feat: Office-Extraktion (docx/xlsx/pptx) fuer files-, mail- und notes-MCP

officeutil.py (analog pdfutil): server-seitige Text-Extraktion, damit das
LLM Inhalte als Text bekommt statt Base64-Blob. Verkabelt in files.read_file,
mail.read_attachment (und notes.read_resource, bereits in e34711d).
Aeltere Binaerformate (.doc/.xls/.ppt) und .rtf bewusst nicht behandelt --
is_office() False -> Fallback auf Rohdaten. requirements-extra: python-pptx.

War bereits deployed und lief (FileTypes-Tests 14 passed), lag aber
uncommittet im Arbeitsverzeichnis; ohne officeutil.py waere ein frischer
Clone seit e34711d kaputt gewesen.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XhHNJKk3RsHgUsJ27WQSom

mail/server.py

@@ -25,6 +25,7 @@ from starlette.routing import Mount
 sys.path.insert(0, os.path.dirname(os.path.dirname(__file__)))
 from common import get_current_user, OAUTH_ROUTES, BearerAuthMiddleware
 from pdfutil import pdf_to_content
+from officeutil import office_to_content, is_office
 
 from common import load_config as _lc
 _cfg = _lc()
@@ -343,7 +344,7 @@ def read_attachment(
     key: Annotated[str, Field(description="Message key from search results")],
     attachment_index: Annotated[int, Field(description="Attachment number from the read_mail attachment list (1-based)")],
 ) -> list[TextContent | ImageContent | EmbeddedResource]:
-    """Read an email attachment. Images shown inline, PDFs as extracted text, text directly, other documents as binary. Get the index from read_mail."""
+    """Read an email attachment. Images shown inline, PDFs and Office documents (docx, xlsx, pptx) as extracted text, text directly, other formats as binary. Get the index from read_mail."""
     _trim_memory()
     user = get_current_user()
     if not user:
@@ -369,6 +370,8 @@ def read_attachment(
         return [TextContent(type="text", text=payload.decode("utf-8", errors="replace")[:100000])]
     if mime == "application/pdf" or att["filename"].lower().endswith(".pdf"):
         return pdf_to_content(payload, att["filename"], mime, uri=f"mail://attachment/{att['filename']}")
+    if is_office(mime, att["filename"]):
+        return office_to_content(payload, att["filename"], mime, uri=f"mail://attachment/{att['filename']}")
     return [EmbeddedResource(type="resource", resource=BlobResourceContents(
         uri=f"mail://attachment/{att['filename']}",
         blob=base64.b64encode(payload).decode(), mimeType=mime))]