feat: refactor summarizer and PDF extraction pipeline

- Split summarizer into summary_generator and summary_persister modules
- Refactor pdf_image_extractor to two-phase pipeline with PicoDet layout detection
- Add layout_detector service for PicoDet-S_layout_3cls integration
- Add exceptions module with ConflictError and NotFoundError
- Improve admin dashboard with better statistics and task management
- Add design review document with system optimization suggestions
- Add new tests for crawler, pdf_downloader, pipeline, and summary_utils
- Update dependencies and configuration
- Clean up dead code and improve error handling

app/services/pdf_downloader.py

@@ -9,6 +9,7 @@ from pathlib import Path
 
 import requests
 
+from app.config import settings
 from app.utils import PAPERS_DIR, TMP_DIR
 
 logger = logging.getLogger(__name__)
@@ -51,6 +52,14 @@ def _get_session() -> requests.Session:
     return _http_session
 
 
+def close_http_session() -> None:
+    """关闭全局 HTTP Session，供应用 shutdown 时调用。"""
+    global _http_session
+    if _http_session is not None:
+        _http_session.close()
+        _http_session = None
+
+
 async def download_pdf(arxiv_id: str, pdf_url: str) -> Path:
     """下载 PDF 到 data/tmp/{arxiv_id}/paper.pdf。"""
     if not pdf_url:
@@ -62,10 +71,16 @@ async def download_pdf(arxiv_id: str, pdf_url: str) -> Path:
 
     try:
         session = _get_session()
-        resp = session.get(pdf_url, timeout=120, allow_redirects=True)
+        resp = session.get(pdf_url, timeout=settings.PDF_DOWNLOAD_TIMEOUT, allow_redirects=True)
         resp.raise_for_status()
         dest.write_bytes(resp.content)
     except Exception as exc:
+        # 清理残留的部分文件
+        if dest.exists():
+            try:
+                dest.unlink()
+            except OSError:
+                pass
         raise PdfDownloadError(f"failed to download PDF for {arxiv_id}: {exc}") from exc
 
     logger.info("Downloaded PDF: %s (%d bytes)", arxiv_id, dest.stat().st_size)