This document provides a brief overview of the Voysis Android SDK. This is an Android library that facilitates sending voice queries to a Voysis instance. The SDK streams audio from the device microphone to the Voysis backend servers when called by the client application.

The full documentation for the voysis api can be found here: Voysis Developer Documentation

minSdkVersion 19

The Service class is the main interface used to process voice recognition requests. It is accessed via the ServiceProvider().make(context , config) method. See demo application for more details. The sdk supports both REST and Websocket connections depending on how the config file is setup. NOTE: It is preferable to use the Websocket connection as it offers more flexibility. The sdk uses Okhttp under the hood for network connections.

To create a service instance you need to provide config information at construction time. The config file is comprised of several fields.

• isVadEnabled - Boolean: [VAD] stands for Voice Activation Detection. If set to true the library will automatically detect when the user has stopped speaking and process requests. This is made possible via an okHttp Websocket connection. See Websocket docs for more info on our Websocket api. If turned off the library will default to a REST interface and the user will need to manually call service.finish() to process requests. See REST docs for more info on our REST api. NOTE: It is preferable to have isVadEnabled set to true as the Websocket provides the user with more functionality and flexibility.

• url - URL: This is the endpoint that network requests will be executed against. The url is provided when your Voysis service is delivered. To sign up for a Voysis service visit our homepage

• refreshToken - String: The refresh token is used for authenticating unique users and for refreshing session tokens. This is all taken care of from within the library once the refreshToken is provided. For information on how to generate a refresh token see here

• userId - String, Optional: This value uniquely identifies a user. This value must be generated by the client and ideally used for the lifetime of that client. For information on userId see here

• The first step is to create a Service instance using the android context and config object.

 val config = DataConfig(isVadEnabled = true, url = URL("INSERT_URL"), refreshToken = "INSERT_TOKEN", userId = "GENERATE_UNIQUE_ID")
 val service = ServiceProvider().make(context, config)

• Next: to make a request, call service.startAudioQuery with the mandatory Callback parameter and optional voysis Context (See context section below for details).

Note: This call issues a network request and should not be done on the main thread. Note also that callbacks may not occur on the main thread.

class ExampleActivity : AppCompatActivity(), Callback {

    private fun onClick() {
        service.startAudioQuery(context = context, callback = this)
    }

    override fun success(response: StreamResponse) {
        // Mandatory: called when final response returned from server.
    }

    override fun failure(error: VoysisException) {
        // Mandatory: called when any error occurs
    }

    override fun recordingStarted() {
	/**
	 * Optional: called when microphone begins recording.
	 * NOTE: It is recommended that when this callback occurs, the user be notified through the ui that they can now       
	 * start speaking. 
	 * Failure to implement proper ui may result in the user speaking before the microphone has opened resulting in       
	 * incomplete audio.
	 */
    }

    override fun queryResponse(query: QueryResponse) {
        //Optional: called when successful connection is made to the server.
    }

    override fun recordingFinished(reason: FinishedReason) {
        //Optional: called when recording stops. Includes finishedReason enum.
    }
    
    override fun audioData(buffer: ByteBuffer) {
        //Optional: returns audio data to the user that can be used generating for dynamic animations, analytics etc.
    }
}

One of the features of the Voysis service is that it can use the StreamResponse.context (Not to be confused with Android context) to refine and improve subsequent requests. In order to avail of this the developer must store the context response from the AudioStreamResponse and send it in the following startAudioQuery(context , callback) request.

Step 1. Add the JitPack repository to your build file. Add it in your root build.gradle at the end of repositories:

allprojects {
	repositories {
		...
		maven { url 'https://jitpack.io' }
	}
}

Step 2. Add the dependency

dependencies {
        implementation 'com.github.voysis:voysis-android:2.x.x'
}

If using proguard you will need add the recommended proguard rules for okhttp

To build the project locally, clone the project and open the root directory in Android Studio.

The user must accept Microphone permissions prior to using this library. Failing to do so will result in an error.