But HA already does that. It’s what Addon Store is, essentially. It just does not exist for people running HA in Docker. If all addons were just third party software, like SSH server or Nginx server, then there would be no question that this is not in generally useful to be republished.
However, those are not the addons that I am talking about. I am talking about addons that add (sometimes essential) functionality to HA. Like Zwave JS. They are not just standalone third-party Docker containers. The addon part configures the third party container in a specific way and it also configures the HA in a way that enables the HA to use that new container. Be it config files, network ports, user credentials, … That is the part that is non-trivial even for very experiences Linux system admins, because it is all HA-specific knowledge that is not documented (for most addons) and is instead hidden deep inside Addon Store or Supervisor configs and scripts.
I am open to contributing the knowledge gap to the documentation, if a programmatic solution is too hard to implement. How the system is working does quite heavily influence what could be the best way to document this. Also it would be best if the documentation could be made in a generic way so that it can be easily transferred to all new addons as well by saying things like “then you look at the X field in the addon definition and that will be the name of the Docker container to run”. And at that point any sane (read - lazy) developer and sysadmin will be asking “why isn’t there a simple Python script that can do this for me?”.
This is an interesting debate, but ultimately this is missing the point. I’ll rephrase in my own words what I think many have already said or implied.
It’s not about easy vs hard, but rather about a specific goal. This makes it fundamentally a design choice. The docker method works the way it does, because it is intended to work that way, and that was a human choice and not a technical reason. While there is parity by all installation methods running HA core, the divergence beyond that is there for a reason. Trying to align more of these would lead to duplication in effort and larger overlaps, which in my view doesn’t seem to be the intention (or valueable use of time of the core devs). With the installation methods that exist, all the bases are covered when it comes to skill levels.
Disclaimer: Note that I run HA core on bare metal in a pyenv. I have never tried or needed the other methods so far, but I’ve been hanging out for a number of years in the forum now and have followed a number of these conversations.
That’s just plain not true, I’m no Linux expert and all of the configuration files, ports mapping, and credentials are all documented.
You mentioned earlier that you haven’t used any addons / set these up yet, but you speak of it like you know how it works which is contradictory.
I’ve never once had to read source code to get things setup and I currently run 20 docker contains half of which connect directly to HA. This thread seems to be going in circles, and it seems to be the general consensus that no source code needs to be read to set these things up.
There is. Its name is supervisor. It’s even written in python. That is the tool the HA team publishes which allows a developer to attach a config file to a stock 3rd party image to add all of this functionality to it
Along with all the other infrastructure things the HA team has identified as needs for HA users (backup, restore, watchdog, update monitoring, log access, network config changes, etc)
If you do a docker installation you are making the explicit choice to run without this tool. Which means any of the things the tool does that you like you have to do yourself.
I feel like we’re going in circles so I guess I’ll stop here. As Taras said much earlier your welcome to make any FR you want but this one seems quite unlikely. At least by the HA team, Community Guides are always welcome.
Going through this thread I really feel maintainers are missing the point. So I created an account just to explain the use-case step-by-step and why it is a real issue.
I built a custom REST API that tracks internal information on my server.
I am running my HA server in a docker container.
I google “Use REST API as sensor in Home-Assistant”
I can’t install it because it’s an addon and I am running HA in a container. There is no docker image for it, I just get the code.
This is not an “expert” feature, this is just an installation with reduced features where I can’t do simple things like pulling data from a REST API.
The installation page (Installation - Home Assistant) should make it clear that HAOS is the right way with stronger wording than:
Home Assistant offers four different installation methods. We recommend using one of the following two methods:
Home Assistant Operating System: Minimal Operating System optimized to power Home Assistant. It comes with Supervisor to manage Home Assistant Core and Add-ons. Recommended installation method.
Home Assistant Container: Standalone container-based installation of Home Assistant Core (e.g. Docker).
Where the tradeoffs are missing entirely. It even lists both approaches as “recommended” with HAOS having a second “Recommended” in it’s description.
Hi !
They RESTful sensor (the thing you linked to) is not an Addon, it’s an integration. Furthermore it’s a core integration shipped directly with every HA installation per default. To use it you really only have to follow the very documentation you linked to. You don’t have to install anything additionally.
It’s not an add-on, it’s an integration (look at the URL). All (official) versions of HA runs the core, which includes all the built-in integrations, such as this one.
Just to be clear, integrations and addons are different things. Whatever you found under the integrations Integrations - Home Assistant (home-assistant.io) page, are all part of the core - you’d get all of those regardless how you install your HA.
Add-ons, on the other hand, are services that are “related” to your home setup, and only applicable to HAOS install method. If you want a different database? That’s available as an add-on. Plex? That’s available as an add-on. You want Node-Red for automation instead of the built-in automation editor/engine? That’s available as an add-on. You want to do Cloudflared so that to connect to your HA services from outside? That’s available as an add-on.
In the background, addons are containers also. They are just packaged specifically for HAOS, so that to smooth out the learning curve for users who choose HAOS.
== An example ==
One might ask: wait a minute, what if I go container-based installation, and want to do Plex? Couldn’t I just install Plex server somewhere else myself?
The answer is sure, of course you could, and that’s the point: you very likely could find equivalents of those HAOS add-ons available somewhere else in a different format, and you could totally do that yourselves if you want to.
Now, what if I want to integrate my HA to my Plex server (that is in another metal box)?
Sure you could do that also. Search “plex” under the aforementioned integrations page, and you will find how you could connect HA and Plex together, and build your own interesting automations.
While that is all entirely true and correct, I think that, based on the context of this thread, that an even finer point needs to be put on it…
Yes, those things are all add-ons and are related to HA OS (and HA Supervised as well) BUT All of those things are also available and can be installed as their own Docker containers even without the use of HA OS or HA Supervised.
Add-ons only provide a way to install and configure those Docker containers from within HA itself.
They are fully available and functional with HA without the Supervisor. You just need to install & configure them yourself.
Apart from the “Supervisor’s” automation for setup, there is another point to consider when choosing HAOS/HA Supervised versus HA Container:
In case of HA Container, the HA ingress functionality does not exist. So, if the addon provides GUI front-end, it can be included in HA, but only as an external web-site via i-frame (I think).
Say, I would have five extra addons.
If I want to be able to use their GUI from public internet (like I do access my HA core), then the requirement for authentication (part of addon? Nginx proxy simple authentication? …), firewall, and port-forwarding significantly increases the complexity of the setup.
I am running HA as a container and I would like to understand how to configure Warema WMS Integration as an addon. HA OS or supervised installations are absolutely not an option for me, so here I am trying to figure it out and learn how to configure and run these containers separately through docker.
There are people like me which are not confident with docker and Linux using it.
I reuse my jeedom atlas box to run HA. But I didn’t know I ll not be able to use adding.
I ve installed hacs but it cannot provide me everything.
And now I want to install a theme but I need to update a yaml configuration file using vscode and I don’t find a solution anywhere.
Just wanted to update this post, as I came here looking for this answer. I run HA in a docker container as well and wanted to try the Wyze addon via HACS. I used the instructions on HACS to get it going, which was a snap. (Download | HACS)
Open a terminal
Go inside the container with docker exec -it homeassistant bash
Run the HACS download script
wget -O - https://get.hacs.xyz | bash -
FWIW I do have a local directory so it persists across containers/host reboots.
Wow, this got heated.
IMHO, the maintainers are making a mistake trying to develop an OS supervisor. The requirements for installing home assistant supervisor are just… draconian and very limiting. Why can’t the supervisor just manage docker containers and host system configurations if that is needed by a container instead of just taking over everything on the host? Why does it require network-manager, for example? Are you trying to develop a web interface that replaces X server or are you developing a home automation system?
I have to agree mostly with the OP. Today’s tendance goes to micro service infrastructures and there are good reasons for that (no virtualization layer, interoperability, compartmentalized concerns & security, etc.).
VMs are still used a lot today because there are a buttload of tools & OSes to manage them, but their usage should slowly decrease for the benefit of containerized applications.
I also agree with @chedim about the supervisor, there will always be dependencies for all applications (at least database, files, network, etc.) and an application developer should try to open his app to allow these dependencies to work with it rather than including them in a home made OS because it might be more and more difficult to maintain.
I originally chose HA over Jeedom because it had a nice Docker image available (that’s the first thing I check when choosing a service), and I hope it’ll stay as powerful as today.
Maybe nothing related but I’m a developer for >20 years like @aigarius
Jumping in here… I’m on the side of the OP on this. I think this is ridiculous that the maintainers here are choosing to build their own operating system for this. This is counter to current technology trends and frankly is probably sucking a lot of time and resources from core features. Let OS developers develop OS’s. You guys are doing NOTHING special that requires an entire OS. If you want to be able to manage docker containers for your environment you guys should look at docker in docker.
Float a core docker image to setup HA, supervisor etc, then run add-ons within that is DIND containers. Problem solved, and if someone wants to dedicate a whole OS you system will still work even on bare metal.
Sounds like you guys have gone way down the wrong path and are making excuses as to why you cannot go back to something easier.
Making an app that only runs on a specific OS, specifically developed for the app is a 1970’s approach to computing.
!