first pass at turn based conversation

vocode/streaming/output_device/base_output_device.py

@@ -0,0 +1,13 @@
+from vocode.streaming.models.audio_encoding import AudioEncoding
+
+
+class BaseOutputDevice:
+    def __init__(self, sampling_rate: int, audio_encoding: AudioEncoding):
+        self.sampling_rate = sampling_rate
+        self.audio_encoding = audio_encoding
+
+    async def send_async(self, chunk):
+        raise NotImplemented
+
+    async def maybe_send_mark_async(self, message):
+        pass

vocode/streaming/output_device/speaker_output.py

@@ -0,0 +1,34 @@
+import sounddevice as sd
+import numpy as np
+
+from .base_output_device import BaseOutputDevice
+from vocode.streaming.models.audio_encoding import AudioEncoding
+
+
+class SpeakerOutput(BaseOutputDevice):
+    DEFAULT_SAMPLING_RATE = 44100
+
+    def __init__(
+        self,
+        device_info: dict,
+        sampling_rate: int = None,
+        audio_encoding: AudioEncoding = AudioEncoding.LINEAR16,
+    ):
+        self.device_info = device_info
+        sampling_rate = sampling_rate or int(
+            self.device_info.get("default_samplerate", self.DEFAULT_SAMPLING_RATE)
+        )
+        super().__init__(sampling_rate, audio_encoding)
+        self.stream = sd.OutputStream(
+            channels=1,
+            samplerate=self.sampling_rate,
+            dtype=np.int16,
+            device=int(self.device_info["index"]),
+        )
+        self.stream.start()
+
+    async def send_async(self, chunk):
+        self.stream.write(np.frombuffer(chunk, dtype=np.int16))
+
+    def terminate(self):
+        self.stream.close()

vocode/streaming/output_device/telephone_output.py

@@ -0,0 +1,7 @@
+from .base_output_device import BaseOutputDevice
+from vocode.streaming.models.audio_encoding import AudioEncoding
+
+
+class TelephoneOutput(BaseOutputDevice):
+    def __init__(self):
+        super().__init__(sampling_rate=8000, audio_encoding=AudioEncoding.MULAW)