fix: keep embedded metadata and group formats on directory upload

The filepath extractor was merged on the right, so a parent folder's name
overrode the title inside the file. Grouping also split a book's formats when
its own folder was the upload root.

backend/src/chitai/services/book.py

@@ -152,15 +152,24 @@ class BookService(SQLAlchemyAsyncRepositoryService[Book]):
         if not data.files:
             raise ValueError("Must upload at least one file")
 
-        # Group book files if they are within the same nested directory
+        # Group the files that make up a single book.
-        books = defaultdict(list)
+        #
+        # `file.filename` carries the browser's webkitRelativePath, whose first
+        # component is whichever folder the user picked. Depth is therefore relative
+        # to the selection, not to the library: picking a book's own folder and
+        # picking the shelf above it submit the same files at different depths.
+        #
+        # Inside a subdirectory the parent directory is unambiguously the book. Sitting
+        # directly in the selected folder it is genuinely ambiguous — "Fluent Python.epub"
+        # alongside "Fluent Python.pdf" is one book in two formats, while "a.epub"
+        # alongside "b.epub" is two books — so the filename stem breaks the tie.
+        books: dict[tuple[Path, str], list[UploadFile]] = defaultdict(list)
         for file in data.files:
             filepath = Path(file.filename)
-            # Books within the root directory should be treated as separate books
             if len(filepath.parent.parts) > 1:
-                books[filepath.parent].append(file)
+                books[(filepath.parent, "")].append(file)
             else:
-                books[filepath].append(file)
+                books[(filepath.parent, filepath.stem)].append(file)
 
         return [
             await self.create_book(

backend/src/chitai/services/metadata_extractor.py

@@ -67,8 +67,11 @@ class Extractor:
         for file in files:
             metadata = FilenameExtractor.extract_metadata(file) | metadata
 
-        # Get metadata from filepath
+        # Get metadata from filepath. Kept on the left so that anything the file
-        metadata = metadata | FilepathExtractor.extract_metadata(files[0], root_path)
+        # itself declared outranks a guess made from its directory names — a folder
+        # called "Fluent Python - Luciano Ramalho" must not overwrite the title the
+        # EPUB already carries.
+        metadata = FilepathExtractor.extract_metadata(files[0], root_path) | metadata
 
         # format the title
         if metadata.get('title', None):

backend/tests/integration/test_book.py

@@ -369,6 +369,61 @@ async def test_create_multiple_books_from_directory(
     assert len(data.get("items") or data.get("data")) >= 1
 
 
+async def test_create_books_from_parent_directory_keeps_embedded_title(
+    authenticated_client: AsyncClient,
+) -> None:
+    """A folder name in the upload path must not override the file's own metadata.
+
+    The browser sends webkitRelativePath, so picking the shelf above a book's folder
+    submits one more path component than picking the folder itself. That extra level
+    used to make the directory name win over the title inside the EPUB.
+    """
+    source = Path("tests/data_files/Metamorphosis - Franz Kafka.epub")
+    files = [
+        (
+            "files",
+            (
+                "Shelf/Metamorphosis - Franz Kafka/Metamorphosis.epub",
+                source.read_bytes(),
+                "application/epub+zip",
+            ),
+        )
+    ]
+
+    response = await authenticated_client.post(
+        "/books/fromFiles?library_id=1", files=files, data={"library_id": 1}
+    )
+
+    assert response.status_code == 201
+
+    books = response.json()["items"]
+    assert len(books) == 1
+    assert books[0]["title"] == "Metamorphosis"
+
+
+async def test_create_books_groups_formats_within_one_folder(
+    authenticated_client: AsyncClient,
+) -> None:
+    """Picking a book's own folder yields one book with both formats, not two books."""
+    epub = Path("tests/data_files/Metamorphosis - Franz Kafka.epub").read_bytes()
+    pdf = Path("tests/data_files/Calculus Made Easy - Silvanus Thompson.pdf").read_bytes()
+
+    files = [
+        ("files", ("Metamorphosis/Metamorphosis.epub", epub, "application/epub+zip")),
+        ("files", ("Metamorphosis/Metamorphosis.pdf", pdf, "application/pdf")),
+    ]
+
+    response = await authenticated_client.post(
+        "/books/fromFiles?library_id=1", files=files, data={"library_id": 1}
+    )
+
+    assert response.status_code == 201
+
+    books = response.json()["items"]
+    assert len(books) == 1
+    assert len(books[0]["files"]) == 2
+
+
 # NOTE: the multi-book ZIP download is covered at the service level, in
 # tests/unit/test_services/test_book_service.py. Driving `/books/download` through
 # AsyncTestClient hangs in fixture teardown: it is the only `Stream` endpoint in the