feat: added an example with speech contexts

python/snippets/stt_streaming_recognize_context.py

@@ -68,7 +68,8 @@ def print_streaming_recognition_responses(responses):
                 print('"' + alternative.transcript + '"')
             print("------------------")
 
-stub = stt_pb2_grpc.SpeechToTextStub(grpc.secure_channel(endpoint, grpc.ssl_channel_credentials()))
+#stub = stt_pb2_grpc.SpeechToTextStub(grpc.secure_channel(endpoint, grpc.ssl_channel_credentials()))
+stub = stt_pb2_grpc.SpeechToTextStub(grpc.insecure_channel(endpoint))
 metadata = authorization_metadata(api_key, secret_key, "tinkoff.cloud.stt")
 responses = stub.StreamingRecognize(generate_repeated_requests(3), metadata=metadata)
 print_streaming_recognition_responses(responses)

python/snippets/stt_streaming_recognize_context_dictionaries.py

@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+
+import sys
+sys.path.append("..")
+
+from tinkoff.cloud.stt.v1 import stt_pb2_grpc, stt_pb2
+from auth import authorization_metadata
+import grpc
+import os
+import wave
+
+endpoint = os.environ.get("VOICEKIT_ENDPOINT") or "api.tinkoff.ai:443"
+api_key = os.environ["VOICEKIT_API_KEY"]
+secret_key = os.environ["VOICEKIT_SECRET_KEY"]
+
+def build_first_request(sample_rate_hertz, num_channels, context):
+    request = stt_pb2.StreamingRecognizeRequest()
+    request.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
+    request.streaming_config.config.sample_rate_hertz = sample_rate_hertz
+    request.streaming_config.config.num_channels = num_channels
+    # ВАЖНО!
+    # Не рекомендуется:
+    # - добавление слов короче 5 символов
+    # - выставление score в слишком большое значение
+    request.streaming_config.config.speech_contexts.append(context)
+    return request
+
+def generate_requests(context):
+    try:
+        with wave.open("../../audio/numbers.wav") as f:
+            yield build_first_request(f.getframerate(), f.getnchannels(), context)
+            frame_samples = f.getframerate()//10 # Send 100ms at a time
+            for data in iter(lambda:f.readframes(frame_samples), b''):
+                request = stt_pb2.StreamingRecognizeRequest()
+                request.audio_content = data
+                yield request
+            # Sending 1 second of silence
+            for i in range(10):
+                request = stt_pb2.StreamingRecognizeRequest()
+                request.audio_content = bytes(frame_samples)
+                yield request
+    except Exception as e:
+        print("Got exception in generate_requests", e)
+        raise
+
+def generate_repeated_requests(times_repeated):
+    context = stt_pb2.SpeechContext(speech_context_dictionary_id = "numbers")
+    null_context = stt_pb2.SpeechContext(phrases = [])
+    # Контекст может быть указан несколько раз во время стриминговой сессии
+    # В этом примере чередуется добавление контекста и его сброс
+    # Стоит отметить, что для сброса контекста нужно переслать SpeechContext с пустым списком фраз
+    for i in range(times_repeated):
+        for request in generate_requests(context if (i % 2 == 1) else null_context):
+            yield request
+
+
+def print_streaming_recognition_responses(responses):
+    for response in responses:
+        for result in response.results:
+            print("Channel", result.recognition_result.channel)
+            print("Phrase start:", result.recognition_result.start_time.ToTimedelta())
+            print("Phrase end:  ", result.recognition_result.end_time.ToTimedelta())
+            for alternative in result.recognition_result.alternatives:
+                print('"' + alternative.transcript + '"')
+            print("------------------")
+
+#stub = stt_pb2_grpc.SpeechToTextStub(grpc.secure_channel(endpoint, grpc.ssl_channel_credentials()))
+stub = stt_pb2_grpc.SpeechToTextStub(grpc.insecure_channel(endpoint))
+metadata = authorization_metadata(api_key, secret_key, "tinkoff.cloud.stt")
+responses = stub.StreamingRecognize(generate_repeated_requests(3), metadata=metadata)
+print_streaming_recognition_responses(responses)

python/tinkoff/cloud/stt/v1/stt_pb2_grpc.py

@@ -1,139 +1,81 @@
 # Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
 import grpc
 
 from tinkoff.cloud.longrunning.v1 import longrunning_pb2 as tinkoff_dot_cloud_dot_longrunning_dot_v1_dot_longrunning__pb2
 from tinkoff.cloud.stt.v1 import stt_pb2 as tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2
 
 
 class SpeechToTextStub(object):
-    """Speech recognition.
-    """
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Recognize = channel.unary_unary(
-                '/tinkoff.cloud.stt.v1.SpeechToText/Recognize',
-                request_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeRequest.SerializeToString,
-                response_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeResponse.FromString,
-                )
-        self.StreamingRecognize = channel.stream_stream(
-                '/tinkoff.cloud.stt.v1.SpeechToText/StreamingRecognize',
-                request_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeRequest.SerializeToString,
-                response_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeResponse.FromString,
-                )
-        self.LongRunningRecognize = channel.unary_unary(
-                '/tinkoff.cloud.stt.v1.SpeechToText/LongRunningRecognize',
-                request_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.LongRunningRecognizeRequest.SerializeToString,
-                response_deserializer=tinkoff_dot_cloud_dot_longrunning_dot_v1_dot_longrunning__pb2.Operation.FromString,
-                )
+  """Speech recognition
+  """
 
+  def __init__(self, channel):
+    """Constructor.
 
-class SpeechToTextServicer(object):
-    """Speech recognition.
+    Args:
+      channel: A grpc.Channel.
     """
+    self.Recognize = channel.unary_unary(
+        '/tinkoff.cloud.stt.v1.SpeechToText/Recognize',
+        request_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeRequest.SerializeToString,
+        response_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeResponse.FromString,
+        )
+    self.StreamingRecognize = channel.stream_stream(
+        '/tinkoff.cloud.stt.v1.SpeechToText/StreamingRecognize',
+        request_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeRequest.SerializeToString,
+        response_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeResponse.FromString,
+        )
+    self.LongRunningRecognize = channel.unary_unary(
+        '/tinkoff.cloud.stt.v1.SpeechToText/LongRunningRecognize',
+        request_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.LongRunningRecognizeRequest.SerializeToString,
+        response_deserializer=tinkoff_dot_cloud_dot_longrunning_dot_v1_dot_longrunning__pb2.Operation.FromString,
+        )
 
-    def Recognize(self, request, context):
-        """Method to recognize whole audio at once: sending complete audio, getting complete recognition result.
-        """
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StreamingRecognize(self, request_iterator, context):
-        """Method for streaming recognition.
-        """
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LongRunningRecognize(self, request, context):
-        """Method to create longrunning recognition operation. Created operation will persist for a limited time and will be deleted after that time has expired.
-        """
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
 
+class SpeechToTextServicer(object):
+  """Speech recognition
+  """
 
-def add_SpeechToTextServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Recognize': grpc.unary_unary_rpc_method_handler(
-                    servicer.Recognize,
-                    request_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeRequest.FromString,
-                    response_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeResponse.SerializeToString,
-            ),
-            'StreamingRecognize': grpc.stream_stream_rpc_method_handler(
-                    servicer.StreamingRecognize,
-                    request_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeRequest.FromString,
-                    response_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeResponse.SerializeToString,
-            ),
-            'LongRunningRecognize': grpc.unary_unary_rpc_method_handler(
-                    servicer.LongRunningRecognize,
-                    request_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.LongRunningRecognizeRequest.FromString,
-                    response_serializer=tinkoff_dot_cloud_dot_longrunning_dot_v1_dot_longrunning__pb2.Operation.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'tinkoff.cloud.stt.v1.SpeechToText', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
+  def Recognize(self, request, context):
+    """Method to recognize whole audio at once: sending complete audio, getting complete recognition result.
+    """
+    context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+    context.set_details('Method not implemented!')
+    raise NotImplementedError('Method not implemented!')
 
- # This class is part of an EXPERIMENTAL API.
-class SpeechToText(object):
-    """Speech recognition.
+  def StreamingRecognize(self, request_iterator, context):
+    """Method for streaming recognition.
     """
+    context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+    context.set_details('Method not implemented!')
+    raise NotImplementedError('Method not implemented!')
 
-    @staticmethod
-    def Recognize(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/tinkoff.cloud.stt.v1.SpeechToText/Recognize',
-            tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeRequest.SerializeToString,
-            tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+  def LongRunningRecognize(self, request, context):
+    """Method to create longrunning recognition operation. Created operation will persist for a limited time and will be deleted after that time has expired.
+    """
+    context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+    context.set_details('Method not implemented!')
+    raise NotImplementedError('Method not implemented!')
 
-    @staticmethod
-    def StreamingRecognize(request_iterator,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.stream_stream(request_iterator, target, '/tinkoff.cloud.stt.v1.SpeechToText/StreamingRecognize',
-            tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeRequest.SerializeToString,
-            tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
 
-    @staticmethod
-    def LongRunningRecognize(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/tinkoff.cloud.stt.v1.SpeechToText/LongRunningRecognize',
-            tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.LongRunningRecognizeRequest.SerializeToString,
-            tinkoff_dot_cloud_dot_longrunning_dot_v1_dot_longrunning__pb2.Operation.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
+def add_SpeechToTextServicer_to_server(servicer, server):
+  rpc_method_handlers = {
+      'Recognize': grpc.unary_unary_rpc_method_handler(
+          servicer.Recognize,
+          request_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeRequest.FromString,
+          response_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.RecognizeResponse.SerializeToString,
+      ),
+      'StreamingRecognize': grpc.stream_stream_rpc_method_handler(
+          servicer.StreamingRecognize,
+          request_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeRequest.FromString,
+          response_serializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.StreamingRecognizeResponse.SerializeToString,
+      ),
+      'LongRunningRecognize': grpc.unary_unary_rpc_method_handler(
+          servicer.LongRunningRecognize,
+          request_deserializer=tinkoff_dot_cloud_dot_stt_dot_v1_dot_stt__pb2.LongRunningRecognizeRequest.FromString,
+          response_serializer=tinkoff_dot_cloud_dot_longrunning_dot_v1_dot_longrunning__pb2.Operation.SerializeToString,
+      ),
+  }
+  generic_handler = grpc.method_handlers_generic_handler(
+      'tinkoff.cloud.stt.v1.SpeechToText', rpc_method_handlers)
+  server.add_generic_rpc_handlers((generic_handler,))