AVA
Desktop client for the AVA vocal assistant.
Installation
To write
Organization
Components architecture
The AVA core is an orchestration of a bunch of components, all responsible of a single concern. These components are executed in different threads, and communicating from one to another by using queue.Queue from the Python standard library. Components have a loose coupling and a tight cohesion, which mean they are not directly connected but use a well defined protocol. Each component is a class extending _BaseComponents and must implement a run method. They may implement a setup method that will be called after the __init__ and before the run. The method run of the component is by default executed in an infinite loop. This allow the component to have a very straight forward approach: execute a piece of code for one element popped out the queue. Some components might want to implement they own way to keep the thread alive: the component must inherit the class RunOneTime from ava/components.py.
Threading components
The components should not thread themselves. They should respond to a single concern. The main thread of the AVA Core is responsible of creating all the required threads and instantiate the component in it. The ComponentManager is the class responsible of it. The API is simple, add_component adds a component to the list, start_all instantiate and run all the components in a thread fashion.
Queues
The queues are singleton objects, each component can freely get the instance of a queue and send messages. All the queues are defined in ava/queues.py. A queue should have a single purpose, and only one consumer. Multiple components can write in a queue, but only one can read from it. Most of components passively read from that queue, and react when data arrives.
Components
Speech to Text
The Speech to Text is responsible to convert the speech to a text format. It listens the chosen input (raw from the computer or through mobile app), then translate it into a string, and sends that string to the QueueCommand queue.
Executor
The Executor is responsible to analyze the command of the user and dispatch the action to the right component. It listens to the QueueCommand queue and write to :
QueuePluginif the command is an action for the plugin manager
Plugin manager
The Plugin manager is responsible to install, configure and forward commands to the plugins. It listens to the QueuePlugin queue for incoming command, then selections the right plugin to execute it. The result of the command can be communicate to the Text to Speech component by writing it to the QueueTtS queue.
Text to Speech
The Text to Speech gives a vocal response to the user. It listen to the QueueTtS for incoming text to read.