@volley/recognition-client-sdk 0.1.707 → 0.1.767

package/dist/index.js

@@ -3747,6 +3747,8 @@ var RecognitionProvider;
   RecognitionProvider2["CARTESIA"] = "cartesia";
   RecognitionProvider2["DASHSCOPE"] = "dashscope";
   RecognitionProvider2["BEDROCK"] = "bedrock";
+  RecognitionProvider2["INWORLD_STT"] = "inworld-stt";
+  RecognitionProvider2["AWS_TRANSCRIBE"] = "aws-transcribe";
   RecognitionProvider2["TEST_ASR_PROVIDER_QUOTA"] = "test-asr-provider-quota";
   RecognitionProvider2["TEST_ASR_STREAMING"] = "test-asr-streaming";
 })(RecognitionProvider || (RecognitionProvider = {}));
@@ -3819,6 +3821,14 @@ var BedrockModel;
   BedrockModel2["VOXTRAL_MINI_3B_2507"] = "mistral.voxtral-mini-3b-2507";
   BedrockModel2["VOXTRAL_SMALL_24B_2507"] = "mistral.voxtral-small-24b-2507";
 })(BedrockModel || (BedrockModel = {}));
+var InworldSttModel;
+(function(InworldSttModel2) {
+  InworldSttModel2["INWORLD_STT_1"] = "inworld/inworld-stt-1";
+})(InworldSttModel || (InworldSttModel = {}));
+var AwsTranscribeModel;
+(function(AwsTranscribeModel2) {
+  AwsTranscribeModel2["DEFAULT"] = "default";
+})(AwsTranscribeModel || (AwsTranscribeModel = {}));
 var SelfServeVllmModel;
 (function(SelfServeVllmModel2) {
   SelfServeVllmModel2["QWEN3_ASR_1_7B"] = "qwen3-asr-1.7b";
@@ -3944,9 +3954,9 @@ var ErrorResultSchemaV1 = z.object({
   // Detailed description
 });
 var ClientControlActionV1;
-(function(ClientControlActionV13) {
-  ClientControlActionV13["READY_FOR_UPLOADING_RECORDING"] = "ready_for_uploading_recording";
-  ClientControlActionV13["STOP_RECORDING"] = "stop_recording";
+(function(ClientControlActionV12) {
+  ClientControlActionV12["READY_FOR_UPLOADING_RECORDING"] = "ready_for_uploading_recording";
+  ClientControlActionV12["STOP_RECORDING"] = "stop_recording";
 })(ClientControlActionV1 || (ClientControlActionV1 = {}));
 var ClientControlActionsV1 = z.nativeEnum(ClientControlActionV1);
 var ClientControlMessageSchemaV1 = z.object({
@@ -3979,6 +3989,8 @@ var AudioMetricsResultSchemaV1 = z.object({
   maxVolume: z.number(),
   minVolume: z.number(),
   avgVolume: z.number(),
+  peakVolumeDb: z.number().nullable(),
+  avgVolumeDb: z.number().nullable(),
   silenceRatio: z.number(),
   clippingRatio: z.number(),
   snrEstimate: z.number().nullable(),
@@ -3995,7 +4007,8 @@ var RecognitionResultSchemaV1 = z.discriminatedUnion("type", [
   // P1 - P2
   FunctionCallResultSchemaV1,
   ClientControlMessageSchemaV1,
-  SessionConfiguredSchemaV1
+  SessionConfiguredSchemaV1,
+  AudioMetricsResultSchemaV1
 ]);
 
 // ../../libs/types/dist/provider-transcription.types.js
@@ -4441,6 +4454,9 @@ var ASRRequestSchemaV1 = z.object({
   prefixMode: z.nativeEnum(PrefixMode).optional().default(PrefixMode.NONE),
   prefixId: z.string().optional(),
   prefixTextToRemove: z.array(z.string()).optional(),
+  // Streaming audio metrics opt-in: when > 0, server emits AudioMetrics results throttled to this interval (ms).
+  // Undefined / 0 disables streaming audio metrics (final metrics still embedded in Metadata).
+  audioMetricsIntervalMs: z.number().optional(),
   // Debug options (FOR DEBUG/TESTING ONLY - not for production use)
   debugCommand: RequestDebugCommandSchema
 });
@@ -5289,6 +5305,7 @@ var MessageHandler = class {
   /**
    * Handle incoming WebSocket message
    */
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
   handleMessage(msg) {
     if (this.callbacks.logger) {
       this.callbacks.logger("debug", "[RecogSDK] Received WebSocket message", {
@@ -5327,6 +5344,9 @@ var MessageHandler = class {
       case RecognitionResultTypeV1.SESSION_CONFIGURED:
         this.callbacks.onSessionConfigured?.(msgData);
         break;
+      case RecognitionResultTypeV1.AUDIO_METRICS:
+        this.callbacks.onAudioMetrics?.(msgData);
+        break;
       default:
         if (this.callbacks.logger) {
           this.callbacks.logger("debug", "[RecogSDK] Unknown message type", { type: msgType });
@@ -5474,6 +5494,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       onMetadata: config.onMetadata || (() => {
       }),
       onSessionConfigured: config.onSessionConfigured,
+      onAudioMetrics: config.onAudioMetrics,
       onError: config.onError || (() => {
       }),
       onConnected: config.onConnected || (() => {
@@ -5502,6 +5523,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       onError: this.config.onError,
       onControlMessage: this.handleControlMessage.bind(this),
       onSessionConfigured: this.config.onSessionConfigured,
+      onAudioMetrics: this.config.onAudioMetrics,
       ...this.config.logger && { logger: this.config.logger }
     });
   }
@@ -5827,6 +5849,10 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
         // Include prefix text to remove if provided (for server-side prefix text removal)
         ...this.config.asrRequestConfig.prefixTextToRemove && {
           prefixTextToRemove: this.config.asrRequestConfig.prefixTextToRemove
+        },
+        // Streaming audio metrics opt-in (ms interval). Server only forwards metrics if > 0.
+        ...this.config.asrRequestConfig.audioMetricsIntervalMs !== void 0 && {
+          audioMetricsIntervalMs: this.config.asrRequestConfig.audioMetricsIntervalMs
         }
       };
       super.sendMessage(
@@ -6181,6 +6207,14 @@ var ConfigBuilder = class {
     this.config.onSessionConfigured = callback;
     return this;
   }
+  /**
+   * Set streaming audio metrics callback (optional).
+   * Only fires when `asrRequestConfig.audioMetricsIntervalMs > 0`.
+   */
+  onAudioMetrics(callback) {
+    this.config.onAudioMetrics = callback;
+    return this;
+  }
   /**
    * Set error callback
    */
@@ -6233,6 +6267,7 @@ var ConfigBuilder = class {
   /**
    * Set logger function
    */
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
   logger(logger) {
     this.config.logger = logger;
     return this;
@@ -6670,6 +6705,7 @@ function createSimplifiedVGFClient(config) {
 }
 export {
   AudioEncoding,
+  AwsTranscribeModel,
   BedrockModel,
   CartesiaModel,
   ClientControlActionV1,