feat(downloads): local file upload ingest — media library backend (E7 S5a)

Add the non-yt-dlp ingest path so a user-uploaded file joins the same download
library and inherits its function set. Any file type is accepted: media
(video/audio) gets the full set (metadata, share, play/edit), a generic file
(pdf/zip/docx/…) is download-only.

- MediaAsset.media_kind ("video"|"audio"|"file") — the stored discriminator that
  gates library actions. Set at every finalize (download probe, editor carries the
  source's kind, upload probe); consolidates the ad-hoc audio-detect that lived only
  in links.py into one invariant (downloads.mediakind.derive). Migration 0076 adds
  the column + backfills existing rows.
- downloads/upload.py: ingest_upload probes the file (ffprobe → codecs/dims/duration),
  places it in a per-user .uploads tree, and creates an already-ready asset + done job
  (the same cache-hit shortcut enqueue uses). Uploads never expire (irreplaceable).
- POST /api/downloads/upload: streaming multipart with the per-file cap
  (download_upload_max_bytes, default 500 MiB) enforced on actual bytes read; quota
  charges the known size up front (quota.check_upload).
- Security: a generic file is always served force-download (attachment +
  application/octet-stream) on both the in-app and public endpoints, so uploaded
  HTML/SVG can't execute inline. Edit + redownload reject non-media / upload sources.
- python-multipart added (first multipart endpoint in the app).

backend/alembic/versions/0076_media_asset_kind.py

@@ -0,0 +1,41 @@
+"""media_assets.media_kind (E7 S5a: the video|audio|file discriminator for the media library)
+
+Adds a stored media-kind to each asset so the library can gate its actions on it (play/edit for
+media, download-only for a generic uploaded file) instead of re-deriving from codecs in prose. The
+column is the single home of the invariant `downloads.mediakind.derive` computes at finalize.
+
+Backfill for existing rows: a stored video codec -> 'video', an audio-only codec -> 'audio', an
+editor derivative (source_kind='edit', which carries no codecs but is always a video clip) ->
+'video'. Rows still pending/errored (no codecs, not an edit) stay NULL — kind is set when they
+finalize. Empty `media_kind` never reaches the UI: only `ready`/`done` rows render actions, and
+those are all backfilled here.
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "0076_media_asset_kind"
+down_revision: Union[str, None] = "0075_playlist_sync_baseline"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column("media_assets", sa.Column("media_kind", sa.String(length=16), nullable=True))
+    op.execute(
+        """
+        UPDATE media_assets
+           SET media_kind = CASE
+               WHEN vcodec IS NOT NULL THEN 'video'
+               WHEN acodec IS NOT NULL THEN 'audio'
+               WHEN source_kind = 'edit' THEN 'video'
+               ELSE NULL
+           END
+        """
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("media_assets", "media_kind")

backend/app/config.py

@@ -208,6 +208,10 @@ class Settings(BaseSettings):
     download_default_max_bytes: int = 5_368_709_120  # 5 GiB
     download_default_max_concurrent: int = 2
     download_default_max_jobs: int = 50
+    # Per-file cap for local uploads (E7). Unlike a download (whose size is unknown until the worker
+    # finishes), an upload's size is known at ingest, so this bounds a single file directly; the
+    # user's overall byte footprint is still governed by download_default_max_bytes.
+    download_upload_max_bytes: int = 524_288_000  # 500 MiB
     # Retention: an asset expires this many days after its last access; the GC job then
     # deletes it (once no job references it) after an additional grace window.
     download_retention_days: int = 30

backend/app/downloads/links.py

@@ -80,13 +80,13 @@ def public_meta(
 
     Resolves the owner's per-download display overrides (custom title/channel) over the asset's
     auto-extracted values, and surfaces the source + any extra reference links as clickable URLs."""
-    from app.downloads import service
+    from app.downloads import mediakind, service
 
     source_url = service.reference_url(job, asset)
-    # Audio-only assets (mp3 rips, extracted audio) carry a codec but no video stream. WebKit (all
-    # iOS browsers) refuses to play them inside a <video> element, so the watch page must render a
-    # native <audio> instead — flag it here from the recorded codecs (a real video always has vcodec).
-    is_audio = asset.vcodec is None and asset.acodec is not None
+    # Audio-only assets (mp3 rips, extracted audio) can't play inside a <video> element on WebKit
+    # (all iOS browsers), so the watch page renders a native <audio> instead. A generic uploaded
+    # file (media_kind='file') has no player at all — the watch page shows file info + a download.
+    # Both branches read the stored media_kind (the single source), not a re-derived codec check.
     return {
         "title": (job and job.display_name) or asset.title,
         "uploader": (job and job.display_uploader) or asset.uploader,
@@ -96,7 +96,10 @@ def public_meta(
         "duration_s": asset.duration_s,
         "width": asset.width,
         "height": asset.height,
-        "audio": is_audio,
+        "audio": asset.media_kind == mediakind.AUDIO,
+        "media_kind": asset.media_kind,
+        "size_bytes": asset.size_bytes,
+        "container": asset.container,
         "allow_download": link.allow_download,
         "file_url": file_url,
         "poster_url": poster_url,

backend/app/downloads/mediakind.py

@@ -0,0 +1,23 @@
+"""The media-kind discriminator for a library asset — the single home of one invariant.
+
+`MediaAsset.media_kind` is `"video" | "audio" | "file"`. It decides which library actions apply:
+media (video/audio) gets play/edit/share; a generic uploaded file gets download + share only, and
+is never rendered inline (see the force-download serving in the file/watch routes). Every finalize
+path computes it HERE — the download worker + the upload ingest via `derive`, the editor by
+carrying the source's kind over — so the rule lives in code, not in a comment that can drift (an
+earlier ad-hoc `vcodec is None and acodec is not None` audio check lived only in links.py).
+"""
+
+VIDEO = "video"
+AUDIO = "audio"
+FILE = "file"
+
+
+def derive(vcodec: str | None, acodec: str | None) -> str:
+    """Media kind from the probed codecs: a video stream -> video, audio-only -> audio, neither
+    (a non-media upload: pdf/zip/docx/…) -> file."""
+    if vcodec:
+        return VIDEO
+    if acodec:
+        return AUDIO
+    return FILE

backend/app/downloads/pipeline.py

@@ -15,7 +15,7 @@ from app.config import settings
 from app.db import SessionLocal
 from app.downloads import edit as editmod
 from app.downloads import errors as dlerrors
-from app.downloads import adapters, credentials, formats, service, storage
+from app.downloads import adapters, credentials, formats, mediakind, service, storage
 from app.downloads.jobdb import (
     _edit_staging_dir,
     _job_status,
@@ -484,6 +484,7 @@ def _download(
                 asset.acodec = (
                     info.get("acodec") if info.get("acodec") not in (None, "none") else None
                 )
+                asset.media_kind = mediakind.derive(asset.vcodec, asset.acodec)
                 asset.title = meta.title
                 asset.uploader = meta.uploader
                 asset.uploader_url = adapter.uploader_url(info)

backend/app/downloads/quota.py

@@ -122,6 +122,22 @@ def check_enqueue(db: Session, user_id: int) -> None:
         raise QuotaExceeded("max_bytes", lim.max_bytes, used)
 
 
+def check_upload(db: Session, user_id: int, add_bytes: int) -> None:
+    """Like check_enqueue, but for an upload whose size is KNOWN at ingest: also reject when adding
+    this file would push the footprint over max_bytes. A download can't pre-check its size (unknown
+    until the worker finishes) so check_enqueue only compares the current footprint; an upload can,
+    so it's charged against the cap before it lands. Raises QuotaExceeded(max_jobs|max_bytes)."""
+    lim = resolve(db, user_id)
+    if lim.unlimited:
+        return
+    n = active_jobs(db, user_id)
+    if n >= lim.max_jobs:
+        raise QuotaExceeded("max_jobs", lim.max_jobs, n)
+    projected = footprint(db, user_id) + add_bytes
+    if projected > lim.max_bytes:
+        raise QuotaExceeded("max_bytes", lim.max_bytes, projected)
+
+
 def usage(db: Session, user_id: int) -> dict:
     """Compact usage snapshot for the UI (footprint + counts vs limits)."""
     lim = resolve(db, user_id)

backend/app/downloads/service.py

@@ -17,7 +17,7 @@ from app.config import settings
 from app.downloads import adapters
 from app.downloads import credentials
 from app.downloads import edit as editmod
-from app.downloads import formats, quota, storage
+from app.downloads import formats, mediakind, quota, storage
 from app.downloads.status import HOLDING, JobStatus
 from app.models import Channel, DownloadJob, MediaAsset, Video
 
@@ -216,6 +216,10 @@ def enqueue_edit(
     src_asset = db.get(MediaAsset, source_job.asset_id) if source_job.asset_id else None
     if src_asset is None or src_asset.status != "ready" or not src_asset.rel_path:
         raise EditError("source_not_ready")
+    # A generic uploaded file (media_kind='file') has no timeline to trim/crop — the editor path is
+    # media-only. Reject here so an upload's file never reaches the ffmpeg edit worker.
+    if src_asset.media_kind == mediakind.FILE:
+        raise EditError("not_media")
 
     spec = editmod.normalize_edit_spec(edit_spec)
     if not spec:
@@ -232,6 +236,9 @@ def enqueue_edit(
         asset.thumbnail_url = src_asset.thumbnail_url
         asset.upload_date = src_asset.upload_date
         asset.duration_s = editmod.clip_duration(spec, src_asset.duration_s)
+        # A clip inherits its source's kind (a video edit stays video, an audio trim stays audio);
+        # the ffmpeg edit worker doesn't re-probe codecs, so set it here where the source is known.
+        asset.media_kind = src_asset.media_kind or mediakind.VIDEO
 
     job = DownloadJob(
         user_id=user_id,

backend/app/downloads/storage.py

@@ -147,6 +147,28 @@ def edit_rel_path(user_id: int, asset_id: int, ext: str) -> str:
     return f".edits/{user_id}/clip_{asset_id}.{ext}"
 
 
+def upload_rel_path(user_id: int, asset_id: int, filename: str) -> str:
+    """On-disk path (relative to DOWNLOAD_ROOT) for a locally-uploaded file.
+
+    Kept in a per-user `.uploads/` tree — device-downloadable via the file endpoint, but NOT part of
+    the Plex library (an upload is arbitrary user content, not a canonical episode; no .nfo/poster
+    episode sidecars). The name is system-decided and bound to the asset id (`_a{id}` keeps two
+    same-named uploads apart); the original filename is display metadata only. The extension is
+    preserved (shell-safe, bounded) so the served file + its Content-Disposition carry a correct
+    suffix; an extensionless upload stays extensionless."""
+    stem = sanitize(Path(filename).stem, 80)
+    ext = _UNSAFE.sub("", Path(filename).suffix.lstrip(".")).lower()[:16]
+    suffix = f".{ext}" if ext else ""
+    return f".uploads/{user_id}/{stem}_a{asset_id}{suffix}"
+
+
+def upload_ext(filename: str) -> str | None:
+    """The stored `container` for an uploaded file: its shell-safe lowercased extension, or None when
+    the upload has no extension (a bare name)."""
+    ext = _UNSAFE.sub("", Path(filename).suffix.lstrip(".")).lower()[:16]
+    return ext or None
+
+
 def download_filename(display_name: str, container: str | None, path: Path) -> str:
     """The Content-Disposition filename for a served download: the cleaned display name carrying a single
     correct extension — the asset container, else the file's own suffix — without doubling it when the

backend/app/downloads/upload.py

@@ -0,0 +1,158 @@
+"""Local file ingest for the download center (E7 S5) — the non-yt-dlp path into the media library.
+
+A user-uploaded file joins the SAME library as a downloaded one, so it inherits the whole function
+set already built (metadata chips, share links, and — for media — in-app play/edit). Unlike a
+download it bypasses yt-dlp AND the worker entirely: the file is already on the host, so ingest
+probes it, places it on disk, and creates an already-`ready` MediaAsset + `done` DownloadJob in one
+step — the same "the file exists, mark it done now" shortcut `service.enqueue` takes on a cache hit.
+
+Kinds (the invariant lives in `mediakind`): a probe with a real video stream -> video, audio-only
+-> audio, neither -> a generic 'file' (pdf/zip/docx/…) that is download-only and never rendered
+inline (see the force-download serving in the file/watch routes). This module is the ingest seam a
+future sub-function (e.g. convert) plugs a new adapter into, mirroring how it sidesteps the worker.
+
+An upload does NOT get an `expires_at`: it's user-provided and may be irreplaceable (there's no
+source to re-download), so it doesn't age out on the retention clock the way a re-fetchable download
+does — it lives until the user deletes it.
+"""
+
+import json
+import secrets
+import subprocess
+from datetime import datetime, timezone
+from pathlib import Path
+
+from sqlalchemy.orm import Session
+
+from app.config import settings
+from app.downloads import edit as editmod
+from app.downloads import mediakind, quota, service, storage
+from app.models import DownloadJob, MediaAsset
+
+
+def _probe(path: Path) -> tuple[str | None, str | None, int | None, int | None, int | None]:
+    """(vcodec, acodec, width, height, duration_s) from one ffprobe call; all None if ffprobe fails
+    or the file isn't media (a file we can't probe is treated as a generic non-media upload).
+
+    A video stream flagged `attached_pic` (embedded cover art in an mp3/m4a) is skipped — it's a
+    thumbnail, not a real video track, so an audio file with album art still classifies as audio."""
+    try:
+        out = subprocess.run(
+            [
+                "ffprobe",
+                "-v",
+                "error",
+                "-show_streams",
+                "-show_format",
+                "-of",
+                "json",
+                str(path),
+            ],
+            capture_output=True,
+            text=True,
+            timeout=30,
+            check=True,
+        ).stdout
+        data = json.loads(out)
+    except (subprocess.TimeoutExpired, subprocess.CalledProcessError, OSError, ValueError):
+        return None, None, None, None, None
+
+    vcodec = acodec = None
+    width = height = None
+    for stream in data.get("streams", []):
+        kind = stream.get("codec_type")
+        if kind == "video" and vcodec is None:
+            if (stream.get("disposition") or {}).get("attached_pic"):
+                continue  # cover art, not a video track
+            vcodec = stream.get("codec_name")
+            width = stream.get("width")
+            height = stream.get("height")
+        elif kind == "audio" and acodec is None:
+            acodec = stream.get("codec_name")
+
+    duration = None
+    try:
+        raw = (data.get("format") or {}).get("duration")
+        if raw is not None:
+            duration = int(float(raw))
+    except (TypeError, ValueError):
+        duration = None
+    return vcodec, acodec, width, height, duration
+
+
+def _title_from_filename(filename: str) -> str:
+    """A human display title from the original upload name: the stem with separators tidied to
+    spaces (kept as-is otherwise — this is display metadata, not the shell-safe on-disk name)."""
+    stem = Path(filename).stem.replace("_", " ").replace(".", " ").strip()
+    return (stem or "Upload")[:255]
+
+
+def ingest_upload(
+    db: Session,
+    user_id: int,
+    *,
+    filename: str,
+    src_path: Path,
+    size: int,
+    display_name: str | None = None,
+) -> DownloadJob:
+    """Ingest an already-uploaded file at `src_path` into `user_id`'s library and return the `done`
+    job. `src_path` is MOVED into place on success (the caller stages it to a temp file first).
+
+    Raises quota.QuotaExceeded (holdings/footprint cap) — checked BEFORE any disk write, so a
+    rejected upload leaves nothing behind and the caller just cleans up its temp file."""
+    quota.check_upload(db, user_id, size)  # the file's size is known → charge it up front
+
+    vcodec, acodec, width, height, duration = _probe(src_path)
+    kind = mediakind.derive(vcodec, acodec)
+
+    # An upload never dedups (each file is its own asset): a random ref makes the cache key unique,
+    # so `format_sig` is a constant marker rather than a format hash.
+    ref = f"{user_id}:{secrets.token_hex(8)}"[:512]
+    asset = MediaAsset(
+        source_kind="upload", source_ref=ref, format_sig="upload", status="pending", ref_count=0
+    )
+    db.add(asset)
+    db.flush()  # need asset.id for the on-disk path
+
+    rel = storage.upload_rel_path(user_id, asset.id, filename)
+    storage.place_file(src_path, settings.download_root, rel)
+    final = storage.abs_path(settings.download_root, rel)
+    actual_size = final.stat().st_size if final.exists() else size
+
+    asset.status = "ready"
+    asset.rel_path = rel
+    asset.size_bytes = actual_size
+    asset.container = storage.upload_ext(filename)
+    asset.vcodec = vcodec
+    asset.acodec = acodec
+    asset.width = width
+    asset.height = height
+    asset.duration_s = duration
+    asset.media_kind = kind
+    asset.title = (display_name or "").strip() or _title_from_filename(filename)
+    asset.last_access_at = datetime.now(timezone.utc)
+    # No expires_at — see the module docstring (an upload is irreplaceable, so it doesn't age out).
+    if kind == mediakind.VIDEO:
+        # A frame poster gives the library row a thumbnail + the share link an og:image. Audio has no
+        # frame and a generic file no player, so neither gets one. Best-effort (ffmpeg decodes server
+        # side even for a codec the browser can't play).
+        asset.poster_path = editmod.ensure_poster(asset, settings.download_root)
+
+    job = DownloadJob(
+        user_id=user_id,
+        asset_id=asset.id,
+        job_kind="upload",
+        source_kind="upload",
+        source_ref=ref,
+        profile_snapshot={},
+        display_name=(display_name or "").strip() or None,
+        status="done",
+        progress=100,
+        queue_pos=service.next_queue_pos(db),
+    )
+    db.add(job)
+    service.incr_ref_count(db, asset.id)
+    db.commit()
+    db.refresh(job)
+    return job

backend/app/models.py

@@ -708,6 +708,12 @@ class MediaAsset(Base, TimestampMixin):
     container: Mapped[str | None] = mapped_column(String(16))
     vcodec: Mapped[str | None] = mapped_column(String(32))
     acodec: Mapped[str | None] = mapped_column(String(32))
+    # "video" | "audio" | "file" — the media-kind discriminator that drives which library actions
+    # apply (play/edit for media; download-only for a generic uploaded file). Null while an asset is
+    # still pending (kind is known once the file is probed at finalize). It is the single stored home
+    # of the invariant `downloads.mediakind.derive` used to compute — never re-derived in prose (an
+    # earlier ad-hoc `vcodec is None and acodec` check lived only in links.py; consolidated here).
+    media_kind: Mapped[str | None] = mapped_column(String(16))
     nfo_written: Mapped[bool] = mapped_column(Boolean, default=False, server_default="false")
     # Set once the GC has warned holders of the upcoming expiry, so it doesn't warn every run.
     gc_notified: Mapped[bool] = mapped_column(Boolean, default=False, server_default="false")

backend/app/routes/downloads.py

@@ -6,11 +6,14 @@ worker is a separate process); this module never downloads — it enqueues, mana
 serves finished files (range-aware, with the user's custom display name), and shares them.
 """
 
+import os
 import re
+import tempfile
 from datetime import datetime, timedelta, timezone
+from pathlib import Path
 from urllib.parse import parse_qs, urlparse
 
-from fastapi import APIRouter, Depends, HTTPException, Request
+from fastapi import APIRouter, Depends, File, Form, HTTPException, Request, UploadFile
 from fastapi.responses import FileResponse
 from pydantic import BaseModel
 from sqlalchemy import func, select
@@ -23,7 +26,9 @@ from app.db import get_db
 from app.downloads import credentials as dlcreds
 from app.downloads import edit as editmod
 from app.downloads import links as linksmod
+from app.downloads import mediakind
 from app.downloads import quota, service, storage
+from app.downloads import upload as uploadmod
 from app.models import (
     DownloadJob,
     DownloadLink,
@@ -123,6 +128,9 @@ def _serialize(job: DownloadJob, asset: MediaAsset | None) -> dict:
             "width": asset.width,
             "height": asset.height,
             "container": asset.container,
+            # "video" | "audio" | "file" — the library gates play/edit on this (media only); a
+            # generic uploaded file is download-only. Null while still in flight (set at finalize).
+            "media_kind": asset.media_kind,
             "uploader_url": asset.uploader_url,
             "thumbnail_url": asset.thumbnail_url,
             "expires_at": iso(asset.expires_at),
@@ -354,6 +362,61 @@ def enqueue_download(
     return _serialize_job(db, job)
 
 
+# Read the upload in 1 MiB chunks so a large file never sits fully in memory, and the size cap is
+# enforced on the ACTUAL bytes read (a client-set Content-Length can't be trusted to gate it).
+_UPLOAD_CHUNK = 1024 * 1024
+
+
+@router.post("/upload")
+def upload_download(
+    file: UploadFile = File(...),
+    display_name: str | None = Form(None),
+    user: User = Depends(require_human),
+    db: Session = Depends(get_db),
+) -> dict:
+    """Ingest a local file into the caller's library (E7 S5). Any file type is accepted; media
+    (video/audio) gets the full function set, a generic file is download-only. The file is streamed
+    to a temp path under the download volume (enforcing the per-file cap as it goes), then handed to
+    the non-yt-dlp ingest which probes + places it and returns an already-`done` job.
+
+    A sync (threadpool) handler on purpose: the ingest does blocking work — an ffprobe of up to 30s
+    plus sync DB writes — so it must NOT run on the event loop. `file.file` is the sync file object
+    Starlette spools the multipart body into."""
+    cap = sysconfig.get_int(db, "download_upload_max_bytes")
+    staging = Path(settings.download_root) / ".staging" / "uploads"
+    staging.mkdir(parents=True, exist_ok=True)
+    fd, tmp = tempfile.mkstemp(dir=str(staging))
+    tmp_path = Path(tmp)
+    try:
+        size = 0
+        with os.fdopen(fd, "wb") as out:
+            while chunk := file.file.read(_UPLOAD_CHUNK):
+                size += len(chunk)
+                if cap and size > cap:
+                    raise HTTPException(
+                        status_code=422, detail={"code": "upload_too_large", "limit": cap}
+                    )
+                out.write(chunk)
+        if size == 0:
+            raise HTTPException(status_code=400, detail={"code": "upload_empty"})
+        try:
+            job = uploadmod.ingest_upload(
+                db,
+                user.id,
+                filename=file.filename or "upload",
+                src_path=tmp_path,
+                size=size,
+                display_name=(display_name or "").strip() or None,
+            )
+        except quota.QuotaExceeded as e:
+            raise HTTPException(status_code=422, detail=_quota_detail(e))
+        return _serialize_job(db, job)
+    finally:
+        # Ingest MOVES the temp file into place on success (unlink is then a no-op); on any early exit
+        # — cap hit, empty, quota, probe/move error — this removes the abandoned staging file.
+        tmp_path.unlink(missing_ok=True)
+
+
 class EnqueueEditIn(BaseModel):
     source_job_id: int = 0
     edit_spec: dict
@@ -544,8 +607,10 @@ def redownload_download(
     never deletes a file another job/user still shares: the old asset is only removed if THIS job
     was its last holder."""
     job = _own_job(db, user, job_id)
-    if job.job_kind == "edit":
-        raise HTTPException(status_code=409, detail="Editor clips can't be re-downloaded.")
+    if job.job_kind in ("edit", "upload"):
+        # Neither has a re-fetchable source: an editor clip is derived locally, an upload IS the
+        # source. Re-running yt-dlp against source_kind='upload'/'edit' would just fail.
+        raise HTTPException(status_code=409, detail="This item can't be re-downloaded.")
     if job.status in ("queued", "running", "paused"):
         raise HTTPException(status_code=409, detail="This download is still in progress.")
     service.redownload(db, job)
@@ -610,7 +675,16 @@ def download_file(
 
     asset.last_access_at = datetime.now(timezone.utc)
     db.commit()
-    # Starlette's FileResponse honours the Range header (206 partial content) for seek/resume.
+    # A generic uploaded file is served as a neutral, force-download attachment — never with a
+    # sniffable content-type — so uploaded HTML/SVG can't execute inline on this origin. Media keeps
+    # its guessed type so the in-app player streams it. Both are range-aware (FileResponse → 206).
+    if asset.media_kind == mediakind.FILE:
+        return FileResponse(
+            path,
+            filename=filename,
+            media_type="application/octet-stream",
+            content_disposition_type="attachment",
+        )
     return FileResponse(path, filename=filename)
 
 

backend/app/routes/public.py

@@ -15,7 +15,7 @@ from sqlalchemy.orm import Session
 from app.config import settings
 from app.db import get_db
 from app.downloads import links as linksmod
-from app.downloads import storage
+from app.downloads import mediakind, storage
 from app.models import DownloadJob, DownloadLink, MediaAsset
 from app.ratelimit import RateLimiter
 from app.security import verify_password
@@ -122,8 +122,18 @@ def watch_file(token: str, g: str | None = None, db: Session = Depends(get_db)):
         raise HTTPException(status_code=404, detail="This video is no longer available.")
 
     filename = storage.download_filename(asset.title or "video", asset.container, path)
+    # A generic uploaded file (media_kind='file') is ALWAYS a force-download attachment with a
+    # neutral type — never served inline — so arbitrary uploaded HTML/SVG can't execute on this
+    # login-free public origin. Media follows the link's own inline/attachment choice so it can play
+    # in the watch page. Both are range-aware (FileResponse → 206) for seeking/resume.
+    if asset.media_kind == mediakind.FILE:
+        return FileResponse(
+            path,
+            filename=filename,
+            media_type="application/octet-stream",
+            content_disposition_type="attachment",
+        )
     disposition = "attachment" if link.allow_download else "inline"
-    # Starlette's FileResponse honours the Range header (206) for seeking/scrubbing.
     return FileResponse(path, filename=filename, content_disposition_type=disposition)
 
 

backend/app/sysconfig.py

@@ -127,6 +127,7 @@ SPECS: tuple[ConfigSpec, ...] = (
         min=1,
         max=100_000,
     ),
+    ConfigSpec("download_upload_max_bytes", "int", "downloads", "download_upload_max_bytes", min=0),
     ConfigSpec(
         "download_retention_days", "int", "downloads", "download_retention_days", min=0, max=3_650
     ),

backend/requirements.lock

@@ -43,6 +43,7 @@ pydantic-settings==2.15.0
 pydantic==2.13.4
 pydantic_core==2.46.4
 python-dotenv==1.2.3
+python-multipart==0.0.32
 requests==2.34.2
 starlette==1.6.0
 typing-inspection==0.4.4

backend/requirements.txt

@@ -1,4 +1,7 @@
 fastapi>=0.115,<1.0
+# Multipart form parsing — required by Starlette/FastAPI to accept file uploads (UploadFile/Form).
+# The download center's local-file Upload (E7) is the only multipart endpoint in the app.
+python-multipart>=0.0.9,<1.0
 uvicorn[standard]>=0.30,<1.0
 sqlalchemy>=2.0,<2.1
 psycopg[binary]>=3.2,<4.0

backend/tests/test_links.py

@@ -65,6 +65,9 @@ def _asset(**over):
         height=720,
         vcodec="h264",
         acodec="aac",
+        media_kind="video",
+        size_bytes=1000,
+        container="mp4",
         source_webpage_url=None,
         source_kind="url",
     )
@@ -74,12 +77,13 @@ def _asset(**over):
 
 class TestPublicMetaAudioFlag:
     """The watch page picks <audio> vs <video> from this flag; WebKit refuses an audio-only mp3 in a
-    <video>, so an audio-only asset (codec present, no video stream) must report audio=True."""
+    <video>, so an audio-only asset (media_kind='audio') must report audio=True. The flag now reads
+    the stored media_kind (the single source), not a re-derived codec check."""
 
     _link = SimpleNamespace(allow_download=True)
 
     def test_audio_only_asset_is_flagged_audio(self):
-        asset = _asset(vcodec=None, acodec="mp3", width=None, height=None)
+        asset = _asset(media_kind="audio", vcodec=None, acodec="mp3", width=None, height=None)
         meta = public_meta(self._link, None, asset, "/f.mp3")
         assert meta["audio"] is True
 
@@ -87,10 +91,11 @@ class TestPublicMetaAudioFlag:
         meta = public_meta(self._link, None, _asset(), "/f.mp4")
         assert meta["audio"] is False
 
-    def test_a_silent_video_is_not_audio(self):
-        # Video present but no audio track — still a video player, not the audio bar.
-        meta = public_meta(self._link, None, _asset(acodec=None), "/f.mp4")
+    def test_a_generic_file_is_not_audio(self):
+        # A non-media upload (media_kind='file') has no player at all — never the audio bar.
+        meta = public_meta(self._link, None, _asset(media_kind="file"), "/f.bin")
         assert meta["audio"] is False
+        assert meta["media_kind"] == "file"
 
 
 class TestCreateLinkDefault:

backend/tests/test_upload.py

@@ -0,0 +1,184 @@
+"""E7 S5a — local file upload ingest (the non-yt-dlp path into the media library).
+
+Covers the ingest spine: media vs generic-file classification, the download-only serving of a
+generic file (force-download so uploaded HTML/SVG can't execute inline), quota accounting with the
+size known up front, the per-file cap enforced while streaming, filename sanitization, and the two
+guards that keep an upload out of paths that assume a re-fetchable source (edit / redownload).
+
+DB-backed (real Postgres) — skipped when no database is reachable. `settings.download_root` is
+pointed at a per-test tmp dir so ingest's on-disk moves are isolated; `_probe` is monkeypatched so
+the tests don't depend on ffprobe (its real behaviour is exercised by the download/edit suites).
+"""
+
+import pytest
+
+from app.config import settings
+from app.downloads import mediakind, quota, storage
+from app.downloads import upload as uploadmod
+from app.models import DownloadQuota, MediaAsset
+
+
+# --- pure: the kind invariant ---------------------------------------------------------------
+
+
+def test_derive_media_kind():
+    assert uploadmod.mediakind.derive("h264", "aac") == mediakind.VIDEO
+    assert uploadmod.mediakind.derive("vp9", None) == mediakind.VIDEO
+    assert uploadmod.mediakind.derive(None, "aac") == mediakind.AUDIO
+    assert uploadmod.mediakind.derive(None, None) == mediakind.FILE
+
+
+# --- ingest (direct) ------------------------------------------------------------------------
+
+
+@pytest.fixture
+def _root(tmp_path, monkeypatch):
+    monkeypatch.setattr(settings, "download_root", str(tmp_path))
+    return tmp_path
+
+
+def _probe(monkeypatch, result):
+    monkeypatch.setattr(uploadmod, "_probe", lambda _p: result)
+
+
+def test_ingest_generic_file_is_download_only(db, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, None, None, None, None))
+    u = make_user()
+    src = _root / "src.bin"
+    src.write_bytes(b"%PDF-1.4 hello")
+    job = uploadmod.ingest_upload(
+        db, u.id, filename="My Notes.pdf", src_path=src, size=src.stat().st_size
+    )
+    asset = db.get(MediaAsset, job.asset_id)
+    assert asset.media_kind == mediakind.FILE
+    assert asset.status == "ready" and job.status == "done" and job.progress == 100
+    assert asset.container == "pdf"
+    assert asset.poster_path is None  # a generic file has no player and no poster
+    assert asset.title == "My Notes"  # display title from the original name
+    assert job.job_kind == "upload" and job.source_kind == "upload"
+    # The file was MOVED into the per-user uploads tree, inside the root, and the temp is gone.
+    assert not src.exists()
+    assert asset.rel_path.startswith(".uploads/")
+    assert storage.safe_abs_path(str(_root), asset.rel_path) is not None
+
+
+def test_ingest_audio_is_media_no_poster(db, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, "aac", None, None, 200))
+    u = make_user()
+    src = _root / "song.m4a"
+    src.write_bytes(b"fake-audio")
+    job = uploadmod.ingest_upload(db, u.id, filename="song.m4a", src_path=src, size=10)
+    asset = db.get(MediaAsset, job.asset_id)
+    assert asset.media_kind == mediakind.AUDIO
+    assert asset.duration_s == 200
+    assert asset.poster_path is None  # audio has no frame to poster
+
+
+def test_ingest_video_records_dims_and_poster(db, make_user, _root, monkeypatch):
+    _probe(monkeypatch, ("h264", "aac", 640, 480, 60))
+    # Stub the frame-cut so the test doesn't shell out to ffmpeg.
+    monkeypatch.setattr(uploadmod.editmod, "ensure_poster", lambda _a, _root: "p/x.jpg")
+    u = make_user()
+    src = _root / "clip.mp4"
+    src.write_bytes(b"fake-video")
+    job = uploadmod.ingest_upload(db, u.id, filename="clip.mp4", src_path=src, size=10)
+    asset = db.get(MediaAsset, job.asset_id)
+    assert asset.media_kind == mediakind.VIDEO
+    assert (asset.width, asset.height) == (640, 480)
+    assert asset.poster_path == "p/x.jpg"
+
+
+def test_ingest_filename_is_sanitized_no_traversal(db, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, None, None, None, None))
+    u = make_user()
+    src = _root / "src"
+    src.write_bytes(b"x")
+    job = uploadmod.ingest_upload(db, u.id, filename="../../etc/passwd", src_path=src, size=1)
+    asset = db.get(MediaAsset, job.asset_id)
+    assert ".." not in asset.rel_path
+    assert asset.rel_path.startswith(f".uploads/{u.id}/")
+    assert storage.safe_abs_path(str(_root), asset.rel_path) is not None
+
+
+def test_ingest_rejects_over_footprint_cap_before_writing(db, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, None, None, None, None))
+    u = make_user()
+    db.add(
+        DownloadQuota(user_id=u.id, max_bytes=1000, max_concurrent=2, max_jobs=50, unlimited=False)
+    )
+    db.commit()
+    src = _root / "big.bin"
+    src.write_bytes(b"x" * 10)
+    with pytest.raises(quota.QuotaExceeded) as e:
+        uploadmod.ingest_upload(db, u.id, filename="big.bin", src_path=src, size=2000)
+    assert e.value.reason == "max_bytes"
+    assert src.exists()  # rejected before any disk move — nothing landed
+    assert db.query(MediaAsset).count() == 0  # and no orphan asset row
+
+
+# --- route ----------------------------------------------------------------------------------
+
+
+def test_upload_route_generic_forces_download(as_user, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, None, None, None, None))
+    c = as_user(make_user())
+    r = c.post(
+        "/api/downloads/upload",
+        files={"file": ("report.html", b"<script>alert(1)</script>", "text/html")},
+    )
+    assert r.status_code == 200, r.text
+    job = r.json()
+    assert job["asset"]["media_kind"] == "file"
+    assert job["can_download"] is True
+    # The served file must be a neutral force-download attachment — never inline text/html.
+    fr = c.get(f"/api/downloads/{job['id']}/file")
+    assert fr.status_code == 200
+    assert fr.headers["content-type"].startswith("application/octet-stream")
+    assert "attachment" in fr.headers["content-disposition"].lower()
+
+
+def test_upload_route_enforces_per_file_cap(as_user, make_user, db, _root, monkeypatch):
+    from app import sysconfig
+
+    sysconfig.set_value(db, "download_upload_max_bytes", 5)
+    db.commit()
+    c = as_user(make_user())
+    r = c.post(
+        "/api/downloads/upload",
+        files={"file": ("big.bin", b"way over the cap", "application/octet-stream")},
+    )
+    assert r.status_code == 422
+    assert r.json()["detail"]["code"] == "upload_too_large"
+
+
+def test_upload_route_rejects_empty_file(as_user, make_user, _root, monkeypatch):
+    c = as_user(make_user())
+    r = c.post(
+        "/api/downloads/upload", files={"file": ("empty.bin", b"", "application/octet-stream")}
+    )
+    assert r.status_code == 400
+    assert r.json()["detail"]["code"] == "upload_empty"
+
+
+def test_redownload_rejects_an_upload(as_user, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, None, None, None, None))
+    c = as_user(make_user())
+    job = c.post(
+        "/api/downloads/upload", files={"file": ("a.bin", b"data", "application/octet-stream")}
+    ).json()
+    r = c.post(f"/api/downloads/{job['id']}/redownload")
+    assert r.status_code == 409
+
+
+def test_edit_rejects_a_generic_upload_source(as_user, make_user, _root, monkeypatch):
+    _probe(monkeypatch, (None, None, None, None, None))
+    c = as_user(make_user())
+    job = c.post(
+        "/api/downloads/upload", files={"file": ("a.bin", b"data", "application/octet-stream")}
+    ).json()
+    r = c.post(
+        "/api/downloads/edit",
+        json={"source_job_id": job["id"], "edit_spec": {"trim": {"start_s": 0, "end_s": 1}}},
+    )
+    assert r.status_code == 400
+    assert r.json()["detail"]["reason"] == "not_media"