Fixed chat file previews and stopped re-reading messages per file

.claude/CLAUDE.md

@@ -31,14 +31,16 @@ The Schuly **Microsoft plugin** is the only caller.
 - `entrance.http_client()` offers only `request`, `post`, `head` and `close`, takes a float timeout
   and reads every body into memory. Anything that must stay capped or be streamed uses
   `IMicrosoftHttpClients.download_client()` (httpx, HTTP/1.1, no redirects by default, no cookies kept,
-  because all accounts share it). Don't call `note.download` or `note.download_shared`: they hold the
+  because all accounts share it). `open_download` opens a pre-authorized link before the answer starts, so a refusal is a real status; it follows
+  redirects only through the host check, passes `Range` on and streams raw. `FileDownloads` keeps a link for 10 minutes, so a video's seeks cost no Graph calls. Don't call `note.download` or `note.download_shared`: they hold the
   whole file with no cap and no host check.
 - `except entrance.HTTPError` works alone, but it is a tuple: unpack it next to other types.
 - `TeamsFiles.download_url` stays: it serves the `shares/...` path chat files need, which OutNote's
   `download_url` can't. A folder is a 404 that says so; Graph's refusal at `/content` keeps its own
   status (403, 404, 415, 429).
 - One `OutNote` per account lives on its broker. A throttled call raises `GraphError` 429 after about
-  60 s, which `AccountRunner` answers as 429 with `Retry-After` (and any rotated token).
+  60 s, or at once with `retry=False`, which `AccountRunner` answers as 429 with `Retry-After` (and any rotated token) and logs as a warning with the masked route (`GraphCallTrail`). Keep ids and content out of that line. The chat file, thumbnail and message reads use `retry=False`: the caller waits, not a worker thread.
+- Graph throttles the single chat message read (`chats/{id}/messages/{id}`) far sooner than the list, so `ChatReader.file_ref` answers from what the list already said, then asks `note.chat_message` (Teams' own API, which falls back to that same read and may answer 429 for it), and when that refuses or lacks the file, the list.
 - `AccountRunner` strips `_native` from every result. No route calls OutNote's native reads today: chat
   messages stay on Graph's list because `ChatReader` and the callers need its attachments, mentions and
   reactions. `emails()` gets an explicit `$select` because the web's list shows `toRecipients`.

requirements.txt

@@ -2,7 +2,7 @@ fastapi==0.122.0
 uvicorn==0.38.0
 httpx==0.28.1
 python-multipart==0.0.20
-outnote==2.0.1
+outnote==2.2.0
 ms-entrance==2.0.0
 wreq==0.13.0
 python-docx==1.2.0

src/api/controllers/chats_controller.py

@@ -39,8 +39,8 @@ async def chat_hosted_content(self, chat_id: str, message_id: str, hosted_id: st
         return FileResponses.image(await self.mediator.send(GetChatHostedContentQuery(token, chat_id, message_id, hosted_id)))
 
     @router.get("/chats/{chat_id}/messages/{message_id}/files/{attachment_id}", tags=["Microsoft 365"], response_class=StreamingResponse, responses={200: {"content": {"application/octet-stream": {}}}})
-    async def chat_file(self, chat_id: str, message_id: str, attachment_id: str, token: str = Header(..., alias=HttpHeaders.REFRESH_TOKEN_HEADER)):
-        return FileResponses.stream(await self.mediator.send(GetChatFileQuery(token, chat_id, message_id, attachment_id)))
+    async def chat_file(self, chat_id: str, message_id: str, attachment_id: str, token: str = Header(..., alias=HttpHeaders.REFRESH_TOKEN_HEADER), byte_range: str | None = Header(None, alias="Range")):
+        return FileResponses.stream(await self.mediator.send(GetChatFileQuery(token, chat_id, message_id, attachment_id, byte_range)))
 
     @router.get("/chats/{chat_id}/messages/{message_id}/files/{attachment_id}/thumbnail", tags=["Microsoft 365"], response_class=Response, responses={200: {"content": {"image/jpeg": {}}}})
     async def chat_file_thumbnail(self, chat_id: str, message_id: str, attachment_id: str, token: str = Header(..., alias=HttpHeaders.REFRESH_TOKEN_HEADER), size: str = Query("medium")):

src/api/controllers/files_controller.py

@@ -78,9 +78,9 @@ async def drive_item_children(self, drive_id: str, item_id: str, response: Respo
         return await self.mediator.send(ListDriveItemChildrenQuery(token, drive_id, item_id, top))
 
     @router.get("/drives/{drive_id}/items/{item_id}/content", tags=["Teams files"], response_class=StreamingResponse, responses={200: {"content": {"application/octet-stream": {}}}})
-    async def drive_item_content(self, drive_id: str, item_id: str, token: str = Header(..., alias=HttpHeaders.REFRESH_TOKEN_HEADER)):
+    async def drive_item_content(self, drive_id: str, item_id: str, token: str = Header(..., alias=HttpHeaders.REFRESH_TOKEN_HEADER), byte_range: str | None = Header(None, alias="Range")):
         """The file's bytes as an attachment (<= 100 MB, no-store)."""
-        return FileResponses.stream(await self.mediator.send(GetDriveItemContentQuery(token, drive_id, item_id)))
+        return FileResponses.stream(await self.mediator.send(GetDriveItemContentQuery(token, drive_id, item_id, byte_range)))
 
     @router.get("/drives/{drive_id}/items/{item_id}/pdf", tags=["Files as PDF"], response_class=StreamingResponse, responses={200: {"content": {"application/pdf": {}}}})
     async def shared_drive_item_pdf(self, drive_id: str, item_id: str, token: str = Header(..., alias=HttpHeaders.REFRESH_TOKEN_HEADER)):

src/api/extensions/chats_extensions.py

@@ -12,7 +12,6 @@
 from src.application.queries.get_chat_photo_query import GetChatPhotoQuery, GetChatPhotoHandler
 from src.application.queries.list_chat_messages_query import ListChatMessagesQuery, ListChatMessagesHandler
 from src.application.queries.list_chats_query import ListChatsQuery, ListChatsHandler
-from src.application.services.chat_file_lookup import ChatFileLookup
 from src.application.services.chat_html_sanitizer import ChatHtmlSanitizer
 from src.infrastructure.teams.chat_reader import ChatReader
 from src.infrastructure.teams.chat_sender import ChatSender
@@ -23,7 +22,6 @@ class ChatsExtensions:
     @staticmethod
     def add_chats(services: ServiceCollection) -> ServiceCollection:
         services.add_singleton(ChatHtmlSanitizer)
-        services.add_singleton(ChatFileLookup)
         services.add_singleton(ITeamsChat, TeamsChatClient)
         services.add_singleton(IChatReader, ChatReader)
         services.add_singleton(IChatSender, ChatSender)

src/api/extensions/files_extensions.py

@@ -21,6 +21,7 @@
 from src.application.services.drive_file_text_reader import DriveFileTextReader
 from src.application.services.drive_item_checks import DriveItemChecks
 from src.infrastructure.files.file_text_extractor import FileTextExtractor
+from src.application.services.file_downloads import FileDownloads
 from src.infrastructure.graph.file_thumbnails import FileThumbnails
 from src.infrastructure.graph.pdf_converter import PdfConverter
 from src.infrastructure.graph.teams_files import TeamsFiles
@@ -34,6 +35,7 @@ def add_files(services: ServiceCollection) -> ServiceCollection:
         services.add_singleton(IFileTextExtractor, FileTextExtractor)
         services.add_singleton(ITeamsFiles, TeamsFiles)
         services.add_singleton(DriveItemChecks)
+        services.add_singleton(FileDownloads)
         services.add_singleton(DriveFileTextReader)
         services.add_singleton(AccountImageReader)
 

src/api/responses/file_responses.py

@@ -21,7 +21,13 @@ def attachment(cls, content: BinaryContent) -> Response:
 
     @classmethod
     def stream(cls, content: FileStream) -> StreamingResponse:
-        return StreamingResponse(content.chunks, media_type=content.content_type, headers=cls._download_headers(content.file_name))
+        """Always an attachment that can't run, whatever the type: a `<video>` plays it all the same and a navigation downloads it."""
+        headers = {**cls._download_headers(content.file_name), "Accept-Ranges": "bytes"}
+        if content.content_length is not None:
+            headers["Content-Length"] = str(content.content_length)
+        if content.content_range:
+            headers["Content-Range"] = content.content_range
+        return StreamingResponse(content.chunks, status_code=content.status_code, media_type=content.content_type, headers=headers)
 
     @classmethod
     def image(cls, content: BinaryContent) -> Response:

src/application/abstractions/i_chat_reader.py

@@ -8,4 +8,7 @@ class IChatReader(ABC):
     """Graph chat messages made ready for the web."""
 
     @abstractmethod
-    def enrich(self, note: IGraphClient, broker: IAccountBroker, messages: list[dict]) -> list[dict]: ...
+    def enrich(self, note: IGraphClient, broker: IAccountBroker, messages: list[dict], chat_id: str | None = None) -> list[dict]: ...
+
+    @abstractmethod
+    def file_ref(self, note: IGraphClient, broker: IAccountBroker, chat_id: str, message_id: str, attachment_id: str) -> dict | None: ...

src/application/abstractions/i_file_thumbnails.py

@@ -24,6 +24,9 @@ def image_type(self, head: bytes) -> str | None: ...
     @abstractmethod
     def safe_graph_id(self, value) -> bool: ...
 
+    @abstractmethod
+    def remember_item(self, broker: IAccountBroker, url: str, item: dict) -> None: ...
+
     @abstractmethod
     def chat_file_thumbnail(self, note: IGraphClient, broker: IAccountBroker, url: str, size: str) -> bytes | None: ...
 

src/application/abstractions/i_graph_client.py

@@ -5,7 +5,7 @@ class IGraphClient(ABC):
     """What the application asks of a per-account Graph client. OutNote is registered as one by the infrastructure."""
 
     @abstractmethod
-    def get(self, path, params=None, client=None, raw=False): ...
+    def get(self, path, params=None, client=None, raw=False, retry=True): ...
 
     @abstractmethod
     def list(self, path, params=None, limit=None, client=None): ...

src/application/abstractions/i_microsoft_http_clients.py

@@ -1,5 +1,7 @@
 from abc import ABC, abstractmethod
-from collections.abc import Iterator
+from collections.abc import Callable
+
+from src.domain.files.file_stream import FileStream
 
 
 class IMicrosoftHttpClients(ABC):
@@ -9,7 +11,7 @@ class IMicrosoftHttpClients(ABC):
     def download_client(self): ...
 
     @abstractmethod
-    def stream_download(self, url: str, limit: int, follow_redirects: bool = True) -> Iterator[bytes]: ...
+    def open_download(self, url: str, limit: int, allowed: Callable[[str], bool], content_type: str, file_name: str, byte_range: str | None = None) -> FileStream: ...
 
     @abstractmethod
     async def post_form(self, url: str, data: dict, client=None) -> dict: ...

src/application/abstractions/i_teams_files.py

@@ -22,6 +22,15 @@ def fetch_bytes(self, url: str, cap: int) -> bytes: ...
     @abstractmethod
     def download_link(self, note: IGraphClient, path: str, item: dict) -> str: ...
 
+    @abstractmethod
+    def cached_download(self, broker: IAccountBroker, key: tuple) -> dict | None: ...
+
+    @abstractmethod
+    def remember_download(self, broker: IAccountBroker, key: tuple, target: dict) -> None: ...
+
+    @abstractmethod
+    def forget_download(self, broker: IAccountBroker, key: tuple) -> None: ...
+
     @abstractmethod
     def item_base(self, drive_id: str | None, item_id: str) -> str: ...
 

src/application/queries/get_chat_file_pdf_query.py

@@ -3,9 +3,9 @@
 from mediatorx import IQuery, IQueryHandler
 
 from src.application.abstractions.i_account_runner import IAccountRunner
+from src.application.abstractions.i_chat_reader import IChatReader
 from src.application.abstractions.i_file_thumbnails import IFileThumbnails
 from src.application.abstractions.i_pdf_converter import IPdfConverter
-from src.application.services.chat_file_lookup import ChatFileLookup
 from src.application.services.drive_item_checks import DriveItemChecks
 from src.application.services.graph_paths import GraphPaths
 from src.domain.errors.service_error import ServiceError
@@ -22,9 +22,9 @@ class GetChatFilePdfQuery(IQuery[SpooledPdf]):
 
 
 class GetChatFilePdfHandler(IQueryHandler[GetChatFilePdfQuery, SpooledPdf]):
-    def __init__(self, accounts: IAccountRunner, lookup: ChatFileLookup, thumbnails: IFileThumbnails, pdf: IPdfConverter, checks: DriveItemChecks):
+    def __init__(self, accounts: IAccountRunner, reader: IChatReader, thumbnails: IFileThumbnails, pdf: IPdfConverter, checks: DriveItemChecks):
         self._accounts = accounts
-        self._lookup = lookup
+        self._reader = reader
         self._thumbnails = thumbnails
         self._pdf = pdf
         self._checks = checks
@@ -35,7 +35,7 @@ async def handle(self, query: GetChatFilePdfQuery) -> SpooledPdf:
         GraphIds.check(query.attachment_id, GraphIds.ATTACHMENT_ID_PATTERN, "attachment_id")
 
         def convert(note, broker):
-            ref = self._lookup.find(note, query.chat_id, query.message_id, query.attachment_id)
+            ref = self._reader.file_ref(note, broker, query.chat_id, query.message_id, query.attachment_id)
             if ref is None:
                 raise ServiceError(404, "File not found in the message.")
             if not self._thumbnails.is_share_url(ref.get("contentUrl")):

src/application/queries/get_chat_file_query.py

@@ -2,12 +2,11 @@
 
 from mediatorx import IQuery, IQueryHandler
 
-from src.application.abstractions.i_account_runner import IAccountRunner
-from src.application.abstractions.i_microsoft_http_clients import IMicrosoftHttpClients
+from src.application.abstractions.i_chat_reader import IChatReader
+from src.application.abstractions.i_file_thumbnails import IFileThumbnails
 from src.application.abstractions.i_teams_files import ITeamsFiles
-from src.application.services.chat_file_lookup import ChatFileLookup
-from src.application.services.content_names import ContentNames
 from src.application.services.drive_item_checks import DriveItemChecks
+from src.application.services.file_downloads import FileDownloads
 from src.application.services.graph_paths import GraphPaths
 from src.domain.errors.service_error import ServiceError
 from src.domain.files.file_limits import FileLimits
@@ -21,34 +20,35 @@ class GetChatFileQuery(IQuery[FileStream]):
     chat_id: str
     message_id: str
     attachment_id: str
+    byte_range: str | None = None
 
 
 class GetChatFileHandler(IQueryHandler[GetChatFileQuery, FileStream]):
-    def __init__(self, accounts: IAccountRunner, lookup: ChatFileLookup, files: ITeamsFiles, http: IMicrosoftHttpClients):
-        self._accounts = accounts
-        self._lookup = lookup
+    def __init__(self, reader: IChatReader, thumbnails: IFileThumbnails, files: ITeamsFiles, downloads: FileDownloads):
+        self._reader = reader
+        self._thumbnails = thumbnails
         self._files = files
-        self._http = http
+        self._downloads = downloads
 
     async def handle(self, query: GetChatFileQuery) -> FileStream:
         GraphIds.check(query.chat_id, GraphIds.CHAT_ID_PATTERN, "chat_id")
         GraphIds.check(query.message_id, GraphIds.MESSAGE_ID_PATTERN, "message_id")
         GraphIds.check(query.attachment_id, GraphIds.ATTACHMENT_ID_PATTERN, "attachment_id")
 
-        def resolve(note, broker):
-            ref = self._lookup.find(note, query.chat_id, query.message_id, query.attachment_id)
+        def read(note, broker):
+            ref = self._reader.file_ref(note, broker, query.chat_id, query.message_id, query.attachment_id)
             if ref is None:
                 raise ServiceError(404, "File not found in the message.")
-            item = note.get(GraphPaths.drive_item_path(ref["contentUrl"], "name,size,file,folder,@microsoft.graph.downloadUrl"))
+            path = GraphPaths.share_item_path(ref["contentUrl"])
+            # Throttled, this answers at once: the caller waits, rather than a worker thread.
+            item = note.get(path, retry=False)
             if item.get("folder") is not None:
                 raise ServiceError(404, DriveItemChecks.FOLDER_DETAIL)
             if (item.get("size") or 0) > FileLimits.MAX_CHAT_FILE_BYTES:
                 raise ServiceError(413, "The file is larger than 100 MB.")
-            url = self._files.download_link(note, f"shares/{GraphPaths.share_id(ref['contentUrl'])}/driveItem", item)
-            if not self._files.is_download_url(url):
-                raise ServiceError(502, "The file has an unexpected download location.")
-            return item, url
-
-        item, url = await self._accounts.run(query.refresh_token, resolve)
-        # The download URL is pre-authorized and short-lived; the bytes stream through without being held.
-        return FileStream(self._http.stream_download(url, FileLimits.MAX_CHAT_FILE_BYTES, follow_redirects=False), ContentNames.safe_content_type((item.get("file") or {}).get("mimeType")), ContentNames.safe_name(item.get("name")))
+            self._thumbnails.remember_item(broker, ref["contentUrl"], item)
+            return item, self._files.download_link(note, path, item)
+
+        key = ("chat", query.chat_id, query.message_id, query.attachment_id)
+        # The link is pre-authorized and short-lived; the bytes stream through without being held.
+        return await self._downloads.open(query.refresh_token, key, read, FileLimits.MAX_CHAT_FILE_BYTES, query.byte_range)

src/application/queries/get_chat_file_thumbnail_query.py

@@ -2,9 +2,9 @@
 
 from mediatorx import IQuery, IQueryHandler
 
+from src.application.abstractions.i_chat_reader import IChatReader
 from src.application.abstractions.i_file_thumbnails import IFileThumbnails
 from src.application.services.account_image_reader import AccountImageReader
-from src.application.services.chat_file_lookup import ChatFileLookup
 from src.domain.files.binary_content import BinaryContent
 from src.domain.validation.graph_ids import GraphIds
 
@@ -19,9 +19,9 @@ class GetChatFileThumbnailQuery(IQuery[BinaryContent]):
 
 
 class GetChatFileThumbnailHandler(IQueryHandler[GetChatFileThumbnailQuery, BinaryContent]):
-    def __init__(self, images: AccountImageReader, lookup: ChatFileLookup, thumbnails: IFileThumbnails):
+    def __init__(self, images: AccountImageReader, reader: IChatReader, thumbnails: IFileThumbnails):
         self._images = images
-        self._lookup = lookup
+        self._reader = reader
         self._thumbnails = thumbnails
 
     async def handle(self, query: GetChatFileThumbnailQuery) -> BinaryContent:
@@ -31,7 +31,7 @@ async def handle(self, query: GetChatFileThumbnailQuery) -> BinaryContent:
         self._images.check_thumbnail_size(query.size)
 
         def fetch(note, broker):
-            ref = self._lookup.find(note, query.chat_id, query.message_id, query.attachment_id)
+            ref = self._reader.file_ref(note, broker, query.chat_id, query.message_id, query.attachment_id)
             if ref is None or not self._thumbnails.is_share_url(ref["contentUrl"]):
                 return None
             return self._thumbnails.chat_file_thumbnail(note, broker, ref["contentUrl"], query.size)