@ondewo/s2t-client-typescript 5.3.0 → 5.5.0

package/api/ondewo/s2t/speech-to-text_pb.d.ts

@@ -28,16 +28,21 @@ export class TranscribeRequestConfig extends jspb.Message {
   hasPyannote(): boolean;
   clearPyannote(): TranscribeRequestConfig;
 
-  getMatchbox(): Matchbox | undefined;
-  setMatchbox(value?: Matchbox): TranscribeRequestConfig;
-  hasMatchbox(): boolean;
-  clearMatchbox(): TranscribeRequestConfig;
-
   getReturnOptions(): TranscriptionReturnOptions | undefined;
   setReturnOptions(value?: TranscriptionReturnOptions): TranscribeRequestConfig;
   hasReturnOptions(): boolean;
   clearReturnOptions(): TranscribeRequestConfig;
 
+  getLanguage(): string;
+  setLanguage(value: string): TranscribeRequestConfig;
+  hasLanguage(): boolean;
+  clearLanguage(): TranscribeRequestConfig;
+
+  getTask(): string;
+  setTask(value: string): TranscribeRequestConfig;
+  hasTask(): boolean;
+  clearTask(): TranscribeRequestConfig;
+
   getOneofLanguageModelNameCase(): TranscribeRequestConfig.OneofLanguageModelNameCase;
 
   getOneofPostProcessingCase(): TranscribeRequestConfig.OneofPostProcessingCase;
@@ -64,8 +69,9 @@ export namespace TranscribeRequestConfig {
     postProcessing?: PostProcessingOptions.AsObject,
     utteranceDetection?: UtteranceDetectionOptions.AsObject,
     pyannote?: Pyannote.AsObject,
-    matchbox?: Matchbox.AsObject,
     returnOptions?: TranscriptionReturnOptions.AsObject,
+    language?: string,
+    task?: string,
   }
 
   export enum OneofLanguageModelNameCase { 
@@ -86,13 +92,22 @@ export namespace TranscribeRequestConfig {
   export enum VoiceActivityDetectionCase { 
     VOICE_ACTIVITY_DETECTION_NOT_SET = 0,
     PYANNOTE = 6,
-    MATCHBOX = 7,
   }
 
   export enum OneofReturnOptionsCase { 
     ONEOF_RETURN_OPTIONS_NOT_SET = 0,
     RETURN_OPTIONS = 8,
   }
+
+  export enum LanguageCase { 
+    _LANGUAGE_NOT_SET = 0,
+    LANGUAGE = 9,
+  }
+
+  export enum TaskCase { 
+    _TASK_NOT_SET = 0,
+    TASK = 10,
+  }
 }
 
 export class TranscriptionReturnOptions extends jspb.Message {
@@ -206,6 +221,126 @@ export namespace PostProcessingOptions {
   }
 }
 
+export class Transcription extends jspb.Message {
+  getTranscription(): string;
+  setTranscription(value: string): Transcription;
+
+  getConfidenceScore(): number;
+  setConfidenceScore(value: number): Transcription;
+
+  getWordsList(): Array<WordDetail>;
+  setWordsList(value: Array<WordDetail>): Transcription;
+  clearWordsList(): Transcription;
+  addWords(value?: WordDetail, index?: number): WordDetail;
+
+  getAlternativesList(): Array<TranscriptionAlternative>;
+  setAlternativesList(value: Array<TranscriptionAlternative>): Transcription;
+  clearAlternativesList(): Transcription;
+  addAlternatives(value?: TranscriptionAlternative, index?: number): TranscriptionAlternative;
+
+  serializeBinary(): Uint8Array;
+  toObject(includeInstance?: boolean): Transcription.AsObject;
+  static toObject(includeInstance: boolean, msg: Transcription): Transcription.AsObject;
+  static serializeBinaryToWriter(message: Transcription, writer: jspb.BinaryWriter): void;
+  static deserializeBinary(bytes: Uint8Array): Transcription;
+  static deserializeBinaryFromReader(message: Transcription, reader: jspb.BinaryReader): Transcription;
+}
+
+export namespace Transcription {
+  export type AsObject = {
+    transcription: string,
+    confidenceScore: number,
+    wordsList: Array<WordDetail.AsObject>,
+    alternativesList: Array<TranscriptionAlternative.AsObject>,
+  }
+}
+
+export class TranscriptionAlternative extends jspb.Message {
+  getTranscript(): string;
+  setTranscript(value: string): TranscriptionAlternative;
+
+  getConfidence(): number;
+  setConfidence(value: number): TranscriptionAlternative;
+
+  getWordsList(): Array<WordDetail>;
+  setWordsList(value: Array<WordDetail>): TranscriptionAlternative;
+  clearWordsList(): TranscriptionAlternative;
+  addWords(value?: WordDetail, index?: number): WordDetail;
+
+  serializeBinary(): Uint8Array;
+  toObject(includeInstance?: boolean): TranscriptionAlternative.AsObject;
+  static toObject(includeInstance: boolean, msg: TranscriptionAlternative): TranscriptionAlternative.AsObject;
+  static serializeBinaryToWriter(message: TranscriptionAlternative, writer: jspb.BinaryWriter): void;
+  static deserializeBinary(bytes: Uint8Array): TranscriptionAlternative;
+  static deserializeBinaryFromReader(message: TranscriptionAlternative, reader: jspb.BinaryReader): TranscriptionAlternative;
+}
+
+export namespace TranscriptionAlternative {
+  export type AsObject = {
+    transcript: string,
+    confidence: number,
+    wordsList: Array<WordDetail.AsObject>,
+  }
+}
+
+export class WordDetail extends jspb.Message {
+  getStartTime(): number;
+  setStartTime(value: number): WordDetail;
+
+  getEndTime(): number;
+  setEndTime(value: number): WordDetail;
+
+  getWord(): string;
+  setWord(value: string): WordDetail;
+
+  getConfidence(): number;
+  setConfidence(value: number): WordDetail;
+
+  getWordAlternativesList(): Array<WordAlternative>;
+  setWordAlternativesList(value: Array<WordAlternative>): WordDetail;
+  clearWordAlternativesList(): WordDetail;
+  addWordAlternatives(value?: WordAlternative, index?: number): WordAlternative;
+
+  serializeBinary(): Uint8Array;
+  toObject(includeInstance?: boolean): WordDetail.AsObject;
+  static toObject(includeInstance: boolean, msg: WordDetail): WordDetail.AsObject;
+  static serializeBinaryToWriter(message: WordDetail, writer: jspb.BinaryWriter): void;
+  static deserializeBinary(bytes: Uint8Array): WordDetail;
+  static deserializeBinaryFromReader(message: WordDetail, reader: jspb.BinaryReader): WordDetail;
+}
+
+export namespace WordDetail {
+  export type AsObject = {
+    startTime: number,
+    endTime: number,
+    word: string,
+    confidence: number,
+    wordAlternativesList: Array<WordAlternative.AsObject>,
+  }
+}
+
+export class WordAlternative extends jspb.Message {
+  getWord(): string;
+  setWord(value: string): WordAlternative;
+
+  getConfidence(): number;
+  setConfidence(value: number): WordAlternative;
+
+  serializeBinary(): Uint8Array;
+  toObject(includeInstance?: boolean): WordAlternative.AsObject;
+  static toObject(includeInstance: boolean, msg: WordAlternative): WordAlternative.AsObject;
+  static serializeBinaryToWriter(message: WordAlternative, writer: jspb.BinaryWriter): void;
+  static deserializeBinary(bytes: Uint8Array): WordAlternative;
+  static deserializeBinaryFromReader(message: WordAlternative, reader: jspb.BinaryReader): WordAlternative;
+}
+
+export namespace WordAlternative {
+  export type AsObject = {
+    word: string,
+    confidence: number,
+  }
+}
+
 export class TranscribeStreamRequest extends jspb.Message {
   getAudioChunk(): Uint8Array | string;
   getAudioChunk_asU8(): Uint8Array;
@@ -240,28 +375,6 @@ export namespace TranscribeStreamRequest {
   }
 }
 
-export class Transcription extends jspb.Message {
-  getTranscription(): string;
-  setTranscription(value: string): Transcription;
-
-  getConfidenceScore(): number;
-  setConfidenceScore(value: number): Transcription;
-
-  serializeBinary(): Uint8Array;
-  toObject(includeInstance?: boolean): Transcription.AsObject;
-  static toObject(includeInstance: boolean, msg: Transcription): Transcription.AsObject;
-  static serializeBinaryToWriter(message: Transcription, writer: jspb.BinaryWriter): void;
-  static deserializeBinary(bytes: Uint8Array): Transcription;
-  static deserializeBinaryFromReader(message: Transcription, reader: jspb.BinaryReader): Transcription;
-}
-
-export namespace Transcription {
-  export type AsObject = {
-    transcription: string,
-    confidenceScore: number,
-  }
-}
-
 export class TranscribeStreamResponse extends jspb.Message {
   getTranscriptionsList(): Array<Transcription>;
   setTranscriptionsList(value: Array<Transcription>): TranscribeStreamResponse;
@@ -356,11 +469,6 @@ export class TranscribeFileResponse extends jspb.Message {
   getTime(): number;
   setTime(value: number): TranscribeFileResponse;
 
-  getWordTimingList(): Array<WordTiming>;
-  setWordTimingList(value: Array<WordTiming>): TranscribeFileResponse;
-  clearWordTimingList(): TranscribeFileResponse;
-  addWordTiming(value?: WordTiming, index?: number): WordTiming;
-
   getAudioUuid(): string;
   setAudioUuid(value: string): TranscribeFileResponse;
 
@@ -376,37 +484,10 @@ export namespace TranscribeFileResponse {
   export type AsObject = {
     transcriptionsList: Array<Transcription.AsObject>,
     time: number,
-    wordTimingList: Array<WordTiming.AsObject>,
     audioUuid: string,
   }
 }
 
-export class WordTiming extends jspb.Message {
-  getWord(): string;
-  setWord(value: string): WordTiming;
-
-  getBegin(): number;
-  setBegin(value: number): WordTiming;
-
-  getEnd(): number;
-  setEnd(value: number): WordTiming;
-
-  serializeBinary(): Uint8Array;
-  toObject(includeInstance?: boolean): WordTiming.AsObject;
-  static toObject(includeInstance: boolean, msg: WordTiming): WordTiming.AsObject;
-  static serializeBinaryToWriter(message: WordTiming, writer: jspb.BinaryWriter): void;
-  static deserializeBinary(bytes: Uint8Array): WordTiming;
-  static deserializeBinaryFromReader(message: WordTiming, reader: jspb.BinaryReader): WordTiming;
-}
-
-export namespace WordTiming {
-  export type AsObject = {
-    word: string,
-    begin: number,
-    end: number,
-  }
-}
-
 export class S2tPipelineId extends jspb.Message {
   getId(): string;
   setId(value: string): S2tPipelineId;
@@ -713,16 +794,6 @@ export class AcousticModels extends jspb.Message {
   getType(): string;
   setType(value: string): AcousticModels;
 
-  getQuartznet(): Quartznet | undefined;
-  setQuartznet(value?: Quartznet): AcousticModels;
-  hasQuartznet(): boolean;
-  clearQuartznet(): AcousticModels;
-
-  getQuartznetTriton(): QuartznetTriton | undefined;
-  setQuartznetTriton(value?: QuartznetTriton): AcousticModels;
-  hasQuartznetTriton(): boolean;
-  clearQuartznetTriton(): AcousticModels;
-
   getWav2vec(): Wav2Vec | undefined;
   setWav2vec(value?: Wav2Vec): AcousticModels;
   hasWav2vec(): boolean;
@@ -754,8 +825,6 @@ export class AcousticModels extends jspb.Message {
 export namespace AcousticModels {
   export type AsObject = {
     type: string,
-    quartznet?: Quartznet.AsObject,
-    quartznetTriton?: QuartznetTriton.AsObject,
     wav2vec?: Wav2Vec.AsObject,
     wav2vecTriton?: Wav2VecTriton.AsObject,
     whisper?: Whisper.AsObject,
@@ -773,6 +842,9 @@ export class Whisper extends jspb.Message {
   getLanguage(): string;
   setLanguage(value: string): Whisper;
 
+  getTask(): string;
+  setTask(value: string): Whisper;
+
   serializeBinary(): Uint8Array;
   toObject(includeInstance?: boolean): Whisper.AsObject;
   static toObject(includeInstance: boolean, msg: Whisper): Whisper.AsObject;
@@ -786,6 +858,7 @@ export namespace Whisper {
     modelPath: string,
     useGpu: boolean,
     language: string,
+    task: string,
   }
 }
 
@@ -802,6 +875,18 @@ export class WhisperTriton extends jspb.Message {
   getCheckStatusTimeout(): number;
   setCheckStatusTimeout(value: number): WhisperTriton;
 
+  getLanguage(): string;
+  setLanguage(value: string): WhisperTriton;
+
+  getTask(): string;
+  setTask(value: string): WhisperTriton;
+
+  getTritonServerHost(): string;
+  setTritonServerHost(value: string): WhisperTriton;
+
+  getTritonServerPort(): number;
+  setTritonServerPort(value: number): WhisperTriton;
+
   serializeBinary(): Uint8Array;
   toObject(includeInstance?: boolean): WhisperTriton.AsObject;
   static toObject(includeInstance: boolean, msg: WhisperTriton): WhisperTriton.AsObject;
@@ -816,6 +901,10 @@ export namespace WhisperTriton {
     tritonModelName: string,
     tritonModelVersion: string,
     checkStatusTimeout: number,
+    language: string,
+    task: string,
+    tritonServerHost: string,
+    tritonServerPort: number,
   }
 }
 
@@ -854,6 +943,12 @@ export class Wav2VecTriton extends jspb.Message {
   getCheckStatusTimeout(): number;
   setCheckStatusTimeout(value: number): Wav2VecTriton;
 
+  getTritonServerHost(): string;
+  setTritonServerHost(value: string): Wav2VecTriton;
+
+  getTritonServerPort(): number;
+  setTritonServerPort(value: number): Wav2VecTriton;
+
   serializeBinary(): Uint8Array;
   toObject(includeInstance?: boolean): Wav2VecTriton.AsObject;
   static toObject(includeInstance: boolean, msg: Wav2VecTriton): Wav2VecTriton.AsObject;
@@ -868,44 +963,8 @@ export namespace Wav2VecTriton {
     tritonModelName: string,
     tritonModelVersion: string,
     checkStatusTimeout: number,
-  }
-}
-
-export class Quartznet extends jspb.Message {
-  getConfigPath(): string;
-  setConfigPath(value: string): Quartznet;
-
-  getLoadType(): string;
-  setLoadType(value: string): Quartznet;
-
-  getPtFiles(): PtFiles | undefined;
-  setPtFiles(value?: PtFiles): Quartznet;
-  hasPtFiles(): boolean;
-  clearPtFiles(): Quartznet;
-
-  getCkptFile(): CkptFile | undefined;
-  setCkptFile(value?: CkptFile): Quartznet;
-  hasCkptFile(): boolean;
-  clearCkptFile(): Quartznet;
-
-  getUseGpu(): boolean;
-  setUseGpu(value: boolean): Quartznet;
-
-  serializeBinary(): Uint8Array;
-  toObject(includeInstance?: boolean): Quartznet.AsObject;
-  static toObject(includeInstance: boolean, msg: Quartznet): Quartznet.AsObject;
-  static serializeBinaryToWriter(message: Quartznet, writer: jspb.BinaryWriter): void;
-  static deserializeBinary(bytes: Uint8Array): Quartznet;
-  static deserializeBinaryFromReader(message: Quartznet, reader: jspb.BinaryReader): Quartznet;
-}
-
-export namespace Quartznet {
-  export type AsObject = {
-    configPath: string,
-    loadType: string,
-    ptFiles?: PtFiles.AsObject,
-    ckptFile?: CkptFile.AsObject,
-    useGpu: boolean,
+    tritonServerHost: string,
+    tritonServerPort: number,
   }
 }
 
@@ -949,32 +1008,6 @@ export namespace CkptFile {
   }
 }
 
-export class QuartznetTriton extends jspb.Message {
-  getConfigPath(): string;
-  setConfigPath(value: string): QuartznetTriton;
-
-  getTritonUrl(): string;
-  setTritonUrl(value: string): QuartznetTriton;
-
-  getTritonModel(): string;
-  setTritonModel(value: string): QuartznetTriton;
-
-  serializeBinary(): Uint8Array;
-  toObject(includeInstance?: boolean): QuartznetTriton.AsObject;
-  static toObject(includeInstance: boolean, msg: QuartznetTriton): QuartznetTriton.AsObject;
-  static serializeBinaryToWriter(message: QuartznetTriton, writer: jspb.BinaryWriter): void;
-  static deserializeBinary(bytes: Uint8Array): QuartznetTriton;
-  static deserializeBinaryFromReader(message: QuartznetTriton, reader: jspb.BinaryReader): QuartznetTriton;
-}
-
-export namespace QuartznetTriton {
-  export type AsObject = {
-    configPath: string,
-    tritonUrl: string,
-    tritonModel: string,
-  }
-}
-
 export class LanguageModels extends jspb.Message {
   getPath(): string;
   setPath(value: string): LanguageModels;
@@ -1095,11 +1128,6 @@ export class VoiceActivityDetection extends jspb.Message {
   hasPyannote(): boolean;
   clearPyannote(): VoiceActivityDetection;
 
-  getMatchbox(): Matchbox | undefined;
-  setMatchbox(value?: Matchbox): VoiceActivityDetection;
-  hasMatchbox(): boolean;
-  clearMatchbox(): VoiceActivityDetection;
-
   serializeBinary(): Uint8Array;
   toObject(includeInstance?: boolean): VoiceActivityDetection.AsObject;
   static toObject(includeInstance: boolean, msg: VoiceActivityDetection): VoiceActivityDetection.AsObject;
@@ -1113,33 +1141,27 @@ export namespace VoiceActivityDetection {
     active: string,
     samplingRate: number,
     pyannote?: Pyannote.AsObject,
-    matchbox?: Matchbox.AsObject,
   }
 }
 
 export class Pyannote extends jspb.Message {
-  getModelPath(): string;
-  setModelPath(value: string): Pyannote;
+  getModelName(): string;
+  setModelName(value: string): Pyannote;
 
   getMinAudioSize(): number;
   setMinAudioSize(value: number): Pyannote;
 
-  getOffset(): number;
-  setOffset(value: number): Pyannote;
-
-  getOnset(): number;
-  setOnset(value: number): Pyannote;
-
-  getLogScale(): boolean;
-  setLogScale(value: boolean): Pyannote;
-
   getMinDurationOff(): number;
   setMinDurationOff(value: number): Pyannote;
 
   getMinDurationOn(): number;
   setMinDurationOn(value: number): Pyannote;
 
-  getOneofLogScaleCase(): Pyannote.OneofLogScaleCase;
+  getTritonServerHost(): string;
+  setTritonServerHost(value: string): Pyannote;
+
+  getTritonServerPort(): number;
+  setTritonServerPort(value: number): Pyannote;
 
   serializeBinary(): Uint8Array;
   toObject(includeInstance?: boolean): Pyannote.AsObject;
@@ -1151,44 +1173,12 @@ export class Pyannote extends jspb.Message {
 
 export namespace Pyannote {
   export type AsObject = {
-    modelPath: string,
+    modelName: string,
     minAudioSize: number,
-    offset: number,
-    onset: number,
-    logScale: boolean,
     minDurationOff: number,
     minDurationOn: number,
-  }
-
-  export enum OneofLogScaleCase { 
-    ONEOF_LOG_SCALE_NOT_SET = 0,
-    LOG_SCALE = 5,
-  }
-}
-
-export class Matchbox extends jspb.Message {
-  getModelConfig(): string;
-  setModelConfig(value: string): Matchbox;
-
-  getEncoderPath(): string;
-  setEncoderPath(value: string): Matchbox;
-
-  getDecoderPath(): string;
-  setDecoderPath(value: string): Matchbox;
-
-  serializeBinary(): Uint8Array;
-  toObject(includeInstance?: boolean): Matchbox.AsObject;
-  static toObject(includeInstance: boolean, msg: Matchbox): Matchbox.AsObject;
-  static serializeBinaryToWriter(message: Matchbox, writer: jspb.BinaryWriter): void;
-  static deserializeBinary(bytes: Uint8Array): Matchbox;
-  static deserializeBinaryFromReader(message: Matchbox, reader: jspb.BinaryReader): Matchbox;
-}
-
-export namespace Matchbox {
-  export type AsObject = {
-    modelConfig: string,
-    encoderPath: string,
-    decoderPath: string,
+    tritonServerHost: string,
+    tritonServerPort: number,
   }
 }
 
@@ -1456,14 +1446,14 @@ export namespace TrainUserLanguageModelRequest {
   }
 }
 
-export enum InferenceBackend { 
-  INFERENCE_BACKEND_UNKNOWN = 0,
-  INFERENCE_BACKEND_PYTORCH = 1,
-  INFERENCE_BACKEND_FLAX = 2,
-}
 export enum Decoding { 
   DEFAULT = 0,
   GREEDY = 1,
   BEAM_SEARCH_WITH_LM = 2,
   BEAM_SEARCH = 3,
 }
+export enum InferenceBackend { 
+  INFERENCE_BACKEND_UNKNOWN = 0,
+  INFERENCE_BACKEND_PYTORCH = 1,
+  INFERENCE_BACKEND_FLAX = 2,
+}