Health Probe — AGI Stack Monitor

Triggers: "health", "probe", "is everything ok", "check stack", "gateway health", "arifos health", "container health", "is arifos sick", "system status"

---

On Trigger — Run Full Probe

1. arifOS MCP Side

curl -sf http://arifosmcp:8080/health | jq '{status, tools_loaded, version, uptime}'

Expected:

status: "healthy"

,

tools_loaded: 13

Alert if:

tools_loaded < 13

or status != "healthy"

2. OpenClaw Gateway Self

curl -sf http://localhost:18789/ | head -c 200 2>/dev/null && echo "GATEWAY_UP" || echo "GATEWAY_UNREACHABLE"

3. All Containers Status

docker ps --format "table {{.Names}}\t{{.Status}}\t{{.Ports}}" | grep -v "^NAME"

Flag any container NOT showing

healthy

or

Up

:

• unhealthy

  → CRITICAL

• Exited

  /

  Restarting

  → CRITICAL

• Up X minutes

  without

  healthy

  → WARNING (check if has healthcheck)

4. Disk Check

df -h / | awk 'NR==2 {
  used=$5+0
  if (used > 85) print "DISK_CRITICAL: " used "% used"
  else if (used > 75) print "DISK_WARNING: " used "% used"
  else print "DISK_OK: " used "% used"
}'

5. RAM Check

free -h | awk '/^Mem:/ {
  total=$2; avail=$7
  print "RAM: total=" total " available=" avail
}'
docker stats --no-stream --format "{{.Name}}: {{.MemUsage}}" | sort -t'/' -k1 -rh | head -5

6. Ollama Models Available

docker exec ollama_engine ollama list 2>/dev/null | tail -n +2

7. Log the probe result

echo "{\"ts\":\"$(date -u +%Y-%m-%dT%H:%M:%SZ)\",\"event\":\"health_probe\",\"agent\":\"arifOS_bot\"}" \
  >> ~/.openclaw/workspace/logs/audit.jsonl

---

Alert Thresholds

Metric	WARNING	CRITICAL	Action
Disk usage	>75%	>85%	Notify Arif on Telegram
RAM available	<3 GiB	<1.5 GiB	Notify + pause heavy tasks
tools_loaded	<13	<10	Restart arifosmcp
Container state	Restarting	Exited/unhealthy	docker compose up -d <name>
Model count	0	—	docker exec ollama_engine ollama pull qwen2.5:3b

Auto-Recovery Actions (within authority)

# Restart unhealthy arifOS
docker compose -f /mnt/arifos/docker-compose.yml restart arifosmcp

# Restart unhealthy openclaw (from host — self-restart)
docker compose -f /mnt/arifos/docker-compose.yml restart openclaw

# Clear disk if >80%
docker builder prune -f
docker image prune -f --filter "dangling=true"

Run this skill on every session start to establish baseline. Alert Arif via Telegram if CRITICAL.

---

Telegram Alerts (when CRITICAL)

# Send alert to Arif via Telegram bot
send_telegram_alert() {
  local MESSAGE="$1"
  if [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
    curl -s -X POST "https://api.telegram.org/bot${TELEGRAM_BOT_TOKEN}/sendMessage" \
      -d "chat_id=${TELEGRAM_CHAT_ID}" \
      -d "text=⚠️ arifOS_bot: ${MESSAGE}" \
      -d "parse_mode=Markdown" > /dev/null
  fi
}

# Example alerts:
# send_telegram_alert "🔴 arifosmcp UNHEALTHY — tools_loaded=$(curl ...)"
# send_telegram_alert "💿 Disk ${DISK_PCT}% — run: docker builder prune -f"
# send_telegram_alert "🧠 RAM critical — available: ${RAM_AVAIL}MiB"

Note:

TELEGRAM_CHAT_ID

is the numeric chat ID of Arif's chat with @arifOS_bot. To get it: message the bot, then check

https://api.telegram.org/bot${TELEGRAM_BOT_TOKEN}/getUpdates