I love pomme 🍏🍎
Honestly, the 60% is a rough estimate, not a precise measurement. I still spend a lot of time reviewing the docs. AI is especially handy for the screenshots and repetitive parts, which would take much longer to put together manually. It just makes that side of the work a lot quicker.
Hey, thx for the kind words about the project, and for taking the time to write this up (˶ᵔ ᵕ ᵔ˶)
Just to clear up the NewPipe and FreeTube part: TypeType doesn’t use them, we run on PipePipeExtractor, and I contribute upstream to PipePipe whenever I can.
PipePipe is a different project with its own services and its own direction, and we have nothing to do with them these days.
On the generated code, I totally get the concern. It’s not a mass of code nobody ever looked at, I read what ships, I test it, and I maintain it myself. These days I also maintain it with other contributors, but I still review what goes in, for example https://github.com/TypeType-Video/TypeType-Frontend/pull/62
The disclosure is just me being honest about the tools involved. I understand why people focus on that first, with all the AI slop out there, but it’s a bit frustrating sometimes. There’s a real history behind the project, with reviews, tests and releases.
I try to stay as transparent as I can. And honestly, at the end of the day I have a great community, friendly and supportive, and that’s enough for me.
I just share the project to offer an alternative. Concretely, I have nothing to gain from it. As long as I’m alive, TypeType will be properly maintained ♡
Hey, sorry for the slow reply, I’ve been really busy lately!
Thanks for taking the time to write all this up, I really appreciate it, the NUC setup sounds genuinely cool, and I hadn’t thought about TypeType being used that way.
I can’t promise I’ll build a dedicated client myself, but I’d love to hear the other ideas you mentioned.
Maybe someone will have fun experimenting with the API and mpv someday :)
Thanks ! :)
On stability: extraction comes from PipePipeExtractor, and I am still involved in that project. I learned how to fix things fast from the creator of PipePipe, who taught me a huge amount and I owe him a lot. So when YouTube breaks something, the fix lands there and I am close to it.
On third parties: yes. The server has a documented API (OpenAPI in the repo), and the playback engine is published on npm and JSR as @typetype/mse:
https://www.npmjs.com/package/@typetype/mse
There are RSS feeds too, and import/export with a dozen other apps and a shit load of other feature on TypeType thx to all the contributors ! The Android client is just another client on the same API, and is by far for me ofc the best client to this day to consume TypeType.
Tell me your use case, I am curious :)
Thank you ! :) Actually there is a one-command installer, and it is the recommended way:
curl -fsSL https://typetype.video/install | bash
It creates the stack folder, generates the secrets, picks free ports when the defaults are taken, sets up the object store for downloads, and starts everything. It is interactive too, so it asks before pulling images and starting containers.
The quick start is here: https://typetype-video.github.io/Docs-TypeType/self-hosting/quick-start
And if you want to understand each part instead of running the script, the manual step-by-step is right there as well: https://typetype-video.github.io/Docs-TypeType/self-hosting/docker-compose#manual-setup
If you are ok with the command line, you will be fine with this :)
Keep in mind this instance follow the beta line, if u want a more stable experience there is this instance : https://watch.typetype.video/
Hey ! :) No YT account needed, it works like FreeTube. Everything stays anonymous, and the extraction happens on your own instance instead of going through a Google login :)
Only thing to keep in mind, same as the other projects: since the requests come from your server and not from your browser, YouTube can sometimes throw a bot check at it. But no account, no Google, nothing to sign into :)
You can try TypeType at : https://experiment.typetype.video/
Hello, as stated here:
> Criticism is welcome.
This document reflects my current position; it is not dogma.
If you think something here is wrong, please open an issue, start a discussion, or contact me. Explain your reasoning. I prefer honest criticism to silent compliance.
Back to the subject, no worries at all !!! I don’t want to get into a confrontation with you, If you don’t like this project, feel free not to like it thank God you still can!
I don’t want to impose my AGENT.md on everyone, everyone should be free to use their own freakingAGENT.md!
So, honestly, I’m not sure what else to say. Regardless of what people tell me, TypeType is a part of my life. I’m proud of our community, our contributors, and our users. I try to be transparent, and I can make mistakes. But the problem with AI is that, quite often, if I respond, I’m wrong; and if I don’t respond, I’m also wrong !
So I’m trying to make a few constructive attempts, but I don’t want to insult or disrespect you. That would be pointless and unkind :(
Honestly, I can only wish you a good day and all the best in life :) !
(And we are a little bit older than 3 month ;p)
Thanks, glad the allow-list sounds useful! I meant the whole stack on one box: the app/server, Token, Postgres, Dragonfly, Downloader and Garage. There isn’t a separate media-proxy container; playback goes through the Server.
I was too optimistic with the 2 GB figure in the post. For a small personal setup, I’d go with 2 vCPU, 4 GiB RAM and about 20 GiB free disk space (SSD preferred, but HDD should work too). For a few users and regular downloads, 4 vCPU, 8 GiB RAM and about 40 GiB free disk space would be a better starting point. Downloads use disk rather than RAM, and need extra room while they’re being processed.
The sizing details are here: https://typetype-video.github.io/Docs-TypeType/self-hosting/prerequisites