feat: add concurrency safety, caption detection, admin enhancements, and performance improvements

app/routes/admin.py

@@ -2,8 +2,10 @@
 
 from __future__ import annotations
 
+import csv
 import hashlib
 import hmac
+import io
 from datetime import date
 
 from fastapi import (
@@ -15,7 +17,7 @@ from fastapi import (
     Query,
     Request,
 )
-from fastapi.responses import RedirectResponse
+from fastapi.responses import RedirectResponse, Response
 from pydantic import BaseModel, field_validator
 from sqlalchemy.orm import Session
 
@@ -298,6 +300,183 @@ async def admin_job_detail(
     return detail
 
 
+# ── 任务监控 ──────────────────────────────────────────────────────────
+
+
+@router.get("/jobs")
+async def admin_jobs(
+    request: Request,
+    _admin: None = Depends(verify_admin),
+    db: Session = Depends(get_db),
+    status: str = Query("all"),
+    job_type: str = Query("all"),
+    page: int = Query(1, ge=1),
+    per_page: int = Query(20, ge=1, le=100),
+):
+    """后台任务监控页。"""
+    jobs, total = admin_svc.query_jobs(
+        db, status=status, job_type=job_type, page=page, per_page=per_page
+    )
+    counts = admin_svc.get_job_status_counts(db)
+
+    def pagination_url(p: int) -> str:
+        params = dict(request.query_params)
+        params["page"] = str(p)
+        return "/admin/jobs?" + "&".join(f"{k}={v}" for k, v in params.items())
+
+    return templates.TemplateResponse(
+        request,
+        "admin_jobs.html",
+        {
+            "jobs": jobs,
+            "total": total,
+            "page": page,
+            "per_page": per_page,
+            "current_status": status,
+            "current_type": job_type,
+            "status_counts": counts,
+            "pagination_url": pagination_url,
+        },
+    )
+
+
+# ── 锁管理 ────────────────────────────────────────────────────────────
+
+
+@router.post("/locks/{lock_id}/release")
+async def admin_release_lock(
+    lock_id: int,
+    _admin: None = Depends(verify_admin),
+    db: Session = Depends(get_db),
+):
+    """强制释放一个卡死的任务锁。"""
+    if not admin_svc.force_release_lock(db, lock_id):
+        raise HTTPException(
+            status_code=404, detail=f"Lock not found or already released: {lock_id}"
+        )
+    return {"status": "success", "lock_id": lock_id}
+
+
+# ── 重抓 ──────────────────────────────────────────────────────────────
+
+
+@router.post("/paper-recrawl/{arxiv_id}")
+async def admin_paper_recrawl(
+    arxiv_id: str,
+    background_tasks: BackgroundTasks,
+    _admin: None = Depends(verify_admin),
+    db: Session = Depends(get_db),
+):
+    """重新抓取单篇已存在论文的完整元数据。"""
+    job = create_job(
+        db, "recrawl_one", owner="admin_recrawl", payload={"arxiv_id": arxiv_id}
+    )
+    enqueue_job(background_tasks, job.id)
+    return {"status": "queued", "job_id": job.id, "arxiv_id": arxiv_id}
+
+
+# ── 索引重建 ──────────────────────────────────────────────────────────
+
+
+class RebuildIndexRequest(BaseModel):
+    target: str  # "fts" / "chroma" / "both"
+
+    @field_validator("target")
+    @classmethod
+    def target_must_be_valid(cls, v: str) -> str:
+        if v not in ("fts", "chroma", "both"):
+            raise ValueError("target must be 'fts', 'chroma' or 'both'")
+        return v
+
+
+@router.post("/rebuild-indexes")
+async def admin_rebuild_indexes(
+    body: RebuildIndexRequest,
+    background_tasks: BackgroundTasks,
+    _admin: None = Depends(verify_admin),
+    db: Session = Depends(get_db),
+):
+    """重建搜索索引（FTS5 / ChromaDB）。"""
+    job_ids: list[int] = []
+    if body.target in ("fts", "both"):
+        job = create_job(db, "reindex_fts", owner="admin_reindex", payload={})
+        enqueue_job(background_tasks, job.id)
+        job_ids.append(job.id)
+    if body.target in ("chroma", "both"):
+        job = create_job(db, "reindex_chroma", owner="admin_reindex", payload={})
+        enqueue_job(background_tasks, job.id)
+        job_ids.append(job.id)
+    return {"status": "queued", "job_ids": job_ids, "target": body.target}
+
+
+# ── 导出 CSV ──────────────────────────────────────────────────────────
+
+
+@router.get("/papers/export.csv")
+async def admin_papers_export(
+    _admin: None = Depends(verify_admin),
+    db: Session = Depends(get_db),
+    q: str = Query(""),
+    date_from: str | None = Query(None),
+    date_to: str | None = Query(None),
+    tag: str = Query(""),
+    summary_status: str = Query("all"),
+    sort: str = Query("date_desc"),
+):
+    """导出当前过滤条件下的论文为 CSV（含 UTF-8 BOM，Excel 友好）。"""
+    papers, _total, statuses = admin_svc.query_papers(
+        db,
+        q=q,
+        date_from=date_from,
+        date_to=date_to,
+        tag=tag,
+        summary_status=summary_status,
+        sort=sort,
+        page=1,
+        per_page=10**6,
+    )
+
+    buf = io.StringIO()
+    buf.write("")  # UTF-8 BOM for Excel
+    writer = csv.writer(buf)
+    writer.writerow(
+        [
+            "arxiv_id",
+            "title_en",
+            "title_zh",
+            "paper_date",
+            "upvotes",
+            "summary_status",
+            "authors",
+            "tags",
+            "pdf_url",
+        ]
+    )
+    for paper in papers:
+        authors = ";".join(a.name for a in paper.authors)
+        tags = ";".join(t.tag for t in paper.tags)
+        writer.writerow(
+            [
+                paper.arxiv_id,
+                paper.title_en or "",
+                paper.title_zh or "",
+                str(paper.paper_date) if paper.paper_date else "",
+                paper.upvotes or 0,
+                statuses.get(paper.arxiv_id, "none"),
+                authors,
+                tags,
+                paper.pdf_url or "",
+            ]
+        )
+
+    filename = f"papers_{today_str().replace('-', '')}.csv"
+    return Response(
+        content=buf.getvalue(),
+        media_type="text/csv; charset=utf-8",
+        headers={"Content-Disposition": f'attachment; filename="{filename}"'},
+    )
+
+
 # ── 日志 ──────────────────────────────────────────────────────────────
 
 
@@ -438,24 +617,25 @@ async def admin_paper_delete(
 
 
 class BatchActionRequest(BaseModel):
-    action: str  # "delete" or "summarize"
+    action: str  # "delete" / "summarize" / "recrawl"
     arxiv_ids: list[str]
 
     @field_validator("action")
     @classmethod
     def action_must_be_valid(cls, v: str) -> str:
-        if v not in ("delete", "summarize"):
-            raise ValueError("action must be 'delete' or 'summarize'")
+        if v not in ("delete", "summarize", "recrawl"):
+            raise ValueError("action must be 'delete', 'summarize' or 'recrawl'")
         return v
 
 
 @router.post("/papers-batch-action")
 async def admin_papers_batch_action(
     body: BatchActionRequest,
+    background_tasks: BackgroundTasks,
     _admin: None = Depends(verify_admin),
     db: Session = Depends(get_db),
 ):
-    """批量操作论文（删除或总结）。"""
+    """批量操作论文（删除 / 总结 / 重抓）。"""
     if not body.arxiv_ids:
         raise HTTPException(status_code=400, detail="arxiv_ids 不能为空")
 
@@ -475,3 +655,18 @@ async def admin_papers_batch_action(
             "message": f"已将 {count} 篇论文重置为待总结",
             "count": count,
         }
+
+    elif body.action == "recrawl":
+        job = create_job(
+            db,
+            "recrawl_batch",
+            owner="admin_recrawl",
+            payload={"arxiv_ids": body.arxiv_ids},
+        )
+        enqueue_job(background_tasks, job.id)
+        return {
+            "status": "queued",
+            "job_id": job.id,
+            "count": len(body.arxiv_ids),
+            "message": f"已将 {len(body.arxiv_ids)} 篇论文加入重抓队列",
+        }