@volley/recognition-client-sdk 0.1.767 → 0.1.782

package/dist/index.bundled.d.ts

@@ -24,6 +24,7 @@ declare enum RecognitionProvider {
     BEDROCK = "bedrock",
     INWORLD_STT = "inworld-stt",
     AWS_TRANSCRIBE = "aws-transcribe",
+    AMAZON_NOVA_SONIC = "amazon-nova-sonic",
     TEST_ASR_PROVIDER_QUOTA = "test-asr-provider-quota",
     TEST_ASR_STREAMING = "test-asr-streaming"
 }
@@ -96,6 +97,7 @@ declare enum ElevenLabsModel {
  * @see https://platform.openai.com/docs/models/gpt-4o-transcribe
  */
 declare enum OpenAIRealtimeModel {
+    GPT_REALTIME_WHISPER = "gpt-realtime-whisper",
     GPT_4O_TRANSCRIBE = "gpt-4o-transcribe",
     GPT_4O_MINI_TRANSCRIBE = "gpt-4o-mini-transcribe"
 }
@@ -146,6 +148,15 @@ declare enum InworldSttModel {
 declare enum AwsTranscribeModel {
     DEFAULT = "default"
 }
+/**
+ * Amazon Nova Sonic bidirectional streaming model (Bedrock).
+ * Speech-to-speech model; we consume the USER FINAL transcript and discard the assistant text/audio output.
+ * @see https://docs.aws.amazon.com/nova/latest/userguide/speech-bidirection.html
+ */
+declare enum AmazonNovaSonicModel {
+    AMAZON_NOVA_SONIC_V1 = "amazon.nova-sonic-v1:0",
+    AMAZON_NOVA_2_SONIC = "amazon.nova-2-sonic-v1:0"
+}
 /**
  * Self-serve vLLM batch transcription models
  * Backed by recognition-inference / RunPod `/transcribe`
@@ -156,125 +167,7 @@ declare enum SelfServeVllmModel {
 /**
  * Type alias for any model from any provider
  */
-type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | InworldSttModel | SelfServeVllmModel | BedrockModel | AwsTranscribeModel | string;
-
-/**
- * Audio encoding types
- */
-declare enum AudioEncoding {
-    ENCODING_UNSPECIFIED = 0,
-    LINEAR16 = 1,
-    OGG_OPUS = 2,
-    FLAC = 3,
-    MULAW = 4,
-    ALAW = 5
-}
-declare namespace AudioEncoding {
-    /**
-     * Convert numeric ID to AudioEncoding enum
-     * @param id - Numeric encoding identifier (0-5)
-     * @returns AudioEncoding enum value or undefined if invalid
-     */
-    function fromId(id: number): AudioEncoding | undefined;
-    /**
-     * Convert string name to AudioEncoding enum
-     * @param nameStr - String name like "linear16", "LINEAR16", "ogg_opus", "OGG_OPUS", etc. (case insensitive)
-     * @returns AudioEncoding enum value or undefined if invalid
-     */
-    function fromName(nameStr: string): AudioEncoding | undefined;
-    /**
-     * Convert AudioEncoding enum to numeric ID
-     * @param encoding - AudioEncoding enum value
-     * @returns Numeric ID (0-5)
-     */
-    function toId(encoding: AudioEncoding): number;
-    /**
-     * Convert AudioEncoding enum to string name
-     * @param encoding - AudioEncoding enum value
-     * @returns String name like "LINEAR16", "MULAW", etc.
-     */
-    function toName(encoding: AudioEncoding): string;
-    /**
-     * Check if a numeric ID is a valid encoding
-     * @param id - Numeric identifier to validate
-     * @returns true if valid encoding ID
-     */
-    function isIdValid(id: number): boolean;
-    /**
-     * Check if a string name is a valid encoding
-     * @param nameStr - String name to validate
-     * @returns true if valid encoding name
-     */
-    function isNameValid(nameStr: string): boolean;
-}
-/**
- * Common sample rates (in Hz)
- */
-declare enum SampleRate {
-    RATE_8000 = 8000,
-    RATE_16000 = 16000,
-    RATE_22050 = 22050,
-    RATE_24000 = 24000,
-    RATE_32000 = 32000,
-    RATE_44100 = 44100,
-    RATE_48000 = 48000
-}
-declare namespace SampleRate {
-    /**
-     * Convert Hz value to SampleRate enum
-     * @param hz - Sample rate in Hz (8000, 16000, etc.)
-     * @returns SampleRate enum value or undefined if invalid
-     */
-    function fromHz(hz: number): SampleRate | undefined;
-    /**
-     * Convert string name to SampleRate enum
-     * @param nameStr - String name like "rate_8000", "RATE_16000", etc. (case insensitive)
-     * @returns SampleRate enum value or undefined if invalid
-     */
-    function fromName(nameStr: string): SampleRate | undefined;
-    /**
-     * Convert SampleRate enum to Hz value
-     * @param rate - SampleRate enum value
-     * @returns Hz value (8000, 16000, etc.)
-     */
-    function toHz(rate: SampleRate): number;
-    /**
-     * Convert SampleRate enum to string name
-     * @param rate - SampleRate enum value
-     * @returns String name like "RATE_8000", "RATE_16000", etc.
-     */
-    function toName(rate: SampleRate): string;
-    /**
-     * Check if a numeric Hz value is a valid sample rate
-     * @param hz - Hz value to validate
-     * @returns true if valid sample rate
-     */
-    function isHzValid(hz: number): boolean;
-    /**
-     * Check if a string name is a valid sample rate
-     * @param nameStr - String name to validate
-     * @returns true if valid sample rate name
-     */
-    function isNameValid(nameStr: string): boolean;
-}
-/**
- * Supported languages for recognition
- * Using BCP-47 language tags
- */
-declare enum Language {
-    ENGLISH_US = "en-US",
-    ENGLISH_GB = "en-GB",
-    SPANISH_ES = "es-ES",
-    SPANISH_MX = "es-MX",
-    FRENCH_FR = "fr-FR",
-    GERMAN_DE = "de-DE",
-    ITALIAN_IT = "it-IT",
-    PORTUGUESE_BR = "pt-BR",
-    JAPANESE_JP = "ja-JP",
-    KOREAN_KR = "ko-KR",
-    CHINESE_CN = "zh-CN",
-    CHINESE_TW = "zh-TW"
-}
+type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | InworldSttModel | SelfServeVllmModel | BedrockModel | AwsTranscribeModel | AmazonNovaSonicModel | string;
 
 /**
  * Recognition Result Types V1
@@ -294,6 +187,16 @@ declare enum RecognitionResultTypeV1 {
     AUDIO_METRICS = "AudioMetrics",
     SESSION_CONFIGURED = "SessionConfigured"
 }
+/**
+ * Source of a phrase detection — what kind of provider feature produced
+ * the hit. Currently only Deepgram's `search` parameter is wired up, so
+ * this enum has one value. New entries (e.g. KEYWORDS, KEYTERMS,
+ * SPEECH_CONTEXTS) get added when other providers join.
+ */
+declare enum DetectionTypeV1 {
+    /** Deepgram phonetic phrase match via the `search=…` request parameter */
+    SEARCH = "search"
+}
 /**
  * Transcription result V1 - contains transcript message
  * In the long run game side should not need to know it. In the short run it is send back to client.
@@ -318,6 +221,25 @@ declare const TranscriptionResultSchemaV1: z.ZodObject<{
     receivedAtMs: z.ZodOptional<z.ZodNumber>;
     accumulatedAudioTimeMs: z.ZodOptional<z.ZodNumber>;
     rawAudioTimeMs: z.ZodOptional<z.ZodNumber>;
+    detections: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        type: z.ZodNativeEnum<typeof DetectionTypeV1>;
+        query: z.ZodString;
+        score: z.ZodNumber;
+        startMs: z.ZodOptional<z.ZodNumber>;
+        endMs: z.ZodOptional<z.ZodNumber>;
+    }, "strip", z.ZodTypeAny, {
+        type: DetectionTypeV1;
+        query: string;
+        score: number;
+        startMs?: number | undefined;
+        endMs?: number | undefined;
+    }, {
+        type: DetectionTypeV1;
+        query: string;
+        score: number;
+        startMs?: number | undefined;
+        endMs?: number | undefined;
+    }>, "many">>;
 }, "strip", z.ZodTypeAny, {
     type: RecognitionResultTypeV1.TRANSCRIPTION;
     audioUtteranceId: string;
@@ -337,6 +259,13 @@ declare const TranscriptionResultSchemaV1: z.ZodObject<{
     receivedAtMs?: number | undefined;
     accumulatedAudioTimeMs?: number | undefined;
     rawAudioTimeMs?: number | undefined;
+    detections?: {
+        type: DetectionTypeV1;
+        query: string;
+        score: number;
+        startMs?: number | undefined;
+        endMs?: number | undefined;
+    }[] | undefined;
 }, {
     type: RecognitionResultTypeV1.TRANSCRIPTION;
     audioUtteranceId: string;
@@ -356,6 +285,13 @@ declare const TranscriptionResultSchemaV1: z.ZodObject<{
     receivedAtMs?: number | undefined;
     accumulatedAudioTimeMs?: number | undefined;
     rawAudioTimeMs?: number | undefined;
+    detections?: {
+        type: DetectionTypeV1;
+        query: string;
+        score: number;
+        startMs?: number | undefined;
+        endMs?: number | undefined;
+    }[] | undefined;
 }>;
 type TranscriptionResultV1 = z.infer<typeof TranscriptionResultSchemaV1>;
 /**
@@ -676,6 +612,124 @@ declare const AudioMetricsResultSchemaV1: z.ZodObject<{
 }>;
 type AudioMetricsResultV1 = z.infer<typeof AudioMetricsResultSchemaV1>;
 
+/**
+ * Audio encoding types
+ */
+declare enum AudioEncoding {
+    ENCODING_UNSPECIFIED = 0,
+    LINEAR16 = 1,
+    OGG_OPUS = 2,
+    FLAC = 3,
+    MULAW = 4,
+    ALAW = 5
+}
+declare namespace AudioEncoding {
+    /**
+     * Convert numeric ID to AudioEncoding enum
+     * @param id - Numeric encoding identifier (0-5)
+     * @returns AudioEncoding enum value or undefined if invalid
+     */
+    function fromId(id: number): AudioEncoding | undefined;
+    /**
+     * Convert string name to AudioEncoding enum
+     * @param nameStr - String name like "linear16", "LINEAR16", "ogg_opus", "OGG_OPUS", etc. (case insensitive)
+     * @returns AudioEncoding enum value or undefined if invalid
+     */
+    function fromName(nameStr: string): AudioEncoding | undefined;
+    /**
+     * Convert AudioEncoding enum to numeric ID
+     * @param encoding - AudioEncoding enum value
+     * @returns Numeric ID (0-5)
+     */
+    function toId(encoding: AudioEncoding): number;
+    /**
+     * Convert AudioEncoding enum to string name
+     * @param encoding - AudioEncoding enum value
+     * @returns String name like "LINEAR16", "MULAW", etc.
+     */
+    function toName(encoding: AudioEncoding): string;
+    /**
+     * Check if a numeric ID is a valid encoding
+     * @param id - Numeric identifier to validate
+     * @returns true if valid encoding ID
+     */
+    function isIdValid(id: number): boolean;
+    /**
+     * Check if a string name is a valid encoding
+     * @param nameStr - String name to validate
+     * @returns true if valid encoding name
+     */
+    function isNameValid(nameStr: string): boolean;
+}
+/**
+ * Common sample rates (in Hz)
+ */
+declare enum SampleRate {
+    RATE_8000 = 8000,
+    RATE_16000 = 16000,
+    RATE_22050 = 22050,
+    RATE_24000 = 24000,
+    RATE_32000 = 32000,
+    RATE_44100 = 44100,
+    RATE_48000 = 48000
+}
+declare namespace SampleRate {
+    /**
+     * Convert Hz value to SampleRate enum
+     * @param hz - Sample rate in Hz (8000, 16000, etc.)
+     * @returns SampleRate enum value or undefined if invalid
+     */
+    function fromHz(hz: number): SampleRate | undefined;
+    /**
+     * Convert string name to SampleRate enum
+     * @param nameStr - String name like "rate_8000", "RATE_16000", etc. (case insensitive)
+     * @returns SampleRate enum value or undefined if invalid
+     */
+    function fromName(nameStr: string): SampleRate | undefined;
+    /**
+     * Convert SampleRate enum to Hz value
+     * @param rate - SampleRate enum value
+     * @returns Hz value (8000, 16000, etc.)
+     */
+    function toHz(rate: SampleRate): number;
+    /**
+     * Convert SampleRate enum to string name
+     * @param rate - SampleRate enum value
+     * @returns String name like "RATE_8000", "RATE_16000", etc.
+     */
+    function toName(rate: SampleRate): string;
+    /**
+     * Check if a numeric Hz value is a valid sample rate
+     * @param hz - Hz value to validate
+     * @returns true if valid sample rate
+     */
+    function isHzValid(hz: number): boolean;
+    /**
+     * Check if a string name is a valid sample rate
+     * @param nameStr - String name to validate
+     * @returns true if valid sample rate name
+     */
+    function isNameValid(nameStr: string): boolean;
+}
+/**
+ * Supported languages for recognition
+ * Using BCP-47 language tags
+ */
+declare enum Language {
+    ENGLISH_US = "en-US",
+    ENGLISH_GB = "en-GB",
+    SPANISH_ES = "es-ES",
+    SPANISH_MX = "es-MX",
+    FRENCH_FR = "fr-FR",
+    GERMAN_DE = "de-DE",
+    ITALIAN_IT = "it-IT",
+    PORTUGUESE_BR = "pt-BR",
+    JAPANESE_JP = "ja-JP",
+    KOREAN_KR = "ko-KR",
+    CHINESE_CN = "zh-CN",
+    CHINESE_TW = "zh-TW"
+}
+
 /**
  * Error Exception Types
  *
@@ -1371,6 +1425,7 @@ declare const ASRRequestSchemaV1: z.ZodObject<{
     prefixId: z.ZodOptional<z.ZodString>;
     prefixTextToRemove: z.ZodOptional<z.ZodArray<z.ZodString, "many">>;
     audioMetricsIntervalMs: z.ZodOptional<z.ZodNumber>;
+    appendSearch: z.ZodOptional<z.ZodBoolean>;
     debugCommand: z.ZodOptional<z.ZodObject<{
         enableDebugLog: z.ZodDefault<z.ZodOptional<z.ZodBoolean>>;
         enableAudioStorage: z.ZodDefault<z.ZodOptional<z.ZodBoolean>>;
@@ -1413,6 +1468,7 @@ declare const ASRRequestSchemaV1: z.ZodObject<{
     prefixId?: string | undefined;
     prefixTextToRemove?: string[] | undefined;
     audioMetricsIntervalMs?: number | undefined;
+    appendSearch?: boolean | undefined;
     debugCommand?: {
         enableDebugLog: boolean;
         enableAudioStorage: boolean;
@@ -1445,6 +1501,7 @@ declare const ASRRequestSchemaV1: z.ZodObject<{
     prefixId?: string | undefined;
     prefixTextToRemove?: string[] | undefined;
     audioMetricsIntervalMs?: number | undefined;
+    appendSearch?: boolean | undefined;
     debugCommand?: {
         enableDebugLog?: boolean | undefined;
         enableAudioStorage?: boolean | undefined;
@@ -1676,6 +1733,29 @@ interface ASRRequestConfig {
      * @example 500
      */
     audioMetricsIntervalMs?: number;
+    /**
+     * Opt-in: round-trip Deepgram `search` phrase hits into the transcript.
+     *
+     * When `true` AND the resolved provider/model is **deepgram nova-2** AND the
+     * GameContext `gamePhase` is `'Solve Puzzle'`, every Deepgram Results event
+     * with a `channel.search` hit at confidence ≥ 0.6 has the original query
+     * prepended to the transcript text delivered to the client. This restores
+     * parity with the legacy Roku→Deepgram WoF Puzzle-Solve path where the
+     * phrase round-trip lets downstream NLU match multi-word puzzle solutions
+     * even when nova-2's primary transcription drifts.
+     *
+     * Default: `false` (no prepend; transcript is whatever nova-2 produces).
+     *
+     * Scope guard rationale:
+     * - nova-2 only: nova-3 / flux do not need this (they handle phrase
+     *   spotting differently and the prepend would only add noise).
+     * - Solve-Puzzle scene only: other WoF scenes (Letter-Guess,
+     *   Bonus-Round, etc.) do NOT want the slotMap phrase prepended — only
+     *   Puzzle-Solve depends on the phrase round-trip.
+     *
+     * @default false
+     */
+    appendSearch?: boolean;
     /**
      * Optional fallback ASR configurations
      *
@@ -1736,7 +1816,10 @@ declare enum GeminiModel {
 }
 
 /**
- * OpenAI Model Types
+ * OpenAI Batch API Model Types (HTTP API)
+ * @see https://platform.openai.com/docs/guides/speech-to-text
+ *
+ * Note: For openai-realtime provider models, see OpenAIRealtimeModel in provider.types.ts
  */
 declare enum OpenAIModel {
     WHISPER_1 = "whisper-1"
@@ -2934,5 +3017,5 @@ declare function getRecognitionConductorHttpBase(stage?: Stage | string | null |
 declare function getRecognitionConductorWsBase(stage?: Stage | string | null | undefined): string;
 declare function getRecognitionConductorHost(stage?: Stage | string | null | undefined): string;
 
-export { AudioEncoding, AwsTranscribeModel, BedrockModel, CartesiaModel, ClientControlActionV1, ClientState, ConfigBuilder, ConnectionError, ControlSignalTypeV1 as ControlSignal, ControlSignalTypeV1, DashScopeModel, DeepgramModel, ElevenLabsModel, ErrorTypeV1, FinalTranscriptStability, FireworksModel, GeminiModel, GladiaModel, GoogleModel, Language, MistralVoxtralModel, OpenAIModel, OpenAIRealtimeModel, RECOGNITION_CONDUCTOR_BASES, RECOGNITION_SERVICE_BASES, RealTimeTwoWayWebSocketRecognitionClient, RecognitionContextTypeV1, RecognitionError, RecognitionProvider, RecognitionResultTypeV1, RecognitionVGFStateSchema, RecordingStatus, STAGES, SampleRate, SelfServeVllmModel, SimplifiedVGFRecognitionClient, TimeoutError, TranscriptionStatus, ValidationError, createClient, createClientWithBuilder, createDefaultASRConfig, createInitialRecognitionState, createSimplifiedVGFClient, getRecognitionConductorBase, getRecognitionConductorHost, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionServiceBase, getRecognitionServiceHost, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getUserFriendlyMessage, isExceptionImmediatelyAvailable, isNormalDisconnection, isValidRecordingStatusTransition, normalizeStage, resetRecognitionVGFState };
+export { AmazonNovaSonicModel, AudioEncoding, AwsTranscribeModel, BedrockModel, CartesiaModel, ClientControlActionV1, ClientState, ConfigBuilder, ConnectionError, ControlSignalTypeV1 as ControlSignal, ControlSignalTypeV1, DashScopeModel, DeepgramModel, ElevenLabsModel, ErrorTypeV1, FinalTranscriptStability, FireworksModel, GeminiModel, GladiaModel, GoogleModel, Language, MistralVoxtralModel, OpenAIModel, OpenAIRealtimeModel, RECOGNITION_CONDUCTOR_BASES, RECOGNITION_SERVICE_BASES, RealTimeTwoWayWebSocketRecognitionClient, RecognitionContextTypeV1, RecognitionError, RecognitionProvider, RecognitionResultTypeV1, RecognitionVGFStateSchema, RecordingStatus, STAGES, SampleRate, SelfServeVllmModel, SimplifiedVGFRecognitionClient, TimeoutError, TranscriptionStatus, ValidationError, createClient, createClientWithBuilder, createDefaultASRConfig, createInitialRecognitionState, createSimplifiedVGFClient, getRecognitionConductorBase, getRecognitionConductorHost, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionServiceBase, getRecognitionServiceHost, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getUserFriendlyMessage, isExceptionImmediatelyAvailable, isNormalDisconnection, isValidRecordingStatusTransition, normalizeStage, resetRecognitionVGFState };
 export type { ASRRequestConfig, ASRRequestV1, AudioMetricsResultV1, AuthenticationException, ConnectionException, ErrorResultV1, FunctionCallResultV1, GameContextV1, IRecognitionClient, IRecognitionClientConfig, IRecognitionClientStats, ISimplifiedVGFRecognitionClient, MetadataResultV1, ProviderException, QuotaExceededException, RealTimeTwoWayWebSocketRecognitionClientConfig, RecognitionCallbackUrl, RecognitionException, RecognitionState, RecordingStatusType, SimplifiedVGFClientConfig, SlotMap, Stage, TimeoutException, TranscriptionResult, TranscriptionResultV1, TranscriptionStatusType, UnknownException, ValidationException };

package/dist/index.d.ts

@@ -11,6 +11,6 @@ export { type RecognitionState, RecognitionVGFStateSchema, RecordingStatus, Tran
 export { resetRecognitionVGFState } from './vgf-recognition-mapper.js';
 export { AudioEncoding } from '@recog/websocket';
 export { type GameContextV1, type SlotMap, RecognitionContextTypeV1, ControlSignalTypeV1, ControlSignalTypeV1 as ControlSignal, // Alias for backward compatibility
-type TranscriptionResultV1, type FunctionCallResultV1, type MetadataResultV1, type AudioMetricsResultV1, type ErrorResultV1, RecognitionResultTypeV1, ClientControlActionV1, type ASRRequestConfig, type ASRRequestV1, FinalTranscriptStability, createDefaultASRConfig, RecognitionProvider, DeepgramModel, ElevenLabsModel, FireworksModel, GladiaModel, GoogleModel, GeminiModel, OpenAIModel, SelfServeVllmModel, OpenAIRealtimeModel, MistralVoxtralModel, CartesiaModel, DashScopeModel, BedrockModel, AwsTranscribeModel, Language, SampleRate, STAGES, type Stage } from '@recog/shared-types';
+type TranscriptionResultV1, type FunctionCallResultV1, type MetadataResultV1, type AudioMetricsResultV1, type ErrorResultV1, RecognitionResultTypeV1, ClientControlActionV1, type ASRRequestConfig, type ASRRequestV1, FinalTranscriptStability, createDefaultASRConfig, RecognitionProvider, DeepgramModel, ElevenLabsModel, FireworksModel, GladiaModel, GoogleModel, GeminiModel, OpenAIModel, SelfServeVllmModel, OpenAIRealtimeModel, MistralVoxtralModel, CartesiaModel, DashScopeModel, BedrockModel, AwsTranscribeModel, AmazonNovaSonicModel, Language, SampleRate, STAGES, type Stage } from '@recog/shared-types';
 export { getRecognitionServiceBase, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getRecognitionServiceHost, getRecognitionConductorBase, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionConductorHost, normalizeStage, RECOGNITION_SERVICE_BASES, RECOGNITION_CONDUCTOR_BASES } from '@recog/shared-config';
 //# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AACA,OAAO,EACL,wCAAwC,EACxC,KAAK,8CAA8C,EACnD,KAAK,mBAAmB,EACxB,qBAAqB,EACtB,MAAM,yBAAyB,CAAC;AAGjC,OAAO,EACL,KAAK,kBAAkB,EACvB,KAAK,wBAAwB,EAC7B,KAAK,uBAAuB,EAC5B,KAAK,sBAAsB,EAC3B,WAAW,EACZ,MAAM,+BAA+B,CAAC;AAGvC,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AAGpD,OAAO,EAAE,YAAY,EAAE,uBAAuB,EAAE,MAAM,cAAc,CAAC;AAGrE,OAAO,EACL,gBAAgB,EAChB,eAAe,EACf,YAAY,EACZ,eAAe,EAChB,MAAM,aAAa,CAAC;AAGrB,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAGlD,YAAY,EACV,oBAAoB,EACpB,mBAAmB,EACnB,gBAAgB,EAChB,mBAAmB,EACnB,uBAAuB,EACvB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EACjB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,+BAA+B,EAC/B,sBAAsB,EACvB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,8BAA8B,EAC9B,yBAAyB,EACzB,KAAK,+BAA+B,EACpC,KAAK,yBAAyB,EAC/B,MAAM,wCAAwC,CAAC;AAEhD,OAAO,EACL,KAAK,gBAAgB,EACrB,yBAAyB,EACzB,eAAe,EACf,mBAAmB,EACnB,KAAK,mBAAmB,EACxB,KAAK,uBAAuB,EAC5B,6BAA6B,EAC7B,gCAAgC,EACjC,MAAM,4BAA4B,CAAC;AAEpC,OAAO,EAAE,wBAAwB,EAAE,MAAM,6BAA6B,CAAC;AAGvE,OAAO,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAGjD,OAAO,EAEL,KAAK,aAAa,EAClB,KAAK,OAAO,EACZ,wBAAwB,EACxB,mBAAmB,EACnB,mBAAmB,IAAI,aAAa,EAAG,mCAAmC;AAG1E,KAAK,qBAAqB,EAC1B,KAAK,oBAAoB,EACzB,KAAK,gBAAgB,EACrB,KAAK,oBAAoB,EACzB,KAAK,aAAa,EAClB,uBAAuB,EACvB,qBAAqB,EAGrB,KAAK,gBAAgB,EACrB,KAAK,YAAY,EACjB,wBAAwB,EACxB,sBAAsB,EACtB,mBAAmB,EACnB,aAAa,EACb,eAAe,EACf,cAAc,EACd,WAAW,EACX,WAAW,EACX,WAAW,EACX,WAAW,EACX,kBAAkB,EAClB,mBAAmB,EACnB,mBAAmB,EACnB,aAAa,EACb,cAAc,EACd,YAAY,EACZ,kBAAkB,EAClB,QAAQ,EACR,UAAU,EAGV,MAAM,EACN,KAAK,KAAK,EACX,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,yBAAyB,EACzB,6BAA6B,EAC7B,2BAA2B,EAC3B,yBAAyB,EACzB,2BAA2B,EAC3B,+BAA+B,EAC/B,6BAA6B,EAC7B,2BAA2B,EAC3B,cAAc,EACd,yBAAyB,EACzB,2BAA2B,EAC5B,MAAM,sBAAsB,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AACA,OAAO,EACL,wCAAwC,EACxC,KAAK,8CAA8C,EACnD,KAAK,mBAAmB,EACxB,qBAAqB,EACtB,MAAM,yBAAyB,CAAC;AAGjC,OAAO,EACL,KAAK,kBAAkB,EACvB,KAAK,wBAAwB,EAC7B,KAAK,uBAAuB,EAC5B,KAAK,sBAAsB,EAC3B,WAAW,EACZ,MAAM,+BAA+B,CAAC;AAGvC,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AAGpD,OAAO,EAAE,YAAY,EAAE,uBAAuB,EAAE,MAAM,cAAc,CAAC;AAGrE,OAAO,EACL,gBAAgB,EAChB,eAAe,EACf,YAAY,EACZ,eAAe,EAChB,MAAM,aAAa,CAAC;AAGrB,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAGlD,YAAY,EACV,oBAAoB,EACpB,mBAAmB,EACnB,gBAAgB,EAChB,mBAAmB,EACnB,uBAAuB,EACvB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EACjB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,+BAA+B,EAC/B,sBAAsB,EACvB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,8BAA8B,EAC9B,yBAAyB,EACzB,KAAK,+BAA+B,EACpC,KAAK,yBAAyB,EAC/B,MAAM,wCAAwC,CAAC;AAEhD,OAAO,EACL,KAAK,gBAAgB,EACrB,yBAAyB,EACzB,eAAe,EACf,mBAAmB,EACnB,KAAK,mBAAmB,EACxB,KAAK,uBAAuB,EAC5B,6BAA6B,EAC7B,gCAAgC,EACjC,MAAM,4BAA4B,CAAC;AAEpC,OAAO,EAAE,wBAAwB,EAAE,MAAM,6BAA6B,CAAC;AAGvE,OAAO,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAGjD,OAAO,EAEL,KAAK,aAAa,EAClB,KAAK,OAAO,EACZ,wBAAwB,EACxB,mBAAmB,EACnB,mBAAmB,IAAI,aAAa,EAAG,mCAAmC;AAG1E,KAAK,qBAAqB,EAC1B,KAAK,oBAAoB,EACzB,KAAK,gBAAgB,EACrB,KAAK,oBAAoB,EACzB,KAAK,aAAa,EAClB,uBAAuB,EACvB,qBAAqB,EAGrB,KAAK,gBAAgB,EACrB,KAAK,YAAY,EACjB,wBAAwB,EACxB,sBAAsB,EACtB,mBAAmB,EACnB,aAAa,EACb,eAAe,EACf,cAAc,EACd,WAAW,EACX,WAAW,EACX,WAAW,EACX,WAAW,EACX,kBAAkB,EAClB,mBAAmB,EACnB,mBAAmB,EACnB,aAAa,EACb,cAAc,EACd,YAAY,EACZ,kBAAkB,EAClB,oBAAoB,EACpB,QAAQ,EACR,UAAU,EAGV,MAAM,EACN,KAAK,KAAK,EACX,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,yBAAyB,EACzB,6BAA6B,EAC7B,2BAA2B,EAC3B,yBAAyB,EACzB,2BAA2B,EAC3B,+BAA+B,EAC/B,6BAA6B,EAC7B,2BAA2B,EAC3B,cAAc,EACd,yBAAyB,EACzB,2BAA2B,EAC5B,MAAM,sBAAsB,CAAC"}

package/dist/index.js

@@ -3749,6 +3749,7 @@ var RecognitionProvider;
   RecognitionProvider2["BEDROCK"] = "bedrock";
   RecognitionProvider2["INWORLD_STT"] = "inworld-stt";
   RecognitionProvider2["AWS_TRANSCRIBE"] = "aws-transcribe";
+  RecognitionProvider2["AMAZON_NOVA_SONIC"] = "amazon-nova-sonic";
   RecognitionProvider2["TEST_ASR_PROVIDER_QUOTA"] = "test-asr-provider-quota";
   RecognitionProvider2["TEST_ASR_STREAMING"] = "test-asr-streaming";
 })(RecognitionProvider || (RecognitionProvider = {}));
@@ -3799,6 +3800,7 @@ var ElevenLabsModel;
 })(ElevenLabsModel || (ElevenLabsModel = {}));
 var OpenAIRealtimeModel;
 (function(OpenAIRealtimeModel2) {
+  OpenAIRealtimeModel2["GPT_REALTIME_WHISPER"] = "gpt-realtime-whisper";
   OpenAIRealtimeModel2["GPT_4O_TRANSCRIBE"] = "gpt-4o-transcribe";
   OpenAIRealtimeModel2["GPT_4O_MINI_TRANSCRIBE"] = "gpt-4o-mini-transcribe";
 })(OpenAIRealtimeModel || (OpenAIRealtimeModel = {}));
@@ -3829,6 +3831,11 @@ var AwsTranscribeModel;
 (function(AwsTranscribeModel2) {
   AwsTranscribeModel2["DEFAULT"] = "default";
 })(AwsTranscribeModel || (AwsTranscribeModel = {}));
+var AmazonNovaSonicModel;
+(function(AmazonNovaSonicModel2) {
+  AmazonNovaSonicModel2["AMAZON_NOVA_SONIC_V1"] = "amazon.nova-sonic-v1:0";
+  AmazonNovaSonicModel2["AMAZON_NOVA_2_SONIC"] = "amazon.nova-2-sonic-v1:0";
+})(AmazonNovaSonicModel || (AmazonNovaSonicModel = {}));
 var SelfServeVllmModel;
 (function(SelfServeVllmModel2) {
   SelfServeVllmModel2["QWEN3_ASR_1_7B"] = "qwen3-asr-1.7b";
@@ -3845,6 +3852,18 @@ var RecognitionResultTypeV1;
   RecognitionResultTypeV12["AUDIO_METRICS"] = "AudioMetrics";
   RecognitionResultTypeV12["SESSION_CONFIGURED"] = "SessionConfigured";
 })(RecognitionResultTypeV1 || (RecognitionResultTypeV1 = {}));
+var DetectionTypeV1;
+(function(DetectionTypeV12) {
+  DetectionTypeV12["SEARCH"] = "search";
+})(DetectionTypeV1 || (DetectionTypeV1 = {}));
+var DetectionV1Schema = z.object({
+  type: z.nativeEnum(DetectionTypeV1),
+  query: z.string(),
+  score: z.number().min(0).max(1),
+  startMs: z.number().optional(),
+  endMs: z.number().optional()
+  // Audio time (ms from stream start) where the hit ends
+});
 var TranscriptionResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.TRANSCRIPTION),
   audioUtteranceId: z.string(),
@@ -3863,8 +3882,9 @@ var TranscriptionResultSchemaV1 = z.object({
   endTimestamp: z.number().optional(),
   receivedAtMs: z.number().optional(),
   accumulatedAudioTimeMs: z.number().optional(),
-  rawAudioTimeMs: z.number().optional()
-  // Total audio duration sent to provider (includes prefix)
+  rawAudioTimeMs: z.number().optional(),
+  detections: z.array(DetectionV1Schema).optional()
+  // Provider-reported phrase detections (query + score, optionally startMs/endMs). Always populated when the provider returns hits, regardless of `appendSearch`. Other providers leave this undefined.
 });
 var FunctionCallResultSchemaV1 = z.object({
   type: z.literal(RecognitionResultTypeV1.FUNCTION_CALL),
@@ -4117,7 +4137,15 @@ var TranscriptMessageSchema = z.object({
    * @example true
    * @default false
    */
-  is_fallback: z.boolean().optional()
+  is_fallback: z.boolean().optional(),
+  /**
+   * Provider-reported phrase detections (query + score, optionally
+   * startMs/endMs). Always populated when the provider returns hits,
+   * regardless of `appendSearch` or scene gating. Other providers leave
+   * this undefined.
+   * @example [{ query: 'justin bieber one time', score: 0.78, startMs: 1200, endMs: 2800 }]
+   */
+  detections: z.array(DetectionV1Schema).optional()
 });
 var VADEndSignalSchema = z.object({
   type: z.literal(ProviderMessageType.VAD_END_SIGNAL),
@@ -4457,6 +4485,9 @@ var ASRRequestSchemaV1 = z.object({
   // 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(),
+  // Opt-in: round-trip Deepgram `search` phrase hits into the transcript.
+  // Active only when (model = deepgram nova-2) AND (GameContext.gamePhase = 'Solve Puzzle'). See ASRRequestConfig.appendSearch in asr-config.types.ts for full semantics.
+  appendSearch: z.boolean().optional(),
   // Debug options (FOR DEBUG/TESTING ONLY - not for production use)
   debugCommand: RequestDebugCommandSchema
 });
@@ -5853,6 +5884,12 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
         // Streaming audio metrics opt-in (ms interval). Server only forwards metrics if > 0.
         ...this.config.asrRequestConfig.audioMetricsIntervalMs !== void 0 && {
           audioMetricsIntervalMs: this.config.asrRequestConfig.audioMetricsIntervalMs
+        },
+        // Opt-in: round-trip Deepgram nova-2 search-phrase hits into the
+        // transcript. Only fires server-side when (model = nova-2) AND
+        // (GameContext.gamePhase = 'Solve Puzzle'). See ASRRequestConfig.appendSearch.
+        ...this.config.asrRequestConfig.appendSearch !== void 0 && {
+          appendSearch: this.config.asrRequestConfig.appendSearch
         }
       };
       super.sendMessage(
@@ -6704,6 +6741,7 @@ function createSimplifiedVGFClient(config) {
   return new SimplifiedVGFRecognitionClient(config);
 }
 export {
+  AmazonNovaSonicModel,
   AudioEncoding,
   AwsTranscribeModel,
   BedrockModel,