add Moss result to db

sentence_api/repository.py

@@ -120,6 +120,20 @@ class VideoRepository:
                     FOREIGN KEY (share_id) REFERENCES dub_shares(share_id) ON DELETE CASCADE,
                     UNIQUE (share_id, sentence_index)
                 );
+
+                CREATE TABLE IF NOT EXISTS moss_segments (
+                    video_hash TEXT NOT NULL,
+                    segment_index INTEGER NOT NULL,
+                    start_seconds REAL NOT NULL,
+                    end_seconds REAL NOT NULL,
+                    start_ms INTEGER NOT NULL,
+                    end_ms INTEGER NOT NULL,
+                    text TEXT NOT NULL,
+                    speaker TEXT,
+                    words_json TEXT,
+                    PRIMARY KEY (video_hash, segment_index),
+                    FOREIGN KEY (video_hash) REFERENCES videos(video_hash) ON DELETE CASCADE
+                );
                 """
             )
 
@@ -247,12 +261,17 @@ class VideoRepository:
         document: SentenceBoundaryDocument,
         transcription: Optional[str],
         ready: bool = True,
+        moss_segments: Optional[List[Dict[str, Any]]] = None,
     ) -> None:
         with self._connect() as connection:
             connection.execute(
                 "DELETE FROM sentences WHERE video_hash = ?",
                 (document.video_hash,),
             )
+            connection.execute(
+                "DELETE FROM moss_segments WHERE video_hash = ?",
+                (document.video_hash,),
+            )
             connection.executemany(
                 """
                 INSERT INTO sentences (
@@ -288,6 +307,19 @@ class VideoRepository:
                     document.video_hash,
                 ),
             )
+            if moss_segments:
+                connection.executemany(
+                    """
+                    INSERT INTO moss_segments (
+                        video_hash, segment_index, start_seconds, end_seconds,
+                        start_ms, end_ms, text, speaker, words_json
+                    ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+                    """,
+                    [
+                        _moss_segment_row(document.video_hash, index, segment)
+                        for index, segment in enumerate(moss_segments)
+                    ],
+                )
 
     def mark_oss_uploaded(self, video_hash: str, remote_url: str) -> None:
         with self._connect() as connection:
@@ -327,6 +359,32 @@ class VideoRepository:
             ).fetchone()
         return dict(row) if row else None
 
+    def get_moss_segments(self, video_hash: str) -> List[Dict[str, Any]]:
+        with self._connect() as connection:
+            rows = connection.execute(
+                """
+                SELECT * FROM moss_segments
+                WHERE video_hash = ?
+                ORDER BY segment_index
+                """,
+                (normalize_video_hash(video_hash),),
+            ).fetchall()
+        segments = []
+        for row in rows:
+            item = dict(row)
+            raw_words = item.pop("words_json", None)
+            words = None
+            if raw_words:
+                try:
+                    parsed = json.loads(raw_words)
+                except json.JSONDecodeError:
+                    parsed = None
+                if isinstance(parsed, list):
+                    words = parsed
+            item["words"] = words
+            segments.append(item)
+        return segments
+
     def get_document(self, video_hash: str) -> Optional[SentenceBoundaryDocument]:
         video = self.get_video(video_hash)
         if video is None or video["duration_ms"] is None:
@@ -1221,3 +1279,22 @@ def _is_cjk_char(char: str) -> bool:
         or 0xF900 <= code <= 0xFAFF
         or 0x20000 <= code <= 0x2CEAF
     )
+
+
+def _moss_segment_row(video_hash: str, index: int, segment: Dict[str, Any]) -> tuple:
+    start_seconds = float(segment["start_seconds"])
+    end_seconds = float(segment["end_seconds"])
+    words = segment.get("words")
+    words_json = json.dumps(words, ensure_ascii=False) if words else None
+    speaker = segment.get("speaker")
+    return (
+        video_hash,
+        index,
+        start_seconds,
+        end_seconds,
+        int(round(start_seconds * 1000)),
+        int(round(end_seconds * 1000)),
+        str(segment.get("text") or "").strip(),
+        str(speaker) if speaker is not None else None,
+        words_json,
+    )