@volley/recognition-client-sdk 0.1.707 → 0.1.767

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -3747,6 +3747,8 @@ var RecognitionProvider;
3747
3747
  RecognitionProvider2["CARTESIA"] = "cartesia";
3748
3748
  RecognitionProvider2["DASHSCOPE"] = "dashscope";
3749
3749
  RecognitionProvider2["BEDROCK"] = "bedrock";
3750
+ RecognitionProvider2["INWORLD_STT"] = "inworld-stt";
3751
+ RecognitionProvider2["AWS_TRANSCRIBE"] = "aws-transcribe";
3750
3752
  RecognitionProvider2["TEST_ASR_PROVIDER_QUOTA"] = "test-asr-provider-quota";
3751
3753
  RecognitionProvider2["TEST_ASR_STREAMING"] = "test-asr-streaming";
3752
3754
  })(RecognitionProvider || (RecognitionProvider = {}));
@@ -3819,6 +3821,14 @@ var BedrockModel;
3819
3821
  BedrockModel2["VOXTRAL_MINI_3B_2507"] = "mistral.voxtral-mini-3b-2507";
3820
3822
  BedrockModel2["VOXTRAL_SMALL_24B_2507"] = "mistral.voxtral-small-24b-2507";
3821
3823
  })(BedrockModel || (BedrockModel = {}));
3824
+ var InworldSttModel;
3825
+ (function(InworldSttModel2) {
3826
+ InworldSttModel2["INWORLD_STT_1"] = "inworld/inworld-stt-1";
3827
+ })(InworldSttModel || (InworldSttModel = {}));
3828
+ var AwsTranscribeModel;
3829
+ (function(AwsTranscribeModel2) {
3830
+ AwsTranscribeModel2["DEFAULT"] = "default";
3831
+ })(AwsTranscribeModel || (AwsTranscribeModel = {}));
3822
3832
  var SelfServeVllmModel;
3823
3833
  (function(SelfServeVllmModel2) {
3824
3834
  SelfServeVllmModel2["QWEN3_ASR_1_7B"] = "qwen3-asr-1.7b";
@@ -3944,9 +3954,9 @@ var ErrorResultSchemaV1 = z.object({
3944
3954
  // Detailed description
3945
3955
  });
3946
3956
  var ClientControlActionV1;
3947
- (function(ClientControlActionV13) {
3948
- ClientControlActionV13["READY_FOR_UPLOADING_RECORDING"] = "ready_for_uploading_recording";
3949
- ClientControlActionV13["STOP_RECORDING"] = "stop_recording";
3957
+ (function(ClientControlActionV12) {
3958
+ ClientControlActionV12["READY_FOR_UPLOADING_RECORDING"] = "ready_for_uploading_recording";
3959
+ ClientControlActionV12["STOP_RECORDING"] = "stop_recording";
3950
3960
  })(ClientControlActionV1 || (ClientControlActionV1 = {}));
3951
3961
  var ClientControlActionsV1 = z.nativeEnum(ClientControlActionV1);
3952
3962
  var ClientControlMessageSchemaV1 = z.object({
@@ -3979,6 +3989,8 @@ var AudioMetricsResultSchemaV1 = z.object({
3979
3989
  maxVolume: z.number(),
3980
3990
  minVolume: z.number(),
3981
3991
  avgVolume: z.number(),
3992
+ peakVolumeDb: z.number().nullable(),
3993
+ avgVolumeDb: z.number().nullable(),
3982
3994
  silenceRatio: z.number(),
3983
3995
  clippingRatio: z.number(),
3984
3996
  snrEstimate: z.number().nullable(),
@@ -3995,7 +4007,8 @@ var RecognitionResultSchemaV1 = z.discriminatedUnion("type", [
3995
4007
  // P1 - P2
3996
4008
  FunctionCallResultSchemaV1,
3997
4009
  ClientControlMessageSchemaV1,
3998
- SessionConfiguredSchemaV1
4010
+ SessionConfiguredSchemaV1,
4011
+ AudioMetricsResultSchemaV1
3999
4012
  ]);
4000
4013
 
4001
4014
  // ../../libs/types/dist/provider-transcription.types.js
@@ -4418,6 +4431,9 @@ var ASRRequestSchemaV1 = z.object({
4418
4431
  prefixMode: z.nativeEnum(PrefixMode).optional().default(PrefixMode.NONE),
4419
4432
  prefixId: z.string().optional(),
4420
4433
  prefixTextToRemove: z.array(z.string()).optional(),
4434
+ // Streaming audio metrics opt-in: when > 0, server emits AudioMetrics results throttled to this interval (ms).
4435
+ // Undefined / 0 disables streaming audio metrics (final metrics still embedded in Metadata).
4436
+ audioMetricsIntervalMs: z.number().optional(),
4421
4437
  // Debug options (FOR DEBUG/TESTING ONLY - not for production use)
4422
4438
  debugCommand: RequestDebugCommandSchema
4423
4439
  });
@@ -5221,6 +5237,7 @@ var MessageHandler = class {
5221
5237
  /**
5222
5238
  * Handle incoming WebSocket message
5223
5239
  */
5240
+ // eslint-disable-next-line @typescript-eslint/no-explicit-any
5224
5241
  handleMessage(msg) {
5225
5242
  if (this.callbacks.logger) {
5226
5243
  this.callbacks.logger("debug", "[RecogSDK] Received WebSocket message", {
@@ -5259,6 +5276,9 @@ var MessageHandler = class {
5259
5276
  case RecognitionResultTypeV1.SESSION_CONFIGURED:
5260
5277
  this.callbacks.onSessionConfigured?.(msgData);
5261
5278
  break;
5279
+ case RecognitionResultTypeV1.AUDIO_METRICS:
5280
+ this.callbacks.onAudioMetrics?.(msgData);
5281
+ break;
5262
5282
  default:
5263
5283
  if (this.callbacks.logger) {
5264
5284
  this.callbacks.logger("debug", "[RecogSDK] Unknown message type", { type: msgType });
@@ -5380,6 +5400,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
5380
5400
  onMetadata: config.onMetadata || (() => {
5381
5401
  }),
5382
5402
  onSessionConfigured: config.onSessionConfigured,
5403
+ onAudioMetrics: config.onAudioMetrics,
5383
5404
  onError: config.onError || (() => {
5384
5405
  }),
5385
5406
  onConnected: config.onConnected || (() => {
@@ -5408,6 +5429,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
5408
5429
  onError: this.config.onError,
5409
5430
  onControlMessage: this.handleControlMessage.bind(this),
5410
5431
  onSessionConfigured: this.config.onSessionConfigured,
5432
+ onAudioMetrics: this.config.onAudioMetrics,
5411
5433
  ...this.config.logger && { logger: this.config.logger }
5412
5434
  });
5413
5435
  }
@@ -5733,6 +5755,10 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
5733
5755
  // Include prefix text to remove if provided (for server-side prefix text removal)
5734
5756
  ...this.config.asrRequestConfig.prefixTextToRemove && {
5735
5757
  prefixTextToRemove: this.config.asrRequestConfig.prefixTextToRemove
5758
+ },
5759
+ // Streaming audio metrics opt-in (ms interval). Server only forwards metrics if > 0.
5760
+ ...this.config.asrRequestConfig.audioMetricsIntervalMs !== void 0 && {
5761
+ audioMetricsIntervalMs: this.config.asrRequestConfig.audioMetricsIntervalMs
5736
5762
  }
5737
5763
  };
5738
5764
  super.sendMessage(