AI on the High Seas (Part 3): Sample Code for Your Boat’s Onboard AI System

This is a continuation from AI on the High Seas (Part 2): Building an Onboard AI System for Your Sailboat

This article includes practical code examples to help you implement key AI features on your sailboat. These snippets work with a Raspberry Pi running Python, Home Assistant, and optionally a local LLM or voice interface.

We have 5 scripts:

1. Intrusion Detection Script
2. Barometric Pressure Drop Warning
3. Maintenance Reminder Based on Days Passed
4. Voice Output (Using Mycroft or pyttsx3)
5. Calling a Local LLM for Advice (Ollama)

1. Intrusion Detection Script

Detects motion via a PIR sensor and sends an alert via Home Assistant.

import RPi.GPIO as GPIO
import time
import requests

PIR_PIN = 17
HA_WEBHOOK = "http://your-homeassistant.local/api/webhook/intruder_alert"

GPIO.setmode(GPIO.BCM)
GPIO.setup(PIR_PIN, GPIO.IN)

try:
    while True:
        if GPIO.input(PIR_PIN):
            print("Motion detected!")
            requests.post(HA_WEBHOOK)
            time.sleep(5)  # delay to avoid spamming
        time.sleep(0.5)
except KeyboardInterrupt:
    GPIO.cleanup()

2. Barometric Pressure Drop Warning

Monitors pressure sensor and alerts if storm conditions may be forming.

import time
from smbus2 import SMBus
from bme280 import BME280

sensor = BME280(i2c_dev=SMBus(1))

pressure_history = []

while True:
    pressure = sensor.get_pressure()
    pressure_history.append(pressure)
    if len(pressure_history) > 6:
        del pressure_history[0]
    
    if len(pressure_history) == 6:
        drop = pressure_history[0] - pressure_history[-1]
        if drop > 4:
            print("Storm warning: Pressure dropping rapidly!")
    time.sleep(600)  # every 10 minutes

3. Maintenance Reminder Based on Days Passed

import datetime
import json

STATE_FILE = "maintenance_log.json"

try:
    with open(STATE_FILE, "r") as f:
        last_check = datetime.datetime.fromisoformat(json.load(f)["last"])
except:
    last_check = datetime.datetime.now()

now = datetime.datetime.now()
days_passed = (now - last_check).days

if days_passed >= 30:
    print("Maintenance due!")
    with open(STATE_FILE, "w") as f:
        json.dump({"last": now.isoformat()}, f)
else:
    print(f"{30 - days_passed} days until next check.")

4. Voice Output (Using Mycroft or pyttsx3)

Text-to-speech notification for system alerts.

import pyttsx3
engine = pyttsx3.init()
engine.say("Freshwater level is below 20 percent. Please refill.")
engine.runAndWait()

5. Calling a Local LLM for Advice (Ollama)

Query your local model via HTTP with curl or Python.

curl http://localhost:11434/api/generate \
  -d '{"model": "llama2", "prompt": "Should I reef the sails with wind at 25 knots?"}'

You can also call this from Python:

import requests

res = requests.post("http://localhost:11434/api/generate", json={
    "model": "llama2",
    "prompt": "Should I reef the sails with wind at 25 knots?"
})
print(res.json()["response"])

Next Steps

These examples give you a solid foundation to build out more advanced automations. Once your system is stable, consider integrating logging, dashboards, and crew interaction tools for a fully responsive onboard AI.