dashscope 1.24.7__py3-none-any.whl → 1.24.9__py3-none-any.whl

dashscope/audio/asr/recognition.py

@@ -189,7 +189,7 @@ class Recognition(BaseApi):
         responses = self.__launch_request()
         for part in responses:
             if part.status_code == HTTPStatus.OK:
-                if len(part.output) == 0:
+                if len(part.output) == 0 or ('finished' in part.output and part.output['finished'] == True):
                     self._on_complete_timestamp = time.time() * 1000
                     logger.debug('last package delay {}'.format(
                         self.get_last_package_delay()))

dashscope/audio/qwen_omni/omni_realtime.py

@@ -4,6 +4,7 @@ import json
 import platform
 import threading
 import time
+from dataclasses import field, dataclass
 from typing import List
 import uuid
 from enum import Enum, unique
@@ -29,6 +30,26 @@ class OmniRealtimeCallback:
         pass
 
 
+@dataclass
+class TranslationParams:
+    """
+    TranslationParams
+    """
+    language: str = field(default=None)
+
+
+@dataclass
+class TranscriptionParams:
+    """
+    TranscriptionParams
+    """
+    language: str = field(default=None)
+    sample_rate: int = field(default=16000)
+    input_audio_format: str = field(default="pcm")
+    corpus: dict = field(default=None)
+    corpus_text: str = field(default=None)
+
+
 @unique
 class AudioFormat(Enum):
     # format, sample_rate, channels, bit_rate, name
@@ -171,7 +192,7 @@ class OmniRealtimeConversation:
 
     def update_session(self,
                        output_modalities: List[MultiModality],
-                       voice: str,
+                       voice: str = None,
                        input_audio_format: AudioFormat = AudioFormat.
                        PCM_16000HZ_MONO_16BIT,
                        output_audio_format: AudioFormat = AudioFormat.
@@ -184,6 +205,8 @@ class OmniRealtimeConversation:
                        turn_detection_threshold: float = 0.2,
                        turn_detection_silence_duration_ms: int = 800,
                        turn_detection_param: dict = None,
+                       translation_params: TranslationParams = None,
+                       transcription_params: TranscriptionParams = None,
                        **kwargs) -> None:
         '''
         update session configuration, should be used before create response
@@ -206,6 +229,13 @@ class OmniRealtimeConversation:
             In a quiet environment, it may be necessary to decrease the threshold to improve sensitivity
         turn_detection_silence_duration_ms: int
             duration of silence in milliseconds to detect turn, range [200, 6000]
+        translation_params: TranslationParams
+            translation params, include language. Only effective with qwen3-livetranslate-flash-realtime model or
+             further models. Do not set this parameter for other models.
+        transcription_params: TranscriptionParams
+            transcription params, include language, sample_rate, input_audio_format, corpus.
+            Only effective with qwen3-asr-flash-realtime model or
+            further models. Do not set this parameter for other models.
         '''
         self.config = {
             'modalities': [m.value for m in output_modalities],
@@ -230,6 +260,20 @@ class OmniRealtimeConversation:
                 self.config['turn_detection'].update(turn_detection_param)
         else:
             self.config['turn_detection'] = None
+        if translation_params is not None:
+            self.config['translation'] = {
+                'language': translation_params.language
+            }
+        if transcription_params is not None:
+            self.config['language'] = transcription_params.language
+            if transcription_params.corpus is not None:
+                self.config['corpus'] = transcription_params.corpus
+            if transcription_params.corpus_text is not None:
+                self.config['corpus'] = {
+                    "text": transcription_params.corpus_text
+                }
+            self.config['input_audio_format'] = transcription_params.input_audio_format
+            self.config['sample_rate']= transcription_params.sample_rate
         self.config.update(kwargs)
         self.__send_str(
             json.dumps({

dashscope/audio/tts_v2/enrollment.py

@@ -68,20 +68,25 @@ class VoiceEnrollmentService(BaseApi):
         logger.debug('>>>>recv', response)
         return response
 
-    def create_voice(self, target_model: str, prefix: str, url: str) -> str:
+    def create_voice(self, target_model: str, prefix: str, url: str, language_hints: List[str] = None) -> str:
         '''
         创建新克隆音色
         param: target_model 克隆音色对应的语音合成模型版本
         param: prefix 音色自定义前缀，仅允许数字和小写字母，小于十个字符。
         param: url 用于克隆的音频文件url
+        param: language_hints 克隆音色目标语言
         return: voice_id
         '''
-        response = self.__call_with_input(input={
+
+        input_params = {
             'action': 'create_voice',
             'target_model': target_model,
             'prefix': prefix,
-            'url': url,
-        }, )
+            'url': url
+        }
+        if language_hints is not None:
+            input_params['language_hints'] = language_hints
+        response = self.__call_with_input(input_params)
         self._last_request_id = response.request_id
         if response.status_code == 200:
             return response.output['voice_id']

dashscope/multimodal/multimodal_request_params.py

@@ -72,7 +72,31 @@ class RequestBodyInput(DashPayloadInput):
             "directive": self.directive,
             "dialog_id": self.dialog_id
         }
+@dataclass
+class AsrPostProcessing:
+    replace_words: list = field(default=None)
+
+    def to_dict(self):
+        if self.replace_words is None:
+            return None
+        if len(self.replace_words) == 0:
+            return None
+        return {
+            "replace_words":  [word.to_dict() for word in self.replace_words]
+        }
+
+@dataclass
+class ReplaceWord:
+    source: str = field(default=None)
+    target: str = field(default=None)
+    match_mode: str = field(default=None)
 
+    def to_dict(self):
+        return {
+            "source": self.source,
+            "target": self.target,
+            "match_mode": self.match_mode
+        }
 
 @dataclass
 class Upstream:
@@ -80,7 +104,9 @@ class Upstream:
     audio_format: str = field(default="pcm")  # 上行语音格式，默认pcm.支持pcm/opus
     type: str = field(default="AudioOnly")  # 上行类型：AudioOnly 仅语音通话; AudioAndVideo 上传视频
     mode: str = field(default="tap2talk")  # 客户端交互模式 push2talk/tap2talk/duplex
-    # sample_rate: int  # 合成音频采样率
+    sample_rate: int = field(default=16000)  # 音频采样率
+    vocabulary_id: str = field(default=None)
+    asr_post_processing: AsrPostProcessing = field(default=None)
     pass_through_params: dict = field(default=None)
 
     def to_dict(self):
@@ -88,8 +114,12 @@ class Upstream:
             "type": self.type,
             "mode": self.mode,
             "audio_format": self.audio_format,
-            # "sample_rate": self.sample_rate
+            "sample_rate": self.sample_rate,
+            "vocabulary_id": self.vocabulary_id,
         }
+        if self.asr_post_processing is not None:
+            upstream["asr_post_processing"] = self.asr_post_processing.to_dict()
+
         if self.pass_through_params is not None:
             upstream.update(self.pass_through_params)
         return upstream

dashscope/version.py

@@ -1,3 +1,3 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
-__version__ = '1.24.7'
+__version__ = '1.24.9'

{dashscope-1.24.7.dist-info → dashscope-1.24.9.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dashscope
-Version: 1.24.7
+Version: 1.24.9
 Summary: dashscope client sdk library
 Home-page: https://dashscope.aliyun.com/
 Author: Alibaba Cloud

{dashscope-1.24.7.dist-info → dashscope-1.24.9.dist-info}/RECORD

@@ -3,7 +3,7 @@ dashscope/cli.py,sha256=64oGkevgX0RHPPmMg0sevXDgaFLQNA_0vdtjQ7Z2pHM,26492
 dashscope/files.py,sha256=vRDQygm3lOqBZR73o7KNHs1iTBVuvLncuwJNxIYjzAU,3981
 dashscope/model.py,sha256=B5v_BtYLPqj6raClejBgdKg6WTGwhH_f-20pvsQqmsk,1491
 dashscope/models.py,sha256=dE4mzXkl85G343qVylSGpURPRdA5pZSqXlx6PcxqC_Q,1275
-dashscope/version.py,sha256=bwLuOw_5t9V66aTp9e4fg40jaUHqhx-IXawXHN8THGg,74
+dashscope/version.py,sha256=MtrJg1M58kTZJEqP85UuTy6JbT8pmsUXF5os_JzMTWI,74
 dashscope/aigc/__init__.py,sha256=kYvYEoRK-NUHyMWpBDNQBz4fVA__uOhHRK2kDTBaWgk,617
 dashscope/aigc/chat_completion.py,sha256=ONlyyssIbfaKKcFo7cEKhHx5OCF2XX810HFzIExW1ho,14813
 dashscope/aigc/code_generation.py,sha256=p_mxDKJLQMW0IjFD46JRlZuEZCRESSVKEfLlAevBtqw,10936
@@ -32,12 +32,12 @@ dashscope/assistants/files.py,sha256=CaQkZK7TFeMaAxtqMi-1rBVJrlKXdehZG9plNZ6zslo
 dashscope/audio/__init__.py,sha256=7e3ejVsDJxEbMHN-9E0nEDfU-CnnQ4JgtgUxqNs0IG4,192
 dashscope/audio/asr/__init__.py,sha256=JoCenJAUVOQXPmAn1toKeFYCfc8BqNn0NKpqjuJvNJc,1055
 dashscope/audio/asr/asr_phrase_manager.py,sha256=vHOLExaKCtjedkihIu7gyfQyarR9rN5JZn79LvlCpco,7693
-dashscope/audio/asr/recognition.py,sha256=b_aAPvOKjpWdSiYhM_hp30sZ06QdmNBSDJwhiv78kHM,20932
+dashscope/audio/asr/recognition.py,sha256=-xMcdwHbjTV1RIQRQguzBzB5pKd1tskrDukLgzsJDi8,20999
 dashscope/audio/asr/transcription.py,sha256=lYzPjh7jJQwjMoxx8-AY0YCMBKNKO0bi7xd5tZGSHPc,9094
 dashscope/audio/asr/translation_recognizer.py,sha256=JgBmhkIl_kqH8uVwop6Fba5KlXccftKFrhaygN9PKjU,39680
 dashscope/audio/asr/vocabulary.py,sha256=N0pMS2x1lDxqJ14FgTGKctfuVkR2_hlEsCNWFcgYpTY,6717
 dashscope/audio/qwen_omni/__init__.py,sha256=MEFxmyxr5H6bW22l_R9073Pl6Ka6knvhrATGT-4UBjI,298
-dashscope/audio/qwen_omni/omni_realtime.py,sha256=eBmoOxuKcfzMHuXsQWCrIIKmso9iEzYylOeYZ5upv-w,14869
+dashscope/audio/qwen_omni/omni_realtime.py,sha256=b7t14nsciA8YcJ4MGr2GzmDxbgBR2wpbml9ZG_GNqiI,16722
 dashscope/audio/qwen_tts/__init__.py,sha256=JS3axY1grqO0aTIJufZ3KS1JsU6yf6y4K2CQlNvUK9I,132
 dashscope/audio/qwen_tts/speech_synthesizer.py,sha256=7LHR-PXhn-VE1cCOp_82Jq0zE9rMc3xy3dszUeyLLNs,2927
 dashscope/audio/qwen_tts_realtime/__init__.py,sha256=vVkmeJr_mEAn_O0Rh5AU3ICg6qIZqppUryJ5lY8VYPo,254
@@ -45,7 +45,7 @@ dashscope/audio/qwen_tts_realtime/qwen_tts_realtime.py,sha256=uMLglxRjUZgol9Z7MT
 dashscope/audio/tts/__init__.py,sha256=xYpMFseUZGgqgj_70zcX2VsLv-L7qxJ3d-bbdj_hO0I,245
 dashscope/audio/tts/speech_synthesizer.py,sha256=vD1xQV-rew8qAsIaAGH5amsNtB0SqdtNhVHhJHGQ-xk,7622
 dashscope/audio/tts_v2/__init__.py,sha256=me9a3_7KsHQxcJ8hx4SeKlY1e_ThHVvGMw7Yn0uoscM,333
-dashscope/audio/tts_v2/enrollment.py,sha256=-nrlywYSOP73Bm9ETTSxNnlp-B8ezJcUmd59mVvyvgk,6361
+dashscope/audio/tts_v2/enrollment.py,sha256=ekeZJz_swhI0OwRANuUwsZjdP0rRoUergSsCUQmsh8E,6577
 dashscope/audio/tts_v2/speech_synthesizer.py,sha256=p764P4TYwLkvvPCpA4VnFwlNbIJbuNbp2d9mxgni7Ws,22047
 dashscope/client/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dashscope/client/base_api.py,sha256=znAJ65DeHiFw1H7FWK0YrkLz1CoNcyqUxF8EJ3gujeY,52523
@@ -73,7 +73,7 @@ dashscope/multimodal/__init__.py,sha256=fyqeolbDLWVn5wSpPZ3nAOnUBRF9k6mlsy6dCmgj
 dashscope/multimodal/dialog_state.py,sha256=CtOdfGWhq0ePG3bc8-7inhespETtPD4QDli1513hd1A,1522
 dashscope/multimodal/multimodal_constants.py,sha256=z_QVq01E43FAqKQnDu9vdf89d1zuYlWyANewWTEXVJM,1282
 dashscope/multimodal/multimodal_dialog.py,sha256=HymlaQYp7SgJdoKbT27SNiviyRRoM91zklNBwTHmm1Q,23939
-dashscope/multimodal/multimodal_request_params.py,sha256=Lbxf_kLnFUkhty8AU9wL7ws9tYbmhHPVmsiXLdynlJg,8402
+dashscope/multimodal/multimodal_request_params.py,sha256=iOnATOdv4aRp5ffU8lY2Gu0UNlz-sBCEun2zFG4saFk,9356
 dashscope/multimodal/tingwu/__init__.py,sha256=Gi9GEM0bdeJlZpvyksSeHOc2--_tG5aF6QAx6TAS2fE,225
 dashscope/multimodal/tingwu/tingwu.py,sha256=01d-QOeuB1QmRhiZqbXJ8pHoGqT0C-xZTjIs_ZBXOyw,2613
 dashscope/multimodal/tingwu/tingwu_realtime.py,sha256=oBeqrZit3uBZHuyI7m9VILz2qaqJRMO0-Nm2eJ5Q63g,20215
@@ -100,9 +100,9 @@ dashscope/tokenizers/tokenizer.py,sha256=3FQVDvMNkCW9ccYeJdjrd_PIMMD3Xv7aNZkaYOE
 dashscope/tokenizers/tokenizer_base.py,sha256=5EJIFuizMWESEmLmbd38yJnfeHmPnzZPwsO4aOGjpl4,707
 dashscope/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dashscope/utils/oss_utils.py,sha256=aZIHlMN2JOfVw6kp0SVrMw_N1MfoTcR_-wiRbJ7DgHw,7501
-dashscope-1.24.7.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-dashscope-1.24.7.dist-info/METADATA,sha256=3YnzYbPbSB-pNzn3CNw_Ka5MTDXM8oc_-ddrkugwav4,7146
-dashscope-1.24.7.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-dashscope-1.24.7.dist-info/entry_points.txt,sha256=e9C3sOf9zDYL0O5ROEGX6FT8w-QK_kaGRWmPZDHAFys,49
-dashscope-1.24.7.dist-info/top_level.txt,sha256=woqavFJK9zas5xTqynmALqOtlafghjsk63Xk86powTU,10
-dashscope-1.24.7.dist-info/RECORD,,
+dashscope-1.24.9.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+dashscope-1.24.9.dist-info/METADATA,sha256=30oaoqG5k4PuhwOh7ndPEvAYH1cFJ3JneX58N8id_qE,7146
+dashscope-1.24.9.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+dashscope-1.24.9.dist-info/entry_points.txt,sha256=e9C3sOf9zDYL0O5ROEGX6FT8w-QK_kaGRWmPZDHAFys,49
+dashscope-1.24.9.dist-info/top_level.txt,sha256=woqavFJK9zas5xTqynmALqOtlafghjsk63Xk86powTU,10
+dashscope-1.24.9.dist-info/RECORD,,