@volley/recognition-client-sdk 0.1.622 → 0.1.670

package/dist/browser.bundled.d.ts

@@ -12,11 +12,14 @@ declare enum RecognitionProvider {
     DEEPGRAM = "deepgram",
     ELEVENLABS = "elevenlabs",
     FIREWORKS = "fireworks",
+    GLADIA = "gladia",
     GOOGLE = "google",
     GEMINI_BATCH = "gemini-batch",
     OPENAI_BATCH = "openai-batch",
+    SELF_SERVE_VLLM = "self-serve-vllm",
     OPENAI_REALTIME = "openai-realtime",
     MISTRAL_VOXTRAL = "mistral-voxtral",
+    CARTESIA = "cartesia",
     DASHSCOPE = "dashscope",
     TEST_ASR_PROVIDER_QUOTA = "test-asr-provider-quota",
     TEST_ASR_STREAMING = "test-asr-streaming"
@@ -67,6 +70,13 @@ declare enum FireworksModel {
     WHISPER_V3 = "whisper-v3",
     WHISPER_V3_TURBO = "whisper-v3-turbo"
 }
+/**
+ * Gladia Solaria realtime transcription models
+ * @see https://docs.gladia.io/api-reference/v2/live/init
+ */
+declare enum GladiaModel {
+    SOLARIA_1 = "solaria-1"
+}
 /**
  * ElevenLabs Scribe models for speech-to-text
  * @see https://elevenlabs.io/blog/introducing-scribe-v2-realtime
@@ -74,8 +84,7 @@ declare enum FireworksModel {
  * @see https://elevenlabs.io/docs/api-reference/speech-to-text/convert
  */
 declare enum ElevenLabsModel {
-    SCRIBE_V2_REALTIME = "scribe_v2_realtime",
-    SCRIBE_V1 = "scribe_v1"
+    SCRIBE_V2_REALTIME = "scribe_v2_realtime"
 }
 /**
  * OpenAI Realtime API transcription models
@@ -94,6 +103,14 @@ declare enum OpenAIRealtimeModel {
 declare enum MistralVoxtralModel {
     VOXTRAL_MINI_REALTIME_2602 = "voxtral-mini-transcribe-realtime-2602"
 }
+/**
+ * Cartesia Ink-Whisper Realtime transcription models
+ * @see https://docs.cartesia.ai/build-with-cartesia/stt-models
+ */
+declare enum CartesiaModel {
+    INK_WHISPER = "ink-whisper",
+    INK_WHISPER_20250604 = "ink-whisper-2025-06-04"
+}
 /**
  * DashScope Qwen-ASR Realtime transcription models
  * @see https://www.alibabacloud.com/help/en/model-studio/qwen-real-time-speech-recognition
@@ -102,10 +119,17 @@ declare enum DashScopeModel {
     QWEN3_ASR_FLASH_REALTIME_2602 = "qwen3-asr-flash-realtime-2026-02-10",
     QWEN3_ASR_FLASH_REALTIME = "qwen3-asr-flash-realtime"
 }
+/**
+ * Self-serve vLLM batch transcription models
+ * Backed by recognition-inference / RunPod `/transcribe`
+ */
+declare enum SelfServeVllmModel {
+    QWEN3_ASR_1_7B = "qwen3-asr-1.7b"
+}
 /**
  * Type alias for any model from any provider
  */
-type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | DashScopeModel | string;
+type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | SelfServeVllmModel | string;
 
 /**
  * Audio encoding types
@@ -369,6 +393,8 @@ declare const MetadataResultSchemaV1: z.ZodObject<{
     rawAudioTimeMs: z.ZodOptional<z.ZodNumber>;
     costInUSD: z.ZodOptional<z.ZodDefault<z.ZodNumber>>;
     apiType: z.ZodOptional<z.ZodNativeEnum<typeof ASRApiType>>;
+    provider: z.ZodOptional<z.ZodString>;
+    model: z.ZodOptional<z.ZodString>;
     asrConfig: z.ZodOptional<z.ZodString>;
     rawAsrMetadata: z.ZodOptional<z.ZodString>;
     transcriptOutcome: z.ZodOptional<z.ZodNativeEnum<typeof TranscriptOutcomeType>>;
@@ -423,6 +449,8 @@ declare const MetadataResultSchemaV1: z.ZodObject<{
     rawAudioTimeMs?: number | undefined;
     costInUSD?: number | undefined;
     apiType?: ASRApiType | undefined;
+    provider?: string | undefined;
+    model?: string | undefined;
     asrConfig?: string | undefined;
     rawAsrMetadata?: string | undefined;
     transcriptOutcome?: TranscriptOutcomeType | undefined;
@@ -453,6 +481,8 @@ declare const MetadataResultSchemaV1: z.ZodObject<{
     rawAudioTimeMs?: number | undefined;
     costInUSD?: number | undefined;
     apiType?: ASRApiType | undefined;
+    provider?: string | undefined;
+    model?: string | undefined;
     asrConfig?: string | undefined;
     rawAsrMetadata?: string | undefined;
     transcriptOutcome?: TranscriptOutcomeType | undefined;

package/dist/index.bundled.d.ts

@@ -12,11 +12,14 @@ declare enum RecognitionProvider {
     DEEPGRAM = "deepgram",
     ELEVENLABS = "elevenlabs",
     FIREWORKS = "fireworks",
+    GLADIA = "gladia",
     GOOGLE = "google",
     GEMINI_BATCH = "gemini-batch",
     OPENAI_BATCH = "openai-batch",
+    SELF_SERVE_VLLM = "self-serve-vllm",
     OPENAI_REALTIME = "openai-realtime",
     MISTRAL_VOXTRAL = "mistral-voxtral",
+    CARTESIA = "cartesia",
     DASHSCOPE = "dashscope",
     TEST_ASR_PROVIDER_QUOTA = "test-asr-provider-quota",
     TEST_ASR_STREAMING = "test-asr-streaming"
@@ -67,6 +70,13 @@ declare enum FireworksModel {
     WHISPER_V3 = "whisper-v3",
     WHISPER_V3_TURBO = "whisper-v3-turbo"
 }
+/**
+ * Gladia Solaria realtime transcription models
+ * @see https://docs.gladia.io/api-reference/v2/live/init
+ */
+declare enum GladiaModel {
+    SOLARIA_1 = "solaria-1"
+}
 /**
  * ElevenLabs Scribe models for speech-to-text
  * @see https://elevenlabs.io/blog/introducing-scribe-v2-realtime
@@ -74,8 +84,7 @@ declare enum FireworksModel {
  * @see https://elevenlabs.io/docs/api-reference/speech-to-text/convert
  */
 declare enum ElevenLabsModel {
-    SCRIBE_V2_REALTIME = "scribe_v2_realtime",
-    SCRIBE_V1 = "scribe_v1"
+    SCRIBE_V2_REALTIME = "scribe_v2_realtime"
 }
 /**
  * OpenAI Realtime API transcription models
@@ -94,6 +103,14 @@ declare enum OpenAIRealtimeModel {
 declare enum MistralVoxtralModel {
     VOXTRAL_MINI_REALTIME_2602 = "voxtral-mini-transcribe-realtime-2602"
 }
+/**
+ * Cartesia Ink-Whisper Realtime transcription models
+ * @see https://docs.cartesia.ai/build-with-cartesia/stt-models
+ */
+declare enum CartesiaModel {
+    INK_WHISPER = "ink-whisper",
+    INK_WHISPER_20250604 = "ink-whisper-2025-06-04"
+}
 /**
  * DashScope Qwen-ASR Realtime transcription models
  * @see https://www.alibabacloud.com/help/en/model-studio/qwen-real-time-speech-recognition
@@ -102,10 +119,17 @@ declare enum DashScopeModel {
     QWEN3_ASR_FLASH_REALTIME_2602 = "qwen3-asr-flash-realtime-2026-02-10",
     QWEN3_ASR_FLASH_REALTIME = "qwen3-asr-flash-realtime"
 }
+/**
+ * Self-serve vLLM batch transcription models
+ * Backed by recognition-inference / RunPod `/transcribe`
+ */
+declare enum SelfServeVllmModel {
+    QWEN3_ASR_1_7B = "qwen3-asr-1.7b"
+}
 /**
  * Type alias for any model from any provider
  */
-type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | DashScopeModel | string;
+type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | SelfServeVllmModel | string;
 
 /**
  * Audio encoding types
@@ -369,6 +393,8 @@ declare const MetadataResultSchemaV1: z.ZodObject<{
     rawAudioTimeMs: z.ZodOptional<z.ZodNumber>;
     costInUSD: z.ZodOptional<z.ZodDefault<z.ZodNumber>>;
     apiType: z.ZodOptional<z.ZodNativeEnum<typeof ASRApiType>>;
+    provider: z.ZodOptional<z.ZodString>;
+    model: z.ZodOptional<z.ZodString>;
     asrConfig: z.ZodOptional<z.ZodString>;
     rawAsrMetadata: z.ZodOptional<z.ZodString>;
     transcriptOutcome: z.ZodOptional<z.ZodNativeEnum<typeof TranscriptOutcomeType>>;
@@ -423,6 +449,8 @@ declare const MetadataResultSchemaV1: z.ZodObject<{
     rawAudioTimeMs?: number | undefined;
     costInUSD?: number | undefined;
     apiType?: ASRApiType | undefined;
+    provider?: string | undefined;
+    model?: string | undefined;
     asrConfig?: string | undefined;
     rawAsrMetadata?: string | undefined;
     transcriptOutcome?: TranscriptOutcomeType | undefined;
@@ -453,6 +481,8 @@ declare const MetadataResultSchemaV1: z.ZodObject<{
     rawAudioTimeMs?: number | undefined;
     costInUSD?: number | undefined;
     apiType?: ASRApiType | undefined;
+    provider?: string | undefined;
+    model?: string | undefined;
     asrConfig?: string | undefined;
     rawAsrMetadata?: string | undefined;
     transcriptOutcome?: TranscriptOutcomeType | undefined;
@@ -2737,5 +2767,5 @@ 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, ClientControlActionV1, ClientState, ConfigBuilder, ConnectionError, ControlSignalTypeV1 as ControlSignal, ControlSignalTypeV1, DashScopeModel, DeepgramModel, ElevenLabsModel, ErrorTypeV1, FinalTranscriptStability, FireworksModel, GeminiModel, GoogleModel, Language, MistralVoxtralModel, OpenAIModel, OpenAIRealtimeModel, RECOGNITION_CONDUCTOR_BASES, RECOGNITION_SERVICE_BASES, RealTimeTwoWayWebSocketRecognitionClient, RecognitionContextTypeV1, RecognitionError, RecognitionProvider, RecognitionResultTypeV1, RecognitionVGFStateSchema, RecordingStatus, STAGES, SampleRate, SimplifiedVGFRecognitionClient, TimeoutError, TranscriptionStatus, ValidationError, createClient, createClientWithBuilder, createDefaultASRConfig, createInitialRecognitionState, createSimplifiedVGFClient, getRecognitionConductorBase, getRecognitionConductorHost, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionServiceBase, getRecognitionServiceHost, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getUserFriendlyMessage, isExceptionImmediatelyAvailable, isNormalDisconnection, isValidRecordingStatusTransition, normalizeStage, resetRecognitionVGFState };
+export { AudioEncoding, 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, 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 ErrorResultV1, RecognitionResultTypeV1, ClientControlActionV1, type ASRRequestConfig, type ASRRequestV1, FinalTranscriptStability, createDefaultASRConfig, RecognitionProvider, DeepgramModel, ElevenLabsModel, FireworksModel, GoogleModel, GeminiModel, OpenAIModel, OpenAIRealtimeModel, MistralVoxtralModel, DashScopeModel, Language, SampleRate, STAGES, type Stage } from '@recog/shared-types';
+type TranscriptionResultV1, type FunctionCallResultV1, type MetadataResultV1, type ErrorResultV1, RecognitionResultTypeV1, ClientControlActionV1, type ASRRequestConfig, type ASRRequestV1, FinalTranscriptStability, createDefaultASRConfig, RecognitionProvider, DeepgramModel, ElevenLabsModel, FireworksModel, GladiaModel, GoogleModel, GeminiModel, OpenAIModel, SelfServeVllmModel, OpenAIRealtimeModel, MistralVoxtralModel, CartesiaModel, DashScopeModel, 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,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,mBAAmB,EACnB,mBAAmB,EACnB,cAAc,EACd,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,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,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

@@ -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;
@@ -3874,6 +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)
@@ -5249,7 +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
@@ -5597,7 +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;
     }
@@ -5637,7 +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");
@@ -5807,7 +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)",
@@ -5929,7 +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;
     }
@@ -6591,6 +6609,7 @@ function createSimplifiedVGFClient(config) {
 }
 export {
   AudioEncoding,
+  CartesiaModel,
   ClientControlActionV1,
   ClientState,
   ConfigBuilder,
@@ -6604,6 +6623,7 @@ export {
   FinalTranscriptStability,
   FireworksModel,
   GeminiModel,
+  GladiaModel,
   GoogleModel,
   Language,
   MistralVoxtralModel,
@@ -6620,6 +6640,7 @@ export {
   RecordingStatus,
   STAGES,
   SampleRate,
+  SelfServeVllmModel,
   SimplifiedVGFRecognitionClient,
   TimeoutError,
   TranscriptionStatus,