Launching today
d3kOS
Open-source AI helm computer for boats on a Raspberry Pi
1 follower
Open-source AI helm computer for boats on a Raspberry Pi
1 follower
Turn a Raspberry Pi 4 into a boat helm computer. Tap any gauge for AI diagnosis from your own manuals, get AIS collision advisories, bow-camera hazard alerts and anchor alarms, plus charts. Works with your existing NMEA 2000. GPL v3, public beta. Gallery: 1 video overview, 2 tap any number for an AI answer, 3 day dashboard, 4 night mode, 5 SafeHelm collision advisory (advisory only), 6 hands-free voice assistant, 7 live weather radar (online), 8 signed OTA updates (free registration).
Free Options
Launch Team / Built With
TinesThe single, secure environment for agents, apps, and automations.
Promoted
Hi Product Hunt, I'm Don, the maker of d3kOS.
I got tired of marine electronics that show you a number and nothing else. When my boat broke down on Lake Erie, the answer was buried in a 500-page, unindexed engine manual written in manufacturer jargon I could barely decode. I wanted a boat that could read that manual for me, explain what my coolant temperature or oil pressure really means, and tell me what to do, so I built one.
d3kOS is a helm computer on a Raspberry Pi that reads your existing NMEA 2000 network and explains what it sees. Tap a gauge, and the AI checks it against your sensor history and your own engine manual. New in this release is SafeHelm, an AIS collision advisory that applies COLREGs rules and speaks up, but never touches the helm.
Before this, I spent years leading IT teams of up to 64 people on national programs: student loans, land registration, major Canadian banks and the IT transition to de Havilland. That’s why I built d3kOS the way I’d want any system I’d trust on a boat: documented, tested and open.
How it’s built and tested: I built d3kOS solo, using a multi-model AI workflow that I document openly as the AAO Methodology, which sets time limits and review gates to keep AI-written code from drifting. The stack is about 3,500 files, so I run automated Playwright regression tests and graph-based code indexing, which found and helped me fix 120+ hidden bugs. It has been tested in my lab and on my own boat, and other boaters are now trying it on their own vessels and sending me feedback.
Where it stands today:
- It’s a public beta (v0.9.9.5). Tested in my lab and on my own boat, and now being tried by other boaters on their own vessels.
- It isn't a certified instrument, and the AI can be wrong. Keep a proper watch.
- Charts, gauges and manual lookup work with no account. Cameras, voice, OTA updates and the mobile app need free registration.
- Forward Watch (my signalk-forward-watch module) alerts you to hazards ahead using a bow camera. It’s a second pair of eyes, not a radar.
- Fish identification is still in development, and there's no autopilot control.
It's all open source (GPL v3) and built on Signal K, AvNav, OpenCPN, Vosk, YOLOv8 and ChromaDB, so you can read, fork and fix any of it. If you can burn an SD card, you can run it. A typical build costs under $999 USD, depending on your screen and optional sensors.
What's the one thing you wish your boat's instruments could tell you that they can't today? That's what I'm building next, and I'll be here all day answering questions.
Download and community: https://atmyboat.com
Code: https://github.com/SkipperDon/d3kOS