TubeSubarr

TubeSubarr is a self-hosted Node.js + TypeScript MVP for managing YouTube or podcast channel subscriptions, downloading new videos with yt-dlp, exposing per-user Jellyfin views via symlinks, and applying cleanup rules.

Project structure

.
├── prisma/
│   ├── schema.prisma        # SQLite schema for users, channels, videos, jobs
│   └── seed.ts              # Seeds local users (Lana, Richard, Shared)
├── src/
│   ├── public/              # Simple subscriptions/videos UI
│   ├── routes/              # REST API routes
│   ├── services/            # Domain services for storage, yt-dlp, Jellyfin
│   ├── workers/             # Queue + poll/download/sync/cleanup workers
│   ├── app.ts               # Express app factory
│   ├── config.ts            # Environment/config loading
│   ├── db.ts                # Prisma client
│   └── server.ts            # Bootstrap + schedulers
├── Dockerfile
├── docker-compose.yml
└── .env.example

MVP capabilities

• Multi-user local accounts with separate YouTube and Podcast Jellyfin views.
• Canonical channel storage under /store/.../channels/<channel-slug> and user-specific symlinked Jellyfin views under /views/.../<user>/<channel-slug>.
• REST API for subscriptions, videos, and Jellyfin webhook ingestion.
• Background workers for channel polling, downloads, Jellyfin watched sync, and cleanup.
• Cleanup rules for watched and stale-unwatched videos with a trash-first delete flow.
• Minimal browser UI for subscriptions and video pin/watch actions.

Prisma schema highlights

• User: local application/Jellyfin user mapping.
• Channel: canonical channel subscription and retention settings.
• ChannelAssignment: many-to-many user/channel ownership.
• Video: download state, Jellyfin item ID, watch state, pin/playlist protection, delete timestamps.
• Job: lightweight SQLite-backed job queue.

API

• GET /subscriptions
• POST /subscriptions
• PATCH /subscriptions/:id
• DELETE /subscriptions/:id
• GET /videos
• PATCH /videos/:id
• POST /webhooks/jellyfin

Run locally

Node.js

cp .env.example .env
npm install
npx prisma generate
npx prisma db push
npm run seed
npm run dev

Open http://localhost:4000.

Docker

docker compose up --build

The container installs ffmpeg and yt-dlp, runs prisma db push, and starts the server on port 4000.

Notes

• Jellyfin libraries should already exist and point to ./data/views/youtube/<user> and ./data/views/podcasts/<user>.
• The worker queue is intentionally simple for MVP scope; a production version should use stronger locking and retries.
• yt-dlp integration uses child processes and JSON probing/download workflows that can be extended for richer metadata or search.