Plex Pass

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Build me the media server I actually need instead of paying for Plex Pass — and read the honest part first, because this one is mostly not about code.

Read this first. Jellyfin is free, open source, and does what a media server does. The three things a subscription buys are apps on televisions and consoles, hardware transcoding, and metadata matching that works on a badly named file. The first of those is the real one: nobody wants to write a television app, and the platforms make it deliberately tedious. Before starting, check that Jellyfin has a client for every screen you actually use. If it does, you are finished in an afternoon and this prompt is about the layer around it.

STACK
- Jellyfin as the server. Do not write a media server, a transcoder or a client
- Node 20+ with Fastify and SQLite through better-sqlite3 for the layer around it: acquisition, organisation, requests, health
- ffmpeg for anything you do yourself
- Caddy in front

WHAT ACTUALLY MATTERS: THE FILES
- The library is a directory tree with a strict naming convention, because every scanner in existence matches on the filename and the folder
- Films: `Title (Year)/Title (Year) [quality].ext`. Series: `Show (Year)/Season 01/Show - S01E02 - Title.ext`
- Get the naming right and metadata matching stops being a problem. Get it wrong and no amount of software rescues it — this is what the paid matcher is compensating for
- A renamer that proposes changes and never applies them silently, with a dry run and a log of everything it did
- Subtitles beside the file with a language code, and the audio and subtitle tracks correctly tagged inside the container

HARDWARE TRANSCODING, WHICH YOU CAN JUST HAVE
- Enable the graphics hardware in the container: on Intel that is Quick Sync through the render device, which is the single most cost-effective feature here and needs no subscription in Jellyfin
- One modest Intel machine will transcode several streams at once. A machine without it will melt on one
- Better still, avoid transcoding: store files in a format your clients can play directly, and check what each of your devices supports. Direct play uses almost no processor at all
- Measure it. Play a film on every screen you own and watch whether the server is transcoding; fix the file rather than buying a bigger machine

THE LAYER WORTH BUILDING
- requests: id, kind, title, year, external_id, requested_by, status, note, created_at, fulfilled_at — a page where the household asks for something rather than telling you
- Approval, and a notification when it arrives
- library_items: id, kind, path, size_bytes, container, video_codec, audio_codec, resolution, bitrate, duration_ms, has_subtitles, added_at, last_played_at — your own index of what you have and in what shape
- A report of files that will force a transcode on your worst client, which is the maintenance list that matters
- health: missing episodes in a series, files that fail to probe, duplicates, orphaned subtitles, folders that do not match the convention
- watch_history mirrored from the server, so 'what have we not finished' survives a rebuild

STORAGE AND KEEPING IT
- One directory tree, on a filesystem you can grow. Media is large and grows forever; plan for that rather than being surprised
- Backup the metadata, the watch history and the configuration nightly, off the machine. The media itself is usually re-acquirable and enormous — decide deliberately whether it is backed up, and write the decision down
- Verification: re-hash a sample on a schedule and report anything that changed
- Never let the disk reach full. A media server with no free space corrupts the library database, which is the one part that is not re-acquirable

ACCESS
- On your own network, over a VPN for away-from-home use. That is the simplest and safest answer
- If it must be public, put it behind proper authentication and rate limiting, and understand that you are running a public video service
- Per-user libraries and parental controls if the household needs them

WHAT NOT TO DO
- Do not write a client for a television. It is months per platform, the tooling is unpleasant, and the result will be worse than the one that already exists
- Do not build your own transcoder
- Do not put your library on a service that scans it

WHAT MATTERS MOST
File naming and hardware transcoding. Fix the naming convention across the whole library before anything else, then enable the graphics hardware and confirm every screen in the house plays without transcoding. Those two afternoons replace most of what the subscription is for.

Give me the compose file for the server with hardware acceleration enabled, the renamer, the request page, the health report, migrations, .env.example, and a README that opens with the client check.