#Ask HA how to spell a word

1 messages ยท Page 1 of 1 (latest)

lime cliff
#

I used to do this relatively often with Alexa and have been missing it, then I just realized how easy it was to implement.

alias: voice spelling
description: ""
triggers:
  - trigger: conversation
    command:
      - spell {word}
      - how do you spell {word}
conditions: []
actions:
  - set_conversation_response: "{{ trigger.slots.word |  list | join(' ') }}"
mode: single

I feel like this is so simple that it should just be a baked in sentence!

brazen glacier
# lime cliff I used to do this relatively often with Alexa and have been missing it, then I j...

you should use triple backticks around code so it shows as a code block

like
this

but to speak directly to your question. I feel adding an intent for this type of thing would be out of scope for the voice intent library in a similar way to how THIS one is explained in the comments there.
however you could open a feature request on the projects GitHub discussions HERE which may spark some discussion about a way to implement it.

lime cliff
#

I guess I was partially just posting so others saw how easy it was, and not complaining that it should be standard

#

There are a few other simple "skills" that I would like, but I'm not sure the best way to implement them (specifically not with AI)

brazen glacier
lime cliff
#

Simple math and unit conversions like in that first link would be awrsome

#

Ah. I'm on mobile, and not a big discord user, so that wasn't super clear to me. Thanks for letting me know

brazen glacier
#

I run a local llm which would handle that sort of thing

lime cliff
#

Are we to assume that Alexa was using an LLM 10 years ago for unit conversions and simple math?

#

Or the logic was just so complex to code that we haven't figured it out?

brazen glacier
#

it was probably baked into an intent system that far back to be fair.

lime cliff
#

Yeah, which makes it seem silly to not have

brazen glacier
# lime cliff Yeah, which makes it seem silly to not have

yeah, I do agree with you. spelling is fairly simple but totally reliant on the STT engine.
maths could get a bit more complicated to do.

remember that amazon was self improving because they were harvesting the data of every use of the device. whereas HA doesn't do that so it has to be build from the ground up considering different ways it may get asked. and STT might respond with different things (e.g. 2/two/to/too) and this has to be considered for multiple languages.

lime cliff
#

Yeah. I guess for my simple spelling example, one wouldn't reasonably expect it to get 2/to/two

brazen glacier
#

whereas a llm would be able to work it out from context most likely

lime cliff
#

I just miss my kitchen measurement conversions ๐Ÿ™‚

#

I don't think I will ever be convinced to have an LLM do something that code should do though...

brazen glacier
#

putting a FR in for "basic commands for maths/spelling/etc..." might bring some attention to the idea specially if it gets a few upvotes on the discussions page.

brazen glacier
lime cliff
#

I also like how simple definitions ended up being:

alias: voice definitions
description: ""
triggers:
  - trigger: conversation
    command: define {word}
conditions: []
actions:
  - action: rest_command.dictionary_api
    metadata: {}
    data:
      word: "{{ trigger.slots.word }}"
    response_variable: dictionary_response
  - set_conversation_response: >-
      {{
      dictionary_response['content'][0]['meanings'][0]['definitions'][0]['definition']
      }}
mode: single

With just this in config

rest_command:
  dictionary_api:
    url: "https://api.dictionaryapi.dev/api/v2/entries/en/{{ word }}"
    method: GET
    timeout: 10
brazen glacier
lime cliff
#

Yeah yeah I get it. I'll leave the support channel ๐Ÿ˜‰ Ok it is late. I'll think about that FR. Thanks for your time!