Release/v0.8.0

[cssquirrel]

Added

• Added new "Enforcement" section to settings. The primary purpose of this is to help filter against messages where the LLM returns a response indicating it is a language model. After doing this, it often fails to remember its personality. It can also be used to filter unwanted content from message responses in general, by prompting a request for a new response.

Changed

• Moved position of button bar and status text to give more room for both.

Fixed

• No longer show "X more text before speaking" in status if voice transcription isn't active.
• Toggling off Speech Synthesis button now prevents Sock from talking out loud when responding.
• "Thinking" visual will continue while speech synthesis is running and end before playback, so
  so the entire "process user text -> go to LLM for response -> convert response to speech" process
  smoothly has one whole thinking phase instead of a visual gap between LLM response and talking.