Interactive demo

Try insightd before you install it.

Explore a sample homelab with Docker, k3s, Proxmox, endpoints, alerts, logs, and diagnosis evidence. No account, no backend, no telemetry — just realistic mock data so you can feel the product first.

Open the demo Install on your homelab

Sample data only. The real insightd runs on your hardware and stores monitoring data locally. This public demo is static and deliberately disconnected from any real system.

demo.insightd.local

Health

Needs attention

Assets

hosts, containers, endpoints

Open alerts

1 critical, 2 warning

Current top insight

critical Jellyfin is restart-spamming after ffmpeg exits with SIGKILL. Memory pressure on media-node is the most likely cause.

Sample workspace

Golden Homelab

Click the sections below to see how insightd moves from “what is broken?” to “why did it happen, what evidence supports it, and what should I try next?”

Docker k3s Proxmox VE HTTP endpoints

Overview

Last updated 24 seconds ago · agent heartbeat normal

5 healthy 2 warning 1 critical

Fleet health 82% media-node pulling score down

CPU load 37% cluster average

Memory 76% media-node at 91%

Alerts routed Slack + ntfy last delivery 2m ago

Attention feed

Jellyfin restart loop

8 restarts in 12 minutes · ffmpeg fatal · likely memory pressure

ZFS pool scrub overdue

tank has not scrubbed in 42 days · Proxmox token read-only

API latency baseline shifted

p95 is 340ms vs 180ms normal · correlated with backup window

What insightd is doing

1 Collect host/container snapshots from agents
2 Compare metrics to robust baselines
3 Mine recent logs for repeated failure patterns
4 Rank likely causes and route calm alerts

Container	Host	Status	Signal
jellyfin	media-node	restarting	8 restarts · ffmpeg SIGKILL
immich-server	media-node	degraded	queue depth 1,284
home-assistant	kingduck-pve	healthy	latency normal
ingress-nginx	k3s-edge-01	healthy	24h uptime

Likely cause

88% confidence

Jellyfin restarts are most likely caused by memory pressure on media-node. Container OOM hints, ffmpeg fatal logs, and host RAM saturation all started within the same window.

Suggested next action

Limit Jellyfin transcode memory, move Frigate inference off the host during peak media scans, or add swap as a short-term guardrail.

recent log patternjellyfin · media-node

[20:41:12] ffmpeg exited with signal: Killed
[20:41:14] transcode worker failed after 00:00:08
[20:41:15] container restart policy triggered
[20:43:02] host memory available: 812 MiB
[20:43:06] detector: restart-spam + memory-pressure matched

Metric evidenceRAM MAD z-score +4.7 vs baseline

Restart evidence8 restarts in 12 minutes

Topology evidenceJellyfin depends on media-node GPU/transcode path

What should this demo make clearer?

I am looking for first-user feedback: what felt trustworthy, what was confusing, and what would make you comfortable installing it on your own homelab?

Leave feedback