It took about 6 months, the repo only went public yesterday. I actually held the release back to run vulnerability testing first. Totally fine if it's not for you.

That stack is for acquiring media, not for streaming it.

I tried them all, but its a PITA for non-English media, so back to qBittorrent.

I don't bother with any of those either.

If you like this project, ping me, maybe we can collaborate.

They don't all support it well, however. My Blu-ray player, for example, mangles filenames about 10% of my media. And my TV just simply doesn't list all the files in a directory, regardless of size.

I spent some time building a .NET service that uses UPNP/DLNA to offer media to local devices. I got it working reasonably well, but the problem is that the client interfaces are absolutely garbage.

The better architecture is to push content over HDMI using an endpoint you control. This doesn't have to be the actual media server. We could invent a new system role and control multiple endpoints centrally. I don't want the TV to do anything more intelligent than display the video and audio frames it is receiving.

This is my take as well.

They're sticky because they stay focused on their specific domain, and they work well.

I also really don't understand the value pitch on "6 apps no one can use" in the post.

I publicly host Seerr and Jellyfin for my family, and they use that combo just fine. They explicitly don't get access to see any of the sausage making under the hood.

Given this doesn't support music yet (according to op) and you still need to host plex or jellyfin... what is the value statement here over seerr?

Over the years we didn't have AI. Maybe it will become abandonware; but maybe the development will keep up thanks to AI.

For example, I've been using a Spotify replacement called Spotifast for a month or so. It's way better than the official Spotify app, and it's all vibecoded by one developer, and it gets lots of updates (it's basically a 1-to-1 clone of Spotify, plus additional features). It wouldn't have been possible before AI, or at least not that easily.

I'm working on one myself, but I probably won't ever release it to the world. I just want one that's simple enough for my wife to use, so we can cancel netflix etc. I need replicate the experience of browsing what's new or popular on her phone, watching trailers for them, and then a single button to download it and getting notified when it's on plex.

https://xkcd.com/927/

That wasn't written by a human. It was written by an AI. AI does this weird thing where it will say things like this - "AI or MCP features. Not in 1.0" - when the user prompting it hasn't actually requested whatever it is the AI has hallucinated as not being included. I've been using Claude Code to update some internal tooling I use (basically a bunch of python scripts I've painfully hacked together for myself that automate stuff) and it often comes out with things like "Not in this version: XYZ" when I have never even asked it to build XYZ.

is this a serious question? It would make it easy for someone to expose an SMS bot to family and friends, for starters. so they can have family request things in a convenient way?

For example my mom, who is tech illiterate, could send an SMS like "please add the new spiderman movie to the media library." And it magically shows up on the apple tv.

Or we could use it to do a "lazy" search on the library. Jellyfin in the year 2026 is still too stupid to understand that when i search "spiderman", the result "spider-man" should be included in the results. God help us lol.

Or maybe you could easily run codex/claude against your media server to help with routine maintenance. Like "hey claude we're a bit low on disk storage. Please find movies/shows that were requested a while ago and are large in size." then if you're happy with the list, ask claude to delete those for you.

Idk...i could think of like 20 more uses cases. Use your imagination. :)

I would use it for a regularly scheduled validation job.

Sometimes a bad copy, mismatch, or low-rated media slips through all the filters. Could use an LLM to fix the media and harden the filters to prevent issues going forward. I’ve probably spent weeks of my life toiling and tweaking over the last 15 years.

Not the OP, but I see a lot of complaints on HN about dumb media player "voice control". I myself had cringe experiences with my car when I wanted to play a particular genre or song.

"What's that movie with the kid who messes with those guys who try to rob his house during Christmas? Do I have that one?"

Natural language searching seems like a good reason.

To generate/translate subtitles.

I want everything AI capable. Just allow me to plug in my own selection of model.

Downloading subtitles with STT fallback via whisper then summary so that you don't even have to watch the show anymore, way more efficient. /s

There is no long term plan, as with anything vibe coded

> Do I keep qBittorrent or SABnzbd? Yes, for now.

What do you mean "for now" if there's one thing i don't want it to replace it's clearly qbittorrent. Trackers have client whitelist and I don't see a brand new vibecoded app being added to any of the serious ones

To be fair, it's an entirely different language and they had an automated test that forced the AI to keep the binary small.

Your assumption that 97% of a binary isn’t usually bloat is not well-founded

It doesn't actually include Prowlarr for starters

This also happens with Sonarr / Radarr as you can't easily get a IMDb API key, and it works automatically because of this

What stands out to me isn't only that the repo was made less than 20 hours ago, but that they've released from 1.0.0 to 1.0.10 since then.

But at the same time the PR numbering is much older, with 1.0.10 coming from #197. It makes me think the repository was private for some time and then made public recently. But still -- 10 releases in a few hours isn't a stability signal. Especially when one of the advertising points is to simplify the update cycle from all the "arr"s.

I don’t get this. So today he can build the experience he wants with little effort (he likely spent lots of tokens). You made no judgement in the quality or on his experience, only on the age of the repo. Why exactly is that?

I get there will be a lot of crapware, but you’re hackers no? Why all the gatekeeping even amongst yourselves? How much time did the openclaw guy take to built it? A weekend?

Sure, you could say “no thanks”. But all the labs certainly thought it was useful enough to replicate. And yes, it was, still is an insecure mess littered with bugs, but why are we all so repulsed?

It’s a genuine need. I build my own version of this, also on go, haven’t shared bc I am not sure how legal it is to distribute, but I have been using it for months and definitely beats running 6+ apps that keep breaking. I run one binary + jellyfin, never going back.

tiresome

Perhaps you will wait for the dry-aged repo.

I believe things will be from now on. People see a gap in the offerings out there and they just whip it up. There will be a certain amount of stability that the LLMs will add by default to all of them and that's it. If you are interested, go ahead and use it.

The repo is AGPLv3.

It's incredibly common to have a CLA because without one accepting PRs is a mess legally. The PR authors have to acknowledge that their specific code is also AGPL every time. Easier to just have a CLA.

Honestly a human can only consume so much TV and music in a day, why these super automated acquisition setups are needed is beyond me. 30min on a Saturday morning manually grabbing magnet links is enough to keep the whole house entertained for a month.

Hi Luis. Cool approach and thanks for sharing - however, I am a bit disturbed by the built-in reliance on your metadata server. Self-hosting/homelabbing would do better to have something the user controls himself/is hosted on something like IPFS.

I'm interested. Does it handle subtitles translation with AI ? This is a must have for me.

First try to obtain subtitles in the locale I need, fallback to using AI to translate the embeded ones, fallback to downloading ones in original language / english and translating.

> I had 6 apps nobody but me could use

That’s how I feel. I’ll give this a try.

How did you review it?

>the image has no OS and no shell What are you (Claude) even trying to say here?

Also, do you genuinely speak in AI-isms, or have you just had claude generate you this too and uncapitalised the first word of each sentence so it looks human?

> reviewed everything

No shot. The initial commit was 551kLOC followed by 100kLOC+ within a day. There's no way you reasonably reviewed _everything_.

> i built it

Claude did

you could have at least written the comment yourself. stopped reading after the 2nd sentence.

> im in the comments

I'm not convinced this is true, could you expand further what you mean here?