test: add first unit test for VAD agent

Mocking audio input probabilities, checking whether it publishes audio data on the output socket.

ref: N25B-213

test/unit/agents/test_vad_streaming.py

@@ -0,0 +1,45 @@
+from unittest.mock import AsyncMock, MagicMock
+
+import numpy as np
+import pytest
+
+from control_backend.agents.vad_agent import Streaming
+
+
+@pytest.fixture
+def audio_in_socket():
+    return AsyncMock()
+
+
+@pytest.fixture
+def audio_out_socket():
+    return AsyncMock()
+
+
+@pytest.fixture
+def streaming(audio_in_socket, audio_out_socket):
+    return Streaming(audio_in_socket, audio_out_socket)
+
+
+@pytest.mark.asyncio
+async def test_voice_activity_detected(audio_in_socket, audio_out_socket, streaming):
+    # After three chunks of audio with speech probability of 1.0, then four chunks of audio with
+    #  speech probability of 0.0, it should send a message over the audio out socket
+    probabilities = [1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0]
+    model_item = MagicMock()
+    model_item.item.side_effect = probabilities
+    streaming.model = MagicMock()
+    streaming.model.return_value = model_item
+
+    audio_in_poller = AsyncMock()
+    audio_in_poller.poll.return_value = np.empty(shape=512, dtype=np.float32)
+    streaming.audio_in_poller = audio_in_poller
+
+    for _ in probabilities:
+        await streaming.run()
+
+    audio_out_socket.send.assert_called_once()
+    data = audio_out_socket.send.call_args[0][0]
+    assert isinstance(data, bytes)
+    # each sample has 512 frames of 4 bytes, expecting 5 chunks (3 with speech, 2 as padding)
+    assert len(data) == 512*4*5