pepperplus/pepperplus-cb
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

refactor: remove constants and put in config file

Browse Source 
removed all constants from all files and put them in src/control_backend/core/config.py
also removed some old mock agents that we don't use anymore

ref: N25B-236
This commit is contained in:
Pim Hutting
2025-11-05 13:43:57 +01:00
parent c6edad0bb4
commit 9e926178da
15 changed files with 136 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
src/control_backend/agents/vad_agent.py
View File

@@ -20,7 +20,11 @@ class SocketPoller[T]:
multiple usages.
"""
def __init__(self, socket: azmq.Socket, timeout_ms: int = 100):
def __init__(
self,
socket: azmq.Socket,
timeout_ms: int = settings.behaviour_settings.socket_poller_timeout_ms,
):
"""
:param socket: The socket to poll and get data from.
:param timeout_ms: A timeout in milliseconds to wait for data.
@@ -49,12 +53,16 @@ class Streaming(CyclicBehaviour):
super().__init__()
self.audio_in_poller = SocketPoller[bytes](audio_in_socket)
self.model, _ = torch.hub.load(
repo_or_dir="snakers4/silero-vad", model="silero_vad", force_reload=False
repo_or_dir=settings.vad_settings.repo_or_dir,
model=settings.vad_settings.model_name,
force_reload=False,
)
self.audio_out_socket = audio_out_socket
self.audio_buffer = np.array([], dtype=np.float32)
self.i_since_speech = 100 # Used to allow small pauses in speech
self.i_since_speech = (
settings.behaviour_settings.vad_initial_since_speech
) # Used to allow small pauses in speech
async def run(self) -> None:
data = await self.audio_in_poller.poll()
@@ -62,15 +70,17 @@ class Streaming(CyclicBehaviour):
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
self.i_since_speech = settings.behaviour_settings.vad_initial_since_speech
return
# copy otherwise Torch will be sad that it's immutable
chunk = np.frombuffer(data, dtype=np.float32).copy()
prob = self.model(torch.from_numpy(chunk), 16000).item()
prob = self.model(torch.from_numpy(chunk), settings.vad_settings.sample_rate_hz).item()
non_speech_patience = settings.behaviour_settings.vad_non_speech_patience_chunks
prob_threshold = settings.behaviour_settings.vad_prob_threshold
if prob > 0.5:
if self.i_since_speech > 3:
if prob > prob_threshold:
if self.i_since_speech > non_speech_patience:
logger.debug("Speech started.")
self.audio_buffer = np.append(self.audio_buffer, chunk)
self.i_since_speech = 0
@@ -78,7 +88,7 @@ class Streaming(CyclicBehaviour):
self.i_since_speech += 1
# prob < 0.5, so speech maybe ended. Wait a bit more before to be more certain
if self.i_since_speech <= 3:
if self.i_since_speech <= non_speech_patience:
self.audio_buffer = np.append(self.audio_buffer, chunk)
return