The New Wonder? DeepSeek AI - Scargill's Tech Blog

Unless you’ve been under a rock in the last few weeks – you’ll be at least vaguelly aware of China’s DeepSeek AI -by all accounts – SERIOUS competition for OpenAi’s ChatGPT and other current AIs.

I’ve spent quite some time playing with DeepSeek now and I have some good news and some not-so-good news for you:

DeepSeek is good – as a simple example – several YouTubers out there have given it some requests to produce Python code. One such YouTuber asked it to produce code for an Arduino board to handle a strip of WS2812b LEDS and added 2 buttons – one to turn the LEDS on/off, the other to sequence through a standard set of colours, one colour change per keypress. “Garage Tinkering” produced a long, detailed YouTube video and I’ll link the YouTube video here. You will be impressed, firstly as he’s a good orator and secondly by the detail into which Deepseek described what it was doing and why.

In summary, not at all complex, this code has to initialise the WS2812b strip, initialise button states, produce a table of “standard” colours, debounce buttons as needed and accept button presses to trigger simple actions on the WS2812b strip. DeepSeek produced the code and was VERY verbal in explaining its “thinking” while developing the code. In the end Deepseek produced code that at first sight didn’t work. The Youtuber figured out that there was a bug in the button debounce code and asked DeepSeek to use a standard debounce library. That done, the code worked PERFECTLY with no other interaction – it just worked.

My Favourite Pub from Way Back

Others have asked for code fragments and had 100% success rate. I asked Deepseek to tell me about my favourite pub, the “Gib” (Gibraltar Rock) in Tynemouth, UK, a favourite haunt when I was in my teens.

I didn’t clarify and just asked it about “The Gib in Tynemouth”. The response was PERFECT. Not only that but Deepseek firstly answered saying “The ‘Gib’ in Tynemouth refers to the Gibbet, a historical structure used for displaying the bodies of executed criminals as a warning to others. What made the Gib in Tynemouth special was its prominent location at the entrance to the River Tyne…” – my fault for not clarifying I was talking about the pub. Once I clarified it gave a great description of the pub and the background including “Ah, you’re referring to The Gibraltar Rock pub, commonly known as “The Gib,” in Tynemouth! This historic pub is a beloved local landmark with a unique charm. What makes it special is its rich history, cozy atmosphere…” and much more.

If I didn’t know better I’d think I was speaking to a human and I’m so old and experienced, I used to play with “Eliza” decades ago – I even wrote a version out of boredom – ok it was useless but fun.

So far, So Good

And now for my first DeepSeek real success – Home Assistant under Debian which in turn would run under Windows WSL in Windows 11 Pro – here is a summary I requested after asking DeepSeek various questions, getting answers and getting the whole lot to work. Throughout, I assumed nothing and had DeepSeek do the lot for me.

But firstly – why? I didn’t want to dedicate an entire PC to Home Assistant -on the other hand my Raspberry Pi (house controller) is starting to slow down a bit so as I’m temporarily in the UK away from my normal IOT kit in Spain, I thought this might be a good, save way to learn on my new UK Windows 11 Pro mini-PC.

Home Assistant under Debian – on WSL for Windows

Firstly, let me say that generating this summary of code and assistance I asked for 2 days ago, took a while… it seems the world and it’s mother are using DeepSeek right now.

Certainly! Below is the updated step-by-step summary that includes setting up Node-RED and Mosquitto MQTT alongside Home Assistant under Docker in your Debian WSL instance, ensuring all services are accessible remotely using your PC’s IP address.

DeepSeek’s Solution…

Certainly! Below is the updated step-by-step summary that includes setting up Node-RED and Mosquitto MQTT alongside Home Assistant under Docker in your Debian WSL instance, ensuring all services are accessible remotely using your PC’s IP address.

Step 1: Set Up WSL on Windows 11

Open PowerShell as Administrator and run:

   wsl --install

This installs WSL and the default Linux distribution (usually Ubuntu).

If WSL is already installed, ensure it’s updated to the latest version:

   wsl --update

Set WSL 2 as the default version:

   wsl --set-default-version 2

Step 2: Install Debian from the Microsoft Store

Open the Microsoft Store and search for Debian.
Click Install to download and install Debian.
Launch Debian from the Start menu and complete the initial setup (create a username and password).

Step 3: Install Docker in Debian WSL

Update your Debian system:

   sudo apt update && sudo apt upgrade -y

Install Docker dependencies:

   sudo apt install apt-transport-https ca-certificates curl software-properties-common -y

Add Docker’s official GPG key:

   curl -fsSL https://download.docker.com/linux/debian/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg

Add the Docker repository:

   echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/debian $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null

Install Docker:

   sudo apt update
   sudo apt install docker-ce docker-ce-cli containerd.io -y

Start and enable the Docker service:

   sudo service docker start
   sudo systemctl enable docker

Add your user to the docker group to run Docker commands without sudo:

   sudo usermod -aG docker $USER

Log out and log back in for this change to take effect.

Step 4: Install Home Assistant, Node-RED, and Mosquitto MQTT Under Docker

Create a directory for your Docker setup:

   mkdir ~/docker-services
   cd ~/docker-services

Create a docker-compose.yml file:

   nano docker-compose.yml

Add the following content to the file:

   version: '3'
   services:
     homeassistant:
       image: ghcr.io/home-assistant/home-assistant:stable
       container_name: homeassistant
       volumes:
         - ./homeassistant/config:/config
       ports:
         - "8123:8123"  # Home Assistant web interface port
       restart: unless-stopped

     nodered:
       image: nodered/node-red
       container_name: nodered
       volumes:
         - ./nodered/data:/data
       ports:
         - "1880:1880"  # Node-RED web interface port
       restart: unless-stopped

     mosquitto:
       image: eclipse-mosquitto
       container_name: mosquitto
       ports:
         - "1883:1883"  # MQTT port
       volumes:
         - ./mosquitto/config:/mosquitto/config
         - ./mosquitto/data:/mosquitto/data
         - ./mosquitto/log:/mosquitto/log
       restart: unless-stopped

Save the file (Ctrl+O, Enter, Ctrl+X).
Start the Docker containers:

   docker-compose up -d

Step 5: Ensure Services Are Accessible Remotely

To access Home Assistant, Node-RED, and Mosquitto MQTT remotely using your PC’s IP address:

1. Find Your PC’s IP Address

Open Command Prompt or PowerShell and run:

  ipconfig

Look for the IPv4 address under your active network adapter (e.g., 192.168.1.x).

2. Configure Port Forwarding on Windows

Open PowerShell as Administrator.
Forward the necessary ports:

   netsh interface portproxy add v4tov4 listenaddress=0.0.0.0 listenport=8123 connectaddress=xxx.xx.xx.xx connectport=8123
   netsh interface portproxy add v4tov4 listenaddress=0.0.0.0 listenport=1880 connectaddress=xxx.xx.xx.xx connectport=1880
   netsh interface portproxy add v4tov4 listenaddress=0.0.0.0 listenport=1883 connectaddress=xxx.xx.xx.xx connectport=1883

Replace xxx.xx.xx.xx with the IP address of your WSL instance (find it using ip addr show eth0 in Debian).

Verify the port forwarding rules:

   netsh interface portproxy show all

Allow the ports through the Windows Firewall:

Open Windows Security > Firewall & Network Protection > Advanced Settings.
Create inbound rules to allow TCP traffic on ports 8123, 1880, and 1883.

3. Access the Services Remotely

Home Assistant:

  http://<your-pc-ip>:8123

Node-RED:

  http://<your-pc-ip>:1880

Mosquitto MQTT:
Use an MQTT client to connect to:

  <your-pc-ip>:1883

Step 6: Automate Docker Containers on Windows Startup

To ensure the Docker containers start automatically when Windows boots:

Open Task Scheduler:

Press Win + S, type Task Scheduler, and open it.

Create a new task:

Click Create Task in the right-hand panel.

Configure the task:

General Tab:
- Name: Start Docker Services
- Description: Optional.
- Select Run whether user is logged on or not and Do not store password.
- Check Run with highest privileges.
Triggers Tab:
- Click New.
- Set Begin the task to At startup.
- Click OK.
Actions Tab:
- Click New.
- Set Action to Start a program.
- In Program/script, enter:
wsl.exe
- In Add arguments (optional), enter:
-d Debian -u root docker-compose -f /home/your-username/docker-services/docker-compose.yml up -d Replace /home/your-username/docker-services/docker-compose.yml with the full path to your docker-compose.yml file.
Conditions Tab:
- Uncheck Start the task only if the computer is on AC power (if applicable).
Settings Tab:
- Check Allow task to be run on demand.
- Uncheck Stop the task if it runs longer than.

Click OK to save the task.

Final Notes

Home Assistant, Node-RED, and Mosquitto MQTT will now run under Docker in your Debian WSL instance and be accessible remotely via your PC’s IP address.
If you want to access these services from outside your local network, consider using a VPN (e.g., Tailscale) or setting up port forwarding on your router (not recommended for security reasons).

Let me know if you need further assistance!

So far, REALLY impressed – but…

After that success which got me a working Home Assistant up and running, I was convinced that DeepSeek was INVINCIBLE…

THEN – I asked it for some help… I have some Home Assistant blocks that need separating – one of the standard blocks is called a MARKDOWN block – text header – quite easy to format but it’s a bit big (high) taking up significant real estate on my phone. DeepSeek had several attempts to adjust the height of the block, even suggesting alternative methods, confidently predicting they’d work – but all of them failed… reminiscent of Claude and ChatGPT. The information these systems have is not bang up-to-date and Home Assistant changes a lot – maybe that doesn’t help!!