Use pyaudio instead of sounddevice (#51)

* Use pyaudio instead of sounddevice

* Change how config dict is merged.

* Update version number in setup.py.

* Update to version 2.

Author: avoliva

requirements.txt

@@ -1,4 +1,4 @@
 symbl_rest >= 1.0.11
 websocket-client >= 0.59.0
-sounddevice >= 0.4.1
+pyaudio ~= 0.2.12
 numpy

setup.py

@@ -20,12 +20,12 @@
 
 setup(
     name="symbl",
-    version="1.0.17",
+    version="2.0.0",
     description="symbl.ai SDK",
     author_email="[email protected]",
     url="",
     keywords=["Symbl.ai SDK"],
-    install_requires=["symbl_rest >= 1.0.11", "websocket-client >= 0.59.0", "sounddevice >= 0.4.1", "numpy"],
+    install_requires=["symbl_rest >= 1.0.11", "websocket-client >= 0.59.0", "pyaudio ~= 0.2.12", "numpy"],
     packages=find_packages(),
     include_package_data=True,
     long_description="""\

symbl/streaming_api/StreamingApi.py

@@ -1,6 +1,7 @@
 from symbl import StreamingConnection
 from symbl.configs.configs import SYMBL_STREAMING_API_FORMAT
 from symbl.AuthenticationToken import get_access_token
+from symbl.utils import Helper
 import websocket
 import base64
 import random
@@ -31,15 +32,19 @@ def start_connection(self, credentials=None, speaker=None, insight_types=None, c
             }
         }
 
-        for key in config.keys():
-            config_object[key] = config[key]
+        merged_config = Helper.merge_two_dicts(config_object, config)
+
+        # Merge subdictionary
+        speechKey = 'speechRecognition'
+        if speechKey in config:
+            merged_config[speechKey] = Helper.merge_two_dicts(config_object[speechKey], config[speechKey])
 
         start_request = {
             "type": "start_request",
             "insightTypes": [] if insight_types == None else [] if type(insight_types) != list else insight_types,
             "speaker": speaker,
             "trackers":trackers,
-            "config": config_object
+            "config": merged_config
         }
 
         return StreamingConnection(url= url, connectionId=id, start_request=start_request)

symbl/streaming_api/StreamingConnection.py

@@ -6,6 +6,7 @@
 import json
 import websocket
 
+
 class StreamingConnection():
 
     def __init__(self, url: str, connectionId: str, start_request: dict):
@@ -63,7 +64,11 @@ def send_audio(self, data):
 
     @wrap_keyboard_interrupt
     def send_audio_from_mic(self, device=None):
-        import sounddevice as sd
-        with sd.InputStream(blocksize=4096, samplerate=44100, channels=1, callback= lambda indata, *args: self.send_audio(indata.copy().tobytes()), dtype='int16', device=device):
-            while True:
-                pass
+        import pyaudio
+        audio = pyaudio.PyAudio()
+        chunk = 4096
+        samplerate = self.start_request['config']['speechRecognition']['sampleRateHertz']
+        stream = audio.open(format=pyaudio.paInt16, channels=1, rate=samplerate, input=True, frames_per_buffer=chunk)
+        while True:
+            data = stream.read(chunk)
+            self.send_audio(data)

symbl/utils/Helper.py

@@ -95,4 +95,10 @@ def parse_entity_response(api_response):
                 entity_res[key]=val
         entity_response.append(entity_res) 
 
-    return dict(entities=entity_response)
+    return dict(entities=entity_response)
+
+# Using copy because we support Python 3.4
+def merge_two_dicts(x, y):
+    z = x.copy()
+    z.update(y) 
+    return z