AI Monitoring on Raspberry Pi

📋 Overview

When you're running a self-hosted AI assistant, monitoring is crucial. In this post, I'll show you how to set up comprehensive monitoring for your AI infrastructure on Raspberry Pi.

🎯 What We're Monitoring

System Resources: CPU, Memory, Temperature, Disk
Services: OpenClaw, nginx, Pi-hole, Cloudflare Tunnel
Network: Proxy status, Internet connectivity
AI Assistant: Response times, API availability

🛠️ Tools Used

Tool	Purpose	Port
Uptime Kuma	Service monitoring & alerts	3001
Pi Dashboard	Real-time system stats	8080
Telegram Bot	Instant notifications	-
systemd timers	Scheduled health checks	-

📊 Uptime Kuma Setup

Installation

# Already installed via PM2
pm2 status uptime-kuma

Adding Monitors

Access Uptime Kuma at http://192.168.51.74:3001

Click "Add New Monitor"
Choose monitor type (HTTP, Ping, Port, etc.)
Configure URL/target
Set interval (I use 60 seconds)
Enable notifications

My Monitor Configuration

Monitor	Type	Target	Interval
Pi Dashboard	HTTP	http://192.168.51.74:8080/admin/	60s
Website	HTTP	http://henryjin8s.xyz:8080/	60s
OpenClaw Gateway	HTTP	http://127.0.0.1:18789/	60s
SSH	TCP Port	192.168.51.74:22	60s
Alibaba Cloud TCP	HTTP	https://dashscope.aliyuncs.com	300s

📱 Telegram Notifications

Setup Steps

Create Telegram bot via @BotFather
Get bot token
In Uptime Kuma: Settings → Notifications
Add Telegram notification
Enter bot token and chat ID
Test notification

Notification Script

I created a custom proxy failover monitor that sends Telegram alerts:

#!/bin/bash
# /home/henry/.openclaw/workspace/telegram-proxy-monitor.sh

TELEGRAM_BOT="YOUR_BOT_TOKEN"
CHAT_ID="YOUR_CHAT_ID"

send_alert() {
    curl -s -X POST "https://api.telegram.org/bot${TELEGRAM_BOT}/sendMessage" \
        -d chat_id="${CHAT_ID}" \
        -d text="$1" \
        -d parse_mode="HTML"
}

🖥️ Pi Dashboard (Custom)

I built a custom dashboard that shows:

Real-time CPU usage
Memory usage
Temperature
Disk usage
Top 5 processes by memory
Uptime Kuma monitor status

Access: http://192.168.51.74:8080/admin/

⚙️ Automated Health Checks

Proxy Failover Monitor

Runs every 2 minutes via systemd timer:

# Check status
systemctl status proxy-failover.timer

# View logs
tail -20 /home/henry/.openclaw/workspace/proxy-failover.log

What It Checks

Primary proxy (192.168.51.2:7890)
Backup proxy (192.168.31.11:7890)
Automatically switches if primary fails
Sends Telegram alert on failover

📈 Monitoring Best Practices

1. Set Appropriate Intervals

Critical services: 60 seconds
External APIs: 300 seconds
System resources: Real-time dashboard

2. Configure Smart Alerts

Don't alert on every blip
Use retry logic (2-3 failures before alert)
Set up escalation (email → Telegram → SMS)

3. Monitor What Matters

Service availability (up/down)
Response times (performance)
Resource usage (capacity planning)
Proxy status (for China users)

🎯 Key Learnings

Start simple: Begin with uptime monitoring, add metrics later
Alert fatigue is real: Only alert on actionable issues
Test your alerts: Make sure notifications actually arrive
Document everything: You'll forget why you set certain thresholds
Monitor the monitor: Make sure Uptime Kuma itself is monitored

📊 Setting Up AI Monitoring on Raspberry Pi