Settings page after first boot. A one-time setup wizard handles the admin password, SMTP, and the command for adding your first agent.

Requirements

Section titled “Requirements”
  • Docker Engine + Compose v2, or a Kubernetes/k3s cluster for the agents
  • ~30 MB RAM for the hub, ~20 MB per agent
  • TCP port 3000 for the web UI (configurable)
  • TCP port 1883 for MQTT (LAN only)