Merge remote-tracking branch 'origin/dev' into feat/transcription-agent

README.md

@@ -33,7 +33,7 @@ uv run --only-group test pytest test/unit
 Or for integration tests:
 
 ```bash
-uv run --only-group integration-test pytest test/integration
+uv run --group integration-test pytest test/integration
 ```
 
 ## GitHooks

pyproject.toml

@@ -7,6 +7,7 @@ requires-python = ">=3.13"
 dependencies = [
     "fastapi[all]>=0.115.6",
     "mlx-whisper>=0.4.3 ; sys_platform == 'darwin'",
+    "numpy>=2.3.3",
     "openai-whisper>=20250625",
     "pyaudio>=0.2.14",
     "pydantic>=2.12.0",
@@ -33,6 +34,7 @@ integration-test = [
     "soundfile>=0.13.1",
 ]
 test = [
+    "numpy>=2.3.3",
     "pytest>=8.4.2",
     "pytest-asyncio>=1.2.0",
     "pytest-cov>=7.0.0",

src/control_backend/agents/vad_agent.py

@@ -59,8 +59,10 @@ class Streaming(CyclicBehaviour):
     async def run(self) -> None:
         data = await self.audio_in_poller.poll()
         if data is None:
-            logger.debug("No audio data received. Discarding buffer until new data arrives.")
-            self.audio_buffer = np.array([], dtype=np.float32)
+            if len(self.audio_buffer) > 0:
+                logger.debug("No audio data received. Discarding buffer until new data arrives.")
+                self.audio_buffer = np.array([], dtype=np.float32)
+                self.i_since_speech = 100
             return
 
         # copy otherwise Torch will be sad that it's immutable

test/unit/agents/test_vad_streaming.py

@@ -18,6 +18,9 @@ def audio_out_socket():
 
 @pytest.fixture
 def streaming(audio_in_socket, audio_out_socket):
+    import torch
+
+    torch.hub.load.return_value = (..., ...)  # Mock
     return Streaming(audio_in_socket, audio_out_socket)
 
 

test/unit/conftest.py

@@ -33,3 +33,13 @@ def pytest_configure(config):
     mock_config_module.settings = MagicMock()
 
     sys.modules["control_backend.core.config"] = mock_config_module
+
+    # --- Mock torch and zmq for VAD ---
+    mock_torch = MagicMock()
+    mock_zmq = MagicMock()
+    mock_zmq.asyncio = mock_zmq
+
+    # In individual tests, these can be imported and the return values changed
+    sys.modules["torch"] = mock_torch
+    sys.modules["zmq"] = mock_zmq
+    sys.modules["zmq.asyncio"] = mock_zmq.asyncio

uv.lock

@@ -1332,6 +1332,7 @@ source = { virtual = "." }
 dependencies = [
     { name = "fastapi", extra = ["all"] },
     { name = "mlx-whisper", marker = "sys_platform == 'darwin'" },
+    { name = "numpy" },
     { name = "openai-whisper" },
     { name = "pyaudio" },
     { name = "pydantic" },
@@ -1358,6 +1359,7 @@ integration-test = [
     { name = "soundfile" },
 ]
 test = [
+    { name = "numpy" },
     { name = "pytest" },
     { name = "pytest-asyncio" },
     { name = "pytest-cov" },
@@ -1368,6 +1370,7 @@ test = [
 requires-dist = [
     { name = "fastapi", extras = ["all"], specifier = ">=0.115.6" },
     { name = "mlx-whisper", marker = "sys_platform == 'darwin'", specifier = ">=0.4.3" },
+    { name = "numpy", specifier = ">=2.3.3" },
     { name = "openai-whisper", specifier = ">=20250625" },
     { name = "pyaudio", specifier = ">=0.2.14" },
     { name = "pydantic", specifier = ">=2.12.0" },
@@ -1392,6 +1395,7 @@ dev = [
 ]
 integration-test = [{ name = "soundfile", specifier = ">=0.13.1" }]
 test = [
+    { name = "numpy", specifier = ">=2.3.3" },
     { name = "pytest", specifier = ">=8.4.2" },
     { name = "pytest-asyncio", specifier = ">=1.2.0" },
     { name = "pytest-cov", specifier = ">=7.0.0" },