fixed a bug

sentence_api/main.py

@@ -2,6 +2,8 @@ import hashlib
 import hmac
 import logging
 import mimetypes
+import shutil
+import subprocess
 import uuid
 from pathlib import Path
 from typing import Any, BinaryIO, Dict, List, Optional
@@ -388,6 +390,7 @@ def create_app(
         video_hash: str = Form(...),
         title: str = Form(default="我的口语配音"),
         segments: str = Form(...),
+        scores: str = Form(default="[]"),
         files: List[UploadFile] = File(...),
     ) -> Dict[str, Any]:
         try:
@@ -398,6 +401,17 @@ def create_app(
             raise HTTPException(status_code=400, detail="At least one dubbing segment is required.")
         if len(segment_items) != len(files):
             raise HTTPException(status_code=400, detail="Segment count does not match audio file count.")
+        try:
+            score_items = json.loads(scores)
+        except (TypeError, ValueError) as exc:
+            raise HTTPException(status_code=400, detail="scores must be valid JSON.") from exc
+        if not isinstance(score_items, list):
+            raise HTTPException(status_code=400, detail="scores must be a JSON array.")
+        scores_by_index = {
+            int(item["sentence_index"]): item
+            for item in score_items
+            if isinstance(item, dict) and "sentence_index" in item
+        }
         document = find_document(video_hash.lower())
         if document is None:
             raise HTTPException(status_code=404, detail="Video or sentence boundaries were not found.")
@@ -418,7 +432,9 @@ def create_app(
                 audio_name = f"{uuid.uuid4().hex}{suffix}"
                 audio_path = service_settings.dub_shares_dir / audio_name
                 await run_in_threadpool(_persist_upload, upload.file, audio_path, 50 * 1024 * 1024)
+                await run_in_threadpool(_trim_leading_silence, audio_path)
                 saved_files.append(audio_path)
+                score = scores_by_index.get(sentence_index, {})
                 prepared_segments.append({
                     "sentence_index": sentence_index,
                     "start_ms": boundary.start_ms,
@@ -426,6 +442,8 @@ def create_app(
                     "text": boundary.text,
                     "audio_filename": audio_name,
                     "audio_size_bytes": audio_path.stat().st_size,
+                    "overall_score": float(score.get("overall_score", 0)),
+                    "recognized_text": str(score.get("recognized_text") or ""),
                 })
             share = video_repository.create_dub_share(
                 video_hash=video_hash.lower(),
@@ -458,12 +476,15 @@ def create_app(
             "video_hash": share["video_hash"],
             "title": share["title"],
             "created_at": share["created_at"],
+            "video_hash": share["video_hash"],
             "segments": [
                 {
                     "index": row["sentence_index"],
                     "start_ms": row["start_ms"],
                     "end_ms": row["end_ms"],
                     "text": row["text"],
+                    "overall_score": row["overall_score"],
+                    "recognized_text": row["recognized_text"],
                     "audio_url": f"/api/v1/dub-shares/{share_id}/audio/{row['audio_filename']}",
                 }
                 for row in share["segments"]
@@ -519,6 +540,34 @@ def _persist_upload(source: BinaryIO, destination: Path, max_bytes: int) -> tupl
     return digest.hexdigest(), size
 
 
+def _trim_leading_silence(audio_path: Path) -> None:
+    ffmpeg = shutil.which("ffmpeg")
+    if ffmpeg is None:
+        return
+    trimmed_path = audio_path.with_name(f"{audio_path.stem}-trimmed{audio_path.suffix}")
+    command = [
+        ffmpeg,
+        "-hide_banner",
+        "-loglevel",
+        "error",
+        "-y",
+        "-i",
+        str(audio_path),
+        "-af",
+        "silenceremove=start_periods=1:start_threshold=-45dB:start_silence=0.08",
+        "-c:a",
+        "aac",
+        str(trimmed_path),
+    ]
+    try:
+        subprocess.run(command, check=True)
+        if trimmed_path.stat().st_size == 0:
+            raise RuntimeError("Trimmed audio is empty.")
+        trimmed_path.replace(audio_path)
+    except (OSError, subprocess.SubprocessError):
+        trimmed_path.unlink(missing_ok=True)
+
+
 async def _persist_request_stream(
     request: Request,
     destination: Path,