@volley/recognition-client-sdk 0.1.800 → 0.1.806

package/dist/index.js

@@ -3769,6 +3769,13 @@ var DeepgramModel;
   DeepgramModel2["NOVA_3"] = "nova-3";
   DeepgramModel2["FLUX_GENERAL_EN"] = "flux-general-en";
 })(DeepgramModel || (DeepgramModel = {}));
+var AssemblyAIModel;
+(function(AssemblyAIModel2) {
+  AssemblyAIModel2["DEFAULT"] = "default";
+  AssemblyAIModel2["UNIVERSAL_STREAMING_ENGLISH"] = "universal-streaming-english";
+  AssemblyAIModel2["UNIVERSAL_STREAMING_MULTILINGUAL"] = "universal-streaming-multilingual";
+  AssemblyAIModel2["U3_RT_PRO"] = "u3-rt-pro";
+})(AssemblyAIModel || (AssemblyAIModel = {}));
 var GoogleModel;
 (function(GoogleModel2) {
   GoogleModel2["CHIRP_3"] = "chirp_3";
@@ -4499,12 +4506,26 @@ var RecognitionContextSchemaV1 = z.discriminatedUnion("type", [
   ASRRequestSchemaV1
 ]);
 
+// ../../libs/types/dist/microphone-source.types.js
+var MicrophoneSourceType;
+(function(MicrophoneSourceType2) {
+  MicrophoneSourceType2["WEB"] = "web";
+  MicrophoneSourceType2["NATIVE"] = "native";
+})(MicrophoneSourceType || (MicrophoneSourceType = {}));
+var DEFAULT_MICROPHONE_SOURCE_TYPE = MicrophoneSourceType.WEB;
+var NAME_TO_ENUM = {
+  web: MicrophoneSourceType.WEB,
+  native: MicrophoneSourceType.NATIVE
+};
+
 // ../../libs/types/dist/recognition-query.types.js
 var RecognitionGameInfoSchema = z.object({
   userId: z.string().optional(),
   gameSessionId: z.string().optional(),
   deviceId: z.string().optional(),
   accountId: z.string().optional(),
+  clientId: z.string().optional(),
+  microphoneSourceType: z.nativeEnum(MicrophoneSourceType).optional(),
   gameId: z.string().optional(),
   gamePhase: z.string().optional(),
   questionAskedId: z.string().optional(),
@@ -4545,7 +4566,7 @@ var AudioFormat;
     [5, AudioFormat2.OPUS],
     [6, AudioFormat2.PCM]
   ]);
-  const NAME_TO_ENUM = /* @__PURE__ */ new Map([
+  const NAME_TO_ENUM2 = /* @__PURE__ */ new Map([
     ["WAV", AudioFormat2.WAV],
     ["MP3", AudioFormat2.MP3],
     ["FLAC", AudioFormat2.FLAC],
@@ -4566,7 +4587,7 @@ var AudioFormat;
   }
   AudioFormat2.fromId = fromId;
   function fromName(nameStr) {
-    return NAME_TO_ENUM.get(nameStr.toUpperCase());
+    return NAME_TO_ENUM2.get(nameStr.toUpperCase());
   }
   AudioFormat2.fromName = fromName;
   function toId(format) {
@@ -4582,7 +4603,7 @@ var AudioFormat;
   }
   AudioFormat2.isIdValid = isIdValid;
   function isNameValid(nameStr) {
-    return NAME_TO_ENUM.has(nameStr.toUpperCase());
+    return NAME_TO_ENUM2.has(nameStr.toUpperCase());
   }
   AudioFormat2.isNameValid = isNameValid;
 })(AudioFormat || (AudioFormat = {}));
@@ -4604,7 +4625,7 @@ var AudioEncoding;
     [4, AudioEncoding2.MULAW],
     [5, AudioEncoding2.ALAW]
   ]);
-  const NAME_TO_ENUM = /* @__PURE__ */ new Map([
+  const NAME_TO_ENUM2 = /* @__PURE__ */ new Map([
     ["ENCODING_UNSPECIFIED", AudioEncoding2.ENCODING_UNSPECIFIED],
     ["LINEAR16", AudioEncoding2.LINEAR16],
     ["OGG_OPUS", AudioEncoding2.OGG_OPUS],
@@ -4625,7 +4646,7 @@ var AudioEncoding;
   }
   AudioEncoding2.fromId = fromId;
   function fromName(nameStr) {
-    return NAME_TO_ENUM.get(nameStr.toUpperCase());
+    return NAME_TO_ENUM2.get(nameStr.toUpperCase());
   }
   AudioEncoding2.fromName = fromName;
   function toId(encoding) {
@@ -4641,7 +4662,7 @@ var AudioEncoding;
   }
   AudioEncoding2.isIdValid = isIdValid;
   function isNameValid(nameStr) {
-    return NAME_TO_ENUM.has(nameStr.toUpperCase());
+    return NAME_TO_ENUM2.has(nameStr.toUpperCase());
   }
   AudioEncoding2.isNameValid = isNameValid;
   function coerce2(value, onStringInput) {
@@ -4681,7 +4702,7 @@ var SampleRate;
     [44100, SampleRate2.RATE_44100],
     [48e3, SampleRate2.RATE_48000]
   ]);
-  const NAME_TO_ENUM = /* @__PURE__ */ new Map([
+  const NAME_TO_ENUM2 = /* @__PURE__ */ new Map([
     ["RATE_8000", SampleRate2.RATE_8000],
     ["RATE_16000", SampleRate2.RATE_16000],
     ["RATE_22050", SampleRate2.RATE_22050],
@@ -4704,7 +4725,7 @@ var SampleRate;
   }
   SampleRate2.fromHz = fromHz;
   function fromName(nameStr) {
-    return NAME_TO_ENUM.get(nameStr.toUpperCase());
+    return NAME_TO_ENUM2.get(nameStr.toUpperCase());
   }
   SampleRate2.fromName = fromName;
   function toHz(rate) {
@@ -4720,7 +4741,7 @@ var SampleRate;
   }
   SampleRate2.isHzValid = isHzValid;
   function isNameValid(nameStr) {
-    return NAME_TO_ENUM.has(nameStr.toUpperCase());
+    return NAME_TO_ENUM2.has(nameStr.toUpperCase());
   }
   SampleRate2.isNameValid = isNameValid;
 })(SampleRate || (SampleRate = {}));
@@ -5190,6 +5211,12 @@ function buildWebSocketUrl(config) {
   if (config.questionAnswerId) {
     url.searchParams.set("questionAnswerId", config.questionAnswerId);
   }
+  if (config.clientId) {
+    url.searchParams.set("clientId", config.clientId);
+  }
+  if (config.microphoneSourceType) {
+    url.searchParams.set("microphoneSourceType", config.microphoneSourceType);
+  }
   if (config.platform) {
     url.searchParams.set("platform", config.platform);
   }
@@ -5535,6 +5562,8 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
       ...config.deviceId && { deviceId: config.deviceId },
       ...config.accountId && { accountId: config.accountId },
       ...config.questionAnswerId && { questionAnswerId: config.questionAnswerId },
+      ...config.clientId && { clientId: config.clientId },
+      ...config.microphoneSourceType && { microphoneSourceType: config.microphoneSourceType },
       ...config.platform && { platform: config.platform },
       ...config.gameContext && { gameContext: config.gameContext },
       ...config.gameId && { gameId: config.gameId },
@@ -6292,6 +6321,21 @@ var ConfigBuilder = class {
     this.config.questionAnswerId = id;
     return this;
   }
+  /**
+   * Set client ID (for downstream tracking/audio-labeling)
+   */
+  clientId(id) {
+    this.config.clientId = id;
+    return this;
+  }
+  /**
+   * Set the client audio capture path (MicrophoneSourceType enum or 'web' | 'native' string).
+   * Defaults to 'web' on the server when omitted.
+   */
+  microphoneSourceType(source) {
+    this.config.microphoneSourceType = source;
+    return this;
+  }
   /**
    * Set platform
    */
@@ -6461,6 +6505,14 @@ var RecognitionVGFStateSchema = z.object({
   // Confidence score for the function call.
   finalFunctionCallTimestamp: z.string().optional(),
   // When the final action after interpreting the transcript was taken. Immutable.
+  // Session identity — when set, the VGF client backfills these into
+  // GameContextV1 if the caller didn't pass a `gameContext` in config.
+  // Lets RecognitionState be the single source of truth: server seeds
+  // `gameId` + `gamePhase` + `promptSlotMap` per player, controller passes
+  // the whole state as `initialState`, no separate `gameContext` needed.
+  // Backward-compatible: if `gameContext` is also passed in config, it wins.
+  gameId: z.string().optional(),
+  gamePhase: z.string().optional(),
   // Support for prompt slot mapping - passed to recognition context when present
   promptSlotMap: z.record(z.string(), z.array(z.string())).optional(),
   // Optional map of slot names to prompt values for recognition context
@@ -6507,6 +6559,9 @@ function createInitialRecognitionState(audioUtteranceId) {
     recognitionActionProcessingState: RecognitionActionProcessingState.NOT_STARTED
   };
 }
+function isTerminal(state) {
+  return state.transcriptionStatus === TranscriptionStatus.FINALIZED || state.transcriptionStatus === TranscriptionStatus.ABORTED || state.transcriptionStatus === TranscriptionStatus.ERROR;
+}
 function isValidRecordingStatusTransition(from, to) {
   const statusOrder = [
     RecordingStatus.NOT_READY,
@@ -6659,6 +6714,13 @@ var SimplifiedVGFRecognitionClient = class {
     }
     this.state = { ...this.state, startRecordingStatus: "READY" };
     this.expectedUuid = this.state.audioUtteranceId;
+    if (!clientConfig.gameContext && this.state.gameId && this.state.gamePhase) {
+      clientConfig.gameContext = {
+        type: RecognitionContextTypeV1.GAME_CONTEXT,
+        gameId: this.state.gameId,
+        gamePhase: this.state.gamePhase
+      };
+    }
     const hasPromptInputs = this.state.promptSlotMap !== void 0 || this.state.promptSTT !== void 0 || this.state.promptSTF !== void 0 || this.state.promptTTF !== void 0;
     if (hasPromptInputs) {
       if (clientConfig.asrRequestConfig) {
@@ -6666,7 +6728,7 @@ var SimplifiedVGFRecognitionClient = class {
       }
       if (!clientConfig.gameContext) {
         if (clientConfig.logger) {
-          clientConfig.logger("warn", "[VGF] prompt inputs found but no gameContext provided. They will not be sent.");
+          clientConfig.logger("warn", "[VGF] prompt inputs found but no gameContext provided and state has no gameId/gamePhase. They will not be sent.");
         }
       } else {
         if (this.state.promptSlotMap !== void 0) {
@@ -6878,11 +6940,8 @@ var SimplifiedVGFRecognitionClient = class {
   getVGFState() {
     return { ...this.state };
   }
-  isTerminalStatus(status) {
-    return status === TranscriptionStatus.FINALIZED || status === TranscriptionStatus.ABORTED || status === TranscriptionStatus.ERROR;
-  }
   notifyStateChange() {
-    if (this.isTerminalStatus(this.state.transcriptionStatus)) {
+    if (isTerminal(this.state)) {
       if (this.lastSentTerminalUuid === this.expectedUuid) {
         if (this.logger) {
           this.logger(
@@ -6908,11 +6967,44 @@ var SimplifiedVGFRecognitionClient = class {
     this.stateChangeCallback({ ...this.state });
   }
 };
+var SimplifiedVGFConfigBuilder = class extends ConfigBuilder {
+  constructor() {
+    super(...arguments);
+    this.vgfConfig = {};
+  }
+  /**
+   * Set the VGF state-change callback, fired whenever the VGF state updates.
+   */
+  onStateChange(callback) {
+    this.vgfConfig.onStateChange = callback;
+    return this;
+  }
+  /**
+   * Set the initial VGF state to restore from a previous session.
+   * The audioUtteranceId is extracted from it when valid.
+   */
+  initialState(state) {
+    this.vgfConfig.initialState = state;
+    return this;
+  }
+  /**
+   * Build the SimplifiedVGFClientConfig (base fields + VGF fields).
+   */
+  build() {
+    return { ...super.build(), ...this.vgfConfig };
+  }
+};
 function createSimplifiedVGFClient(config) {
   return new SimplifiedVGFRecognitionClient(config);
 }
+function createSimplifiedVGFClientWithBuilder(configure) {
+  const builder = new SimplifiedVGFConfigBuilder();
+  const config = configure(builder).build();
+  return new SimplifiedVGFRecognitionClient(config);
+}
 export {
   AmazonNovaSonicModel,
+  AssemblyAIModel,
   AudioEncoding,
   AwsTranscribeModel,
   BedrockModel,
@@ -6933,6 +7025,7 @@ export {
   GladiaModel,
   GoogleModel,
   Language,
+  MicrophoneSourceType,
   MistralVoxtralModel,
   OpenAIModel,
   OpenAIRealtimeModel,
@@ -6948,6 +7041,7 @@ export {
   STAGES,
   SampleRate,
   SelfServeVllmModel,
+  SimplifiedVGFConfigBuilder,
   SimplifiedVGFRecognitionClient,
   TimeoutError,
   TranscriptionStatus,
@@ -6957,6 +7051,7 @@ export {
   createDefaultASRConfig,
   createInitialRecognitionState,
   createSimplifiedVGFClient,
+  createSimplifiedVGFClientWithBuilder,
   getRecognitionConductorBase,
   getRecognitionConductorHost,
   getRecognitionConductorHttpBase,
@@ -6968,6 +7063,7 @@ export {
   getUserFriendlyMessage,
   isExceptionImmediatelyAvailable,
   isNormalDisconnection,
+  isTerminal,
   isValidRecordingStatusTransition,
   normalizeStage,
   resetRecognitionVGFState