I’m running Music Assistant and Home Assistant to manage my smart home, automations and music playback at home. Both systems integrate nicely with each other through various integrations. I can press a button on a light switch and make Home Assistant turn on a smart outlet, some lights and play music from Spotify, Apple Music, YouTube Music, or even files from my local file system on my Sonos speakers. That’s great!

I just always had a hard time finding the right URI or address to include in my automations to trigger the right album or playlist to be sent from Home Assistant to Music Assistant.

Here is some more information about how to get the URI.

Search Music Assistant from Home Assistant

When you open the Developer Tools => Actions menu in Home Assistant, you can select an action called Music Assistant: Search Music Assistant. You can pre-fill this with some information, for example:

action: music_assistant.search
data:
  name: Down Under
  artist: Men at Work
  limit: 25
  library_only: true

It’ll then complain that the config_entry_id is missing. This information was a bit hard to find.

Get Music Assistant config_entry_id

In Home Assistant, open the Music Assistant integration, click on the devices or entities link, and then you’ll find the config_entry_id in your browser’s address bar. You can then plug the config ID into your Music Assistant Developer Actions to interact with Music Assistant, like so:

action: music_assistant.search
data:
  config_entry_id: 01JEfoobarbaz
  name: Down Under
  artist: Men at Work
  limit: 25
  library_only: true

To achieve this, it’s necessary to create a new Docker bridge network and connect containers to it. Unfortunately, it’s also required to adjust the routing table on the host system to make everything work together, but it’s relatively simple.

Docker Setup: compose.yaml

Here’s an example compose.yaml file. I have defined two alpine test containers that both run wget to icanhazip.com to print out the IP address they are using for outgoing traffic to make it more clear.

The networks part at the end of the file is important for creating the new Docker network. This network and the chosen IP address space are just for the Docker containers. For the first Alpine service, you can see that I configured vlan3_network as the network to use for this container.

services:
  alpine:
    image: alpine
    command: "wget -qO- https://icanhazip.com"
    networks:
      - vlan3_network

  alpine2:
    image: alpine
    command: "wget -qO- https://icanhazip.com"

networks:
  vlan3_network:
    driver: bridge
    driver_opts:
      com.docker.network.bridge.name: docker_vlan3
    ipam:
      config:
        - subnet: 192.168.100.0/24

Linux Routing Configuration

echo "101 vlan3_table" | tee -a /etc/iproute2/rt_tables
ip rule add from 192.168.100.0/24 lookup vlan3_table
ip route add default via 192.168.3.1 dev enp2s0.vlan3 table vlan3_table

Essentially, we’re creating a new routing table with the name vlan3_table and adding a new default route so that all traffic from the subnet 192.168.100.0/24 goes through the host network interfaceenp2s0.vlan3.

Test

Now, when you run docker compose up, you should see the output from both containers, ideally displaying a different public IP address.

Persist Configuration

The ip commands from above are not saved when the system reboots. My Ubuntu installation is configured through systemd-networkd, so I’m using the following configuration to make the changes permanent:

echo "101 vlan3_table" | tee -a /etc/iproute2/rt_tables

# /etc/systemd/network/21-enp2s0.vlan3.netdev
[NetDev]
Name=enp2s0.vlan3
Kind=vlan

[VLAN]
Id=3

# /etc/systemd/network/31-enp2s0.vlan3.network
[Match]
Name=enp2s0.vlan3

[Network]
DHCP=ipv4
LinkLocalAddressing=no
IPv6AcceptRouterAdvertisements=false
IPv6AcceptRA=no
IPv6SendRA=no

[RoutingPolicyRule]
From=192.168.100.0/24
Table=101

[Route]
Table=101
Destination=0.0.0.0/0
Gateway=192.168.3.1

After writing about how to use perplexity.ai as the search engine of choice in Alfred.app, I’m describing the same for ChatGPT now. I also switched over to ChatGPT for searching instead of perplexity since it surfaces the sources in a more useful way for me, and it also feels like the results are better. YMMV.

When you copy the following URL, open it in your browsers address bar and have Alfred installed, it should automatically open Alfred and add ChatGPT as a custom search. I called it oa for OpenAI, but since I’m using it also as a default search or fallback, the shortcut doesn’t really matter.

alfred://customsearch/chatgpt%20search/oa/utf8/nospace/https://chatgpt.com/?q=%7Bquery%7D&hints=search

Note: The URL how to start a chat with search enabled by default is this one:

https://chatgpt.com/?q={query}&hints=search

Initially, I wanted to always start the search in a temporary chat, but this doesn’t work at the moment. Apparently this should be the URL:

https://chatgpt.com/?q={query}&hints=search&temporary-chat=true

After the custom search is installed, open the Alfred Preferences, go to Features in the sidebar => Default Results, and on the bottom of the page, Setup fallback results

On macOS, I usually do everything with Alfred, and the default fallback search engine in Alfred is also set to Google. I’ve set up a custom search in Alfred for perplexity.ai a long time ago, but since Google is the default fallback when Alfred doesn’t resolve to a file or application, I end up on Google and barely remember to use my custom perplexity.ai shortcut.

It didn’t occur to me until now that it’s totally possible to change the default fallback search to the custom perplexity web search I’ve set up such a long time ago!

Add perplexity.ai as a Custom Web Search in Alfred

When you copy and paste the following block into your browsers URL bar, it should open Alfred and add the custom search.

alfred://customsearch/perplexity.ai/p/utf8/nospace/https://www.perplexity.ai/search?q=%7Bquery%7D

Change the Fallback Results in Alfred

1. On the Alfred Defaults Results settings page, select Setup fallback results
2. Find your custom perplexity web search in the dropdown and move it to the top
3. Profit

However, I had some issues flashing ESPHome onto this device since you can’t just connect it using a USB adapter. I’m trying to document how I was finally able to flash ESPhome onto the NSPanel. For everything else, please just use the existing guide(s).

What You Need:

• usb to serial adapter, or UART adapter to connect the panel through USB
• jumper wires, well, to actually connect your computer and the panel

How to Wire Everything Together:

This is actually the whole reason for this blog post, everything I found out there didn’t work for me. So, this is what worked for me and my device.

Here’s a pinout of the NSPanel to get a better understanding:

Connect The wires between the UART Adapter and the NSPanel like this

• put the UART Adapter in 3V mode

USB adapter to NSPanel:

• 3v to 3v
• gnd to gnd
• RX to TX
• TX to RX

Note that you need to cross RX and TX! This article helped me with that.

On the NSPanel itself, connect:

• IO0 to gnd on the bottom gnd pin on the right side

Now you should be able to connect the usb adapter to your computer and use the web based esphome flasher.

I’ve been playing around with hosting my own Mastodon instance lately. If I continue this journey, I might write more about it. However, this post focuses solely on setting up email notifications between Amazon SES and Mastodon running within a Docker environment.

I had some trouble sending out emails successfully and it took a while to figure things out. With this post I just want to add to the confusion by providing my working configuration.

I have the following SMTP_ parameter in my .env file:

SMTP_AUTH_METHOD=plain
SMTP_ENABLE_STARTTLS=never
SMTP_ENABLE_STARTTLS_AUTO=false
SMTP_FROM_ADDRESS="Mastodon <notifications@example.com>"
SMTP_LOGIN=username
SMTP_PASSWORD=password
SMTP_OPENSSL_VERIFY_MODE=none
SMTP_PORT=465
SMTP_SERVER=email-smtp.us-east-1.amazonaws.com
SMTP_TLS=true

I was looking for a quick and easy way to save links from the page I’m currently on and save them to a single file in Obsidian. Ideally with some tags, so I can find them later on. While I like the idea of saving the whole web page as well, export it as a PDF or PNG so it can be read when I’m offline or the page is gone. While this sounds good, I figured that an offline archive of a website or project probably doesn’t help me ultimately, since, well, it’s offline and I rather want to use things which are still updated and maintained. Just my two cents on this.

So, instead of a fully featured bookmarking sync service with tons of features and offline reading capabilities, I decided that I just need the url, website title and an option to add hashtags at the end of the text to find it later on.

Let’s save all of this in a single big file in Obsidian. Ultimately, it’s just a plain text markdown file. I can move it around, modify it with other utilities and go about my day.

Long story short, it just requires this bookmarklet:

You might want to use something like this bookmarklet maker to make it usable as a bookmark in your browser’s bookmark bar.

It’ll take the website title, give you an option to modify it, let you add some tags, and put it into your Obsidian vault pointing to a specific note. It’s just appending the contents to the file if it exists already.

You need to update lines two and three with your vault’s name and the name of the note you want to create/use for it.

The bookmarklet also works in Safari on iOS, no need for an additional shortcut.

Here’s a short Keyboard Maestro Macro which allows you to select file(s) in macOS Finder and rename them to lowercased filenames. There’s not much else to it otherwise. I configured it as a selection within the Keyboard Maestro Menu Bar icon, or to use it via the global trigger.

Now that I’m using Cloudflare Tunnel or cloudflared as my ingress for Kubernetes, I wanted to gather some metrics from these pods, just to get an overview of what’s going on there. cloudflared exposes a metrics endpoint which can be scraped by Prometheus and fed into Grafana to have some fancy graphs.

Here’s a guide on Grafana.com how the two tools work together.

Prometheus Stack

I’m using the prometheus-community/kube-prometheus-stack helm chart to deploy the initial monitoring stack to get Prometheus and Grafana up and running.

It’s best to follow the instructions on GitHub to install it, as they also have the latest information there on how to do this.

PodMonitor for cloudflared

When Prometheus and Grafana are up, you can configure a PodMonitor to tell Prometheus to scrape all the information from the individual cloudflared pods. To do so, I make the following assumptions about the Kubernetes environment:

• Namespace for Prometheus / Grafana: monitoring
• release label: monitoring
• The cloudflared deployment has the following labels: app=cloudflared‌and release: monitoring
• The metrics port name is set tohttp-metrics

Here’s a snippet to configure the deployment, secret and PodMonitor.

The snippet is a single YAML file which contains the cloudflared Deployment, the Secret to authenticate cloudflared, and the PodMonitor for Prometheus.

Query Data in Grafana

With the kube-prometheus-stack everything’s connected already, meaning data from Prometheus can be queried directly in Grafana. To do so, just get started in the Explore tab, select Prometheus as the data source and put in a query like so:

sum (rate(cloudflared_tunnel_total_requests {job="default/cf-tunnel"}[$__rate_interval]))

There are also autocomplete suggestions depending on your environment, names for the deployment, pods etc.

With the rate() function you can calculate the average rate of requests in a given timeframe. There are many more functions available to make useful Dashboards.

I got an E Ink Display sometime last year and connected it to a Raspberry Pi Zero to use it as a low-power device to show the weather information for the day, perhaps some other stuff like calendar appointments. The list goes on.

There are a few projects out there to get started with this whole thing.

I even got the PiSugar to connect the Pi Zero with a battery. I managed to keep the Pi Zero powered with the battery for about 30 days on a schedule to turn it on once a day, fetch the information and update the display. After recharging it once, interest faded at some point and the thing ended up in a closet.

The initial plan was to power the E Ink display with an ESP32 Driver board for months or even years at a time without ever recharging. Since the ESP32 controllers can go in a deep-sleep mode and use basically no power at all. All this Arduino and ESP32 stuff failed on me since I couldn’t get the right C code working, that’s why I went for the whole Raspberry Pi route to get something running.

At some point during the holidays I finally got the stuff back out wanting to try all this again. Started the Arduino IDE and tried to gather all the C code I need to update the E Ink display with some basic information from a remote API.

Shouldn’t be too hard, right? Connect to Wi-Fi, fetch some JSON, place it in the right spot, update the display, and go back to sleep for a day.

Since weather information is boring, always wrong anyway, and there are way better methods out there to get a report, I decided to display a random quote from The Office to help me get through the day.

Display and ESP32 Board

For beginners like myself, I just got the E-Ink display and ESP32 driver board with the correct interfaces out of the box, so there’s no need to solder anything or do some other fancy stuff. This board can directly connect to the display and has USB for power and data. Below is what I got and used for this project, but there are tons of different brands and versions out there:

• e-Paper ESP32 Driver Board
• 800×480, 7.5inch E-Ink display

Set-Up

I gathered everything in this GitHub Repository, the only thing required here is to install the Arduino IDE, update the Wi-Fi credentials and perhaps change the display library to another one if you’ve got different hardware.

I’m using the GxEPD2 Display Library which covers a lot of different e-Paper/E-Ink displays and has the model numbers listed (Good Display and Waveshare models seem to be the same). The code on my GitHub repo is set up to the use 800x480 7.5" display I linked above.

tl;dr version:

1. clone the GitHub repository
2. add it to your Arduino IDE
3. update the Wi-Fi credentials
4. (update the display configuration if you’re using another one)
5. upload it to your ESP32 through the Arduino IDE

Case and Battery

This is the frame / case I got from thingiverse to store everything and put it on the wall. You can then get yourself a battery holder with some jumper wires or solder this stuff on, or use the USB plug. There’s a nice calculator available as well to figure out how long the system might be lasting.

Everything should fit into the case, so it’ll look like a picture frame on the wall.

In my case the display and board require about ~500 mW when it’s updating and ~80–100 mW in deep sleep. So if it’s updated once a day with a 2500 mAh battery connected, it should last about 1.5 years, which is fine, I guess.

More Things to Do

Since there’s still a lot of space left on the display, there are a few things one could add, perhaps the date, a reminder, whatever. For me this was mostly to finally learn at least a bit about how to do all this with the Arduino IDE, writing C and uploading this to a super low powered device.

More Resources

• Arduino forum talk about Waveshare e-paper displays with SPI
• Adafruit GFX Graphics Library for any further modifications or things you can display on the screen