Merge pull request #88 from anam-org/alpha

feat: disable input audio option

Author: ao-anam

src/AnamClient.ts

@@ -45,7 +45,11 @@ export default class AnamClient {
       options,
     );
     if (configError) {
-      throw new Error(configError);
+      throw new ClientError(
+        configError,
+        ErrorCode.CLIENT_ERROR_CODE_CONFIGURATION_ERROR,
+        400,
+      );
     }
 
     this.personaConfig = personaConfig;
@@ -114,6 +118,9 @@ export default class AnamClient {
 
     // Validate voice detection configuration
     if (options?.voiceDetection) {
+      if (options.disableInputAudio) {
+        return 'Voice detection is disabled because input audio is disabled. Please set disableInputAudio to false to enable voice detection.';
+      }
       // End of speech sensitivity must be a number between 0 and 1
       if (options.voiceDetection.endOfSpeechSensitivity !== undefined) {
         if (typeof options.voiceDetection.endOfSpeechSensitivity !== 'number') {
@@ -183,8 +190,11 @@ export default class AnamClient {
           iceServers,
           inputAudio: {
             inputAudioState: this.inputAudioState,
-            userProvidedMediaStream: userProvidedAudioStream,
+            userProvidedMediaStream: this.clientOptions?.disableInputAudio
+              ? undefined
+              : userProvidedAudioStream,
             audioDeviceId: this.clientOptions?.audioDeviceId,
+            disableInputAudio: this.clientOptions?.disableInputAudio,
           },
         },
         this.publicEventEmitter,
@@ -203,9 +213,9 @@ export default class AnamClient {
     return sessionId;
   }
 
-  private async startSessionIfNeeded(userProvidedMediaStream?: MediaStream) {
+  private async startSessionIfNeeded(userProvidedAudioStream?: MediaStream) {
     if (!this.sessionId || !this.streamingClient) {
-      await this.startSession(userProvidedMediaStream);
+      await this.startSession(userProvidedAudioStream);
 
       if (!this.sessionId || !this.streamingClient) {
         throw new ClientError(
@@ -223,6 +233,11 @@ export default class AnamClient {
   public async stream(
     userProvidedAudioStream?: MediaStream,
   ): Promise<MediaStream[]> {
+    if (this.clientOptions?.disableInputAudio && userProvidedAudioStream) {
+      console.warn(
+        'AnamClient:Input audio is disabled. User provided audio stream will be ignored.',
+      );
+    }
     await this.startSessionIfNeeded(userProvidedAudioStream);
     if (this._isStreaming) {
       throw new Error('Already streaming');
@@ -261,10 +276,15 @@ export default class AnamClient {
   public async streamToVideoAndAudioElements(
     videoElementId: string,
     audioElementId: string,
-    userProvidedMediaStream?: MediaStream,
+    userProvidedAudioStream?: MediaStream,
   ): Promise<void> {
+    if (this.clientOptions?.disableInputAudio && userProvidedAudioStream) {
+      console.warn(
+        'AnamClient:Input audio is disabled. User provided audio stream will be ignored.',
+      );
+    }
     try {
-      await this.startSessionIfNeeded(userProvidedMediaStream);
+      await this.startSessionIfNeeded(userProvidedAudioStream);
     } catch (error) {
       if (error instanceof ClientError) {
         throw error;
@@ -340,14 +360,24 @@ export default class AnamClient {
   }
 
   public getInputAudioState(): InputAudioState {
+    if (this.clientOptions?.disableInputAudio) {
+      console.warn(
+        'AnamClient: Audio state will not be used because input audio is disabled.',
+      );
+    }
     // if streaming client is available, make sure our state is up to date
     if (this.streamingClient) {
       this.inputAudioState = this.streamingClient.getInputAudioState();
     }
     return this.inputAudioState;
   }
   public muteInputAudio(): InputAudioState {
-    if (this.streamingClient) {
+    if (this.clientOptions?.disableInputAudio) {
+      console.warn(
+        'AnamClient: Input audio is disabled. Muting input audio will have no effect.',
+      );
+    }
+    if (this.streamingClient && !this.clientOptions?.disableInputAudio) {
       this.inputAudioState = this.streamingClient.muteInputAudio();
     } else {
       this.inputAudioState = {
@@ -359,7 +389,12 @@ export default class AnamClient {
   }
 
   public unmuteInputAudio(): InputAudioState {
-    if (this.streamingClient) {
+    if (this.clientOptions?.disableInputAudio) {
+      console.warn(
+        'AnamClient: Input audio is disabled. Unmuting input audio will have no effect.',
+      );
+    }
+    if (this.streamingClient && !this.clientOptions?.disableInputAudio) {
       this.inputAudioState = this.streamingClient.unmuteInputAudio();
     } else {
       this.inputAudioState = {

src/lib/ClientError.ts

@@ -7,6 +7,7 @@ export enum ErrorCode {
   CLIENT_ERROR_CODE_SERVICE_BUSY = 'CLIENT_ERROR_CODE_SERVICE_BUSY',
   CLIENT_ERROR_CODE_NO_PLAN_FOUND = 'CLIENT_ERROR_CODE_NO_PLAN_FOUND',
   CLIENT_ERROR_CODE_UNKNOWN_ERROR = 'CLIENT_ERROR_CODE_UNKNOWN_ERROR',
+  CLIENT_ERROR_CODE_CONFIGURATION_ERROR = 'CLIENT_ERROR_CODE_CONFIGURATION_ERROR',
 }
 
 // TODO: Move to CoreApiRestClient if we have a pattern for not exposing this

src/modules/StreamingClient.ts

@@ -38,6 +38,7 @@ export class StreamingClient {
   private audioStream: MediaStream | null = null;
   private inputAudioState: InputAudioState = { isMuted: false };
   private audioDeviceId: string | undefined;
+  private disableInputAudio: boolean;
 
   constructor(
     sessionId: string,
@@ -53,6 +54,7 @@ export class StreamingClient {
     if (options.inputAudio.userProvidedMediaStream) {
       this.inputAudioStream = options.inputAudio.userProvidedMediaStream;
     }
+    this.disableInputAudio = options.inputAudio.disableInputAudio === true;
     // register event handlers
     this.internalEventEmitter.addListener(
       InternalEvent.WEB_SOCKET_OPEN,
@@ -413,42 +415,45 @@ export class StreamingClient {
      * Audio
      *
      * If the user hasn't provided an audio stream, capture the audio stream from the user's microphone and send it to the peer connection
+     * If input audio is disabled we don't send any audio to the peer connection
      */
-    if (this.inputAudioStream) {
-      // verify the user provided stream has audio tracks
-      if (!this.inputAudioStream.getAudioTracks().length) {
-        throw new Error(
-          'StreamingClient - setupDataChannels: user provided stream does not have audio tracks',
-        );
-      }
-    } else {
-      const audioConstraints: MediaTrackConstraints = {
-        echoCancellation: true,
-      };
-
-      // If an audio device ID is provided in the options, use it
-      if (this.audioDeviceId) {
-        audioConstraints.deviceId = {
-          exact: this.audioDeviceId,
+    if (!this.disableInputAudio) {
+      if (this.inputAudioStream) {
+        // verify the user provided stream has audio tracks
+        if (!this.inputAudioStream.getAudioTracks().length) {
+          throw new Error(
+            'StreamingClient - setupDataChannels: user provided stream does not have audio tracks',
+          );
+        }
+      } else {
+        const audioConstraints: MediaTrackConstraints = {
+          echoCancellation: true,
         };
-      }
 
-      this.inputAudioStream = await navigator.mediaDevices.getUserMedia({
-        audio: audioConstraints,
-      });
-    }
+        // If an audio device ID is provided in the options, use it
+        if (this.audioDeviceId) {
+          audioConstraints.deviceId = {
+            exact: this.audioDeviceId,
+          };
+        }
 
-    // mute the audio tracks if the user has muted the microphone
-    if (this.inputAudioState.isMuted) {
-      this.muteAllAudioTracks();
+        this.inputAudioStream = await navigator.mediaDevices.getUserMedia({
+          audio: audioConstraints,
+        });
+      }
+
+      // mute the audio tracks if the user has muted the microphone
+      if (this.inputAudioState.isMuted) {
+        this.muteAllAudioTracks();
+      }
+      const audioTrack = this.inputAudioStream.getAudioTracks()[0];
+      this.peerConnection.addTrack(audioTrack, this.inputAudioStream);
+      // pass the stream to the callback if it exists
+      this.publicEventEmitter.emit(
+        AnamEvent.INPUT_AUDIO_STREAM_STARTED,
+        this.inputAudioStream,
+      );
     }
-    const audioTrack = this.inputAudioStream.getAudioTracks()[0];
-    this.peerConnection.addTrack(audioTrack, this.inputAudioStream);
-    // pass the stream to the callback if it exists
-    this.publicEventEmitter.emit(
-      AnamEvent.INPUT_AUDIO_STREAM_STARTED,
-      this.inputAudioStream,
-    );
 
     /**
      * Text
@@ -462,9 +467,7 @@ export class StreamingClient {
     dataChannel.onopen = () => {
       this.dataChannel = dataChannel ?? null;
     };
-    dataChannel.onclose = () => {
-      // TODO: should we set the data channel to null here?
-    };
+    dataChannel.onclose = () => {};
     // pass text message to the message history client
     dataChannel.onmessage = (event) => {
       const messageEvent = JSON.parse(event.data) as WebRtcTextMessageEvent;

src/types/AnamPublicClientOptions.ts

@@ -5,4 +5,5 @@ export interface AnamPublicClientOptions {
   api?: CoreApiRestClientOptions;
   voiceDetection?: VoiceDetectionOptions;
   audioDeviceId?: string;
+  disableInputAudio?: boolean;
 }

src/types/streaming/InputAudioOptions.ts

@@ -4,4 +4,5 @@ export interface InputAudioOptions {
   inputAudioState: InputAudioState;
   userProvidedMediaStream?: MediaStream;
   audioDeviceId?: string;
+  disableInputAudio?: boolean;
 }