feat(downloads): M1 — schema, config, worker container, deps

- models: MediaAsset (shared file cache, unique on source+format_sig), DownloadJob
  (per-user request), DownloadProfile (presets), DownloadShare (ACL), DownloadQuota
- migrations 0036 (5 tables) + 0037 (6 builtin presets)
- sysconfig 'downloads' group (per-user defaults, retention/GC, layout, worker concurrency)
- config: DOWNLOAD_ROOT / WORKER_ENABLED env + download defaults
- deps: yt-dlp in requirements, ffmpeg in Dockerfile runtime stage
- worker: dedicated container (python -m app.worker), thin idle entry (loop lands in M2)
- localdev compose: worker service + downloads bind-mount; gitignore downloads/
This commit is contained in:
npeter83 2026-07-02 23:57:40 +02:00
parent 9ca90f515c
commit 317750e491
10 changed files with 496 additions and 0 deletions

View file

@ -131,6 +131,31 @@ class Settings(BaseSettings):
# ("was_live") are real watchable content and stay visible.
feed_default_hidden_live: str = "live,upcoming"
# --- Download center (yt-dlp) ---
# Root directory (inside the container) for downloaded media — a Docker volume/bind mount.
# Env-only (a mount path can't live in the DB). The worker lays a Plex-style tree under it.
download_root: str = "/downloads"
# Whether THIS process runs the download worker loop. On for the dedicated worker
# container, off for the API (which keeps the scheduler). Mirrors scheduler_enabled.
worker_enabled: bool = False
# How many downloads the worker runs concurrently (claimed via FOR UPDATE SKIP LOCKED).
download_worker_concurrency: int = 2
# How often (empty pending queue) the worker polls for a new job, in seconds.
download_poll_seconds: int = 5
# Default per-user quota (admin-tunable; a per-user download_quotas row overrides).
download_default_max_bytes: int = 5_368_709_120 # 5 GiB
download_default_max_concurrent: int = 2
download_default_max_jobs: int = 50
# 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
download_gc_grace_days: int = 2
download_gc_minutes: int = 360
# On-disk layout: "plex" (Channel/Season YYYY/… [id].ext + .nfo/poster) or "flat".
download_layout: str = "plex"
# Default SponsorBlock (skip sponsor segments) for new custom profiles.
sponsorblock_default: bool = False
@property
def allowed_email_set(self) -> set[str]:
return {e.strip().lower() for e in self.allowed_emails.split(",") if e.strip()}

View file

@ -8,6 +8,7 @@ from sqlalchemy import (
DateTime,
Float,
ForeignKey,
Index,
Integer,
JSON,
String,
@ -688,6 +689,160 @@ class Message(Base):
)
class MediaAsset(Base, TimestampMixin):
"""A physical downloaded media file — the SHARED cache layer of the download center.
Mirrors the app's "shared catalog, private per-user state" model: one row per unique
``(source_kind, source_ref, format_sig)`` is the global cache. When a second user requests
the same video with a format that hashes to the same `format_sig`, their `download_job`
links to this ready asset instantly no re-download, no cost. Only the SOURCE download is
shared; per-user edited derivatives (phase 2) are never cached here.
`rel_path` is relative to `settings.download_root` (the mount), so the physical filename is
system-decided and bound to the source id never renamed (the user's custom name is display
metadata on `download_job`, applied only when downloading to their own machine). The naming
fields (`title`/`uploader`/`upload_date`) are denormalized so the Plex-tree path can be built
uniformly for both catalog videos and ad-hoc URLs. `ref_count` (active jobs pointing here)
plus `last_access_at`/`expires_at` drive retention + LRU eviction in the GC job."""
__tablename__ = "media_assets"
__table_args__ = (
UniqueConstraint(
"source_kind", "source_ref", "format_sig", name="uq_media_asset_cache_key"
),
)
id: Mapped[int] = mapped_column(primary_key=True)
source_kind: Mapped[str] = mapped_column(String(16)) # "youtube" | "url"
source_ref: Mapped[str] = mapped_column(String(512)) # video id, or full URL for ad-hoc
format_sig: Mapped[str] = mapped_column(String(64)) # hash of the output-affecting spec
status: Mapped[str] = mapped_column(
String(16), default="pending", server_default="pending", index=True
) # "pending" | "ready" | "error"
rel_path: Mapped[str | None] = mapped_column(Text) # relative to download_root
storyboard_path: Mapped[str | None] = mapped_column(Text) # phase-2 filmstrip mosaic
size_bytes: Mapped[int | None] = mapped_column(BigInteger)
duration_s: Mapped[int | None] = mapped_column(Integer)
width: Mapped[int | None] = mapped_column(Integer)
height: Mapped[int | None] = mapped_column(Integer)
container: Mapped[str | None] = mapped_column(String(16))
vcodec: Mapped[str | None] = mapped_column(String(32))
acodec: Mapped[str | None] = mapped_column(String(32))
nfo_written: Mapped[bool] = mapped_column(
Boolean, default=False, server_default="false"
)
# Denormalized naming/metadata for the Plex tree + ad-hoc display.
title: Mapped[str | None] = mapped_column(Text)
uploader: Mapped[str | None] = mapped_column(String(255))
upload_date: Mapped[date | None] = mapped_column(Date)
thumbnail_url: Mapped[str | None] = mapped_column(Text)
error: Mapped[str | None] = mapped_column(Text)
ref_count: Mapped[int] = mapped_column(Integer, default=0, server_default="0")
last_access_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
expires_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
class DownloadProfile(Base, TimestampMixin):
"""A named download preset — a reusable format profile.
`user_id` NULL + `is_builtin` marks the shipped presets (Best MP4, 1080p, audio-only, );
a user's saved custom profiles carry their `user_id`. `spec` is the format definition
(mode av/v/a, max_height, container, audio_format, codec prefs, embed subs/chapters/
thumbnail, sponsorblock) that `downloads.formats` turns into a yt-dlp selector + a cache
`format_sig`."""
__tablename__ = "download_profiles"
id: Mapped[int] = mapped_column(primary_key=True)
user_id: Mapped[int | None] = mapped_column(
ForeignKey("users.id", ondelete="CASCADE"), index=True
) # NULL for a builtin preset
name: Mapped[str] = mapped_column(String(80))
spec: Mapped[dict] = mapped_column(JSON)
is_builtin: Mapped[bool] = mapped_column(
Boolean, default=False, server_default="false"
)
sort_order: Mapped[int] = mapped_column(Integer, default=0, server_default="0")
class DownloadJob(Base, TimestampMixin, UpdatedAtMixin):
"""A user's request to download one source — the per-user layer over `MediaAsset`.
The worker claims `queued` jobs with ``FOR UPDATE SKIP LOCKED`` and writes live
`progress`/`speed_bps`/`eta_s`/`phase` here (the frontend polls via useLiveQuery the
worker is a separate process and can't reach the in-process WS registry). `profile_snapshot`
freezes the spec at enqueue so later edits/deletion of the profile don't change this job.
`display_name` is the user's own name for the file (display + the Content-Disposition on
local download); the on-disk asset is never renamed. `asset_id` links to the shared file
(SET NULL if the asset is GC'd — the job then shows as expired)."""
__tablename__ = "download_jobs"
__table_args__ = (Index("ix_download_jobs_claim", "status", "queue_pos"),)
id: Mapped[int] = mapped_column(primary_key=True)
user_id: Mapped[int] = mapped_column(
ForeignKey("users.id", ondelete="CASCADE"), index=True
)
asset_id: Mapped[int | None] = mapped_column(
ForeignKey("media_assets.id", ondelete="SET NULL"), index=True
)
source_kind: Mapped[str] = mapped_column(String(16)) # "youtube" | "url"
source_ref: Mapped[str] = mapped_column(String(512))
profile_id: Mapped[int | None] = mapped_column(
ForeignKey("download_profiles.id", ondelete="SET NULL")
)
profile_snapshot: Mapped[dict] = mapped_column(JSON)
display_name: Mapped[str | None] = mapped_column(String(255))
status: Mapped[str] = mapped_column(
String(16), default="queued", server_default="queued", index=True
) # queued | running | paused | done | error | canceled
progress: Mapped[int] = mapped_column(Integer, default=0, server_default="0")
speed_bps: Mapped[int | None] = mapped_column(BigInteger)
eta_s: Mapped[int | None] = mapped_column(Integer)
phase: Mapped[str | None] = mapped_column(String(32))
error: Mapped[str | None] = mapped_column(Text)
queue_pos: Mapped[int] = mapped_column(Integer, default=0, server_default="0")
class DownloadShare(Base, TimestampMixin):
"""A user→user grant: `to_user` may access `from_user`'s downloaded file (private by
default). Since assets are already deduped/shared on disk, a share is just an ACL row."""
__tablename__ = "download_shares"
__table_args__ = (
UniqueConstraint("job_id", "to_user_id", name="uq_download_share"),
)
id: Mapped[int] = mapped_column(primary_key=True)
job_id: Mapped[int] = mapped_column(
ForeignKey("download_jobs.id", ondelete="CASCADE"), index=True
)
from_user_id: Mapped[int] = mapped_column(
ForeignKey("users.id", ondelete="CASCADE"), index=True
)
to_user_id: Mapped[int] = mapped_column(
ForeignKey("users.id", ondelete="CASCADE"), index=True
)
class DownloadQuota(Base, UpdatedAtMixin):
"""Per-user download limits (admin-managed). A row exists only when an admin customizes a
user; otherwise the sysconfig `download_default_*` values apply. `unlimited` bypasses the
byte cap entirely (e.g. for the admin's own account)."""
__tablename__ = "download_quotas"
user_id: Mapped[int] = mapped_column(
ForeignKey("users.id", ondelete="CASCADE"), primary_key=True
)
max_bytes: Mapped[int] = mapped_column(BigInteger)
max_concurrent: Mapped[int] = mapped_column(Integer)
max_jobs: Mapped[int] = mapped_column(Integer)
unlimited: Mapped[bool] = mapped_column(
Boolean, default=False, server_default="false"
)
class MessageKey(Base, TimestampMixin):
"""A user's end-to-end-encryption key material for direct messaging (phase 2).

View file

@ -66,6 +66,15 @@ SPECS: tuple[ConfigSpec, ...] = (
ConfigSpec("google_client_secret", "str", "google", "google_client_secret", secret=True),
# --- Access / registration ---
ConfigSpec("allow_registration", "bool", "access", "allow_registration"),
# --- Download center (yt-dlp) — per-user defaults + retention/GC + layout ---
ConfigSpec("download_default_max_bytes", "int", "downloads", "download_default_max_bytes", min=0),
ConfigSpec("download_default_max_concurrent", "int", "downloads", "download_default_max_concurrent", min=1, max=20),
ConfigSpec("download_default_max_jobs", "int", "downloads", "download_default_max_jobs", min=1, max=100_000),
ConfigSpec("download_retention_days", "int", "downloads", "download_retention_days", min=0, max=3_650),
ConfigSpec("download_gc_grace_days", "int", "downloads", "download_gc_grace_days", min=0, max=3_650),
ConfigSpec("download_layout", "str", "downloads", "download_layout"),
ConfigSpec("download_worker_concurrency", "int", "downloads", "download_worker_concurrency", min=1, max=16),
ConfigSpec("sponsorblock_default", "bool", "downloads", "sponsorblock_default"),
)
_BY_KEY: dict[str, ConfigSpec] = {s.key: s for s in SPECS}

50
backend/app/worker.py Normal file
View file

@ -0,0 +1,50 @@
"""Download worker process entry point.
Runs in a DEDICATED container (`command: python -m app.worker`, `WORKER_ENABLED=1`), separate
from the API so long yt-dlp downloads + ffmpeg postprocessing never block web requests. It
claims `queued` download jobs from the DB with FOR UPDATE SKIP LOCKED and runs them, writing
live progress back to the job row (the frontend polls it the worker is a separate process and
can't reach the API's in-process WebSocket registry).
M1: thin, idle entry point (keeps the container healthy). The claim loop + yt-dlp integration
land in M2.
"""
import logging
import signal
import time
from app.config import settings
log = logging.getLogger("siftlode.worker")
_stop = False
def _handle_signal(signum, frame):
global _stop
_stop = True
def main() -> None:
logging.basicConfig(level=logging.INFO)
signal.signal(signal.SIGTERM, _handle_signal)
signal.signal(signal.SIGINT, _handle_signal)
if not settings.worker_enabled:
log.info("WORKER_ENABLED is off; worker exiting (nothing to do).")
return
log.info(
"Download worker started (concurrency=%d, root=%s). Claim loop lands in M2.",
settings.download_worker_concurrency,
settings.download_root,
)
while not _stop:
# M2: claim queued jobs (FOR UPDATE SKIP LOCKED) and download them.
time.sleep(settings.download_poll_seconds)
log.info("Download worker stopped.")
if __name__ == "__main__":
main()