Using ChatGPT In Your Smart Home for $0.01 a MONTH?!  

You should really look into wyoming

On my list!

Thank you for the kind words!! That’s exactly what I was going for.

Love it! Hope you enjoyed it!

Thank you so much!

Thank you!!

Awesome to hear. So glad it prompted some ideas!

Thanks for the tips! Where did you see the API key? I’ll make sure to regenerate! Edit: I found the spot I missed! Thanks for the tip.

Thank you!!

Glad you liked it! It can be a bit intimidating. But you’ll get the hang of it!

💯

Thank you! Thanks for watching!

Glad you enjoyed it!

This is a great point that I never tried! Will be adjusting mine as well for the future!

Can you detail a little bit how this was done? I'm only seeing how you would pass text along. Thanks!

@@jpcdn80 That's the neat part: It's all just text. From the video at around 12:18, in the Conversation.Process service, you can drop the whole "Current Conditions are below" with all the brackets. Instead, have the text: part just be the response variable from the "get forecasts" service call. Also, in the Integaration settings for Open AI, I have it configured just to deal with the weather. My system prompt is "Parse Home Assistant weather data and describe today's weather forecast using simple English. Don't describe tomorrow's weather or any other future date forecasts. Only use US measurement values. When describing temperatures, only use fahrenheit. Don't use celsius or kilometer measurements. When describing wind speed, round to the nearest whole mile per hour. Be fun, and use occasional jokes or sarcasm to make the response seem more genuine and unique." There are a few more tweaks I did that are hard to type out in a comment here but hopefully this gets the gist of it.

@@TheSlowestZombie would you mind copy/pasting what you have in the text block of the conversation.process?

@@welbeschikbaa Fingers crossed that the formatting works out. Here is my entire script: alias: Weather Report sequence: - service: alexa_media.update_last_called data: {} - service: weather.get_forecasts metadata: {} data: type: daily target: entity_id: weather.forecast_neptune response_variable: dailyforecast - variables: weatherparse: | {{ dailyforecast['weather.forecast_neptune'].forecast[0] }} - service: conversation.process data: agent_id: conversation.openai_conversation text: "{{ weatherparse }}" response_variable: openairesponse enabled: true - variables: plainresponse: | {{ openairesponse['response'].speech.plain.speech }} enabled: true - delay: hours: 0 minutes: 0 seconds: 0 milliseconds: 500 - service: notify.alexa_media_last_called data: message: "{{ plainresponse }}" enabled: true mode: single icon: mdi:weather-sunny-alert

Glad to hear it, Thanks for watching!

Thank you for watching! I’ll definitely do more videos like this in the future. And yes, you could create an automation like this. You can also have a conversation with it about your house and all of the sensors.

@@FutureSmartHome Thanks. That would be a great tool to have for HA. Not surprisingly I have entities that go offline from time to time and getting a notification right away would be extremely helpful.

It could totally do a morning briefing of your calendar. It’s not local. And this was before 4o

Thanks for watching! Check out the pinned comment on this video - I link to my substack which has some snipped.

Thank you! Here is a list of models that are available and the corresponding value that you can type. platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4

@@FutureSmartHome Thank you so much!

I might do that too! Thank you!

@@FutureSmartHome is it also possible to give the OpenAI Integration controll over several entitys to control them?

Yes. In the examples I showed, I’m sending data from multiple different systems.

@@FutureSmartHome Ah sorry i just watches the Kids weather forecast. Great, thanks!

Mobile notifications are limited in the amount of text you can send to the device via push. No way to change that.

Totally valid thing to be concerned about. Send me some of those stats if you come across them. I’ve heard training one of the models is the equivalent of 6 cars worth of lifetime emissions.

@@FutureSmartHome I tried to link some stuff but RU-vid deleted my comments, probably because it though it was spam. You can just search for "ChatGPT water" and you will find several Articles from reputable sources likes Forbes about the topic. Also you can search for Google intent to put a server farm on Chile where the access to water was already a problem. The training cost that you says is what server farms can do, nothing even remotely possible on a home environment AFAIK.

I’m exploring ollama for local GPT. Will probably do a video on it.

Both sensors. But check to see if your provider has an integration! Phyn (water) and Sense (electricity)

Good catch, thank you! I even tried to blur it out but must’ve cut the blur one frame too soon 😂

And thanks for watching!

Hello! So glad you liked the video. There are a few good water sensors out there!

Yes! I’m actually working on a video about this and some of the pros and cons.

Yes. It’s nice to be able to track usage per type. And I can set them up independently with different criteria.

That seems neat, but not nearly as approachable as using GPT. The minimum requirement of ~8GB of ram and multiple gigs of storage makes it a little steep for the average person. Once my GPT credit runs out, I'll look at switching to that myself.

I will be looking into this!

Also most open source models are not even close to as useful and "smart" as gpt4

@@gabynevada Llama 3 is on par with gpt 4. Haven't tested against gpt 4o yet.

@@Techguru604 Think I had only tested until Llama 2, gonna give Llama 3 a try then. Thanks!

I’m not totally sure. You’ll have to check k the home assistant logs to see more. Should be in settings.

There are actually some projects out there that completely swap the boards in Google home mini with a custom system that could indeed do this!

I’ll check it out!

Thank you and thanks for watching. The GPT 3.5 model, which is mostly what I was using, was one cent a month for all of my usage in March.

Can also see where you live for the same reason. I mean it doesn't really matter, and you can redo the API keys, but just FYI. :)

I did regenerate the API key but what do you mean you can see where I live?

Actually, don’t respond i found it thank you!

@@FutureSmartHome NP!

Thanks for watching! A totally fair (& common) point of view. I feel ok with this because I am choosing to send the data when I want to and it’s not unfettered access. I wrote the code that sends the information so, that gives me some comfort.

@@FutureSmartHome Sure. Its good that you have the skills to write the code. That;s awesome! Have subscribed and will follow the channel. I was just making the point having had smart home tech now well over a decade. I know where this can lead. I started with control4 and hated their "dealer only" model. That's why I love and switched to HA. Whilst I understand everyone needs to make a buck out of us home users, we need to be very wary what goes on behind the scenes and under the hood. HA is not like that (now) but could very well head that way (if we aren't careful). Just like FB.. starting out all groovy and "free" no subscriber model.. then 10 years on they are selling our data. I understand nabu casa has a very small subscriber fee for remote access.. that;s fine.. but will that be enough? Isn't "big data" where "the money is"? Probably out of the scope of this discussion, but doesn't Microsoft have their tentacles directly in the HA pie as well...? Thru an "interest" in github? This company (and their true motives) needs to be on everyone’s home automation "watch". That's all I'm saying 🙂

In Ollama is free!

That's fair concern. Not to say that I 100% agree, but if that is bothering you, you can always self host the Ai. Know that it might cost you a bit to get a computer that can respond quickly. If it interests you, check for ollama

​@@wapphigh5250nowadays you don't even need the coding skill for this, just ask the AI to write it for you, and tell it to explain what the code does step-by-step. AI doesn't lie willingly

Oh noooo😢😢😢 openAI knows i used 50 gallons of water this week😢😢😢😢

I agree that is the core ethos of HA and I try to follow it. But while we don’t have good, readily accessible, local LLMs, I think this is an avenue for exploration. This isn’t a HACS integration, it’s the official & sanctioned integration with OpenAI.

@@ChrizzeeB send all the internal data from all the sensors to OpenAI… really? Just to read the same notifications in a different way?

No, HA is setup for whatever you want it to do. If you choose to share it with OpenAI, that’s your choice. If you choose not to share that info, that’s also your choice.

@@thrawnis it is, yes, but "local control, local data" is the core value.

A bit hard to troubleshoot via RU-vid comments BUT I’m guessing at the time you are calling the text to speech you haven’t set the response variable of the OpenAI conversation process to value “agent”. Also make sure you are indeed doing the conversation process action.

@@FutureSmartHome I have followed your video instructions, creating the temporary action in the automation to determine the agent ID. I have noticed that I get an additional tick box to set, labelled 'conversation ID', which you don't seem to get in your setup.

@@danthetube5707 I am a few home assistant versions behind so I’ll have to update and check it out.

@@danthetube5707 Me also I have been trying this for a couple days. I have the response if I use dev tools, but I can seem to find it past there. I don't recive a # agent_id.

@@FutureSmartHome you ever find the issue for no agent_id:#. When I run the conversation.process to get the id it just returns conversation.openai.conversation. Also when I use the dev tool I can get the response, but I have no clue where to find that to send thru. I get either the timestamp as a message, agent, or the “Simon says” reply. Any help is appreciated