@volley/recognition-client-sdk 0.1.622 → 0.1.689

package/dist/recog-client-sdk.browser.js

@@ -3737,11 +3737,14 @@ var RecognitionProvider;
   RecognitionProvider2["DEEPGRAM"] = "deepgram";
   RecognitionProvider2["ELEVENLABS"] = "elevenlabs";
   RecognitionProvider2["FIREWORKS"] = "fireworks";
+  RecognitionProvider2["GLADIA"] = "gladia";
   RecognitionProvider2["GOOGLE"] = "google";
   RecognitionProvider2["GEMINI_BATCH"] = "gemini-batch";
   RecognitionProvider2["OPENAI_BATCH"] = "openai-batch";
+  RecognitionProvider2["SELF_SERVE_VLLM"] = "self-serve-vllm";
   RecognitionProvider2["OPENAI_REALTIME"] = "openai-realtime";
   RecognitionProvider2["MISTRAL_VOXTRAL"] = "mistral-voxtral";
+  RecognitionProvider2["CARTESIA"] = "cartesia";
   RecognitionProvider2["DASHSCOPE"] = "dashscope";
   RecognitionProvider2["TEST_ASR_PROVIDER_QUOTA"] = "test-asr-provider-quota";
   RecognitionProvider2["TEST_ASR_STREAMING"] = "test-asr-streaming";
@@ -3783,10 +3786,13 @@ var FireworksModel;
   FireworksModel2["WHISPER_V3"] = "whisper-v3";
   FireworksModel2["WHISPER_V3_TURBO"] = "whisper-v3-turbo";
 })(FireworksModel || (FireworksModel = {}));
+var GladiaModel;
+(function(GladiaModel2) {
+  GladiaModel2["SOLARIA_1"] = "solaria-1";
+})(GladiaModel || (GladiaModel = {}));
 var ElevenLabsModel;
 (function(ElevenLabsModel2) {
   ElevenLabsModel2["SCRIBE_V2_REALTIME"] = "scribe_v2_realtime";
-  ElevenLabsModel2["SCRIBE_V1"] = "scribe_v1";
 })(ElevenLabsModel || (ElevenLabsModel = {}));
 var OpenAIRealtimeModel;
 (function(OpenAIRealtimeModel2) {
@@ -3797,11 +3803,20 @@ var MistralVoxtralModel;
 (function(MistralVoxtralModel2) {
   MistralVoxtralModel2["VOXTRAL_MINI_REALTIME_2602"] = "voxtral-mini-transcribe-realtime-2602";
 })(MistralVoxtralModel || (MistralVoxtralModel = {}));
+var CartesiaModel;
+(function(CartesiaModel2) {
+  CartesiaModel2["INK_WHISPER"] = "ink-whisper";
+  CartesiaModel2["INK_WHISPER_20250604"] = "ink-whisper-2025-06-04";
+})(CartesiaModel || (CartesiaModel = {}));
 var DashScopeModel;
 (function(DashScopeModel2) {
   DashScopeModel2["QWEN3_ASR_FLASH_REALTIME_2602"] = "qwen3-asr-flash-realtime-2026-02-10";
   DashScopeModel2["QWEN3_ASR_FLASH_REALTIME"] = "qwen3-asr-flash-realtime";
 })(DashScopeModel || (DashScopeModel = {}));
+var SelfServeVllmModel;
+(function(SelfServeVllmModel2) {
+  SelfServeVllmModel2["QWEN3_ASR_1_7B"] = "qwen3-asr-1.7b";
+})(SelfServeVllmModel || (SelfServeVllmModel = {}));
 
 // ../../libs/types/dist/recognition-result-v1.types.js
 var RecognitionResultTypeV1;
@@ -3812,6 +3827,7 @@ var RecognitionResultTypeV1;
   RecognitionResultTypeV12["ERROR"] = "Error";
   RecognitionResultTypeV12["CLIENT_CONTROL_MESSAGE"] = "ClientControlMessage";
   RecognitionResultTypeV12["AUDIO_METRICS"] = "AudioMetrics";
+  RecognitionResultTypeV12["SESSION_CONFIGURED"] = "SessionConfigured";
 })(RecognitionResultTypeV1 || (RecognitionResultTypeV1 = {}));
 var TranscriptionResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.TRANSCRIPTION),
@@ -3874,6 +3890,9 @@ var MetadataResultSchemaV1 = z.object({
   costInUSD: z.number().default(0).optional(),
   // ASR API Type
   apiType: z.nativeEnum(ASRApiType).optional(),
+  // Provider identification
+  provider: z.string().optional(),
+  model: z.string().optional(),
   // ASR configuration as JSON string (no type validation)
   asrConfig: z.string().optional(),
   // Raw ASR metadata payload as provided by the provider (stringified if needed)
@@ -3930,6 +3949,22 @@ var ClientControlMessageSchemaV1 = z.object({
   action: ClientControlActionsV1
   // The control action to perform
 });
+var SessionConfiguredSchemaV1 = z.object({
+  type: z.literal(RecognitionResultTypeV1.SESSION_CONFIGURED),
+  audioUtteranceId: z.string(),
+  // Provider identification
+  provider: z.string().optional(),
+  model: z.string().optional(),
+  sampleRate: z.number().optional(),
+  encoding: z.string().optional(),
+  apiType: z.nativeEnum(ASRApiType).optional(),
+  isFallback: z.boolean().optional(),
+  // Original ASR request as JSON string (includes prefixMode, prefixId, etc.)
+  asrRequest: z.string().optional(),
+  // Provider-specific config as JSON string (raw config sent to provider API, e.g. Deepgram's punctuate/endpointing)
+  providerConfig: z.string().optional()
+  // Stringified JSON of provider-specific config
+});
 var AudioMetricsResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.AUDIO_METRICS),
   valid: z.boolean(),
@@ -3953,7 +3988,8 @@ var RecognitionResultSchemaV1 = z.discriminatedUnion("type", [
   ErrorResultSchemaV1,
   // P1 - P2
   FunctionCallResultSchemaV1,
-  ClientControlMessageSchemaV1
+  ClientControlMessageSchemaV1,
+  SessionConfiguredSchemaV1
 ]);
 
 // ../../libs/types/dist/provider-transcription.types.js
@@ -4396,8 +4432,9 @@ var RecognitionGameInfoSchema = z.object({
   questionAskedId: z.string().optional(),
   /** @deprecated Use questionAskedId instead. Kept for backward compatibility during migration. */
   questionAnswerId: z.string().optional(),
-  platform: z.string().optional()
-  // Platform for audio recording device (use the definition of platform teams)
+  platform: z.string().optional(),
+  experimentCohort: z.enum(["treatment", "control"]).optional()
+  // Experiment cohort, defaults to 'control' if not provided
 });
 var RecognitionQueryMetadataSchema = z.object({
   audioUtteranceId: z.string(),
@@ -5024,6 +5061,9 @@ function buildWebSocketUrl(config) {
   if (config.gameContext?.gamePhase) {
     url.searchParams.set("gamePhase", config.gameContext.gamePhase);
   }
+  if (config.experimentCohort) {
+    url.searchParams.set("experimentCohort", config.experimentCohort);
+  }
   return url.toString();
 }
 
@@ -5181,7 +5221,7 @@ var MessageHandler = class {
     }
     if (msg.data && typeof msg.data !== "object") {
       if (this.callbacks.logger) {
-        this.callbacks.logger("error", "[RecogSDK] Received primitive msg.data from server", {
+        this.callbacks.logger("warn", "[RecogSDK] Received primitive msg.data from server", {
           dataType: typeof msg.data,
           data: msg.data,
           fullMessage: msg
@@ -5206,6 +5246,9 @@ var MessageHandler = class {
       case RecognitionResultTypeV1.CLIENT_CONTROL_MESSAGE:
         this.callbacks.onControlMessage(msgData);
         break;
+      case RecognitionResultTypeV1.SESSION_CONFIGURED:
+        this.callbacks.onSessionConfigured?.(msgData);
+        break;
       default:
         if (this.callbacks.logger) {
           this.callbacks.logger("debug", "[RecogSDK] Unknown message type", { type: msgType });
@@ -5291,7 +5334,8 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       ...config.questionAnswerId && { questionAnswerId: config.questionAnswerId },
       ...config.platform && { platform: config.platform },
       ...config.gameContext && { gameContext: config.gameContext },
-      ...config.gameId && { gameId: config.gameId }
+      ...config.gameId && { gameId: config.gameId },
+      ...config.experimentCohort && { experimentCohort: config.experimentCohort }
     });
     super({
       url,
@@ -5324,6 +5368,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       }),
       onMetadata: config.onMetadata || (() => {
       }),
+      onSessionConfigured: config.onSessionConfigured,
       onError: config.onError || (() => {
       }),
       onConnected: config.onConnected || (() => {
@@ -5351,6 +5396,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       onMetadata: this.config.onMetadata,
       onError: this.config.onError,
       onControlMessage: this.handleControlMessage.bind(this),
+      onSessionConfigured: this.config.onSessionConfigured,
       ...this.config.logger && { logger: this.config.logger }
     });
   }
@@ -5503,7 +5549,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       blobToArrayBuffer(audioData).then((arrayBuffer) => {
         this.sendAudioInternal(arrayBuffer);
       }).catch((error) => {
-        this.log("error", "Failed to convert Blob to ArrayBuffer", error);
+        this.log("warn", "Failed to convert Blob to ArrayBuffer", error);
       });
       return;
     }
@@ -5543,7 +5589,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
    */
   async stopRecording() {
     if (this.state !== "ready" /* READY */) {
-      this.log("warn", "stopRecording called but not in READY state", { state: this.state });
+      this.log("info", "stopRecording called but not in READY state", { state: this.state });
       return;
     }
     this.log("debug", "Stopping recording");
@@ -5713,7 +5759,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
     if (this.state === "stopping" /* STOPPING */) {
       this.state = "stopped" /* STOPPED */;
     } else if (this.state === "connected" /* CONNECTED */ || this.state === "ready" /* READY */ || this.state === "connecting" /* CONNECTING */) {
-      this.log("error", "[DIAGNOSTIC] Unexpected disconnection", {
+      this.log("warn", "[DIAGNOSTIC] Unexpected disconnection", {
         code,
         codeDescription: closeCodeDescription,
         reason: reason || "(empty)",
@@ -5835,7 +5881,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       blobToArrayBuffer(audioData).then((arrayBuffer) => {
         this.sendPrefixAudioInternal(arrayBuffer);
       }).catch((error) => {
-        this.log("error", "Failed to convert Blob to ArrayBuffer for prefix audio", error);
+        this.log("warn", "Failed to convert Blob to ArrayBuffer for prefix audio", error);
       });
       return;
     }