Cleaned up and documented some of the constants and their requirements in SpeechToText.jsx - That'd be a great area to refactor in the future. More than a few places relied on the provider's name to match the key.

I made this a little more explicit by consolidating the SpeechToTextType, but the various accessors probably should just be looking at the key or the name and not both.

STILL NEEDED:

• I am confident nothing broke with the OpenAI nor AssemblyAI implementations, but I don't have credentials to test manually and I didn't see a testing framework in place to assert against.

• SpeechToText.jsx
  import localAISVG from '@/assets/images/robot.png'
  This is a random picture that was in the @/assets/images/ dir, change to a LocalAI image

• The SpeechToTextType is duplicated in both ui/src/ui-component/extended/SpeechToText.jsx and in components/src/speechToText.ts -- not sure where this ought to live. I couldn't import from one to the other.

Addressed as a known duplication with comment explaining such in 197ec72

• The model for LocalAISTT defaults to whisper if not provided - I assume this value comes from OpenAI's client.

SUGGESTIONS/QUESTIONS:

• Should the placeholder be the default value when filling out a configuration? Or just serve as reminder text?
• Specifcally for LocalAISTT, if a model isn't provided should it default to whisper-1 (the default model in the AIO packages of LocalAI)

Screenshots: