A self-hosted web app that turns YouTube channels and playlists into audio podcasts.
- You deploy Poddla on your own server
- You create podcasts in Poddla's web interface
- You subscribe to those podcasts in your podcast player
- Your podcast player lets you know when new episodes appear
- You listen without having to deal with YouTube
Poddla only ever fetches data from YT when requested to do so by a client. There are no update schedules etc. This means no unnecessary resources will be used. Additionally, Poddla minimizes concurrent requests to YT. If a podcast player downloads two uncached episodes concurrently, they will be downloaded from YT one after the other.
This is all quite overkill when I could have simply created or vibed a simple python script for downloading episodes and run it as a cronjob. Actually, there are probably a lot of those on GitHub already. My motivation for building this app wasn't the app itself. I wanted to experiment with some tech, set up some patterns for how to do things while finishing the project in a reasonable amount of time. I have some more interesting ideas I'd like to work on next, and this project will serve as a starting template for those. That said, it is quite a bit nicer than a cronjob and I'm happy I created it.
I've written a little bit about it on my blog
This applies to most Podcast players, but I'm sure there are exceptions: Poddla's podcast RSS feeds need to be accessible on the internet for podcast players to use them. Most podcast player apps don't directly check podcast feeds. They instead offload that work to the app maker's servers. Having to make the feeds accessible over the internet makes hosting Poddla slightly more serious than if it would only be available on one's home network.
Using Poddla requires a login that is created on first startup, you'll see mentions of this in
deploy/docker-compose.yml. The RSS feed and audio file downloads do not require this login,
because that would not work with podcast players. Instead those two use personal "secret feed
tokens" that are hard to guess and are automatically included in links, you should never have to
manually type them. This is security by obscurity, and how private podcast feeds are generally
published. If you expose the RSS feeds on the internet and somebody gets a hold of one of them, they
will be able to access it without any extra login. That applies to the audio download URLs too. I
recommend only exposing the RSS feed on the internet, not the audio downloads or the rest of
Poddla for that matter.
I host my Poddla instance in my homelab and have configured my web server to only allow traffic to Poddla from my home network, except for the RSS feeds which are open to the internet. I have to be on my home network or on my VPN to download episodes on my podcast player. Other options include finding and using a podcast player that fetches feeds directly or running a cronjob that mirrors the feeds to a separate server/CDN on the internet, but that's not as convenient.
Have you tried naming things? :)
Podda is Swedish slang for "to podcast", and the extra l is there to make it even sillier.
I have not published a docker image, but you can build one yourself. devenv is required for that,
see the Development section for details about devenv. If you don't want to deal with Nix and
Devenv on your computer, you can use a VM or create your own Dockerfile.
This will build a docker image and load it into your Docker under the name poddla:latest.
I've only tested this with Podman, but it should work with Docker as well. See .env.example for
what to change if you use Podman or Docker.
devenv shell
docker-build-loadPlease see the deploy/ directory for a sample docker-compose.yml file that will run a Poddla
instance on https://localhost:4444/. It may serve as a starting point for your own hosting setup.
The sample uses localhost, because I don't want to make it too easy to set this up on the
internet. Please read the "Security" section above and protect your deployment.
There's also a command to build and publish docker images (amd64 and arm64) to a registry, I use it with my forgejo instance.
devenv shell
docker-publishSee .env.example for how to configure the registry URL.
The environment assumes an installation of Devenv. Devenv sets up the correct Python, Valkey, dependencies and locks the versions down. This means you get the same version of everything in the docker image as in your development environment and any other developer's environment. Devenv is also able to create devcontainers but I haven't tried that yet.
If you don't want to deal with Nix and Devenv on your computer, you can set it up in a VM, or install everything needed manually.
When you've Installed Devenv, you can run devenv shell in
the root of this repo. It will install all the dependencies and make all devenv scripts available in
your shell.
See the devenv.nix file and/or run devenv info for details.
- Poddla is a Django project.
- Datastar is used to make the hypermedia driven real-time UI: when relevant state changes, new HTML is sent to the browser.
- Open Props UI is used as the basis of the CSS.
- Valkey with Valkey-GLIDE is used as a message bus and cache.
- SQLite is used as the database, through Django's ORM.
- Granian as the ASGI HTTP2 (h2c) server
- yt-dlp Is used to interact with YT
Create your own .env file based on .env.example.
Start Valkey and Caddy. This will start process-compose in the active terminal, you'll want to leave it running.
devenv up valkey caddyStart web in another terminal:
devenv shell
webPoddla is now available on https://localhost/ .
Django's manage script is available as manage inside devenv shell.
devenv shell
manage check
manage makemigrations
# etc.devenv testThe project uses uv to manage python dependencies. While inside the devenv shell, you just use
uv normally.
devenv shell
uv add your-new-dependency
uv sync
# etc.You may notice there's a package.json with vite, prettier and open props (UI). pnpm is set
up in devenv.nix. At this point the reasons for this are:
- Install and bundle open props and open props UI.
- Prettier for formatting.
devenv shell
pnpm install
# etc.An LLM has been used for:
- Coming up with ideas for solutions and rubber ducking
- Annoying debugging problems
- The HTML and CSS have been mostly wire-framed up by an LLM according to my instructions
- The tests are mostly LLM generated
- This text has been checked for typos by an LLM
Poddla's features, structure, and technical design decisions are mine.
I don't expect to get any PRs for this project, but if you do want to submit one, I urge you to understand what you submit.