update: 数据资源文件上传

backend/src/backend/resources.py

@@ -55,15 +55,11 @@ def resource_payload(
     resource: DataResources,
     storage_object: StorageObjects,
 ) -> dict[str, Any]:
-    # ``object_key`` is stored as ``{ws_id}/{user_id}/.resources/{file_name}``
+    # ``object_key`` now follows ``{ws_id}/{user_id}/{target_path}/{file_name}``
-    # (see ``backend/services/storage.py:create_upload_record``). Two
+    # (target_path may be empty). Legacy objects still live under
-    # Jupyter-facing path views:
+    # ``.resources/{file_name}`` and must remain readable. Both shapes share
-    #   - ``jupyter_accessible_path``: per-user Jupyter-relative path. The
+    # the same derivation: strip the workspace/user prefix and use the rest
-    #     Jupyter root_dir is already ``workspace/{ws_id}/{user_id}/``, so
+    # as the Jupyter-relative path.
-    #     we return the tail ``.resources/{file_name}`` for copy-paste use.
-    #   - ``absolute_path``: full filesystem path inside the Jupyter
-    #     container (the rclone mount root). Shape:
-    #     ``{workspaces_root}/{ws_id}/{user_id}/.resources/{file_name}``.
     workspace_prefix = f"{resource.workspace_id}/"
     user_prefix = f"{resource.owner_user_id}/"
     jupyter_accessible_path = ""
@@ -73,7 +69,7 @@ def resource_payload(
     ):
         remainder = storage_object.object_key[len(workspace_prefix):]
         if remainder.startswith(user_prefix):
-            tail = remainder[len(user_prefix):]  # ``.resources/{file_name}``
+            tail = remainder[len(user_prefix):]
             jupyter_accessible_path = tail
             absolute_path = (
                 workspaces_root()
@@ -141,11 +137,7 @@ async def resource_jupyter_relative_path(
         raise HTTPException(422, "script_path must be a clean Jupyter-relative POSIX path")
 
     resource_path = resource_payload(resource, storage_object)["jupyter_accessible_path"]
-    if not resource_path.startswith(".resources/"):
+    # Both legacy ``.resources/`` files and new flat/nested paths are valid.
-        raise HTTPException(
-            status.HTTP_409_CONFLICT,
-            "resource is not stored under .resources/",
-        )
     relative = compute_jupyter_relative_path(script_path, resource_path)
     return {
         "request_id": context.request_id,
@@ -176,6 +168,7 @@ async def create_resource_upload(
             expected_size_bytes=payload.expected_size_bytes,
             expected_hash=payload.expected_hash,
             idempotency_key=idempotency_key,
+            target_path=payload.target_path,
         ),
         session,
         request,

backend/src/backend/schemas.py

@@ -9,6 +9,18 @@ class CreateResourceUploadRequest(StrictModel):
     content_type: str = Field(min_length=1, max_length=255)
     expected_size_bytes: int = Field(ge=0, le=100 * 1024 * 1024)
     expected_hash: str | None = Field(default=None, min_length=64, max_length=64)
+    target_path: str = Field(default="", max_length=1024)
+
+    @field_validator("target_path")
+    @classmethod
+    def validate_target_path(cls, value: str) -> str:
+        # POSIX 相对路径，不能含 ..、绝对前缀、控制字符
+        if value and (value.startswith("/") or "\\" in value
+                      or any(seg == ".." for seg in value.split("/"))
+                      or any(ord(c) < 0x20 or ord(c) == 0x7F for c in value)):
+            raise ValueError("target_path must be a clean relative POSIX path")
+        # 去前导 /；允许尾部 /
+        return value.strip("/")
 
     @field_validator("expected_hash")
     @classmethod

backend/src/backend/services/storage.py

@@ -82,6 +82,14 @@ def _safe_file_name(value: str) -> str:
     return name
 
 
+def _safe_path_segment(segment: str) -> str:
+    """Single path segment: keep alnum / dash / underscore / dot; collapse the rest."""
+    cleaned = "".join(
+        c if c.isalnum() or c in "-_." else "_" for c in segment
+    ).strip("._")
+    return cleaned or "untitled"
+
+
 def _utcnow_naive() -> Any:
     from datetime import datetime, UTC
     return datetime.now(UTC).replace(tzinfo=None)
@@ -92,6 +100,41 @@ def _hash_bytes(value: str) -> bytes:
     return _h.sha256(value.encode("utf-8")).digest()
 
 
+async def _resolve_unique_object_key(
+    session: AsyncSession,
+    object_key: str,
+) -> str:
+    """Append a ULID suffix when an available object already occupies ``object_key``.
+
+    Keeps the original file name in metadata; only the on-disk key gets
+    the disambiguation suffix so repeated uploads of the same name never
+    overwrite existing bytes.
+    """
+    key = object_key
+    for _ in range(5):
+        key_hash = _hash_bytes(key)
+        existing = await session.scalar(
+            select(StorageObjects).where(
+                StorageObjects.object_key_hash == key_hash,
+                StorageObjects.object_status == "available",
+            )
+        )
+        if existing is None:
+            return key
+        suffix = new_ulid().lower()
+        parts = key.rsplit("/", 1)
+        dir_part = parts[0] if len(parts) > 1 else ""
+        name = parts[-1]
+        # Preserve extension if present (ignore leading dot / hidden files).
+        if "." in name[1:]:
+            stem, ext = name.rsplit(".", 1)
+            new_name = f"{stem}-{suffix}.{ext}"
+        else:
+            new_name = f"{name}-{suffix}"
+        key = f"{dir_part}/{new_name}" if dir_part else new_name
+    return key
+
+
 def _build_storage_object(
     *,
     upload: UploadSessions,
@@ -192,15 +235,25 @@ async def create_upload_record(
             payload.usage_type,
             workspace_artifact_bucket=workspace.artifact_bucket,
         )
-        # Use the sanitized file_name as the on-disk basename (no ULID). The
-        # extension is still derived for StorageObjects.file_extension.
         safe_name = _safe_file_name(payload.file_name)
-        file_extension = PurePosixPath(safe_name).suffix.lower()
+        # Defense-in-depth: schemas already validate target_path, but the
-        # Use the sanitized file_name as the on-disk basename (no ULID).
+        # helper is called directly from some callers so re-validate here.
-        # DataResources files live under a per-user ``.resources`` subdir
+        target_path = (payload.target_path or "").strip("/")
-        # so they don't collide with workspace-tree scripts/directories
+        if target_path and (target_path.startswith("/") or "\\" in target_path
-        # the user authors directly in Jupyter.
+                            or any(seg == ".." for seg in target_path.split("/"))
-        object_key = f"{payload.workspace_id}/{payload.user_id}/.resources/{safe_name}"
+                            or any(ord(c) < 0x20 or ord(c) == 0x7F for c in target_path)):
+            raise HTTPException(
+                status.HTTP_422_UNPROCESSABLE_ENTITY,
+                "target_path must be a clean relative POSIX path",
+            )
+        clean_target = "/".join(
+            _safe_path_segment(seg) for seg in target_path.split("/") if seg
+        )
+        if clean_target:
+            object_key = f"{payload.workspace_id}/{payload.user_id}/{clean_target}/{safe_name}"
+        else:
+            object_key = f"{payload.workspace_id}/{payload.user_id}/{safe_name}"
+        object_key = await _resolve_unique_object_key(session, object_key)
         upload = UploadSessions(
             upload_id=new_ulid(),
             workspace_id=payload.workspace_id,