@volley/recognition-client-sdk 0.1.621 → 0.1.670

package/dist/index.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;
@@ -3826,6 +3841,7 @@ var TranscriptionResultSchemaV1 = z.object({
   voiceStart: z.number().optional(),
   voiceDuration: z.number().optional(),
   voiceEnd: z.number().optional(),
+  lastNonSilence: z.number().optional(),
   startTimestamp: z.number().optional(),
   endTimestamp: z.number().optional(),
   receivedAtMs: z.number().optional(),
@@ -3873,6 +3889,9 @@ 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)
@@ -5248,7 +5267,7 @@ 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
@@ -5529,7 +5548,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
             const timeout = setTimeout(() => {
               if (settled) return;
               settled = true;
-              this.log("warn", "Connection timeout", { timeout: connectionTimeout, attempt });
+              this.log("warn", `Connection timeout url=${this.config.url}`, { timeout: connectionTimeout, attempt });
               this.state = "failed" /* FAILED */;
               reject(new Error(`Connection timeout after ${connectionTimeout}ms`));
             }, connectionTimeout);
@@ -5551,7 +5570,7 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
               if (settled) return;
               settled = true;
               clearTimeout(timeout);
-              this.log("warn", "Connection error", { error, attempt });
+              this.log("warn", `Connection error url=${this.config.url}`, { error, attempt });
               this.state = "failed" /* FAILED */;
               reject(error);
             };
@@ -5566,14 +5585,14 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
           lastError = error;
           if (attempt < maxAttempts) {
             const logLevel = attempt < 3 ? "info" : "warn";
-            this.log(logLevel, `Connection attempt ${attempt} failed, retrying after ${delayMs}ms`, {
+            this.log(logLevel, `Connection attempt ${attempt} failed, retrying after ${delayMs}ms url=${this.config.url}`, {
               error: lastError.message,
               nextAttempt: attempt + 1
             });
             this.state = "initial" /* INITIAL */;
             await new Promise((resolve) => setTimeout(resolve, delayMs));
           } else {
-            this.log("warn", `All ${maxAttempts} connection attempts failed`, {
+            this.log("warn", `All ${maxAttempts} connection attempts failed url=${this.config.url}`, {
               error: lastError.message
             });
           }
@@ -5596,7 +5615,7 @@ 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;
     }
@@ -5636,7 +5655,7 @@ 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");
@@ -5806,7 +5825,7 @@ 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)",
@@ -5928,7 +5947,7 @@ 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;
     }
@@ -6187,6 +6206,11 @@ var RecognitionVGFStateSchema = z.object({
   finalTranscript: z.string().optional(),
   // Full finalized transcript for the utterance. Will not change.
   finalConfidence: z.number().optional(),
+  // Voice timing (ms from stream start, prefix-adjusted)
+  voiceEnd: z.number().optional(),
+  // voice end time identified by ASR
+  lastNonSilence: z.number().optional(),
+  // last non-silence sample time from PCM analysis
   // Tracking-only metadata
   asrConfig: z.string().optional(),
   // Json format of the ASR config
@@ -6275,6 +6299,12 @@ function mapTranscriptionResultToState(currentState, result, isRecording) {
         newState.finalConfidence = result.finalTranscriptConfidence;
       }
     }
+    if (result.voiceEnd !== void 0) {
+      newState.voiceEnd = result.voiceEnd;
+    }
+    if (result.lastNonSilence !== void 0) {
+      newState.lastNonSilence = result.lastNonSilence;
+    }
   } else {
     newState.transcriptionStatus = TranscriptionStatus.FINALIZED;
     newState.finalTranscript = result.finalTranscript || "";
@@ -6282,6 +6312,12 @@ function mapTranscriptionResultToState(currentState, result, isRecording) {
       newState.finalConfidence = result.finalTranscriptConfidence;
     }
     newState.finalTranscriptionTimestamp = (/* @__PURE__ */ new Date()).toISOString();
+    if (result.voiceEnd !== void 0) {
+      newState.voiceEnd = result.voiceEnd;
+    }
+    if (result.lastNonSilence !== void 0) {
+      newState.lastNonSilence = result.lastNonSilence;
+    }
     newState.pendingTranscript = "";
     newState.pendingConfidence = void 0;
   }
@@ -6317,7 +6353,9 @@ function resetRecognitionVGFState(currentState) {
     transcriptionStatus: TranscriptionStatus.NOT_STARTED,
     startRecordingStatus: RecordingStatus.READY,
     recognitionActionProcessingState: RecognitionActionProcessingState.NOT_STARTED,
-    finalTranscript: void 0
+    finalTranscript: void 0,
+    voiceEnd: void 0,
+    lastNonSilence: void 0
   };
 }
 function generateUUID() {
@@ -6571,6 +6609,7 @@ function createSimplifiedVGFClient(config) {
 }
 export {
   AudioEncoding,
+  CartesiaModel,
   ClientControlActionV1,
   ClientState,
   ConfigBuilder,
@@ -6584,10 +6623,12 @@ export {
   FinalTranscriptStability,
   FireworksModel,
   GeminiModel,
+  GladiaModel,
   GoogleModel,
   Language,
   MistralVoxtralModel,
   OpenAIModel,
+  OpenAIRealtimeModel,
   RECOGNITION_CONDUCTOR_BASES,
   RECOGNITION_SERVICE_BASES,
   RealTimeTwoWayWebSocketRecognitionClient,
@@ -6599,6 +6640,7 @@ export {
   RecordingStatus,
   STAGES,
   SampleRate,
+  SelfServeVllmModel,
   SimplifiedVGFRecognitionClient,
   TimeoutError,
   TranscriptionStatus,