Monitor Docker host health via SSH with GPT-4o-mini and alerts to Discord

This n8n template builds an automated health monitoring dashboard for your homelab Docker host.

It SSHs into your server, collects 30+ system and container metrics, analyzes trends with AI, and delivers a structured multi-embed dashboard to Discord -- plus real-time critical alerts when things go wrong.

Stop SSHing into your server every morning to check if everything's still running. AI reads your metrics and tells you exactly what needs attention, with copy-paste fix commands.

Good to know

- Estimated cost is ~$0.003 per daily run using GPT-4o-mini. Compatible with Claude or any OpenAI-compatible LLM -- swap the model sub-node to switch providers. See the setup notes inside the workflow for Claude configuration. - Uses Google Sheets for 7-day metric history and trend analysis. A one-click setup trigger auto-creates a formatted tracking spreadsheet with frozen headers and conditional formatting. - The critical alert path runs every 5 minutes with a lightweight check (configurable). The daily digest runs once in the morning. Both schedules are adjustable.

How it works

- Daily schedule trigger SSHs into your Docker host and collects system metrics (real CPU % from /proc/stat, memory, all filesystems, swap, network I/O, top processes, zombie processes, failed services) and Docker metrics (container status, CPU, memory, restarts, health checks, disk usage, dangling images) in ~2 seconds - A 100-point health score is calculated from weighted metrics across CPU, memory, disk, swap, containers, and system health - 7 days of historical data is loaded from Google Sheets for trend comparison - AI analyzes current vs. historical metrics and returns structured JSON with severity-tagged findings, CLI fix commands, trend analysis, and a top recommendation - A 4-embed Discord dashboard is delivered: status header with inline metrics, actionable findings, Docker ecosystem overview with trends, and a footer with timing and API cost - Today's metrics are stored in Google Sheets for future trend tracking - A separate lightweight path runs every 5 minutes checking critical thresholds (disk > 90%, memory > 95%, inodes > 90%, containers down) and fires immediate alerts

How to use