yandexcloud 0.348.0__py3-none-any.whl → 0.350.0__py3-none-any.whl

yandex/cloud/ai/stt/v3/stt_pb2.py

@@ -22,9 +22,10 @@ _runtime_version.ValidateProtobufRuntimeVersion(
 _sym_db = _symbol_database.Default()
 
 
+from google.protobuf import struct_pb2 as google_dot_protobuf_dot_struct__pb2
 
 
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n yandex/cloud/ai/stt/v3/stt.proto\x12\x10speechkit.stt.v3\"\xe2\x03\n\x18TextNormalizationOptions\x12X\n\x12text_normalization\x18\x01 \x01(\x0e\x32<.speechkit.stt.v3.TextNormalizationOptions.TextNormalization\x12\x18\n\x10profanity_filter\x18\x02 \x01(\x08\x12\x17\n\x0fliterature_text\x18\x03 \x01(\x08\x12]\n\x15phone_formatting_mode\x18\x04 \x01(\x0e\x32>.speechkit.stt.v3.TextNormalizationOptions.PhoneFormattingMode\"x\n\x11TextNormalization\x12\"\n\x1eTEXT_NORMALIZATION_UNSPECIFIED\x10\x00\x12\x1e\n\x1aTEXT_NORMALIZATION_ENABLED\x10\x01\x12\x1f\n\x1bTEXT_NORMALIZATION_DISABLED\x10\x02\"`\n\x13PhoneFormattingMode\x12%\n!PHONE_FORMATTING_MODE_UNSPECIFIED\x10\x00\x12\"\n\x1ePHONE_FORMATTING_MODE_DISABLED\x10\x01\"\xce\x01\n\x14\x44\x65\x66\x61ultEouClassifier\x12\x43\n\x04type\x18\x01 \x01(\x0e\x32\x35.speechkit.stt.v3.DefaultEouClassifier.EouSensitivity\x12\'\n\x1fmax_pause_between_words_hint_ms\x18\x02 \x01(\x03\"H\n\x0e\x45ouSensitivity\x12\x1f\n\x1b\x45OU_SENSITIVITY_UNSPECIFIED\x10\x00\x12\x0b\n\x07\x44\x45\x46\x41ULT\x10\x01\x12\x08\n\x04HIGH\x10\x02\"\x17\n\x15\x45xternalEouClassifier\"\xb2\x01\n\x14\x45ouClassifierOptions\x12\x44\n\x12\x64\x65\x66\x61ult_classifier\x18\x01 \x01(\x0b\x32&.speechkit.stt.v3.DefaultEouClassifierH\x00\x12\x46\n\x13\x65xternal_classifier\x18\x02 \x01(\x0b\x32\'.speechkit.stt.v3.ExternalEouClassifierH\x00\x42\x0c\n\nClassifier\"\xd3\x01\n\x15RecognitionClassifier\x12\x12\n\nclassifier\x18\x01 \x01(\t\x12\x45\n\x08triggers\x18\x02 \x03(\x0e\x32\x33.speechkit.stt.v3.RecognitionClassifier.TriggerType\"_\n\x0bTriggerType\x12 \n\x18TRIGGER_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\x10\n\x0cON_UTTERANCE\x10\x01\x12\x0c\n\x08ON_FINAL\x10\x02\x12\x0e\n\nON_PARTIAL\x10\x03\"\\\n\x1cRecognitionClassifierOptions\x12<\n\x0b\x63lassifiers\x18\x01 \x03(\x0b\x32\'.speechkit.stt.v3.RecognitionClassifier\"\x88\x01\n\x15SpeechAnalysisOptions\x12\x1f\n\x17\x65nable_speaker_analysis\x18\x01 \x01(\x08\x12$\n\x1c\x65nable_conversation_analysis\x18\x02 \x01(\x08\x12(\n descriptive_statistics_quantiles\x18\x03 \x03(\x01\"\xc7\x01\n\x08RawAudio\x12@\n\x0e\x61udio_encoding\x18\x01 \x01(\x0e\x32(.speechkit.stt.v3.RawAudio.AudioEncoding\x12\x19\n\x11sample_rate_hertz\x18\x02 \x01(\x03\x12\x1b\n\x13\x61udio_channel_count\x18\x03 \x01(\x03\"A\n\rAudioEncoding\x12\x1e\n\x1a\x41UDIO_ENCODING_UNSPECIFIED\x10\x00\x12\x10\n\x0cLINEAR16_PCM\x10\x01\"\xbf\x01\n\x0e\x43ontainerAudio\x12Q\n\x14\x63ontainer_audio_type\x18\x01 \x01(\x0e\x32\x33.speechkit.stt.v3.ContainerAudio.ContainerAudioType\"Z\n\x12\x43ontainerAudioType\x12$\n CONTAINER_AUDIO_TYPE_UNSPECIFIED\x10\x00\x12\x07\n\x03WAV\x10\x01\x12\x0c\n\x08OGG_OPUS\x10\x02\x12\x07\n\x03MP3\x10\x03\"\x91\x01\n\x12\x41udioFormatOptions\x12/\n\traw_audio\x18\x01 \x01(\x0b\x32\x1a.speechkit.stt.v3.RawAudioH\x00\x12;\n\x0f\x63ontainer_audio\x18\x02 \x01(\x0b\x32 .speechkit.stt.v3.ContainerAudioH\x00\x42\r\n\x0b\x41udioFormat\"\xf7\x01\n\x1aLanguageRestrictionOptions\x12^\n\x10restriction_type\x18\x01 \x01(\x0e\x32\x44.speechkit.stt.v3.LanguageRestrictionOptions.LanguageRestrictionType\x12\x15\n\rlanguage_code\x18\x02 \x03(\t\"b\n\x17LanguageRestrictionType\x12)\n%LANGUAGE_RESTRICTION_TYPE_UNSPECIFIED\x10\x00\x12\r\n\tWHITELIST\x10\x01\x12\r\n\tBLACKLIST\x10\x02\"\xb2\x03\n\x17RecognitionModelOptions\x12\r\n\x05model\x18\x01 \x01(\t\x12:\n\x0c\x61udio_format\x18\x02 \x01(\x0b\x32$.speechkit.stt.v3.AudioFormatOptions\x12\x46\n\x12text_normalization\x18\x03 \x01(\x0b\x32*.speechkit.stt.v3.TextNormalizationOptions\x12J\n\x14language_restriction\x18\x04 \x01(\x0b\x32,.speechkit.stt.v3.LanguageRestrictionOptions\x12\\\n\x15\x61udio_processing_type\x18\x05 \x01(\x0e\x32=.speechkit.stt.v3.RecognitionModelOptions.AudioProcessingType\"Z\n\x13\x41udioProcessingType\x12%\n!AUDIO_PROCESSING_TYPE_UNSPECIFIED\x10\x00\x12\r\n\tREAL_TIME\x10\x01\x12\r\n\tFULL_DATA\x10\x02\"\xde\x01\n\x16SpeakerLabelingOptions\x12R\n\x10speaker_labeling\x18\x01 \x01(\x0e\x32\x38.speechkit.stt.v3.SpeakerLabelingOptions.SpeakerLabeling\"p\n\x0fSpeakerLabeling\x12 \n\x1cSPEAKER_LABELING_UNSPECIFIED\x10\x00\x12\x1c\n\x18SPEAKER_LABELING_ENABLED\x10\x01\x12\x1d\n\x19SPEAKER_LABELING_DISABLED\x10\x02\"\xee\x02\n\x10StreamingOptions\x12\x44\n\x11recognition_model\x18\x01 \x01(\x0b\x32).speechkit.stt.v3.RecognitionModelOptions\x12>\n\x0e\x65ou_classifier\x18\x02 \x01(\x0b\x32&.speechkit.stt.v3.EouClassifierOptions\x12N\n\x16recognition_classifier\x18\x03 \x01(\x0b\x32..speechkit.stt.v3.RecognitionClassifierOptions\x12@\n\x0fspeech_analysis\x18\x04 \x01(\x0b\x32\'.speechkit.stt.v3.SpeechAnalysisOptions\x12\x42\n\x10speaker_labeling\x18\x05 \x01(\x0b\x32(.speechkit.stt.v3.SpeakerLabelingOptions\"\x1a\n\nAudioChunk\x12\x0c\n\x04\x64\x61ta\x18\x01 \x01(\x0c\"#\n\x0cSilenceChunk\x12\x13\n\x0b\x64uration_ms\x18\x01 \x01(\x03\"\x05\n\x03\x45ou\"\xe8\x01\n\x10StreamingRequest\x12=\n\x0fsession_options\x18\x01 \x01(\x0b\x32\".speechkit.stt.v3.StreamingOptionsH\x00\x12-\n\x05\x63hunk\x18\x02 \x01(\x0b\x32\x1c.speechkit.stt.v3.AudioChunkH\x00\x12\x37\n\rsilence_chunk\x18\x03 \x01(\x0b\x32\x1e.speechkit.stt.v3.SilenceChunkH\x00\x12$\n\x03\x65ou\x18\x04 \x01(\x0b\x32\x15.speechkit.stt.v3.EouH\x00\x42\x07\n\x05\x45vent\"\xe3\x02\n\x14RecognizeFileRequest\x12\x11\n\x07\x63ontent\x18\x01 \x01(\x0cH\x00\x12\r\n\x03uri\x18\x02 \x01(\tH\x00\x12\x44\n\x11recognition_model\x18\x03 \x01(\x0b\x32).speechkit.stt.v3.RecognitionModelOptions\x12N\n\x16recognition_classifier\x18\x04 \x01(\x0b\x32..speechkit.stt.v3.RecognitionClassifierOptions\x12@\n\x0fspeech_analysis\x18\x05 \x01(\x0b\x32\'.speechkit.stt.v3.SpeechAnalysisOptions\x12\x42\n\x10speaker_labeling\x18\x06 \x01(\x0b\x32(.speechkit.stt.v3.SpeakerLabelingOptionsB\r\n\x0b\x41udioSource\"@\n\x04Word\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\x15\n\rstart_time_ms\x18\x02 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x03 \x01(\x03\"@\n\x12LanguageEstimation\x12\x15\n\rlanguage_code\x18\x01 \x01(\t\x12\x13\n\x0bprobability\x18\x02 \x01(\x01\"\xbb\x01\n\x0b\x41lternative\x12%\n\x05words\x18\x01 \x03(\x0b\x32\x16.speechkit.stt.v3.Word\x12\x0c\n\x04text\x18\x02 \x01(\t\x12\x15\n\rstart_time_ms\x18\x03 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x04 \x01(\x03\x12\x12\n\nconfidence\x18\x05 \x01(\x01\x12\x37\n\tlanguages\x18\x06 \x03(\x0b\x32$.speechkit.stt.v3.LanguageEstimation\"\x1c\n\tEouUpdate\x12\x0f\n\x07time_ms\x18\x02 \x01(\x03\"a\n\x11\x41lternativeUpdate\x12\x33\n\x0c\x61lternatives\x18\x01 \x03(\x0b\x32\x1d.speechkit.stt.v3.Alternative\x12\x17\n\x0b\x63hannel_tag\x18\x02 \x01(\tB\x02\x18\x01\"\x99\x01\n\x0c\x41udioCursors\x12\x18\n\x10received_data_ms\x18\x01 \x01(\x03\x12\x15\n\rreset_time_ms\x18\x02 \x01(\x03\x12\x17\n\x0fpartial_time_ms\x18\x03 \x01(\x03\x12\x15\n\rfinal_time_ms\x18\x04 \x01(\x03\x12\x13\n\x0b\x66inal_index\x18\x05 \x01(\x03\x12\x13\n\x0b\x65ou_time_ms\x18\x06 \x01(\x03\"n\n\x0f\x46inalRefinement\x12\x13\n\x0b\x66inal_index\x18\x01 \x01(\x03\x12>\n\x0fnormalized_text\x18\x02 \x01(\x0b\x32#.speechkit.stt.v3.AlternativeUpdateH\x00\x42\x06\n\x04Type\"L\n\nStatusCode\x12-\n\tcode_type\x18\x01 \x01(\x0e\x32\x1a.speechkit.stt.v3.CodeType\x12\x0f\n\x07message\x18\x02 \x01(\t\"4\n\x0bSessionUuid\x12\x0c\n\x04uuid\x18\x01 \x01(\t\x12\x17\n\x0fuser_request_id\x18\x02 \x01(\t\"K\n\x0fPhraseHighlight\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\x15\n\rstart_time_ms\x18\x02 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x03 \x01(\x03\"?\n\x1aRecognitionClassifierLabel\x12\r\n\x05label\x18\x01 \x01(\t\x12\x12\n\nconfidence\x18\x02 \x01(\x01\"\xa6\x01\n\x1bRecognitionClassifierResult\x12\x12\n\nclassifier\x18\x01 \x01(\t\x12\x35\n\nhighlights\x18\x02 \x03(\x0b\x32!.speechkit.stt.v3.PhraseHighlight\x12<\n\x06labels\x18\x03 \x03(\x0b\x32,.speechkit.stt.v3.RecognitionClassifierLabel\"\xc7\x02\n\x1bRecognitionClassifierUpdate\x12M\n\x0bwindow_type\x18\x01 \x01(\x0e\x32\x38.speechkit.stt.v3.RecognitionClassifierUpdate.WindowType\x12\x15\n\rstart_time_ms\x18\x02 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x03 \x01(\x03\x12H\n\x11\x63lassifier_result\x18\x04 \x01(\x0b\x32-.speechkit.stt.v3.RecognitionClassifierResult\"c\n\nWindowType\x12\x1f\n\x17WINDOW_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\x12\n\x0eLAST_UTTERANCE\x10\x01\x12\x0e\n\nLAST_FINAL\x10\x02\x12\x10\n\x0cLAST_PARTIAL\x10\x03\"\xbb\x01\n\x15\x44\x65scriptiveStatistics\x12\x0b\n\x03min\x18\x01 \x01(\x01\x12\x0b\n\x03max\x18\x02 \x01(\x01\x12\x0c\n\x04mean\x18\x03 \x01(\x01\x12\x0b\n\x03std\x18\x04 \x01(\x01\x12\x43\n\tquantiles\x18\x05 \x03(\x0b\x32\x30.speechkit.stt.v3.DescriptiveStatistics.Quantile\x1a(\n\x08Quantile\x12\r\n\x05level\x18\x01 \x01(\x01\x12\r\n\x05value\x18\x02 \x01(\x01\"D\n\x16\x41udioSegmentBoundaries\x12\x15\n\rstart_time_ms\x18\x01 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x02 \x01(\x03\"\x87\x06\n\x0fSpeakerAnalysis\x12\x13\n\x0bspeaker_tag\x18\x01 \x01(\t\x12\x41\n\x0bwindow_type\x18\x02 \x01(\x0e\x32,.speechkit.stt.v3.SpeakerAnalysis.WindowType\x12\x43\n\x11speech_boundaries\x18\x03 \x01(\x0b\x32(.speechkit.stt.v3.AudioSegmentBoundaries\x12\x17\n\x0ftotal_speech_ms\x18\x04 \x01(\x03\x12\x14\n\x0cspeech_ratio\x18\x05 \x01(\x01\x12\x18\n\x10total_silence_ms\x18\x06 \x01(\x03\x12\x15\n\rsilence_ratio\x18\x07 \x01(\x01\x12\x13\n\x0bwords_count\x18\x08 \x01(\x03\x12\x15\n\rletters_count\x18\t \x01(\x03\x12\x41\n\x10words_per_second\x18\n \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x43\n\x12letters_per_second\x18\x0b \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x44\n\x13words_per_utterance\x18\x0c \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x46\n\x15letters_per_utterance\x18\r \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x17\n\x0futterance_count\x18\x0e \x01(\x03\x12N\n\x1dutterance_duration_estimation\x18\x0f \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\"L\n\nWindowType\x12\x1f\n\x17WINDOW_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\t\n\x05TOTAL\x10\x01\x12\x12\n\x0eLAST_UTTERANCE\x10\x02\"\x85\x06\n\x14\x43onversationAnalysis\x12I\n\x17\x63onversation_boundaries\x18\x01 \x01(\x0b\x32(.speechkit.stt.v3.AudioSegmentBoundaries\x12.\n&total_simultaneous_silence_duration_ms\x18\x02 \x01(\x03\x12(\n total_simultaneous_silence_ratio\x18\x03 \x01(\x01\x12Y\n(simultaneous_silence_duration_estimation\x18\x04 \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12-\n%total_simultaneous_speech_duration_ms\x18\x05 \x01(\x03\x12\'\n\x1ftotal_simultaneous_speech_ratio\x18\x06 \x01(\x01\x12X\n\'simultaneous_speech_duration_estimation\x18\x07 \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12W\n\x12speaker_interrupts\x18\x08 \x03(\x0b\x32;.speechkit.stt.v3.ConversationAnalysis.InterruptsEvaluation\x12 \n\x18total_speech_duration_ms\x18\t \x01(\x03\x12\x1a\n\x12total_speech_ratio\x18\n \x01(\x01\x1a\xa3\x01\n\x14InterruptsEvaluation\x12\x13\n\x0bspeaker_tag\x18\x01 \x01(\t\x12\x18\n\x10interrupts_count\x18\x02 \x01(\x03\x12\x1e\n\x16interrupts_duration_ms\x18\x03 \x01(\x03\x12<\n\ninterrupts\x18\x04 \x03(\x0b\x32(.speechkit.stt.v3.AudioSegmentBoundaries\"\xa5\x05\n\x11StreamingResponse\x12\x33\n\x0csession_uuid\x18\x01 \x01(\x0b\x32\x1d.speechkit.stt.v3.SessionUuid\x12\x35\n\raudio_cursors\x18\x02 \x01(\x0b\x32\x1e.speechkit.stt.v3.AudioCursors\x12\x1d\n\x15response_wall_time_ms\x18\x03 \x01(\x03\x12\x36\n\x07partial\x18\x04 \x01(\x0b\x32#.speechkit.stt.v3.AlternativeUpdateH\x00\x12\x34\n\x05\x66inal\x18\x05 \x01(\x0b\x32#.speechkit.stt.v3.AlternativeUpdateH\x00\x12\x31\n\neou_update\x18\x06 \x01(\x0b\x32\x1b.speechkit.stt.v3.EouUpdateH\x00\x12=\n\x10\x66inal_refinement\x18\x07 \x01(\x0b\x32!.speechkit.stt.v3.FinalRefinementH\x00\x12\x33\n\x0bstatus_code\x18\x08 \x01(\x0b\x32\x1c.speechkit.stt.v3.StatusCodeH\x00\x12J\n\x11\x63lassifier_update\x18\n \x01(\x0b\x32-.speechkit.stt.v3.RecognitionClassifierUpdateH\x00\x12=\n\x10speaker_analysis\x18\x0b \x01(\x0b\x32!.speechkit.stt.v3.SpeakerAnalysisH\x00\x12G\n\x15\x63onversation_analysis\x18\x0c \x01(\x0b\x32&.speechkit.stt.v3.ConversationAnalysisH\x00\x12\x13\n\x0b\x63hannel_tag\x18\t \x01(\tB\x07\n\x05\x45vent\"0\n\x18\x44\x65leteRecognitionRequest\x12\x14\n\x0coperation_id\x18\x01 \x01(\t\"Y\n\x15StreamingResponseList\x12@\n\x13streaming_responses\x18\x01 \x03(\x0b\x32#.speechkit.stt.v3.StreamingResponse*O\n\x08\x43odeType\x12\x1d\n\x15\x43ODE_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\x0b\n\x07WORKING\x10\x01\x12\x0b\n\x07WARNING\x10\x02\x12\n\n\x06\x43LOSED\x10\x03\x42\\\n\x1ayandex.cloud.api.ai.stt.v3Z>github.com/yandex-cloud/go-genproto/yandex/cloud/ai/stt/v3;sttb\x06proto3')
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n yandex/cloud/ai/stt/v3/stt.proto\x12\x10speechkit.stt.v3\x1a\x1cgoogle/protobuf/struct.proto\"\xe2\x03\n\x18TextNormalizationOptions\x12X\n\x12text_normalization\x18\x01 \x01(\x0e\x32<.speechkit.stt.v3.TextNormalizationOptions.TextNormalization\x12\x18\n\x10profanity_filter\x18\x02 \x01(\x08\x12\x17\n\x0fliterature_text\x18\x03 \x01(\x08\x12]\n\x15phone_formatting_mode\x18\x04 \x01(\x0e\x32>.speechkit.stt.v3.TextNormalizationOptions.PhoneFormattingMode\"x\n\x11TextNormalization\x12\"\n\x1eTEXT_NORMALIZATION_UNSPECIFIED\x10\x00\x12\x1e\n\x1aTEXT_NORMALIZATION_ENABLED\x10\x01\x12\x1f\n\x1bTEXT_NORMALIZATION_DISABLED\x10\x02\"`\n\x13PhoneFormattingMode\x12%\n!PHONE_FORMATTING_MODE_UNSPECIFIED\x10\x00\x12\"\n\x1ePHONE_FORMATTING_MODE_DISABLED\x10\x01\"\xce\x01\n\x14\x44\x65\x66\x61ultEouClassifier\x12\x43\n\x04type\x18\x01 \x01(\x0e\x32\x35.speechkit.stt.v3.DefaultEouClassifier.EouSensitivity\x12\'\n\x1fmax_pause_between_words_hint_ms\x18\x02 \x01(\x03\"H\n\x0e\x45ouSensitivity\x12\x1f\n\x1b\x45OU_SENSITIVITY_UNSPECIFIED\x10\x00\x12\x0b\n\x07\x44\x45\x46\x41ULT\x10\x01\x12\x08\n\x04HIGH\x10\x02\"\x17\n\x15\x45xternalEouClassifier\"\xb2\x01\n\x14\x45ouClassifierOptions\x12\x44\n\x12\x64\x65\x66\x61ult_classifier\x18\x01 \x01(\x0b\x32&.speechkit.stt.v3.DefaultEouClassifierH\x00\x12\x46\n\x13\x65xternal_classifier\x18\x02 \x01(\x0b\x32\'.speechkit.stt.v3.ExternalEouClassifierH\x00\x42\x0c\n\nClassifier\"\xd3\x01\n\x15RecognitionClassifier\x12\x12\n\nclassifier\x18\x01 \x01(\t\x12\x45\n\x08triggers\x18\x02 \x03(\x0e\x32\x33.speechkit.stt.v3.RecognitionClassifier.TriggerType\"_\n\x0bTriggerType\x12 \n\x18TRIGGER_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\x10\n\x0cON_UTTERANCE\x10\x01\x12\x0c\n\x08ON_FINAL\x10\x02\x12\x0e\n\nON_PARTIAL\x10\x03\"\\\n\x1cRecognitionClassifierOptions\x12<\n\x0b\x63lassifiers\x18\x01 \x03(\x0b\x32\'.speechkit.stt.v3.RecognitionClassifier\"\x88\x01\n\x15SpeechAnalysisOptions\x12\x1f\n\x17\x65nable_speaker_analysis\x18\x01 \x01(\x08\x12$\n\x1c\x65nable_conversation_analysis\x18\x02 \x01(\x08\x12(\n descriptive_statistics_quantiles\x18\x03 \x03(\x01\"\xc7\x01\n\x08RawAudio\x12@\n\x0e\x61udio_encoding\x18\x01 \x01(\x0e\x32(.speechkit.stt.v3.RawAudio.AudioEncoding\x12\x19\n\x11sample_rate_hertz\x18\x02 \x01(\x03\x12\x1b\n\x13\x61udio_channel_count\x18\x03 \x01(\x03\"A\n\rAudioEncoding\x12\x1e\n\x1a\x41UDIO_ENCODING_UNSPECIFIED\x10\x00\x12\x10\n\x0cLINEAR16_PCM\x10\x01\"\xbf\x01\n\x0e\x43ontainerAudio\x12Q\n\x14\x63ontainer_audio_type\x18\x01 \x01(\x0e\x32\x33.speechkit.stt.v3.ContainerAudio.ContainerAudioType\"Z\n\x12\x43ontainerAudioType\x12$\n CONTAINER_AUDIO_TYPE_UNSPECIFIED\x10\x00\x12\x07\n\x03WAV\x10\x01\x12\x0c\n\x08OGG_OPUS\x10\x02\x12\x07\n\x03MP3\x10\x03\"\x91\x01\n\x12\x41udioFormatOptions\x12/\n\traw_audio\x18\x01 \x01(\x0b\x32\x1a.speechkit.stt.v3.RawAudioH\x00\x12;\n\x0f\x63ontainer_audio\x18\x02 \x01(\x0b\x32 .speechkit.stt.v3.ContainerAudioH\x00\x42\r\n\x0b\x41udioFormat\"\xf7\x01\n\x1aLanguageRestrictionOptions\x12^\n\x10restriction_type\x18\x01 \x01(\x0e\x32\x44.speechkit.stt.v3.LanguageRestrictionOptions.LanguageRestrictionType\x12\x15\n\rlanguage_code\x18\x02 \x03(\t\"b\n\x17LanguageRestrictionType\x12)\n%LANGUAGE_RESTRICTION_TYPE_UNSPECIFIED\x10\x00\x12\r\n\tWHITELIST\x10\x01\x12\r\n\tBLACKLIST\x10\x02\"5\n\nJsonSchema\x12\'\n\x06schema\x18\x01 \x01(\x0b\x32\x17.google.protobuf.Struct\"\x8a\x01\n\x15SummarizationProperty\x12\x13\n\x0binstruction\x18\x01 \x01(\t\x12\x15\n\x0bjson_object\x18\x02 \x01(\x08H\x00\x12\x33\n\x0bjson_schema\x18\x03 \x01(\x0b\x32\x1c.speechkit.stt.v3.JsonSchemaH\x00\x42\x10\n\x0eResponseFormat\"f\n\x14SummarizationOptions\x12\x11\n\tmodel_uri\x18\x01 \x01(\t\x12;\n\nproperties\x18\x02 \x03(\x0b\x32\'.speechkit.stt.v3.SummarizationProperty\"/\n\x1bSummarizationPropertyResult\x12\x10\n\x08response\x18\x01 \x01(\t\"\xb2\x03\n\x17RecognitionModelOptions\x12\r\n\x05model\x18\x01 \x01(\t\x12:\n\x0c\x61udio_format\x18\x02 \x01(\x0b\x32$.speechkit.stt.v3.AudioFormatOptions\x12\x46\n\x12text_normalization\x18\x03 \x01(\x0b\x32*.speechkit.stt.v3.TextNormalizationOptions\x12J\n\x14language_restriction\x18\x04 \x01(\x0b\x32,.speechkit.stt.v3.LanguageRestrictionOptions\x12\\\n\x15\x61udio_processing_type\x18\x05 \x01(\x0e\x32=.speechkit.stt.v3.RecognitionModelOptions.AudioProcessingType\"Z\n\x13\x41udioProcessingType\x12%\n!AUDIO_PROCESSING_TYPE_UNSPECIFIED\x10\x00\x12\r\n\tREAL_TIME\x10\x01\x12\r\n\tFULL_DATA\x10\x02\"\xde\x01\n\x16SpeakerLabelingOptions\x12R\n\x10speaker_labeling\x18\x01 \x01(\x0e\x32\x38.speechkit.stt.v3.SpeakerLabelingOptions.SpeakerLabeling\"p\n\x0fSpeakerLabeling\x12 \n\x1cSPEAKER_LABELING_UNSPECIFIED\x10\x00\x12\x1c\n\x18SPEAKER_LABELING_ENABLED\x10\x01\x12\x1d\n\x19SPEAKER_LABELING_DISABLED\x10\x02\"\xad\x03\n\x10StreamingOptions\x12\x44\n\x11recognition_model\x18\x01 \x01(\x0b\x32).speechkit.stt.v3.RecognitionModelOptions\x12>\n\x0e\x65ou_classifier\x18\x02 \x01(\x0b\x32&.speechkit.stt.v3.EouClassifierOptions\x12N\n\x16recognition_classifier\x18\x03 \x01(\x0b\x32..speechkit.stt.v3.RecognitionClassifierOptions\x12@\n\x0fspeech_analysis\x18\x04 \x01(\x0b\x32\'.speechkit.stt.v3.SpeechAnalysisOptions\x12\x42\n\x10speaker_labeling\x18\x05 \x01(\x0b\x32(.speechkit.stt.v3.SpeakerLabelingOptions\x12=\n\rsummarization\x18\x06 \x01(\x0b\x32&.speechkit.stt.v3.SummarizationOptions\"\x1a\n\nAudioChunk\x12\x0c\n\x04\x64\x61ta\x18\x01 \x01(\x0c\"#\n\x0cSilenceChunk\x12\x13\n\x0b\x64uration_ms\x18\x01 \x01(\x03\"\x05\n\x03\x45ou\"\xe8\x01\n\x10StreamingRequest\x12=\n\x0fsession_options\x18\x01 \x01(\x0b\x32\".speechkit.stt.v3.StreamingOptionsH\x00\x12-\n\x05\x63hunk\x18\x02 \x01(\x0b\x32\x1c.speechkit.stt.v3.AudioChunkH\x00\x12\x37\n\rsilence_chunk\x18\x03 \x01(\x0b\x32\x1e.speechkit.stt.v3.SilenceChunkH\x00\x12$\n\x03\x65ou\x18\x04 \x01(\x0b\x32\x15.speechkit.stt.v3.EouH\x00\x42\x07\n\x05\x45vent\"\xa2\x03\n\x14RecognizeFileRequest\x12\x11\n\x07\x63ontent\x18\x01 \x01(\x0cH\x00\x12\r\n\x03uri\x18\x02 \x01(\tH\x00\x12\x44\n\x11recognition_model\x18\x03 \x01(\x0b\x32).speechkit.stt.v3.RecognitionModelOptions\x12N\n\x16recognition_classifier\x18\x04 \x01(\x0b\x32..speechkit.stt.v3.RecognitionClassifierOptions\x12@\n\x0fspeech_analysis\x18\x05 \x01(\x0b\x32\'.speechkit.stt.v3.SpeechAnalysisOptions\x12\x42\n\x10speaker_labeling\x18\x06 \x01(\x0b\x32(.speechkit.stt.v3.SpeakerLabelingOptions\x12=\n\rsummarization\x18\x07 \x01(\x0b\x32&.speechkit.stt.v3.SummarizationOptionsB\r\n\x0b\x41udioSource\"@\n\x04Word\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\x15\n\rstart_time_ms\x18\x02 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x03 \x01(\x03\"@\n\x12LanguageEstimation\x12\x15\n\rlanguage_code\x18\x01 \x01(\t\x12\x13\n\x0bprobability\x18\x02 \x01(\x01\"\xbb\x01\n\x0b\x41lternative\x12%\n\x05words\x18\x01 \x03(\x0b\x32\x16.speechkit.stt.v3.Word\x12\x0c\n\x04text\x18\x02 \x01(\t\x12\x15\n\rstart_time_ms\x18\x03 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x04 \x01(\x03\x12\x12\n\nconfidence\x18\x05 \x01(\x01\x12\x37\n\tlanguages\x18\x06 \x03(\x0b\x32$.speechkit.stt.v3.LanguageEstimation\"\x1c\n\tEouUpdate\x12\x0f\n\x07time_ms\x18\x02 \x01(\x03\"a\n\x11\x41lternativeUpdate\x12\x33\n\x0c\x61lternatives\x18\x01 \x03(\x0b\x32\x1d.speechkit.stt.v3.Alternative\x12\x17\n\x0b\x63hannel_tag\x18\x02 \x01(\tB\x02\x18\x01\"\x99\x01\n\x0c\x41udioCursors\x12\x18\n\x10received_data_ms\x18\x01 \x01(\x03\x12\x15\n\rreset_time_ms\x18\x02 \x01(\x03\x12\x17\n\x0fpartial_time_ms\x18\x03 \x01(\x03\x12\x15\n\rfinal_time_ms\x18\x04 \x01(\x03\x12\x13\n\x0b\x66inal_index\x18\x05 \x01(\x03\x12\x13\n\x0b\x65ou_time_ms\x18\x06 \x01(\x03\"n\n\x0f\x46inalRefinement\x12\x13\n\x0b\x66inal_index\x18\x01 \x01(\x03\x12>\n\x0fnormalized_text\x18\x02 \x01(\x0b\x32#.speechkit.stt.v3.AlternativeUpdateH\x00\x42\x06\n\x04Type\"L\n\nStatusCode\x12-\n\tcode_type\x18\x01 \x01(\x0e\x32\x1a.speechkit.stt.v3.CodeType\x12\x0f\n\x07message\x18\x02 \x01(\t\"4\n\x0bSessionUuid\x12\x0c\n\x04uuid\x18\x01 \x01(\t\x12\x17\n\x0fuser_request_id\x18\x02 \x01(\t\"K\n\x0fPhraseHighlight\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\x15\n\rstart_time_ms\x18\x02 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x03 \x01(\x03\"?\n\x1aRecognitionClassifierLabel\x12\r\n\x05label\x18\x01 \x01(\t\x12\x12\n\nconfidence\x18\x02 \x01(\x01\"\xa6\x01\n\x1bRecognitionClassifierResult\x12\x12\n\nclassifier\x18\x01 \x01(\t\x12\x35\n\nhighlights\x18\x02 \x03(\x0b\x32!.speechkit.stt.v3.PhraseHighlight\x12<\n\x06labels\x18\x03 \x03(\x0b\x32,.speechkit.stt.v3.RecognitionClassifierLabel\"\xc7\x02\n\x1bRecognitionClassifierUpdate\x12M\n\x0bwindow_type\x18\x01 \x01(\x0e\x32\x38.speechkit.stt.v3.RecognitionClassifierUpdate.WindowType\x12\x15\n\rstart_time_ms\x18\x02 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x03 \x01(\x03\x12H\n\x11\x63lassifier_result\x18\x04 \x01(\x0b\x32-.speechkit.stt.v3.RecognitionClassifierResult\"c\n\nWindowType\x12\x1f\n\x17WINDOW_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\x12\n\x0eLAST_UTTERANCE\x10\x01\x12\x0e\n\nLAST_FINAL\x10\x02\x12\x10\n\x0cLAST_PARTIAL\x10\x03\"\xbb\x01\n\x15\x44\x65scriptiveStatistics\x12\x0b\n\x03min\x18\x01 \x01(\x01\x12\x0b\n\x03max\x18\x02 \x01(\x01\x12\x0c\n\x04mean\x18\x03 \x01(\x01\x12\x0b\n\x03std\x18\x04 \x01(\x01\x12\x43\n\tquantiles\x18\x05 \x03(\x0b\x32\x30.speechkit.stt.v3.DescriptiveStatistics.Quantile\x1a(\n\x08Quantile\x12\r\n\x05level\x18\x01 \x01(\x01\x12\r\n\x05value\x18\x02 \x01(\x01\"D\n\x16\x41udioSegmentBoundaries\x12\x15\n\rstart_time_ms\x18\x01 \x01(\x03\x12\x13\n\x0b\x65nd_time_ms\x18\x02 \x01(\x03\"\x87\x06\n\x0fSpeakerAnalysis\x12\x13\n\x0bspeaker_tag\x18\x01 \x01(\t\x12\x41\n\x0bwindow_type\x18\x02 \x01(\x0e\x32,.speechkit.stt.v3.SpeakerAnalysis.WindowType\x12\x43\n\x11speech_boundaries\x18\x03 \x01(\x0b\x32(.speechkit.stt.v3.AudioSegmentBoundaries\x12\x17\n\x0ftotal_speech_ms\x18\x04 \x01(\x03\x12\x14\n\x0cspeech_ratio\x18\x05 \x01(\x01\x12\x18\n\x10total_silence_ms\x18\x06 \x01(\x03\x12\x15\n\rsilence_ratio\x18\x07 \x01(\x01\x12\x13\n\x0bwords_count\x18\x08 \x01(\x03\x12\x15\n\rletters_count\x18\t \x01(\x03\x12\x41\n\x10words_per_second\x18\n \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x43\n\x12letters_per_second\x18\x0b \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x44\n\x13words_per_utterance\x18\x0c \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x46\n\x15letters_per_utterance\x18\r \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12\x17\n\x0futterance_count\x18\x0e \x01(\x03\x12N\n\x1dutterance_duration_estimation\x18\x0f \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\"L\n\nWindowType\x12\x1f\n\x17WINDOW_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\t\n\x05TOTAL\x10\x01\x12\x12\n\x0eLAST_UTTERANCE\x10\x02\"\x85\x06\n\x14\x43onversationAnalysis\x12I\n\x17\x63onversation_boundaries\x18\x01 \x01(\x0b\x32(.speechkit.stt.v3.AudioSegmentBoundaries\x12.\n&total_simultaneous_silence_duration_ms\x18\x02 \x01(\x03\x12(\n total_simultaneous_silence_ratio\x18\x03 \x01(\x01\x12Y\n(simultaneous_silence_duration_estimation\x18\x04 \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12-\n%total_simultaneous_speech_duration_ms\x18\x05 \x01(\x03\x12\'\n\x1ftotal_simultaneous_speech_ratio\x18\x06 \x01(\x01\x12X\n\'simultaneous_speech_duration_estimation\x18\x07 \x01(\x0b\x32\'.speechkit.stt.v3.DescriptiveStatistics\x12W\n\x12speaker_interrupts\x18\x08 \x03(\x0b\x32;.speechkit.stt.v3.ConversationAnalysis.InterruptsEvaluation\x12 \n\x18total_speech_duration_ms\x18\t \x01(\x03\x12\x1a\n\x12total_speech_ratio\x18\n \x01(\x01\x1a\xa3\x01\n\x14InterruptsEvaluation\x12\x13\n\x0bspeaker_tag\x18\x01 \x01(\t\x12\x18\n\x10interrupts_count\x18\x02 \x01(\x03\x12\x1e\n\x16interrupts_duration_ms\x18\x03 \x01(\x03\x12<\n\ninterrupts\x18\x04 \x03(\x0b\x32(.speechkit.stt.v3.AudioSegmentBoundaries\"Z\n\x0c\x43ontentUsage\x12\x19\n\x11input_text_tokens\x18\x01 \x01(\x03\x12\x19\n\x11\x63ompletion_tokens\x18\x02 \x01(\x03\x12\x14\n\x0ctotal_tokens\x18\x03 \x01(\x03\"\x86\x01\n\rSummarization\x12>\n\x07results\x18\x01 \x03(\x0b\x32-.speechkit.stt.v3.SummarizationPropertyResult\x12\x35\n\rcontent_usage\x18\x02 \x01(\x0b\x32\x1e.speechkit.stt.v3.ContentUsage\"\xdf\x05\n\x11StreamingResponse\x12\x33\n\x0csession_uuid\x18\x01 \x01(\x0b\x32\x1d.speechkit.stt.v3.SessionUuid\x12\x35\n\raudio_cursors\x18\x02 \x01(\x0b\x32\x1e.speechkit.stt.v3.AudioCursors\x12\x1d\n\x15response_wall_time_ms\x18\x03 \x01(\x03\x12\x36\n\x07partial\x18\x04 \x01(\x0b\x32#.speechkit.stt.v3.AlternativeUpdateH\x00\x12\x34\n\x05\x66inal\x18\x05 \x01(\x0b\x32#.speechkit.stt.v3.AlternativeUpdateH\x00\x12\x31\n\neou_update\x18\x06 \x01(\x0b\x32\x1b.speechkit.stt.v3.EouUpdateH\x00\x12=\n\x10\x66inal_refinement\x18\x07 \x01(\x0b\x32!.speechkit.stt.v3.FinalRefinementH\x00\x12\x33\n\x0bstatus_code\x18\x08 \x01(\x0b\x32\x1c.speechkit.stt.v3.StatusCodeH\x00\x12J\n\x11\x63lassifier_update\x18\n \x01(\x0b\x32-.speechkit.stt.v3.RecognitionClassifierUpdateH\x00\x12=\n\x10speaker_analysis\x18\x0b \x01(\x0b\x32!.speechkit.stt.v3.SpeakerAnalysisH\x00\x12G\n\x15\x63onversation_analysis\x18\x0c \x01(\x0b\x32&.speechkit.stt.v3.ConversationAnalysisH\x00\x12\x38\n\rsummarization\x18\r \x01(\x0b\x32\x1f.speechkit.stt.v3.SummarizationH\x00\x12\x13\n\x0b\x63hannel_tag\x18\t \x01(\tB\x07\n\x05\x45vent\"0\n\x18\x44\x65leteRecognitionRequest\x12\x14\n\x0coperation_id\x18\x01 \x01(\t\"Y\n\x15StreamingResponseList\x12@\n\x13streaming_responses\x18\x01 \x03(\x0b\x32#.speechkit.stt.v3.StreamingResponse*O\n\x08\x43odeType\x12\x1d\n\x15\x43ODE_TYPE_UNSPECIFIED\x10\x00\x1a\x02\x08\x01\x12\x0b\n\x07WORKING\x10\x01\x12\x0b\n\x07WARNING\x10\x02\x12\n\n\x06\x43LOSED\x10\x03\x42\\\n\x1ayandex.cloud.api.ai.stt.v3Z>github.com/yandex-cloud/go-genproto/yandex/cloud/ai/stt/v3;sttb\x06proto3')
 
 _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
@@ -42,110 +43,122 @@ if not _descriptor._USE_C_DESCRIPTORS:
   _globals['_RECOGNITIONCLASSIFIERUPDATE_WINDOWTYPE'].values_by_name["WINDOW_TYPE_UNSPECIFIED"]._serialized_options = b'\010\001'
   _globals['_SPEAKERANALYSIS_WINDOWTYPE'].values_by_name["WINDOW_TYPE_UNSPECIFIED"]._loaded_options = None
   _globals['_SPEAKERANALYSIS_WINDOWTYPE'].values_by_name["WINDOW_TYPE_UNSPECIFIED"]._serialized_options = b'\010\001'
-  _globals['_CODETYPE']._serialized_start=8018
-  _globals['_CODETYPE']._serialized_end=8097
-  _globals['_TEXTNORMALIZATIONOPTIONS']._serialized_start=55
-  _globals['_TEXTNORMALIZATIONOPTIONS']._serialized_end=537
-  _globals['_TEXTNORMALIZATIONOPTIONS_TEXTNORMALIZATION']._serialized_start=319
-  _globals['_TEXTNORMALIZATIONOPTIONS_TEXTNORMALIZATION']._serialized_end=439
-  _globals['_TEXTNORMALIZATIONOPTIONS_PHONEFORMATTINGMODE']._serialized_start=441
-  _globals['_TEXTNORMALIZATIONOPTIONS_PHONEFORMATTINGMODE']._serialized_end=537
-  _globals['_DEFAULTEOUCLASSIFIER']._serialized_start=540
-  _globals['_DEFAULTEOUCLASSIFIER']._serialized_end=746
-  _globals['_DEFAULTEOUCLASSIFIER_EOUSENSITIVITY']._serialized_start=674
-  _globals['_DEFAULTEOUCLASSIFIER_EOUSENSITIVITY']._serialized_end=746
-  _globals['_EXTERNALEOUCLASSIFIER']._serialized_start=748
-  _globals['_EXTERNALEOUCLASSIFIER']._serialized_end=771
-  _globals['_EOUCLASSIFIEROPTIONS']._serialized_start=774
-  _globals['_EOUCLASSIFIEROPTIONS']._serialized_end=952
-  _globals['_RECOGNITIONCLASSIFIER']._serialized_start=955
-  _globals['_RECOGNITIONCLASSIFIER']._serialized_end=1166
-  _globals['_RECOGNITIONCLASSIFIER_TRIGGERTYPE']._serialized_start=1071
-  _globals['_RECOGNITIONCLASSIFIER_TRIGGERTYPE']._serialized_end=1166
-  _globals['_RECOGNITIONCLASSIFIEROPTIONS']._serialized_start=1168
-  _globals['_RECOGNITIONCLASSIFIEROPTIONS']._serialized_end=1260
-  _globals['_SPEECHANALYSISOPTIONS']._serialized_start=1263
-  _globals['_SPEECHANALYSISOPTIONS']._serialized_end=1399
-  _globals['_RAWAUDIO']._serialized_start=1402
-  _globals['_RAWAUDIO']._serialized_end=1601
-  _globals['_RAWAUDIO_AUDIOENCODING']._serialized_start=1536
-  _globals['_RAWAUDIO_AUDIOENCODING']._serialized_end=1601
-  _globals['_CONTAINERAUDIO']._serialized_start=1604
-  _globals['_CONTAINERAUDIO']._serialized_end=1795
-  _globals['_CONTAINERAUDIO_CONTAINERAUDIOTYPE']._serialized_start=1705
-  _globals['_CONTAINERAUDIO_CONTAINERAUDIOTYPE']._serialized_end=1795
-  _globals['_AUDIOFORMATOPTIONS']._serialized_start=1798
-  _globals['_AUDIOFORMATOPTIONS']._serialized_end=1943
-  _globals['_LANGUAGERESTRICTIONOPTIONS']._serialized_start=1946
-  _globals['_LANGUAGERESTRICTIONOPTIONS']._serialized_end=2193
-  _globals['_LANGUAGERESTRICTIONOPTIONS_LANGUAGERESTRICTIONTYPE']._serialized_start=2095
-  _globals['_LANGUAGERESTRICTIONOPTIONS_LANGUAGERESTRICTIONTYPE']._serialized_end=2193
-  _globals['_RECOGNITIONMODELOPTIONS']._serialized_start=2196
-  _globals['_RECOGNITIONMODELOPTIONS']._serialized_end=2630
-  _globals['_RECOGNITIONMODELOPTIONS_AUDIOPROCESSINGTYPE']._serialized_start=2540
-  _globals['_RECOGNITIONMODELOPTIONS_AUDIOPROCESSINGTYPE']._serialized_end=2630
-  _globals['_SPEAKERLABELINGOPTIONS']._serialized_start=2633
-  _globals['_SPEAKERLABELINGOPTIONS']._serialized_end=2855
-  _globals['_SPEAKERLABELINGOPTIONS_SPEAKERLABELING']._serialized_start=2743
-  _globals['_SPEAKERLABELINGOPTIONS_SPEAKERLABELING']._serialized_end=2855
-  _globals['_STREAMINGOPTIONS']._serialized_start=2858
-  _globals['_STREAMINGOPTIONS']._serialized_end=3224
-  _globals['_AUDIOCHUNK']._serialized_start=3226
-  _globals['_AUDIOCHUNK']._serialized_end=3252
-  _globals['_SILENCECHUNK']._serialized_start=3254
-  _globals['_SILENCECHUNK']._serialized_end=3289
-  _globals['_EOU']._serialized_start=3291
-  _globals['_EOU']._serialized_end=3296
-  _globals['_STREAMINGREQUEST']._serialized_start=3299
-  _globals['_STREAMINGREQUEST']._serialized_end=3531
-  _globals['_RECOGNIZEFILEREQUEST']._serialized_start=3534
-  _globals['_RECOGNIZEFILEREQUEST']._serialized_end=3889
-  _globals['_WORD']._serialized_start=3891
-  _globals['_WORD']._serialized_end=3955
-  _globals['_LANGUAGEESTIMATION']._serialized_start=3957
-  _globals['_LANGUAGEESTIMATION']._serialized_end=4021
-  _globals['_ALTERNATIVE']._serialized_start=4024
-  _globals['_ALTERNATIVE']._serialized_end=4211
-  _globals['_EOUUPDATE']._serialized_start=4213
-  _globals['_EOUUPDATE']._serialized_end=4241
-  _globals['_ALTERNATIVEUPDATE']._serialized_start=4243
-  _globals['_ALTERNATIVEUPDATE']._serialized_end=4340
-  _globals['_AUDIOCURSORS']._serialized_start=4343
-  _globals['_AUDIOCURSORS']._serialized_end=4496
-  _globals['_FINALREFINEMENT']._serialized_start=4498
-  _globals['_FINALREFINEMENT']._serialized_end=4608
-  _globals['_STATUSCODE']._serialized_start=4610
-  _globals['_STATUSCODE']._serialized_end=4686
-  _globals['_SESSIONUUID']._serialized_start=4688
-  _globals['_SESSIONUUID']._serialized_end=4740
-  _globals['_PHRASEHIGHLIGHT']._serialized_start=4742
-  _globals['_PHRASEHIGHLIGHT']._serialized_end=4817
-  _globals['_RECOGNITIONCLASSIFIERLABEL']._serialized_start=4819
-  _globals['_RECOGNITIONCLASSIFIERLABEL']._serialized_end=4882
-  _globals['_RECOGNITIONCLASSIFIERRESULT']._serialized_start=4885
-  _globals['_RECOGNITIONCLASSIFIERRESULT']._serialized_end=5051
-  _globals['_RECOGNITIONCLASSIFIERUPDATE']._serialized_start=5054
-  _globals['_RECOGNITIONCLASSIFIERUPDATE']._serialized_end=5381
-  _globals['_RECOGNITIONCLASSIFIERUPDATE_WINDOWTYPE']._serialized_start=5282
-  _globals['_RECOGNITIONCLASSIFIERUPDATE_WINDOWTYPE']._serialized_end=5381
-  _globals['_DESCRIPTIVESTATISTICS']._serialized_start=5384
-  _globals['_DESCRIPTIVESTATISTICS']._serialized_end=5571
-  _globals['_DESCRIPTIVESTATISTICS_QUANTILE']._serialized_start=5531
-  _globals['_DESCRIPTIVESTATISTICS_QUANTILE']._serialized_end=5571
-  _globals['_AUDIOSEGMENTBOUNDARIES']._serialized_start=5573
-  _globals['_AUDIOSEGMENTBOUNDARIES']._serialized_end=5641
-  _globals['_SPEAKERANALYSIS']._serialized_start=5644
-  _globals['_SPEAKERANALYSIS']._serialized_end=6419
-  _globals['_SPEAKERANALYSIS_WINDOWTYPE']._serialized_start=6343
-  _globals['_SPEAKERANALYSIS_WINDOWTYPE']._serialized_end=6419
-  _globals['_CONVERSATIONANALYSIS']._serialized_start=6422
-  _globals['_CONVERSATIONANALYSIS']._serialized_end=7195
-  _globals['_CONVERSATIONANALYSIS_INTERRUPTSEVALUATION']._serialized_start=7032
-  _globals['_CONVERSATIONANALYSIS_INTERRUPTSEVALUATION']._serialized_end=7195
-  _globals['_STREAMINGRESPONSE']._serialized_start=7198
-  _globals['_STREAMINGRESPONSE']._serialized_end=7875
-  _globals['_DELETERECOGNITIONREQUEST']._serialized_start=7877
-  _globals['_DELETERECOGNITIONREQUEST']._serialized_end=7925
-  _globals['_STREAMINGRESPONSELIST']._serialized_start=7927
-  _globals['_STREAMINGRESPONSELIST']._serialized_end=8016
+  _globals['_CODETYPE']._serialized_start=8810
+  _globals['_CODETYPE']._serialized_end=8889
+  _globals['_TEXTNORMALIZATIONOPTIONS']._serialized_start=85
+  _globals['_TEXTNORMALIZATIONOPTIONS']._serialized_end=567
+  _globals['_TEXTNORMALIZATIONOPTIONS_TEXTNORMALIZATION']._serialized_start=349
+  _globals['_TEXTNORMALIZATIONOPTIONS_TEXTNORMALIZATION']._serialized_end=469
+  _globals['_TEXTNORMALIZATIONOPTIONS_PHONEFORMATTINGMODE']._serialized_start=471
+  _globals['_TEXTNORMALIZATIONOPTIONS_PHONEFORMATTINGMODE']._serialized_end=567
+  _globals['_DEFAULTEOUCLASSIFIER']._serialized_start=570
+  _globals['_DEFAULTEOUCLASSIFIER']._serialized_end=776
+  _globals['_DEFAULTEOUCLASSIFIER_EOUSENSITIVITY']._serialized_start=704
+  _globals['_DEFAULTEOUCLASSIFIER_EOUSENSITIVITY']._serialized_end=776
+  _globals['_EXTERNALEOUCLASSIFIER']._serialized_start=778
+  _globals['_EXTERNALEOUCLASSIFIER']._serialized_end=801
+  _globals['_EOUCLASSIFIEROPTIONS']._serialized_start=804
+  _globals['_EOUCLASSIFIEROPTIONS']._serialized_end=982
+  _globals['_RECOGNITIONCLASSIFIER']._serialized_start=985
+  _globals['_RECOGNITIONCLASSIFIER']._serialized_end=1196
+  _globals['_RECOGNITIONCLASSIFIER_TRIGGERTYPE']._serialized_start=1101
+  _globals['_RECOGNITIONCLASSIFIER_TRIGGERTYPE']._serialized_end=1196
+  _globals['_RECOGNITIONCLASSIFIEROPTIONS']._serialized_start=1198
+  _globals['_RECOGNITIONCLASSIFIEROPTIONS']._serialized_end=1290
+  _globals['_SPEECHANALYSISOPTIONS']._serialized_start=1293
+  _globals['_SPEECHANALYSISOPTIONS']._serialized_end=1429
+  _globals['_RAWAUDIO']._serialized_start=1432
+  _globals['_RAWAUDIO']._serialized_end=1631
+  _globals['_RAWAUDIO_AUDIOENCODING']._serialized_start=1566
+  _globals['_RAWAUDIO_AUDIOENCODING']._serialized_end=1631
+  _globals['_CONTAINERAUDIO']._serialized_start=1634
+  _globals['_CONTAINERAUDIO']._serialized_end=1825
+  _globals['_CONTAINERAUDIO_CONTAINERAUDIOTYPE']._serialized_start=1735
+  _globals['_CONTAINERAUDIO_CONTAINERAUDIOTYPE']._serialized_end=1825
+  _globals['_AUDIOFORMATOPTIONS']._serialized_start=1828
+  _globals['_AUDIOFORMATOPTIONS']._serialized_end=1973
+  _globals['_LANGUAGERESTRICTIONOPTIONS']._serialized_start=1976
+  _globals['_LANGUAGERESTRICTIONOPTIONS']._serialized_end=2223
+  _globals['_LANGUAGERESTRICTIONOPTIONS_LANGUAGERESTRICTIONTYPE']._serialized_start=2125
+  _globals['_LANGUAGERESTRICTIONOPTIONS_LANGUAGERESTRICTIONTYPE']._serialized_end=2223
+  _globals['_JSONSCHEMA']._serialized_start=2225
+  _globals['_JSONSCHEMA']._serialized_end=2278
+  _globals['_SUMMARIZATIONPROPERTY']._serialized_start=2281
+  _globals['_SUMMARIZATIONPROPERTY']._serialized_end=2419
+  _globals['_SUMMARIZATIONOPTIONS']._serialized_start=2421
+  _globals['_SUMMARIZATIONOPTIONS']._serialized_end=2523
+  _globals['_SUMMARIZATIONPROPERTYRESULT']._serialized_start=2525
+  _globals['_SUMMARIZATIONPROPERTYRESULT']._serialized_end=2572
+  _globals['_RECOGNITIONMODELOPTIONS']._serialized_start=2575
+  _globals['_RECOGNITIONMODELOPTIONS']._serialized_end=3009
+  _globals['_RECOGNITIONMODELOPTIONS_AUDIOPROCESSINGTYPE']._serialized_start=2919
+  _globals['_RECOGNITIONMODELOPTIONS_AUDIOPROCESSINGTYPE']._serialized_end=3009
+  _globals['_SPEAKERLABELINGOPTIONS']._serialized_start=3012
+  _globals['_SPEAKERLABELINGOPTIONS']._serialized_end=3234
+  _globals['_SPEAKERLABELINGOPTIONS_SPEAKERLABELING']._serialized_start=3122
+  _globals['_SPEAKERLABELINGOPTIONS_SPEAKERLABELING']._serialized_end=3234
+  _globals['_STREAMINGOPTIONS']._serialized_start=3237
+  _globals['_STREAMINGOPTIONS']._serialized_end=3666
+  _globals['_AUDIOCHUNK']._serialized_start=3668
+  _globals['_AUDIOCHUNK']._serialized_end=3694
+  _globals['_SILENCECHUNK']._serialized_start=3696
+  _globals['_SILENCECHUNK']._serialized_end=3731
+  _globals['_EOU']._serialized_start=3733
+  _globals['_EOU']._serialized_end=3738
+  _globals['_STREAMINGREQUEST']._serialized_start=3741
+  _globals['_STREAMINGREQUEST']._serialized_end=3973
+  _globals['_RECOGNIZEFILEREQUEST']._serialized_start=3976
+  _globals['_RECOGNIZEFILEREQUEST']._serialized_end=4394
+  _globals['_WORD']._serialized_start=4396
+  _globals['_WORD']._serialized_end=4460
+  _globals['_LANGUAGEESTIMATION']._serialized_start=4462
+  _globals['_LANGUAGEESTIMATION']._serialized_end=4526
+  _globals['_ALTERNATIVE']._serialized_start=4529
+  _globals['_ALTERNATIVE']._serialized_end=4716
+  _globals['_EOUUPDATE']._serialized_start=4718
+  _globals['_EOUUPDATE']._serialized_end=4746
+  _globals['_ALTERNATIVEUPDATE']._serialized_start=4748
+  _globals['_ALTERNATIVEUPDATE']._serialized_end=4845
+  _globals['_AUDIOCURSORS']._serialized_start=4848
+  _globals['_AUDIOCURSORS']._serialized_end=5001
+  _globals['_FINALREFINEMENT']._serialized_start=5003
+  _globals['_FINALREFINEMENT']._serialized_end=5113
+  _globals['_STATUSCODE']._serialized_start=5115
+  _globals['_STATUSCODE']._serialized_end=5191
+  _globals['_SESSIONUUID']._serialized_start=5193
+  _globals['_SESSIONUUID']._serialized_end=5245
+  _globals['_PHRASEHIGHLIGHT']._serialized_start=5247
+  _globals['_PHRASEHIGHLIGHT']._serialized_end=5322
+  _globals['_RECOGNITIONCLASSIFIERLABEL']._serialized_start=5324
+  _globals['_RECOGNITIONCLASSIFIERLABEL']._serialized_end=5387
+  _globals['_RECOGNITIONCLASSIFIERRESULT']._serialized_start=5390
+  _globals['_RECOGNITIONCLASSIFIERRESULT']._serialized_end=5556
+  _globals['_RECOGNITIONCLASSIFIERUPDATE']._serialized_start=5559
+  _globals['_RECOGNITIONCLASSIFIERUPDATE']._serialized_end=5886
+  _globals['_RECOGNITIONCLASSIFIERUPDATE_WINDOWTYPE']._serialized_start=5787
+  _globals['_RECOGNITIONCLASSIFIERUPDATE_WINDOWTYPE']._serialized_end=5886
+  _globals['_DESCRIPTIVESTATISTICS']._serialized_start=5889
+  _globals['_DESCRIPTIVESTATISTICS']._serialized_end=6076
+  _globals['_DESCRIPTIVESTATISTICS_QUANTILE']._serialized_start=6036
+  _globals['_DESCRIPTIVESTATISTICS_QUANTILE']._serialized_end=6076
+  _globals['_AUDIOSEGMENTBOUNDARIES']._serialized_start=6078
+  _globals['_AUDIOSEGMENTBOUNDARIES']._serialized_end=6146
+  _globals['_SPEAKERANALYSIS']._serialized_start=6149
+  _globals['_SPEAKERANALYSIS']._serialized_end=6924
+  _globals['_SPEAKERANALYSIS_WINDOWTYPE']._serialized_start=6848
+  _globals['_SPEAKERANALYSIS_WINDOWTYPE']._serialized_end=6924
+  _globals['_CONVERSATIONANALYSIS']._serialized_start=6927
+  _globals['_CONVERSATIONANALYSIS']._serialized_end=7700
+  _globals['_CONVERSATIONANALYSIS_INTERRUPTSEVALUATION']._serialized_start=7537
+  _globals['_CONVERSATIONANALYSIS_INTERRUPTSEVALUATION']._serialized_end=7700
+  _globals['_CONTENTUSAGE']._serialized_start=7702
+  _globals['_CONTENTUSAGE']._serialized_end=7792
+  _globals['_SUMMARIZATION']._serialized_start=7795
+  _globals['_SUMMARIZATION']._serialized_end=7929
+  _globals['_STREAMINGRESPONSE']._serialized_start=7932
+  _globals['_STREAMINGRESPONSE']._serialized_end=8667
+  _globals['_DELETERECOGNITIONREQUEST']._serialized_start=8669
+  _globals['_DELETERECOGNITIONREQUEST']._serialized_end=8717
+  _globals['_STREAMINGRESPONSELIST']._serialized_start=8719
+  _globals['_STREAMINGRESPONSELIST']._serialized_end=8808
 # @@protoc_insertion_point(module_scope)

yandex/cloud/ai/stt/v3/stt_pb2.pyi

@@ -9,6 +9,7 @@ import google.protobuf.descriptor
 import google.protobuf.internal.containers
 import google.protobuf.internal.enum_type_wrapper
 import google.protobuf.message
+import google.protobuf.struct_pb2
 import sys
 import typing
 
@@ -426,6 +427,102 @@ class LanguageRestrictionOptions(google.protobuf.message.Message):
 
 global___LanguageRestrictionOptions = LanguageRestrictionOptions
 
+@typing.final
+class JsonSchema(google.protobuf.message.Message):
+    """Represents the expected structure of the model's response using a JSON Schema."""
+
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    SCHEMA_FIELD_NUMBER: builtins.int
+    @property
+    def schema(self) -> google.protobuf.struct_pb2.Struct:
+        """The JSON Schema that the model's output must conform to."""
+
+    def __init__(
+        self,
+        *,
+        schema: google.protobuf.struct_pb2.Struct | None = ...,
+    ) -> None: ...
+    def HasField(self, field_name: typing.Literal["schema", b"schema"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["schema", b"schema"]) -> None: ...
+
+global___JsonSchema = JsonSchema
+
+@typing.final
+class SummarizationProperty(google.protobuf.message.Message):
+    """Represents summarization entry for transcription."""
+
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    INSTRUCTION_FIELD_NUMBER: builtins.int
+    JSON_OBJECT_FIELD_NUMBER: builtins.int
+    JSON_SCHEMA_FIELD_NUMBER: builtins.int
+    instruction: builtins.str
+    """Summarization instruction for model."""
+    json_object: builtins.bool
+    """When set to true, the model will respond with a valid JSON object.
+    Be sure to explicitly ask the model for JSON.
+    Otherwise, it may generate excessive whitespace and run indefinitely until it reaches the token limit.
+    """
+    @property
+    def json_schema(self) -> global___JsonSchema:
+        """Enforces a specific JSON structure for the model's response based on a provided schema."""
+
+    def __init__(
+        self,
+        *,
+        instruction: builtins.str = ...,
+        json_object: builtins.bool = ...,
+        json_schema: global___JsonSchema | None = ...,
+    ) -> None: ...
+    def HasField(self, field_name: typing.Literal["ResponseFormat", b"ResponseFormat", "json_object", b"json_object", "json_schema", b"json_schema"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["ResponseFormat", b"ResponseFormat", "instruction", b"instruction", "json_object", b"json_object", "json_schema", b"json_schema"]) -> None: ...
+    def WhichOneof(self, oneof_group: typing.Literal["ResponseFormat", b"ResponseFormat"]) -> typing.Literal["json_object", "json_schema"] | None: ...
+
+global___SummarizationProperty = SummarizationProperty
+
+@typing.final
+class SummarizationOptions(google.protobuf.message.Message):
+    """Represents transcription summarization options."""
+
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    MODEL_URI_FIELD_NUMBER: builtins.int
+    PROPERTIES_FIELD_NUMBER: builtins.int
+    model_uri: builtins.str
+    """The [ID of the model](/docs/foundation-models/concepts/yandexgpt/models) to be used for completion generation."""
+    @property
+    def properties(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___SummarizationProperty]:
+        """A list of suimmarizations to perform with transcription."""
+
+    def __init__(
+        self,
+        *,
+        model_uri: builtins.str = ...,
+        properties: collections.abc.Iterable[global___SummarizationProperty] | None = ...,
+    ) -> None: ...
+    def ClearField(self, field_name: typing.Literal["model_uri", b"model_uri", "properties", b"properties"]) -> None: ...
+
+global___SummarizationOptions = SummarizationOptions
+
+@typing.final
+class SummarizationPropertyResult(google.protobuf.message.Message):
+    """Represents summarization response entry for transcription."""
+
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    RESPONSE_FIELD_NUMBER: builtins.int
+    response: builtins.str
+    """Summarization response text."""
+    def __init__(
+        self,
+        *,
+        response: builtins.str = ...,
+    ) -> None: ...
+    def ClearField(self, field_name: typing.Literal["response", b"response"]) -> None: ...
+
+global___SummarizationPropertyResult = SummarizationPropertyResult
+
 @typing.final
 class RecognitionModelOptions(google.protobuf.message.Message):
     DESCRIPTOR: google.protobuf.descriptor.Descriptor
@@ -530,6 +627,7 @@ class StreamingOptions(google.protobuf.message.Message):
     RECOGNITION_CLASSIFIER_FIELD_NUMBER: builtins.int
     SPEECH_ANALYSIS_FIELD_NUMBER: builtins.int
     SPEAKER_LABELING_FIELD_NUMBER: builtins.int
+    SUMMARIZATION_FIELD_NUMBER: builtins.int
     @property
     def recognition_model(self) -> global___RecognitionModelOptions:
         """Configuration for speech recognition model."""
@@ -550,6 +648,10 @@ class StreamingOptions(google.protobuf.message.Message):
     def speaker_labeling(self) -> global___SpeakerLabelingOptions:
         """Configuration for speaker labeling"""
 
+    @property
+    def summarization(self) -> global___SummarizationOptions:
+        """Summarization options"""
+
     def __init__(
         self,
         *,
@@ -558,9 +660,10 @@ class StreamingOptions(google.protobuf.message.Message):
         recognition_classifier: global___RecognitionClassifierOptions | None = ...,
         speech_analysis: global___SpeechAnalysisOptions | None = ...,
         speaker_labeling: global___SpeakerLabelingOptions | None = ...,
+        summarization: global___SummarizationOptions | None = ...,
     ) -> None: ...
-    def HasField(self, field_name: typing.Literal["eou_classifier", b"eou_classifier", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis"]) -> builtins.bool: ...
-    def ClearField(self, field_name: typing.Literal["eou_classifier", b"eou_classifier", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis"]) -> None: ...
+    def HasField(self, field_name: typing.Literal["eou_classifier", b"eou_classifier", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis", "summarization", b"summarization"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["eou_classifier", b"eou_classifier", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis", "summarization", b"summarization"]) -> None: ...
 
 global___StreamingOptions = StreamingOptions
 
@@ -666,6 +769,7 @@ class RecognizeFileRequest(google.protobuf.message.Message):
     RECOGNITION_CLASSIFIER_FIELD_NUMBER: builtins.int
     SPEECH_ANALYSIS_FIELD_NUMBER: builtins.int
     SPEAKER_LABELING_FIELD_NUMBER: builtins.int
+    SUMMARIZATION_FIELD_NUMBER: builtins.int
     content: builtins.bytes
     """Bytes with data"""
     uri: builtins.str
@@ -686,6 +790,10 @@ class RecognizeFileRequest(google.protobuf.message.Message):
     def speaker_labeling(self) -> global___SpeakerLabelingOptions:
         """Configuration for speaker labeling"""
 
+    @property
+    def summarization(self) -> global___SummarizationOptions:
+        """Summarization options"""
+
     def __init__(
         self,
         *,
@@ -695,9 +803,10 @@ class RecognizeFileRequest(google.protobuf.message.Message):
         recognition_classifier: global___RecognitionClassifierOptions | None = ...,
         speech_analysis: global___SpeechAnalysisOptions | None = ...,
         speaker_labeling: global___SpeakerLabelingOptions | None = ...,
+        summarization: global___SummarizationOptions | None = ...,
     ) -> None: ...
-    def HasField(self, field_name: typing.Literal["AudioSource", b"AudioSource", "content", b"content", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis", "uri", b"uri"]) -> builtins.bool: ...
-    def ClearField(self, field_name: typing.Literal["AudioSource", b"AudioSource", "content", b"content", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis", "uri", b"uri"]) -> None: ...
+    def HasField(self, field_name: typing.Literal["AudioSource", b"AudioSource", "content", b"content", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis", "summarization", b"summarization", "uri", b"uri"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["AudioSource", b"AudioSource", "content", b"content", "recognition_classifier", b"recognition_classifier", "recognition_model", b"recognition_model", "speaker_labeling", b"speaker_labeling", "speech_analysis", b"speech_analysis", "summarization", b"summarization", "uri", b"uri"]) -> None: ...
     def WhichOneof(self, oneof_group: typing.Literal["AudioSource", b"AudioSource"]) -> typing.Literal["content", "uri"] | None: ...
 
 global___RecognizeFileRequest = RecognizeFileRequest
@@ -740,7 +849,7 @@ class LanguageEstimation(google.protobuf.message.Message):
     LANGUAGE_CODE_FIELD_NUMBER: builtins.int
     PROBABILITY_FIELD_NUMBER: builtins.int
     language_code: builtins.str
-    """Language code in ISO 639-1 format."""
+    """Language tag in IETF BCP 47 format, consisting of ISO 639-1 language code and ISO 3166-1 country code (e.g., en-US, ru-RU)."""
     probability: builtins.float
     """Estimation of language probability."""
     def __init__(
@@ -1342,6 +1451,57 @@ class ConversationAnalysis(google.protobuf.message.Message):
 
 global___ConversationAnalysis = ConversationAnalysis
 
+@typing.final
+class ContentUsage(google.protobuf.message.Message):
+    """An object representing the number of content [tokens](/docs/foundation-models/concepts/yandexgpt/tokens) used by the completion model."""
+
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    INPUT_TEXT_TOKENS_FIELD_NUMBER: builtins.int
+    COMPLETION_TOKENS_FIELD_NUMBER: builtins.int
+    TOTAL_TOKENS_FIELD_NUMBER: builtins.int
+    input_text_tokens: builtins.int
+    """The number of tokens in the textual part of the model input."""
+    completion_tokens: builtins.int
+    """The number of tokens in the generated completion."""
+    total_tokens: builtins.int
+    """The total number of tokens, including all input tokens and all generated tokens."""
+    def __init__(
+        self,
+        *,
+        input_text_tokens: builtins.int = ...,
+        completion_tokens: builtins.int = ...,
+        total_tokens: builtins.int = ...,
+    ) -> None: ...
+    def ClearField(self, field_name: typing.Literal["completion_tokens", b"completion_tokens", "input_text_tokens", b"input_text_tokens", "total_tokens", b"total_tokens"]) -> None: ...
+
+global___ContentUsage = ContentUsage
+
+@typing.final
+class Summarization(google.protobuf.message.Message):
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    RESULTS_FIELD_NUMBER: builtins.int
+    CONTENT_USAGE_FIELD_NUMBER: builtins.int
+    @property
+    def results(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___SummarizationPropertyResult]:
+        """A list of summarizations of transcription."""
+
+    @property
+    def content_usage(self) -> global___ContentUsage:
+        """A set of statistics describing the number of content tokens used by the completion model."""
+
+    def __init__(
+        self,
+        *,
+        results: collections.abc.Iterable[global___SummarizationPropertyResult] | None = ...,
+        content_usage: global___ContentUsage | None = ...,
+    ) -> None: ...
+    def HasField(self, field_name: typing.Literal["content_usage", b"content_usage"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["content_usage", b"content_usage", "results", b"results"]) -> None: ...
+
+global___Summarization = Summarization
+
 @typing.final
 class StreamingResponse(google.protobuf.message.Message):
     """Responses from server.
@@ -1363,6 +1523,7 @@ class StreamingResponse(google.protobuf.message.Message):
     CLASSIFIER_UPDATE_FIELD_NUMBER: builtins.int
     SPEAKER_ANALYSIS_FIELD_NUMBER: builtins.int
     CONVERSATION_ANALYSIS_FIELD_NUMBER: builtins.int
+    SUMMARIZATION_FIELD_NUMBER: builtins.int
     CHANNEL_TAG_FIELD_NUMBER: builtins.int
     response_wall_time_ms: builtins.int
     """Wall clock on server side. This is time when server wrote results to stream"""
@@ -1414,6 +1575,10 @@ class StreamingResponse(google.protobuf.message.Message):
     def conversation_analysis(self) -> global___ConversationAnalysis:
         """Conversation statistics"""
 
+    @property
+    def summarization(self) -> global___Summarization:
+        """Summary"""
+
     def __init__(
         self,
         *,
@@ -1428,11 +1593,12 @@ class StreamingResponse(google.protobuf.message.Message):
         classifier_update: global___RecognitionClassifierUpdate | None = ...,
         speaker_analysis: global___SpeakerAnalysis | None = ...,
         conversation_analysis: global___ConversationAnalysis | None = ...,
+        summarization: global___Summarization | None = ...,
         channel_tag: builtins.str = ...,
     ) -> None: ...
-    def HasField(self, field_name: typing.Literal["Event", b"Event", "audio_cursors", b"audio_cursors", "classifier_update", b"classifier_update", "conversation_analysis", b"conversation_analysis", "eou_update", b"eou_update", "final", b"final", "final_refinement", b"final_refinement", "partial", b"partial", "session_uuid", b"session_uuid", "speaker_analysis", b"speaker_analysis", "status_code", b"status_code"]) -> builtins.bool: ...
-    def ClearField(self, field_name: typing.Literal["Event", b"Event", "audio_cursors", b"audio_cursors", "channel_tag", b"channel_tag", "classifier_update", b"classifier_update", "conversation_analysis", b"conversation_analysis", "eou_update", b"eou_update", "final", b"final", "final_refinement", b"final_refinement", "partial", b"partial", "response_wall_time_ms", b"response_wall_time_ms", "session_uuid", b"session_uuid", "speaker_analysis", b"speaker_analysis", "status_code", b"status_code"]) -> None: ...
-    def WhichOneof(self, oneof_group: typing.Literal["Event", b"Event"]) -> typing.Literal["partial", "final", "eou_update", "final_refinement", "status_code", "classifier_update", "speaker_analysis", "conversation_analysis"] | None: ...
+    def HasField(self, field_name: typing.Literal["Event", b"Event", "audio_cursors", b"audio_cursors", "classifier_update", b"classifier_update", "conversation_analysis", b"conversation_analysis", "eou_update", b"eou_update", "final", b"final", "final_refinement", b"final_refinement", "partial", b"partial", "session_uuid", b"session_uuid", "speaker_analysis", b"speaker_analysis", "status_code", b"status_code", "summarization", b"summarization"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing.Literal["Event", b"Event", "audio_cursors", b"audio_cursors", "channel_tag", b"channel_tag", "classifier_update", b"classifier_update", "conversation_analysis", b"conversation_analysis", "eou_update", b"eou_update", "final", b"final", "final_refinement", b"final_refinement", "partial", b"partial", "response_wall_time_ms", b"response_wall_time_ms", "session_uuid", b"session_uuid", "speaker_analysis", b"speaker_analysis", "status_code", b"status_code", "summarization", b"summarization"]) -> None: ...
+    def WhichOneof(self, oneof_group: typing.Literal["Event", b"Event"]) -> typing.Literal["partial", "final", "eou_update", "final_refinement", "status_code", "classifier_update", "speaker_analysis", "conversation_analysis", "summarization"] | None: ...
 
 global___StreamingResponse = StreamingResponse
 

yandex/cloud/ai/translate/v2/package_options_pb2.py

@@ -0,0 +1,35 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: yandex/cloud/ai/translate/v2/package_options.proto
+# Protobuf Python Version: 5.29.0
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    5,
+    29,
+    0,
+    '',
+    'yandex/cloud/ai/translate/v2/package_options.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n2yandex/cloud/ai/translate/v2/package_options.proto\x12\x1cyandex.cloud.ai.translate.v2BLZJgithub.com/yandex-cloud/go-genproto/yandex/cloud/ai/translate/v2;translateb\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'yandex.cloud.ai.translate.v2.package_options_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  _globals['DESCRIPTOR']._loaded_options = None
+  _globals['DESCRIPTOR']._serialized_options = b'ZJgithub.com/yandex-cloud/go-genproto/yandex/cloud/ai/translate/v2;translate'
+# @@protoc_insertion_point(module_scope)

yandex/cloud/ai/translate/v2/package_options_pb2.pyi

@@ -0,0 +1,8 @@
+"""
+@generated by mypy-protobuf.  Do not edit manually!
+isort:skip_file
+"""
+
+import google.protobuf.descriptor
+
+DESCRIPTOR: google.protobuf.descriptor.FileDescriptor