diff --git a/app/controllers/file_admin_controller.py b/app/controllers/file_admin_controller.py
index cb0bae5..4caba02 100644
--- a/app/controllers/file_admin_controller.py
+++ b/app/controllers/file_admin_controller.py
@@ -43,7 +43,8 @@ class BatchDeleteResult(BaseModel):
     "",
     response_model=FileListResponse,
     summary="列出已上传文件（需鉴权）",
-    description="供文件浏览页使用；字段同公开 /api/files，但要求 docs Basic Auth。支持分页。",
+    description="供文件浏览页使用；字段同公开 /api/files，但要求 docs Basic Auth。支持分页。"
+    "返回前会触发磁盘扫描（频率限制 3s），把 SFTP 上传但未入库的文件补录进来。",
 )
 def list_files(
     limit: int = Query(100, ge=1, le=10000, description="每页条数，1-10000"),
@@ -51,6 +52,7 @@ def list_files(
     service: UploadService = Depends(_service),
     _: str = Depends(require_docs_auth),
 ) -> FileListResponse:
+    service.scan_sftp_files()  # 频率限制内置：3s 内重复只返回 DB 缓存
     total, items = service.list_files(limit=limit, offset=offset)
     return FileListResponse(total=total, items=items)
 
diff --git a/app/dao/uploaded_file_dao.py b/app/dao/uploaded_file_dao.py
index 3df2581..ec13292 100644
--- a/app/dao/uploaded_file_dao.py
+++ b/app/dao/uploaded_file_dao.py
@@ -29,6 +29,11 @@ class UploadedFileDAO:
         """返回数据库中文件总条数。"""
         return self.db.scalar(select(func.count()).select_from(UploadedFile)) or 0
 
+    def list_storage_paths(self) -> set[str]:
+        """返回所有已记录的 storage_path 集合（供磁盘扫描比对，识别未入库文件）。"""
+        stmt = select(UploadedFile.storage_path)
+        return {row for row in self.db.scalars(stmt).all()}
+
     def list(self, limit: int = 100, offset: int = 0) -> list[UploadedFile]:
         stmt = (
             select(UploadedFile)
diff --git a/app/services/upload_service.py b/app/services/upload_service.py
index b405ae6..0c448be 100644
--- a/app/services/upload_service.py
+++ b/app/services/upload_service.py
@@ -10,6 +10,7 @@ from __future__ import annotations
 import hashlib
 import logging
 import os
+import time
 import uuid
 from collections.abc import Iterator
 from datetime import datetime, timezone
@@ -24,6 +25,10 @@ from ..schemas.file import FileUploadResponse, UploadedFileOut
 
 logger = logging.getLogger("zikai.upload")
 
+# 磁盘扫描频率限制：两次扫描至少间隔此秒数，否则跳过（只返回 DB 缓存）。
+_SCAN_MIN_INTERVAL = 3.0
+_last_scan_time: float = 0.0
+
 
 def hash_stream(chunks: Iterator[bytes]) -> tuple[int, str]:
     """对一段字节块序列流式计算 (size, sha256)。供多条上传路径复用。"""
@@ -69,6 +74,72 @@ class UploadService:
 
     # ---------------- 查询 ----------------
 
+    def scan_sftp_files(self, force: bool = False) -> int:
+        """扫描 uploads/ 目录，把磁盘上有但 DB 未记录的文件补录入库。
+
+        SFTP 上传的文件直接落盘到 upload_root（chroot 根），不经过 HTTP 路径，
+        因此没有 DB 记录。本方法遍历磁盘文件，与 DB 已有 storage_path 比对，
+        为缺失项创建记录（source="sftp"，计算 size + sha256）。
+
+        频率限制：两次扫描间隔 < _SCAN_MIN_INTERVAL（3s）则跳过，force=True 强制扫。
+        返回本次新补录的条数（跳过时返回 0）。
+        """
+        global _last_scan_time
+        now = time.monotonic()
+        if not force and (now - _last_scan_time) < _SCAN_MIN_INTERVAL:
+            return 0
+        _last_scan_time = now
+
+        known = self.dao.list_storage_paths()
+        new_count = 0
+        # 排除分片会话暂存目录与 .part 残品
+        skip_dirs = {".work"}
+        for abs_path in self.upload_root.rglob("*"):
+            if not abs_path.is_file():
+                continue
+            if abs_path.suffix == ".part":
+                continue
+            # 跳过 .work 目录下的任何文件
+            rel = abs_path.relative_to(self.upload_root)
+            if rel.parts and rel.parts[0] in skip_dirs:
+                continue
+            storage_path = str(rel).replace("\\", "/")
+            if storage_path in known:
+                continue
+            # 磁盘有、DB 无：补录
+            try:
+                size = abs_path.stat().st_size
+                sha256 = self._hash_file(abs_path)
+            except Exception as exc:
+                logger.warning("扫描文件失败 path=%s: %s", abs_path, exc)
+                continue
+            entity = UploadedFile(
+                storage_path=storage_path,
+                original_filename=abs_path.name,
+                content_type="",
+                size_bytes=size,
+                sha256=sha256,
+                source="sftp",
+                uploaded_by="sftp",
+            )
+            try:
+                self.dao.create(entity)
+                new_count += 1
+                known.add(storage_path)
+            except Exception as exc:
+                logger.warning("补录 SFTP 文件失败 path=%s: %s", storage_path, exc)
+        if new_count:
+            logger.info("SFTP 文件扫描补录 %d 个", new_count)
+        return new_count
+
+    def _hash_file(self, path: Path) -> str:
+        """流式计算文件 sha256（避免大文件一次性读入内存）。"""
+        h = hashlib.sha256()
+        with path.open("rb") as f:
+            while chunk := f.read(self.chunk_bytes):
+                h.update(chunk)
+        return h.hexdigest()
+
     def list_files(self, limit: int = 100, offset: int = 0) -> tuple[int, list[UploadedFileOut]]:
         total = self.dao.count()
         rows = self.dao.list(limit=limit, offset=offset)