# Upgrade guide
-----------

The best way to upgrade Parrot.PY to make sure nothing breaks is to download this repository as a zipfile and extract it as a seperate folder to your current version.
Then to copy over the data directory and change the config/config.py to your previous values.

Version 0.12.1
-----

As of this moment, the TalonVoice beta offers first party support of Parrot models. 
As such, the functionality to run Parrot alongside TalonVoice by communicating through the Talon repl has been removed.
Documentation for the integration has been rewritten.

If you are running the old style communication, it is recommended to get access to the Talon beta instead as the first party integration is vastly superior to the communication bridge we had before.

Version 0.12.0
-----

* Added support for MacOS and Linux ( X11 )
  It is now possible to run Parrot on Mac and Linux, seperate installation instructions have been added to each operating system
* Machinelearning improvements
  Improvements have been made across the board for training AudioNets. Including more user friendly training, robuster ensembles and an improved feature layer.
  Now your AudioNets will automatically be combined when training them.
* Simplified recording
  I've removed several options that were largely unused, so that the only thresholds you have to fill in for recording are power and frequency.
* Improved configuration debugging
  I've added several more error messages that will make it easier to debug why Parrot won't start with your current configuration. This will hopefully make less people get stuck on running play.py

To upgrade from 0.11 up, you need to move your current config file and your personal modes to data/code, anything else should stay the same.
Also, make sure you have pytorch installed as it is required by default.


Version 0.11.0
-----

* Added IPC options to communicate with Parrot and other programs
  Using shared memory, it is now possible to pause, resume, quit, switch modes and so much more from outside and inside of Parrot.
* Improved the performance of overlay switching.
  Now file switching uses shared memory rather than being file based. This was a major bottleneck when switching overlays a lot.
* Added microphone reconnection.
  When your microphone gets disconnected during play time, it will attempt to reconnect every once in a while to find your microphone again without crashing Parrot.
* Improved mode and classifier switching. 
  Now allows for classifiers with different audio settings like 44.1kHz or 16kHz, and different feature engineering settings, to run seemlessly after one another if a switch is requested.
* Added Talon Voice communication code to connect your own noises to Talon actions

Version 0.10.1
-----

* Added DirectX key inputs
* Added setting to delay key releases for games and emulators that have a hard time detecting key presses

To upgrade from 0.10.0, simply install pydirectinput using:

```
pip3 install pydirectinput
```

Version 0.10.0
-----
* Added BaseMode and new and improved pattern detector
* Fixed length of audio recordings and added a source file for complete recording
* Added improved feature engineering strategies
* Added audio conversion menu for advanced usecases
* Removed some former models that are no longer in use
* Implemented a structure for the classifiers to make upgrading less prone to breaking on updating the Parrot.PY version
* Improved documentation and tutorials

Upgrading from before 0.10.0
----

Note that because in the versions before 0.10.0 there was little concern for backwards compatibility, your current models might break when run in the 0.10.0 and above version.
Your recorded data might also not match up with the length of the files when recorded now. 
That means you might have to rerecord your audio and retrain your models.

My sincerest apologies for this inconvenience. As of version 0.10.0, there is a much better system in place that should avoid these problems in the future.

Legacy classifiers are kept in the legacy_models folder in case you need them for older trained models.
Simply replace them where you see fit in the corresponding files they refer to.