add volcengine OSS support

sentence_api/processing.py

@@ -14,6 +14,7 @@ from .audio_metrics import (
 from .config import Settings
 from .generate_boundaries import ALGORITHM_VERSION, make_entry
 from .models import SentenceBoundary, SentenceBoundaryDocument
+from .oss import VolcanoOSSUploader
 from .repository import VideoRepository
 from .transcription import Transcript, Transcriber, split_sentences_at_punctuation
 
@@ -27,10 +28,12 @@ class VideoProcessor:
         settings: Settings,
         repository: VideoRepository,
         transcriber: Transcriber,
+        oss_uploader: Optional[VolcanoOSSUploader] = None,
     ):
         self.settings = settings
         self.repository = repository
         self.transcriber = transcriber
+        self.oss_uploader = oss_uploader
 
     def process(self, video_hash: str) -> None:
         video = self.repository.get_video(video_hash)
@@ -57,7 +60,7 @@ class VideoProcessor:
                 )
                 if not document.sentences:
                     raise RuntimeError("MOSS returned no timestamped speech segments.")
-                self.repository.save_processing_result(document, transcript.text)
+                self._finish_processing(document, transcript.text)
             else:
                 entry, _ = make_entry(media_path, video_hash=video_hash)
                 document = SentenceBoundaryDocument(
@@ -66,7 +69,7 @@ class VideoProcessor:
                     algorithm_version=ALGORITHM_VERSION,
                     sentences=entry["sentences"],
                 )
-                self.repository.save_processing_result(document, None)
+                self._finish_processing(document, None)
         except Exception as exc:
             self.repository.mark_failed(video_hash, str(exc))
             raise
@@ -74,6 +77,28 @@ class VideoProcessor:
             if work_path is not None:
                 work_path.unlink(missing_ok=True)
 
+    def _finish_processing(
+        self,
+        document: SentenceBoundaryDocument,
+        transcription: Optional[str],
+    ) -> None:
+        if self.oss_uploader is None or not self.oss_uploader.enabled:
+            self.repository.save_processing_result(document, transcription, ready=True)
+            return
+
+        # Keep the course invisible while the mandatory OSS upload is running.
+        self.repository.save_processing_result(document, transcription, ready=False)
+        video = self.repository.get_video(document.video_hash)
+        if video is None:
+            raise ValueError(f"Unknown video: {document.video_hash}")
+        remote_url = self.oss_uploader.upload(
+            video_hash=document.video_hash,
+            stored_filename=video["stored_filename"],
+            local_path=self.settings.videos_dir / video["stored_filename"],
+            content_type=video.get("content_type"),
+        )
+        self.repository.mark_oss_uploaded(document.video_hash, remote_url)
+
 
 def extract_audio(media_path: Path, output_path: Path) -> None:
     ffmpeg = shutil.which("ffmpeg")