Summary:

This pull request introduces several enhancements to the Interactive mode and configuration page. Notable changes include the addition of a "Push to Talk" mode, tweaks to the noise logic, and bug fixes. The configuration page has also been updated for a cleaner user experience.

Key Points:

Added support for Safari to use the new Interact mode.
Added "Push to Talk" mode to Interactive mode
Added new interact mode configuration options
- Added toggle for enabling/disabling Push to Talk mode.
- Added toggle for using Whisper for transcription
- Added slider for setting the Whisper model temperature
- Added audio playback speed setting
Fixed bug causing speech loop check to continue running after exiting Interactive mode
Fixed bug causing temperature slider values to not be loaded correctly on startup for GPT and Claude models
Added additional parameters to whisper model requests for increased accuracy.
Tweaked noise logic for improved performance
Updated configuration page to:
- Not unselect the current model when selecting the General Config menu item
- Collapse all config sections by default for a cleaner look
Replaced long press gesture with double tap action to activate context menu on mobile for improved consistency
Updated configuration page to remove the associated input box for every slider control. This provides a nice cleanup of the UI. Replaced with a centered label also displaying the current value for a clean minimal look.
Reduced max width of messages so they can't completely fill the width of screen. Previously when that happened it would lead to somewhat confusing UI experience.

Benefits:

Improved user experience in Interactive mode with "Push to Talk" feature
Increased flexibility with optional Whisper transcription
Enhanced performance and reliability with bug fixes and noise logic tweaks
Simplified configuration page navigation and reduced clutter
Improved mobile usability with consistent double tap gesture for context menu activation
Improved Interact Mode browser compatibility

UI Changes

Auto Generated Notes (Do Not Change)

Contents (#135)

Uncategorised!

Added a dynamic noise floor averaging logic so that the speech detection threshold is the same across different microphone sensitivies
Added a push to talk mode for interact mode that can be enabled/disabled in general settings.
Default config sections to be closed for a cleaner default look.
Update version info
Update the noise floor average to be a rolling average of 2500 rms value samples
Add addition interact mode config settings for tts-model selection and audio playback speed.
Removed the input box present with every slider control. and replaced with a more minimal label with the current value in text beside it. This greatly reduces the clutter in the configuration page in general.
Slight color and list item tweaks in ConversationDialog
Added additional model settings for whisper requests to improve accuracy.

fingerthief / minimal-chat

Interactive Mode Updates and Config Page Refinements #135

UI Changes

Auto Generated Notes (Do Not Change)

Contents (#135)

Uncategorised!