(for Python) Support realtime reading from microphone and auth with client_id + client_secret.

recognition/v1/python3/README.md

@@ -1,6 +1,6 @@
 # Installation
 
-    $ pip3 install grpcio-tools
+    $ pip3 install grpcio-tools requests pyaudio
     $ python3 -m grpc_tools.protoc -I .. -I ../../../task/v1 --python_out=. --grpc_python_out=. ../recognition.proto ../../../task/v1/{storage,task}.proto
 
 # Usage

recognition/v1/python3/recognize.py

@@ -9,6 +9,11 @@
 import recognition_pb2
 import recognition_pb2_grpc
 
+import requests
+import pyaudio
+import wave
+import uuid
+import base64
 
 CHUNK_SIZE = 2048
 SLEEP_TIME = 0.1
@@ -25,22 +30,60 @@
 }
 
 
+def auth(client_id, client_secret):
+    auth_code = base64.b64encode(bytes(f"{client_id}:{client_secret}", "ascii")).decode('ascii')
+    headers = {
+        "Content-Type": "application/x-www-form-urlencoded",
+        "RqUID": str(uuid.uuid4()),
+        "Authorization": f"Basic {auth_code}"
+        }
+    res = requests.post('https://salute.online.sberbank.ru:9443/api/v2/oauth', headers=headers, data="scope=SBER_SPEECH")
+    if res.ok:
+        data = res.json()
+        return res.json().get("access_token", None)
+    else:
+        print(f"Error {res.status_code}, {res.text}")
+        return None
+
 def try_printing_request_id(md):
     for m in md:
         if m.key == 'x-request-id':
             print('RequestID:', m.value)
 
 
-def generate_audio_chunks(path, chunk_size=CHUNK_SIZE, sleep_time=SLEEP_TIME):
-    with open(path, 'rb') as f:
-        for data in iter(lambda: f.read(chunk_size), b''):
+def generate_audio_chunks(path, mic, recognition_options, chunk_size=CHUNK_SIZE, sleep_time=SLEEP_TIME):
+    if path:
+        with open(path, 'rb') as f:
+            for data in iter(lambda: f.read(chunk_size), b''):
+                yield recognition_pb2.RecognitionRequest(audio_chunk=data)
+                time.sleep(sleep_time)
+    elif mic:
+        audio = pyaudio.PyAudio()
+        stream = audio.open(format=pyaudio.paInt16, channels=1,
+            rate=recognition_options.sample_rate, input=True,
+            frames_per_buffer=CHUNK_SIZE)
+        print("Recording...")
+        while True:
+            data = stream.read(CHUNK_SIZE)
             yield recognition_pb2.RecognitionRequest(audio_chunk=data)
-            time.sleep(sleep_time)
+    else:
+        print("User --file or --mic argument")
+        exit(1)
 
 
 def recognize(args):
+    token = None
+    if not hasattr(args, "token"):
+        if not hasattr(args, "client_id") and not hasattr(args, "client_secret"):
+            print("Use token or client_id + client_secret arguments.")
+            exit(1)
+        else:
+            token = auth(args.client_id, args.client_secret)
+    else:
+        token = args.token
+
     ssl_cred = grpc.ssl_channel_credentials()
-    token_cred = grpc.access_token_call_credentials(args.token)
+    token_cred = grpc.access_token_call_credentials(token)
 
     channel = grpc.secure_channel(
         args.host,
@@ -51,7 +94,7 @@ def recognize(args):
 
     con = stub.Recognize(itertools.chain(
         (recognition_pb2.RecognitionRequest(options=args.recognition_options),),
-        generate_audio_chunks(args.file),
+        generate_audio_chunks(args.file, args.mic, args.recognition_options),
     ))
 
     try:
@@ -82,7 +125,7 @@ def recognize(args):
 
 
 class Arguments:
-    NOT_RECOGNITION_OPTIONS = {'host', 'token', 'file', 'normalized_result', 'emotions_result'}
+    NOT_RECOGNITION_OPTIONS = {'host', 'token', 'client_id', 'client_secret', 'file', 'mic', 'normalized_result', 'emotions_result'}
     NOT_RECOGNITION_OPTIONS.update({'ca', 'cert', 'key'})  # diff
     DURATIONS = {'no_speech_timeout', 'max_speech_timeout', 'eou_timeout'}
     REPEATED = {'words'}
@@ -114,8 +157,11 @@ def create_parser():
     parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
 
     parser.add_argument('--host', default='smartspeech.sber.ru', help='host:port of gRPC endpoint')
-    parser.add_argument('--token', required=True, default=argparse.SUPPRESS, help='access token')
-    parser.add_argument('--file', required=True, default=argparse.SUPPRESS, help='audio file for recognition')
+    parser.add_argument('--token', required=False, default=argparse.SUPPRESS, help='access token')
+    parser.add_argument('--client_id', required=False, default=argparse.SUPPRESS, help='client id (instead of token)')
+    parser.add_argument('--client_secret', required=False, default=argparse.SUPPRESS, help='client id (instead of token)')
+    parser.add_argument('--file', required=False, default="", help='audio file for recognition')
+    parser.add_argument('--mic', action='store_true', help=' ')
 
     parser.add_argument('--normalized-result', action='store_true', help='show normalized text')
     parser.add_argument('--emotions-result', action='store_true', help='show emotions result')