@volley/recognition-client-sdk 0.1.297 → 0.1.381

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

@@ -1,4 +1,4 @@
-// ../../node_modules/.pnpm/zod@3.22.5/node_modules/zod/lib/index.mjs
+// ../../node_modules/.pnpm/zod@3.22.4/node_modules/zod/lib/index.mjs
 var util;
 (function(util2) {
   util2.assertEqual = (val) => val;
@@ -3604,14 +3604,14 @@ var ostring = () => stringType().optional();
 var onumber = () => numberType().optional();
 var oboolean = () => booleanType().optional();
 var coerce = {
-  string: ((arg) => ZodString.create({ ...arg, coerce: true })),
-  number: ((arg) => ZodNumber.create({ ...arg, coerce: true })),
-  boolean: ((arg) => ZodBoolean.create({
+  string: (arg) => ZodString.create({ ...arg, coerce: true }),
+  number: (arg) => ZodNumber.create({ ...arg, coerce: true }),
+  boolean: (arg) => ZodBoolean.create({
     ...arg,
     coerce: true
-  })),
-  bigint: ((arg) => ZodBigInt.create({ ...arg, coerce: true })),
-  date: ((arg) => ZodDate.create({ ...arg, coerce: true }))
+  }),
+  bigint: (arg) => ZodBigInt.create({ ...arg, coerce: true }),
+  date: (arg) => ZodDate.create({ ...arg, coerce: true })
 };
 var NEVER = INVALID;
 var z = /* @__PURE__ */ Object.freeze({
@@ -3735,9 +3735,12 @@ var RecognitionProvider;
 (function(RecognitionProvider2) {
   RecognitionProvider2["ASSEMBLYAI"] = "assemblyai";
   RecognitionProvider2["DEEPGRAM"] = "deepgram";
+  RecognitionProvider2["ELEVENLABS"] = "elevenlabs";
+  RecognitionProvider2["FIREWORKS"] = "fireworks";
   RecognitionProvider2["GOOGLE"] = "google";
   RecognitionProvider2["GEMINI_BATCH"] = "gemini-batch";
   RecognitionProvider2["OPENAI_BATCH"] = "openai-batch";
+  RecognitionProvider2["OPENAI_REALTIME"] = "openai-realtime";
 })(RecognitionProvider || (RecognitionProvider = {}));
 var RecognitionMode;
 (function(RecognitionMode2) {
@@ -3757,17 +3760,34 @@ var DeepgramModel;
 })(DeepgramModel || (DeepgramModel = {}));
 var GoogleModel;
 (function(GoogleModel2) {
+  GoogleModel2["CHIRP_3"] = "chirp_3";
+  GoogleModel2["CHIRP_2"] = "chirp_2";
+  GoogleModel2["CHIRP"] = "chirp";
   GoogleModel2["LATEST_LONG"] = "latest_long";
   GoogleModel2["LATEST_SHORT"] = "latest_short";
   GoogleModel2["TELEPHONY"] = "telephony";
   GoogleModel2["TELEPHONY_SHORT"] = "telephony_short";
-  GoogleModel2["MEDICAL_DICTATION"] = "medical_dictation";
-  GoogleModel2["MEDICAL_CONVERSATION"] = "medical_conversation";
   GoogleModel2["DEFAULT"] = "default";
   GoogleModel2["COMMAND_AND_SEARCH"] = "command_and_search";
   GoogleModel2["PHONE_CALL"] = "phone_call";
   GoogleModel2["VIDEO"] = "video";
 })(GoogleModel || (GoogleModel = {}));
+var FireworksModel;
+(function(FireworksModel2) {
+  FireworksModel2["ASR_V1"] = "fireworks-asr-large";
+  FireworksModel2["ASR_V2"] = "fireworks-asr-v2";
+  FireworksModel2["WHISPER_V3"] = "whisper-v3";
+  FireworksModel2["WHISPER_V3_TURBO"] = "whisper-v3-turbo";
+})(FireworksModel || (FireworksModel = {}));
+var ElevenLabsModel;
+(function(ElevenLabsModel2) {
+  ElevenLabsModel2["SCRIBE_V2_REALTIME"] = "scribe_v2_realtime";
+  ElevenLabsModel2["SCRIBE_V1"] = "scribe_v1";
+})(ElevenLabsModel || (ElevenLabsModel = {}));
+var OpenAIRealtimeModel;
+(function(OpenAIRealtimeModel2) {
+  OpenAIRealtimeModel2["GPT_4O_MINI_TRANSCRIBE"] = "gpt-4o-mini-transcribe";
+})(OpenAIRealtimeModel || (OpenAIRealtimeModel = {}));
 
 // ../../libs/types/dist/recognition-result-v1.types.js
 var RecognitionResultTypeV1;
@@ -3781,68 +3801,44 @@ var RecognitionResultTypeV1;
 var TranscriptionResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.TRANSCRIPTION),
   audioUtteranceId: z.string(),
-  // audio utterance id for this transcription session
   finalTranscript: z.string(),
-  // transcript that is final and won't be overwritten by ASR. default empty string.
   finalTranscriptConfidence: z.number().min(0).max(1).optional(),
-  // confidence score for the final transcript (0-1). Optional.
   pendingTranscript: z.string().optional(),
-  // transcript that may be overwritten by ASR. Optional.
   pendingTranscriptConfidence: z.number().min(0).max(1).optional(),
-  // confidence score for the pending transcript (0-1). Optional.
   is_finished: z.boolean(),
-  // entire transcription finished. This will be the last message in the transcription.
   voiceStart: z.number().optional(),
-  // voice start time identified by ASR (in milliseconds from stream start). Optional.
   voiceDuration: z.number().optional(),
-  // voice duration identified by ASR (in milliseconds, calculated as maxVoiceEnd - voiceStart). Optional.
   voiceEnd: z.number().optional(),
-  // voice end time identified by ASR (in milliseconds from stream start). Optional.
   startTimestamp: z.number().optional(),
-  // start timestamp of the transcription in milliseconds. Optional.
   endTimestamp: z.number().optional(),
-  // end timestamp of the transcription in milliseconds. Optional.
   receivedAtMs: z.number().optional(),
-  // server timestamp when this transcript was received (ms since epoch). Optional.
   accumulatedAudioTimeMs: z.number().optional()
   // accumulated audio time watermark in milliseconds. Total duration of all audio chunks sent. Optional.
 });
 var FunctionCallResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.FUNCTION_CALL),
   audioUtteranceId: z.string(),
-  // audio utterance id for this function call
   functionName: z.string(),
   functionArgJson: z.string()
 });
 var MetadataResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.METADATA),
   audioUtteranceId: z.string(),
-  // audio utterance id for this metadata
   // Timing information
   recordingStartMs: z.number().optional(),
-  // When recording started
   recordingEndMs: z.number().optional(),
-  // When recording ended
   transcriptEndMs: z.number().optional(),
-  // When final transcript was received
   socketCloseAtMs: z.number().optional(),
-  // When socket/connection closed
   // Audio Quality Metrics
   duration: z.number().optional(),
-  // Duration of the audio in milliseconds
   volume: z.number().optional(),
-  // Volume of the audio in decibels
   accumulatedAudioTimeMs: z.number().optional(),
-  // Accumulated audio time watermark in milliseconds. Total duration of all audio chunks sent. Optional.
   // Cost Information
   costInUSD: z.number().default(0).optional(),
-  // Estimated cost in USD for this recognition session. Default: 0
   // ASR API Type
   apiType: z.nativeEnum(ASRApiType).optional(),
-  // ASR API type from job (streaming vs file-based)
   // ASR configuration as JSON string (no type validation)
   asrConfig: z.string().optional(),
-  // Stringified JSON of the ASR provider config
   // Raw ASR metadata payload as provided by the provider (stringified if needed)
   rawAsrMetadata: z.string().optional()
 });
@@ -3859,13 +3855,9 @@ var ErrorTypeV1;
 var ErrorResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.ERROR),
   audioUtteranceId: z.string(),
-  // audio utterance id for this error
   errorType: z.nativeEnum(ErrorTypeV1).optional(),
-  // error type category
   message: z.string().optional(),
-  // Error message
   code: z.union([z.string(), z.number()]).optional(),
-  // Error code
   description: z.string().optional()
   // Detailed description
 });
@@ -3877,9 +3869,7 @@ var ClientControlActionV1;
 var ClientControlActionsV1 = z.nativeEnum(ClientControlActionV1);
 var ClientControlMessageSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.CLIENT_CONTROL_MESSAGE),
-  // Message type discriminator
   audioUtteranceId: z.string(),
-  // audio utterance id for this control message
   action: ClientControlActionsV1
   // The control action to perform
 });
@@ -4017,7 +4007,6 @@ var MetadataMessageSchema = z.object({
   type: z.literal(ProviderMessageType.METADATA),
   provider: z.nativeEnum(RecognitionProvider),
   asrConfig: z.string().optional(),
-  // ASR config we sent to provider
   data: z.string().optional()
   // Raw metadata from provider (Deepgram only)
 });
@@ -4141,7 +4130,6 @@ var BaseRecognitionExceptionSchema = z.object({
 var AuthenticationExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.AUTHENTICATION_ERROR),
   isImmediatelyAvailable: z.literal(false),
-  // System issue, not user-facing
   /** Which service failed authentication (e.g., 'deepgram', 'google') */
   service: z.string().optional(),
   /** Authentication method that failed (e.g., 'api_key', 'oauth') */
@@ -4150,7 +4138,6 @@ var AuthenticationExceptionSchema = BaseRecognitionExceptionSchema.extend({
 var ValidationExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.VALIDATION_ERROR),
   isImmediatelyAvailable: z.literal(true),
-  // Can show to user
   /** Field name that failed validation */
   field: z.string().optional(),
   /** Expected value format/type */
@@ -4161,7 +4148,6 @@ var ValidationExceptionSchema = BaseRecognitionExceptionSchema.extend({
 var ProviderExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.PROVIDER_ERROR),
   isImmediatelyAvailable: z.literal(false),
-  // Provider issue, not user-facing
   /** Provider name (e.g., 'deepgram', 'assemblyai', 'google') */
   provider: z.string().optional(),
   /** Provider-specific error code */
@@ -4172,7 +4158,6 @@ var ProviderExceptionSchema = BaseRecognitionExceptionSchema.extend({
 var TimeoutExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.TIMEOUT_ERROR),
   isImmediatelyAvailable: z.literal(true),
-  // Can tell user to try again
   /** Timeout duration in milliseconds */
   timeoutMs: z.number().optional(),
   /** What operation timed out (e.g., 'connection', 'transcription', 'response') */
@@ -4181,7 +4166,6 @@ var TimeoutExceptionSchema = BaseRecognitionExceptionSchema.extend({
 var QuotaExceededExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.QUOTA_EXCEEDED),
   isImmediatelyAvailable: z.literal(true),
-  // Can tell user to wait
   /** Quota type that was exceeded (e.g., 'rate_limit', 'daily_quota', 'concurrent_requests') */
   quotaType: z.string().optional(),
   /** When quota resets (Unix timestamp in ms) */
@@ -4192,7 +4176,6 @@ var QuotaExceededExceptionSchema = BaseRecognitionExceptionSchema.extend({
 var ConnectionExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.CONNECTION_ERROR),
   isImmediatelyAvailable: z.literal(true),
-  // Can tell user about network issues
   /** Number of connection attempts made */
   attempts: z.number().optional(),
   /** URL that failed to connect */
@@ -4203,7 +4186,6 @@ var ConnectionExceptionSchema = BaseRecognitionExceptionSchema.extend({
 var UnknownExceptionSchema = BaseRecognitionExceptionSchema.extend({
   errorType: z.literal(ErrorTypeV1.UNKNOWN_ERROR),
   isImmediatelyAvailable: z.literal(false),
-  // Unknown cause, not user-facing
   /** Stack trace for debugging (sanitized) */
   stack: z.string().optional(),
   /** Additional context for debugging */
@@ -4235,16 +4217,11 @@ var SlotMapSchema = z.record(z.string(), z.array(z.string()));
 var GameContextSchemaV1 = z.object({
   type: z.literal(RecognitionContextTypeV1.GAME_CONTEXT),
   gameId: z.string(),
-  // Unique identifier for the game
   gamePhase: z.string(),
-  // Current phase of the game
   // Prompt fields for different processing stages
   promptSTT: z.string().optional(),
-  // Speech-to-Text: Keywords/phrases to optimize ASR (e.g., for Deepgram keywords)
   promptSTF: z.string().optional(),
-  // Speech-to-Function: Direct speech to function mapping instructions
   promptTTF: z.string().optional(),
-  // Text-to-Function: Text to function mapping instructions after transcription
   // Slot map for entity extraction
   slotMap: SlotMapSchema.optional()
   // Map of slot names to possible values to enhance prompting and keyword matching.
@@ -4271,7 +4248,6 @@ var ASRRequestSchemaV1 = z.object({
   type: z.literal(RecognitionContextTypeV1.ASR_REQUEST),
   // Session identification
   audioUtteranceId: z.string().optional(),
-  // Optional utterance ID (prefer query param)
   // ASR configuration
   provider: z.string(),
   model: z.string().optional(),
@@ -4280,12 +4256,9 @@ var ASRRequestSchemaV1 = z.object({
   encoding: z.number(),
   // Recognition options
   interimResults: z.boolean().optional().default(false),
-  // Enable interim/partial results
   useContext: z.boolean().optional().default(false),
-  // If true, requires GameContext before session creation
   // Final transcript stability mode (timeout for fallback final transcript)
   finalTranscriptStability: z.string().optional(),
-  // Stability mode: 'aggressive' | 'balanced' | 'conservative'
   // Debug options (FOR DEBUG/TESTING ONLY - not for production use)
   debugCommand: RequestDebugCommandSchema
 });
@@ -4298,31 +4271,22 @@ var RecognitionContextSchemaV1 = z.discriminatedUnion("type", [
 // ../../libs/types/dist/recognition-query.types.js
 var RecognitionGameInfoSchema = z.object({
   userId: z.string().optional(),
-  // User identifier
   gameSessionId: z.string().optional(),
-  // Volly session identifier . Called "sessionId" in Platform and most games.
   deviceId: z.string().optional(),
-  // Device identifier
   accountId: z.string().optional(),
-  // Account identifier
   gameId: z.string().optional(),
-  // Game identifier
   gamePhase: z.string().optional(),
-  // Game phase or scene. A string that describes the current state of the game for client team to optimize specific turn.
   questionAnswerId: z.string().optional(),
-  // Question answer identifier for tracking Q&A sessions
   platform: z.string().optional()
   // Platform for audio recording device (use the definition of platform teams)
 });
 var RecognitionQueryMetadataSchema = z.object({
   audioUtteranceId: z.string(),
-  // Audio utterance ID to query
   recognitionGameInfo: RecognitionGameInfoSchema.optional()
   // Optional game info for security
 });
 var UtteranceResourceResponseSchema = z.object({
   audioUtteranceId: z.string(),
-  // Audio utterance ID
   results: z.array(z.any())
   // Array of RecognitionResult (version-specific)
 });
@@ -4579,10 +4543,7 @@ var GeminiModel;
   GeminiModel2["GEMINI_2_5_FLASH"] = "gemini-2.5-flash";
   GeminiModel2["GEMINI_2_5_FLASH_LITE"] = "gemini-2.5-flash-lite";
   GeminiModel2["GEMINI_2_0_FLASH_LATEST"] = "gemini-2.0-flash-latest";
-  GeminiModel2["GEMINI_2_0_FLASH"] = "gemini-2.0-flash-002";
   GeminiModel2["GEMINI_2_0_FLASH_EXP"] = "gemini-2.0-flash-exp";
-  GeminiModel2["GEMINI_1_5_FLASH"] = "gemini-1.5-flash";
-  GeminiModel2["GEMINI_1_5_PRO"] = "gemini-1.5-pro";
 })(GeminiModel || (GeminiModel = {}));
 var GeminiApiVersion;
 (function(GeminiApiVersion2) {
@@ -4779,7 +4740,7 @@ var WebSocketAudioClient = class {
 // ../../libs/websocket/dist/core/audio-upload-websocket-server.js
 import { WebSocketServer, WebSocket as WebSocket2 } from "ws";
 
-// ../../node_modules/.pnpm/uuid@11.1.0/node_modules/uuid/dist/esm-browser/stringify.js
+// ../../node_modules/.pnpm/uuid@11.0.0/node_modules/uuid/dist/esm-browser/stringify.js
 var byteToHex = [];
 for (let i = 0; i < 256; ++i) {
   byteToHex.push((i + 256).toString(16).slice(1));
@@ -4788,7 +4749,7 @@ function unsafeStringify(arr, offset = 0) {
   return (byteToHex[arr[offset + 0]] + byteToHex[arr[offset + 1]] + byteToHex[arr[offset + 2]] + byteToHex[arr[offset + 3]] + "-" + byteToHex[arr[offset + 4]] + byteToHex[arr[offset + 5]] + "-" + byteToHex[arr[offset + 6]] + byteToHex[arr[offset + 7]] + "-" + byteToHex[arr[offset + 8]] + byteToHex[arr[offset + 9]] + "-" + byteToHex[arr[offset + 10]] + byteToHex[arr[offset + 11]] + byteToHex[arr[offset + 12]] + byteToHex[arr[offset + 13]] + byteToHex[arr[offset + 14]] + byteToHex[arr[offset + 15]]).toLowerCase();
 }
 
-// ../../node_modules/.pnpm/uuid@11.1.0/node_modules/uuid/dist/esm-browser/rng.js
+// ../../node_modules/.pnpm/uuid@11.0.0/node_modules/uuid/dist/esm-browser/rng.js
 var getRandomValues;
 var rnds8 = new Uint8Array(16);
 function rng() {
@@ -4801,27 +4762,21 @@ function rng() {
   return getRandomValues(rnds8);
 }
 
-// ../../node_modules/.pnpm/uuid@11.1.0/node_modules/uuid/dist/esm-browser/native.js
+// ../../node_modules/.pnpm/uuid@11.0.0/node_modules/uuid/dist/esm-browser/native.js
 var randomUUID = typeof crypto !== "undefined" && crypto.randomUUID && crypto.randomUUID.bind(crypto);
 var native_default = { randomUUID };
 
-// ../../node_modules/.pnpm/uuid@11.1.0/node_modules/uuid/dist/esm-browser/v4.js
+// ../../node_modules/.pnpm/uuid@11.0.0/node_modules/uuid/dist/esm-browser/v4.js
 function v4(options, buf, offset) {
   if (native_default.randomUUID && !buf && !options) {
     return native_default.randomUUID();
   }
   options = options || {};
-  const rnds = options.random ?? options.rng?.() ?? rng();
-  if (rnds.length < 16) {
-    throw new Error("Random bytes length must be >= 16");
-  }
+  const rnds = options.random || (options.rng || rng)();
   rnds[6] = rnds[6] & 15 | 64;
   rnds[8] = rnds[8] & 63 | 128;
   if (buf) {
     offset = offset || 0;
-    if (offset < 0 || offset + 16 > buf.length) {
-      throw new RangeError(`UUID byte range ${offset}:${offset + 15} is out of buffer bounds`);
-    }
     for (let i = 0; i < 16; ++i) {
       buf[offset + i] = rnds[i];
     }
@@ -5546,16 +5501,65 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
     this.config.onConnected();
   }
   onDisconnected(code, reason) {
-    this.log("debug", "WebSocket disconnected", { code, reason, previousState: this.state });
+    const closeCodeDescription = this.getCloseCodeDescription(code);
+    const is1006 = code === 1006;
+    this.log("debug", "[DIAGNOSTIC] WebSocket disconnected", {
+      code,
+      codeDescription: closeCodeDescription,
+      reason: reason || "(empty)",
+      previousState: this.state,
+      is1006Abnormal: is1006,
+      audioChunksSent: this.audioChunksSent,
+      audioBytesSent: this.audioBytesSent,
+      bufferStats: this.audioBuffer.getStats()
+    });
     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", "Unexpected disconnection", { code, reason });
+      this.log("error", "[DIAGNOSTIC] Unexpected disconnection", {
+        code,
+        codeDescription: closeCodeDescription,
+        reason: reason || "(empty)",
+        is1006,
+        possibleCauses: is1006 ? [
+          "Network connection lost",
+          "Server process crashed",
+          "Provider (Deepgram/AssemblyAI) WebSocket closed abnormally",
+          "Firewall/proxy terminated connection",
+          "Browser/tab suspended (mobile)"
+        ] : []
+      });
       this.state = "failed" /* FAILED */;
     }
     this.cleanup();
     this.config.onDisconnected(code, reason);
   }
+  /**
+   * Get human-readable description for WebSocket close code
+   */
+  getCloseCodeDescription(code) {
+    const descriptions = {
+      1e3: "Normal Closure",
+      1001: "Going Away",
+      1002: "Protocol Error",
+      1003: "Unsupported Data",
+      1005: "No Status Received",
+      1006: "Abnormal Closure (no close frame received)",
+      1007: "Invalid Frame Payload",
+      1008: "Policy Violation",
+      1009: "Message Too Big",
+      1010: "Mandatory Extension",
+      1011: "Internal Server Error",
+      1012: "Service Restart",
+      1013: "Try Again Later",
+      4e3: "Auth Required",
+      4001: "Auth Failed",
+      4002: "Rate Limit Exceeded",
+      4003: "Invalid Session",
+      4004: "Session Expired"
+    };
+    return descriptions[code] || `Unknown (${code})`;
+  }
   onError(error) {
     this.state = "failed" /* FAILED */;
     const errorResult = {