init

services/asr.py

@@ -0,0 +1,54 @@
+from __future__ import annotations
+
+import logging
+from typing import Optional
+
+import numpy as np
+
+logger = logging.getLogger(__name__)
+
+
+class ASRService:
+    def __init__(self) -> None:
+        self._model = None
+        self._post = None
+        self._ready = False
+        self._attempted = False
+        self._init_error: Optional[str] = None
+
+    def _ensure_loaded(self) -> None:
+        if self._ready or self._attempted:
+            return
+        self._attempted = True
+        try:
+            from funasr import AutoModel
+            from funasr.utils.postprocess_utils import rich_transcription_postprocess
+
+            self._model = AutoModel(
+                model="FunAudioLLM/SenseVoiceSmall",
+                device="cuda:0",
+                hub="hf",
+            )
+            self._post = rich_transcription_postprocess
+            self._ready = True
+        except Exception as exc:  # pragma: no cover
+            self._init_error = str(exc)
+            logger.warning("ASR fallback mode: %s", exc)
+
+    async def transcribe(self, audio_16k: np.ndarray) -> str:
+        self._ensure_loaded()
+        if self._ready and self._model is not None and self._post is not None:
+            res = self._model.generate(
+                input=audio_16k,
+                cache={},
+                language="zh",
+                use_itn=True,
+            )
+            return self._post(res[0]["text"])
+
+        # Fallback for initial bring-up
+        return "（ASR降级）语音已接收"
+
+    @property
+    def health(self) -> dict:
+        return {"ready": self._ready, "attempted": self._attempted, "error": self._init_error}