feat: configurable audio quality

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "speechd"
-version = "1.2.5"
+version = "1.2.6"
 description = "Speech-to-Text daemon with Groq Whisper API"
 readme = "README.md"
 requires-python = ">=3.11"

src/speechd/config.py

@@ -7,6 +7,7 @@ model = "whisper-large-v3-turbo"
 language = "ru"
 sample_rate = 16000
 timeout = 300
+audio_quality = 0.8
 """
 
 
@@ -18,6 +19,7 @@ class Config:
     sample_rate: int
     timeout_seconds: int
     runtime_dir: str
+    audio_quality: float
 
     @classmethod
     def load(cls) -> "Config":
@@ -47,4 +49,5 @@ class Config:
             sample_rate=data.get("sample_rate", 16000),
             timeout_seconds=data.get("timeout", 300),
             runtime_dir=os.environ.get("XDG_RUNTIME_DIR", "/tmp"),
+            audio_quality=data.get("audio_quality", 0.8),
         )

src/speechd/daemon.py

@@ -28,6 +28,7 @@ class SpeechDaemon:
             model=config.model,
             language=config.language,
             sample_rate=config.sample_rate,
+            audio_quality=config.audio_quality,
         )
         self.recording = False
         self.frames: list[np.ndarray] = []

src/speechd/transcribe.py

@@ -17,11 +17,14 @@ class TranscriptionResult:
 
 
 class Transcriber:
-    def __init__(self, api_key: str, model: str, language: str, sample_rate: int):
+    def __init__(
+        self, api_key: str, model: str, language: str, sample_rate: int, audio_quality: float
+    ):
         self.client = Groq(api_key=api_key)
         self.model = model
         self.language = language
         self.sample_rate = sample_rate
+        self.audio_quality = audio_quality
 
     def transcribe(self, audio_data: np.ndarray) -> TranscriptionResult:
         if len(audio_data) == 0:
@@ -53,6 +56,6 @@ class Transcriber:
             self.sample_rate,
             format="OGG",
             subtype="OPUS",
-            compression_level=0.8,
+            compression_level=self.audio_quality,
         )
         return buf.getvalue()