feat: remote Whisper via whisper.base_url — OpenAI-compatible upload

transcription.py

@@ -1,4 +1,5 @@
 import asyncio
+import httpx
 
 
 class TranscriptionEngine:
@@ -23,6 +24,27 @@ class TranscriptionEngine:
         language: str = "de",
         model_name: str = "large-v3",
         device: str = "auto",
+        base_url: str = "",
+    ) -> str:
+        if base_url:
+            return await self._transcribe_remote(audio_path, language, model_name, base_url)
+        return await self._transcribe_local(audio_path, language, model_name, device)
+
+    async def _transcribe_remote(
+        self, audio_path: str, language: str, model_name: str, base_url: str
+    ) -> str:
+        async with httpx.AsyncClient(timeout=300) as client:
+            with open(audio_path, "rb") as f:
+                r = await client.post(
+                    f"{base_url}/v1/audio/transcriptions",
+                    files={"file": ("audio.wav", f, "audio/wav")},
+                    data={"model": model_name, "language": language},
+                )
+            r.raise_for_status()
+            return r.json()["text"]
+
+    async def _transcribe_local(
+        self, audio_path: str, language: str, model_name: str, device: str
     ) -> str:
         loop = asyncio.get_event_loop()
         model = self._get_model(model_name, device)