The Green Shed

Went to the Grand Canyon today, which was an audible on original plans, and it was exactly the right decision. Amazing weather, amazing light up there. Very few people. I was able to just sit and look, in silence, for long stretches at a time. And I took a zillion photos.

Home Assistant Config for a Basic JSON REST Device

After much effort (and help from the forums), I was able to get a config working for my backyard weather station.

My station is custom built hardware, and returns a very simple JSON object with its current state when queried.

The json it returns looks like this:

{"timestamp":"2022-03-12 09:35:36 -0700",
 "temperature_c":15.8720368908,
 "temperature_f":60.56966640344,
 "humidity":27.009330285,
 "pressure":965.103717974,
 "dewpoint_c":-3.0690760974329714,
 "dewpoint_f":26.47566302462065}

And this is the Home Assistant configuration I needed to use to get the data into HASS:

rest:
  - resource: http://weather.local
    scan_interval: 60
    sensor:
      - name: Weather Station - Temperature
        value_template: '{{ value_json.temperature_f | round(1) }}'
        unit_of_measurement: '°F'
        device_class: temperature
        state_class: measurement
      - name: Weather Station - Humidity
        value_template: '{{ value_json.humidity | round(1) }}'
        unit_of_measurement: '%'
        device_class: humidity
        state_class: measurement
      - name: Weather Station - Pressure
        value_template: '{{ value_json.pressure | round(1) }}'
        unit_of_measurement: 'mbar'
        device_class: atmospheric_pressure
        state_class: measurement
      - name: Weather Station - Dewpoint
        value_template: '{{ value_json.dewpoint_f | round(1) }}'
        unit_of_measurement: '°F'
        device_class: temperature
        state_class: measurement

Home Assistant is very powerful, but goodness is the learning curve steep. Documentation exists, but much of it is not helpful enough. I should probably contribute back some additional examples, at the least.

(Interesting side note: I asked ChatGPT to help me put together a configuration for this setup. It was able to do so, but the result was inefficient and used an older style of REST sensor setup. Human help in the forums led me to this solution.)

Thought experiment: a lossy compression algorithm that includes noise in the decompression phase. Every decompression is unique. 🤔

Home Assistant API Call for Setting Dimmable Lamp Brightness

I had to play around with the Home Assistant API to figure out how to set the brighness of a dimmable lamp. This is what ended up working.

My lamp was setup with brightness as a Number.

curl -H "Authorization: Bearer your_token_goes_here" -H "Content-Type: application/json" -d '{"entity_id": "number.my_lamp", "value": 100}' http://homeassistant.local:8123/api/services/number/set_value

If the CEO of OpenAI thinks that you are a stochastic parrot, then that means that he doesn’t really recognize you as a person. We have a word for that kind of systemic lack of empathy and that word is “psychopath”.

www.jwz.org/blog/2023…

Spent way too much time trying to get my ESP32 board to send data to my mqtt broker. Still haven’t got it working. Probably should have just stuck with REST. Too clever for my own good.

Use AWS_PROFILE when working with AWS SDKs of all kinds

It turns out that if you set an environment variable named AWS_PROFILE, all AWS SDKs (including the CLI) will respect that value when looking for credentials for the current command.

Thanks to my brother, Ben, for pointing this out to me. It cleaned up a ton of boilerplate in some Ruby code I’ve been writing.

Finished reading: Abandon by Blake Crouch 📚

Well written but too hopelessly dark for my taste.

ChatGPT and its ilk may well result in a better experience for searching than Google — for a while. But if you don’t think the shitty incentives that led to Google’s quality dropping off aren’t going to affect these LLMs in the exact same ways, I’ve got a bridge to sell you.

Docker is just terrible. The default options always seem wrong. The error output is often inscrutible. Debugging is a nightmare. This is just poorly designed software.