NOTE: I recently changed the main thread functionality so it does not end until the TTS is completed, this has probably made some of my previous code redundant/overly complex. I am merging this now as the current main branch has major issues which are fixed in this branch.
Cleaned up (kinda) the canceling functionality, you can now cancel the assistant at any point while it is responding or waiting to respond
Now if you cut off the assistant while it is speaking the string --> USER CUT OFF RESPONSE <-- is added to that point in the prompt and only up to that point in the message is added to the message history-- This is so the AI does not have text in its message history that the user did not hear
Other fixes around canceling TTS and completions.
moved config file into git ignore
added sanitize_text text function that removed character that are not suitable for TTS and replaces some with words example: '/'->'slash'
Updated trim messages function to ensure the first message is always a user message, this is a requirement of the anthropic api
Attempted to improve the prompt and make it use the clipboard less, unless asked.
NOTE: I recently changed the main thread functionality so it does not end until the TTS is completed, this has probably made some of my previous code redundant/overly complex. I am merging this now as the current main branch has major issues which are fixed in this branch.
Cleaned up (kinda) the canceling functionality, you can now cancel the assistant at any point while it is responding or waiting to respond
Now if you cut off the assistant while it is speaking the string
--> USER CUT OFF RESPONSE <--is added to that point in the prompt and only up to that point in the message is added to the message history-- This is so the AI does not have text in its message history that the user did not hearOther fixes around canceling TTS and completions.
moved config file into git ignore
added sanitize_text text function that removed character that are not suitable for TTS and replaces some with words example: '/'->'slash'
Updated trim messages function to ensure the first message is always a user message, this is a requirement of the anthropic api
Attempted to improve the prompt and make it use the clipboard less, unless asked.