Home Assistance automation vs custom scripting

[Falcosc]

Thank you for all your contributions to this project.
My question is about CLI vs Home Assistance. Nearly no integration has an CLI, the closest non CLI type integration is HTTP server bound to localhost as proxy for outgoing requests.

I have natural gas heated water and AC cooling with heating option in winter, but not enough solar power to heat my home and our electricity is 2 times more expensive than my old natural gas heating or even 2.7 times compared to a more efficient gas burner.

I am a bit hesitated setting up Home Assistance, because it's too large to avoid unexpected issues with complexity, unwanted open access or unwanted breakdown during updates.

Let me explain what input my automation needs to explain why home assistance feels too complex:

• smart meter values to detect having too much solar energy
• battery soc to estimate energy need before having too much solar energy
• connectlife air conditioner cooling to use unused solar energy in summer
• connectlife air conditioner heating - disable if outside temperature is too low for efficient electric heating to rely only on gas heating
• smart heating thermostat to reduce gas powered heating if outside temperature allows more efficient air con electric heating

I already have CLI access to smart meter and battery soc, but connect life ac and thermostats are missing because I didn't decide which AC and which thermostats I want to buy yet.
Hisense got recommended by my contact person, but the newer models require this connect life project. That's why I did open the discussion here.

Which means I basically only need 3 sources: smart meter+battery CLI, connect life and radiator thermostat. I don't count the internet for my outside temperature data source.
Does it make sense to set up home assistance for 3 sources with all the application management which comes with it, not to talk about all the issues I am not aware of if a nobody just installs home assistance.

I would be more comfortable with a bunch of scripts in crontab with my own hysteresis implementation, instead of managing to have home assistance services and add-ons online without breaking on update and without putting Home assistance accidentally open into internet.

Would it make sense to get started with home assistance, or is it overkill or too much effort to avoid home assistance issues?
It is difficult to judge, since people which are not comfortable with non gui driven automation exist in way larger numbers and might not care about potential issues which such a complex solution.

I hope this isn't too far off-topic, my main question is in the direction of why HTTP server instead of CLI?

Kind Regards
Falco

[oyvindwe]

I'm not sure what you mean with HTTP server in this context? Most integration uses an upstream Python library to integrate with Home Assistant, including this one.

Could you use the connectlife library to build a Python script or your own CLI?

The library is available on PyPI as well.

pip install connectlife

[Falcosc]

Oh, yeah, I forgot some references. Thanks for your response, glad to see your care about the project.

This https://github.com/bilan/connectlife-api-connector is something between setting raw parameters in

async def update_appliance(self, puid: str, properties: dict[str, str]):

https://github.com/oyvindwe/connectlife/blob/fe56f6ea2350ac13ab75938cd8544db67437868b/connectlife/api.py#L89

and a full home assistant installation.

What is the difference between your project and the connectlife-api-connector?

Would you share any concerns about home assistant's complexity, or is it stable enough to just keep an old installation without maintenance as long as no new devices get connected? Not having a CLI for scripting and so many HA plugin development feels like I misjudge how many risks a home assistant installation has compared to a small setup of scripts which has less code to review.

Without knowing anything about Home Assistance, HA feels to me like installing an PHP Server just for triggering a single line command and wondering in 5 years why it doesn't work anymore after being hacked by a low effort attack caused by a miss configuration of the default server setup.
I mean, I just fear endless efforts of harding the HA installation or having to deal with deactivation of auto updates to avoid getting feed with a "lets open my HA homepage to the web" feature update.

Or it is easy to get a local only no updates HA installation done?

Thank you for your fast answer.

[oyvindwe]

That context was helpful. connectlife-api-connector runs as a standalone process in a Docker container, and communicates with Home Assistant using MQTT. This is called an "add-on" in the Home Assistant ecosystem.

I really only wanted to integrate my dishwasher to be able to monitor remaining time and get a notification when it's done, but I struggled to customize the add-on to support it, for various reasons. I also wanted to build a native integration in Home Assistant, so I wouldn't have to manage yet another docker image (I'm not running HA OS or Supervisor). From there this project grew, and unfortunately I have not been able to support it as much as I'd like to recently.

I've been running Home Assisatant several years. I find it quite stable, but I update monthly as development is quite rapid, and updating severral versions at once can be painful wrt. breaking changes. They are announced, but you do need to read through the release notes carefully. I've alse heard about people that once they have a functioning Home Assistant installation, they are happy to never touch it, but I don't think this is a good idea if you use any cloud integrations, e.g. the ConnectLife API, as integrations are usually updated to support API changes.

Your use case using solar energy and batteries to control appliances should be fairly easy to implement in Home Assistant. I've seen several discussions on the topic. E.g., in Norway there is a lot of people using Home Assistant to optimize electrical consumption based on hourly spot prices while also limiting max consumption per hour.

Also note that Home Assistant is implemented in Python, not PHP. It relies on async/await, and runs quite fast on limited hardware.

But again, if you just would like some simple scripts to control your ConnectLife appliances, you can write them in Python using the library.

[Falcosc]

I just mentioned PHP because it got so popular that old installations become a well known issue because of all the security concerns and missing hardening.

I have read from the breaking connect life changes, which didn't have to do anything with API and just was a new license agreement. It would be nice to just update the broken API in case the connected cloud server changes.

Having anything more than just a script has always the risk to end in dependency hell. That's why I want to get a feeling for the situation before I commit.

Is it possible to update the ConnectLife parts without updating home assist or if I need to maintain the HA installation, is it at least ok to keep it outdated as long as I don't need any update. I just hope that the default installation is already good enough because I don't want to go the "maintain a public server and harden it" route. Would be nice to just have a local network only installation with lots of security issues nobody is able to exploit.

Fresh HA installations if stuck in dependency hell would also be ok for me, since I only have so little interfaces to care of as long as the HA script language is stable enough to migrate across many versions.

From the outside, it feels all so wrong, but seeing so many users on it doesn't confirm my concerns about the risks and efforts needed to maintain HA.
If there is an easy path, I would commit to HA, just need a little push from somebody like you to get a feeling of what to expect.

[Falcosc]

Wouldn't hurt to limit HA to local host only and use SSH tunnel to temporarily connect for maintenance.
Having such a well known system could lead to situations where you get attackers into your local network because a different application got exploited and is looking for HA installations to have fun with it 😅
So wouldn't mind to setup a localhost bound HA if it's easy to be done. At least, I hope that's the least amount of hardening effort I can spend there, since I don't trust authentication solutions of outdated systems.

[oyvindwe]

Yes, I was running HA without external access for a long time. For remote access I used a HomeKit bridge that connected to my Apple TV. (Apple only allows remote access to HomeKit if you have enabled 2FA on your Apple ID, and it's still an explicit choice.)

Only connecting from remote using SSH or VPN is quite sensible, and also quite common AFAIK. I now use NabuCasa for remote connection, but it's a bit more of a hassle to set up 2FA in HA. Also HA does not have secure multi-user support IMO, but that is necessarily not a problem.

[Falcosc]

Yeah, remote is already too much of a hassle for me. I meant blocking LAN access since this is maybe a bit uncommon and could cause issues if the server interface is bound to localhost instead of LAN if some UI solutions don't expect it.