feat/lang_detection_plugin (#220)

ovos_plugin_manager/audio_transformers.py

@@ -1,7 +1,8 @@
-from ovos_plugin_manager.utils import PluginTypes, PluginConfigTypes
-from ovos_plugin_manager.templates.transformers import AudioTransformer
 from ovos_utils.log import LOG
 
+from ovos_plugin_manager.templates.transformers import AudioTransformer, AudioLanguageDetector
+from ovos_plugin_manager.utils import PluginTypes, PluginConfigTypes
+
 
 def find_plugins(*args, **kwargs):
     # TODO: Deprecate in 0.1.0
@@ -57,3 +58,13 @@ def get_audio_transformer_module_configs(module_name: str):
     """
     from ovos_plugin_manager.utils.config import load_plugin_configs
     return load_plugin_configs(module_name, PluginConfigTypes.AUDIO_TRANSFORMER)
+
+
+def find_audio_lang_detector_plugins() -> dict:
+    """
+    Find all installed audio language detector plugins
+    @return: dict plugin names to entrypoints
+    """
+    from ovos_plugin_manager.utils import find_plugins
+    return {k: p for k, p in find_plugins(PluginTypes.AUDIO_TRANSFORMER).items()
+            if issubclass(p, AudioLanguageDetector)}

ovos_plugin_manager/templates/transformers.py

@@ -1,7 +1,10 @@
-from typing import List, Tuple
+import abc
+from typing import List, Tuple, Optional
 
-from ovos_config.config import Configuration
 from ovos_bus_client.util import get_mycroft_bus
+from ovos_config.config import Configuration
+from ovos_config.locale import get_default_lang
+from ovos_utils.log import LOG
 
 from ovos_plugin_manager.utils import ReadWriteStream
 
@@ -233,3 +236,22 @@ def transform(self, wav_file: str, context: dict = None) -> Tuple[str, dict]:
     def default_shutdown(self):
         """ perform any shutdown actions """
         pass
+
+
+class AudioLanguageDetector(AudioTransformer):
+
+    @property
+    def valid_langs(self) -> List[str]:
+        return list(
+            set([get_default_lang()] + Configuration().get("secondary_langs", []))
+        )
+
+    @abc.abstractmethod
+    def detect(self, audio_data: bytes, valid_langs: Optional[List] = None) -> Tuple[str, float]:
+        raise NotImplementedError
+
+    # plugin api
+    def transform(self, audio_data: bytes):
+        lang, prob = self.detect(audio_data)
+        LOG.info(f"Detected speech language '{lang}' with probability {prob}")
+        return audio_data, {"stt_lang": lang, "lang_probability": prob}