Overview

Fixed a fundamental issue in the Markov analysis. Added a whole bunch of UI and control logic for switching between training and generation modes.

Fixed Fundamental Markov Problems

• There was an error in the way I was randomly selecting an initial state to use upon sequence generation. Since everything is written in JavaScript, I am able to create unit tests in node.js allowing me to much more quickly test components and find issues like this.

UI Features

• Added a bunch of UI buttons:
  • for enabling / disabling auto “response”
  • for continuously generating a response on demand when button is held
  • for enabling / disabling when training is happening
  • for training on demand when button is held
  • for enabling / disabling MIDI passthrough (this is not really useful currently because there is a delay)

Issues

• Because of limitations of the Max for Live API, the plugin device will require that particular clips are set up on the track with particular “follow mode” attributes. If this was exposed in the API I could programmatically set this up, but currently it is not available.