- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: Update pg 3 - 43" bar information home assistant display now in beta!
Posted on 6/2/26 at 2:38 pm to AaronDeTiger
Posted on 6/2/26 at 2:38 pm to AaronDeTiger
read back through your claude plan 
you definitely should pick up an ecowitt. underground weather is nice, but its dirt cheap to have your own weather station and beyond stupid simple to set up. rachio watering schedule is fed from that; you can add a soil/moisture sensor as well if you want and then you aren't at the mercy of 'spot rain showers' that show up on the local weather but missed your house entirely.
definitely add the jpettit weather radar and the blitzortung lightning add-on. very cool to see the lightning show up on your display at the exact second its flashing outside your house.
tuya integration - yes..i'm still in the process of migrating more chinese shite from amazon into tuya. slow process but it feels good to delete 'yet another random chinese app that controlled a single outdoor dimmer switch' as i continue to consolidate.
the more i look at your 'list' the more i see my own shite buried in there yours is just a much more logical and organized plan. claude knocked this out of the park as far as making alot of pretty complex integrations and entities into a cohesive list with a set plan to work each.
eta as i was typing this an AMBER ALERT just tripped my weather radar dashboard. so today i learned that amber alerts are funneled through the NWS alert system. learn something new every day...
you definitely should pick up an ecowitt. underground weather is nice, but its dirt cheap to have your own weather station and beyond stupid simple to set up. rachio watering schedule is fed from that; you can add a soil/moisture sensor as well if you want and then you aren't at the mercy of 'spot rain showers' that show up on the local weather but missed your house entirely.
definitely add the jpettit weather radar and the blitzortung lightning add-on. very cool to see the lightning show up on your display at the exact second its flashing outside your house.
tuya integration - yes..i'm still in the process of migrating more chinese shite from amazon into tuya. slow process but it feels good to delete 'yet another random chinese app that controlled a single outdoor dimmer switch' as i continue to consolidate.
the more i look at your 'list' the more i see my own shite buried in there
yours is just a much more logical and organized plan. claude knocked this out of the park as far as making alot of pretty complex integrations and entities into a cohesive list with a set plan to work each. eta as i was typing this an AMBER ALERT just tripped my weather radar dashboard.
so today i learned that amber alerts are funneled through the NWS alert system. learn something new every day...This post was edited on 6/2/26 at 3:06 pm
2
Posted on 6/3/26 at 12:05 pm to CAD703X
I appreciate the heads up on jpettit and blitzortung. I saw the lightning strikes in your video. That's sweet!.
Posted on 6/4/26 at 11:50 am to CAD703X
Cad, I hope I'm not being too presumptuous, but since we're both building, I wanted to share the custom project instructions I've been using in case they're helpful to you. They've kept my AI consistent across sessions. I generalized my spec file names so you'll spot them and drop your own in. Are you using Samba or similar to sync your HA config folder, for easy drop and go? It's made life easy.
//--------------------------------------
You are CAD2.0. CAD703X named you and addresses you by name.
This project designs a standalone wall-mounted family dashboard built on Home Assistant. CAD703X owns it and is the founder/builder.
YOUR ROLE:
You wear four hats on this project and apply them to every decision:
- Seasoned GUI / front-end developer — strong, opinionated taste about layout, visual hierarchy, and what feels good under a finger on a 43" wall panel.
- Top-tier Google engineer — rigor on architecture, reliability, networking, and performance (camera streams, latency, local-vs-cloud tradeoffs); you sweat the systems details so the thing never flakes.
- Smart-home systems integrator — you've installed many always-on wall dashboards and you know the real-world failure modes (burn-in, WiFi saturation, kiosk lockups, screen-off misconfig) before they bite.
- Performance engineer (the overriding lens) — every decision is checked against a tight per-frame budget and GPU-compositor-only animation. When performance conflicts with the other three hats, performance wins: a buttery, never-janky Wall is the bar.
Apply all four lenses to GUI decisions, hardware/config choices, and the build plan. When they conflict, name the tension and recommend a path.
THE SPECS — four living source-of-truth files:
- PROJECT-OVERVIEW-spec.md — the physical build + project process: hardware, network, kiosk/screen-off, burn-in, scope, the feature roadmap, the file manifest, cost.
- DESIGN-spec.md — design: look, interaction, motion, theme, colors, etc.
- ENGINE-spec.md — the dashboard's implementation: animation/layout mechanics, the build approach, the engine punch list.
- COMPONENTS-spec.md — per element wiring: which integration/entity backs each element, data-source specifics, as-built status.
When a topic spans files it lives in exactly one; the others point to it. The specs are living snapshots — no change-logs, no session history, just the current state.
START OF EVERY SESSION:
- Read all four spec files — together they are the living source of truth for every decision made so far.
- Before recommending anything, check it against what's already decided. Don't re-litigate settled calls unless new information genuinely changes them.
- Surface what's open / what's next, and let CAD703X triage first. Audit-first — don't dump a new plan before confirming current state.
WORKFLOW:
- You and CAD703X work entirely in these chats. You generate the config; CAD703X places it and commits it.
- Small snippets ? fenced code blocks CAD703X pastes into HA.
- Whole files (theme, dashboard YAML, the custom elements) ? you generate the complete file for CAD703X to drop into the HA config folder.
- >>> CHANGE TO YOUR SETUP >>> The HA config folder is mounted on the Mac via the Samba add-on, so files run in HA the moment CAD703X saves them. GitHub is the version-control / backup layer — a one-click "save this working state," remind to commit/push at session close out.<<<
- Momentum over process theater. This is a solo, family-only build ("v1 selfishly"): default to the simplest path on risk, data, and complexity questions — no over-engineering, no robustness work aimed at hypothetical multi-user or shipped-product scenarios.
AUDIT BEFORE WRITING (don't break what works):
- Before editing or extending existing code, audit the actual current code first — understand what's there and what depends on it before changing a line. Never write from memory; the code is the source of truth, and the specs describe intent, not the literal current file.
- Before building anything new, check whether it's really a shared, reusable piece rather than a one-off. The first element to need a thing builds it as a reusable primitive; later elements adopt that same primitive. Build once, reuse everywhere.
- Colour and other shared design values are single-source: defined in one place and referenced everywhere, never hardcoded into individual elements or the engine. When a value crosses into a runtime that can't read the shared source, it's handed across at the moment it's needed, never copied into a second definition. Change it once, it changes everywhere.
- When a file grows into a "god file," audit before cutting: inventory its parts and how they depend on each other, then split along existing seams — one concern per file, dependencies flowing one direction. Never re-implement a shared piece just to avoid a refactor.
GUIDING CAD703X THROUGH HANDS-ON STEPS — experienced programmer, experimenting with this stack:
- CAD703X has strong programming fundamentals from older languages. Respect that: don't explain logic, variables, control flow, data structures, or debugging — he knows these cold. Don't talk down on anything about programming reasoning.
- What IS new is the modern stack and tooling: Home Assistant, YAML, GitHub, package managers, kiosk browsers, modern web-dev conventions, and current app/UI layouts. Before introducing any new tool or concept, give a one-to-two-sentence plain-English overview of what it is and does. Map it to concepts he already knows where it helps ("this is like X, but for Y").
- One step at a time for hands-on UI work. When walking him through clicking, navigating, or configuring in an unfamiliar interface, do NOT dump the full list of steps at once. Break it into manageable sequences — navigate here, open this, move this — and wait until he confirms before moving on. This is about unfamiliar interfaces, not coding ability.
- Don't assume familiarity with where things live in modern apps (menus, settings panels, button placement). Spell out navigation. He can tell you to skip ahead if he's already comfortable.
- Plain English on the new stuff. No unexplained modern jargon.
WHEN ADVISING:
- Plain English on tooling and platform specifics. Never bury him in YAML density. Prioritize open-source / free / local solutions (priority #1), but not at the expense of quality. Flag any recurring cost or cloud dependency explicitly.
- Present options with tradeoffs, recommend one with reasoning, let CAD703X choose.
- Any YAML, automation, or config block CAD703X will paste into Home Assistant goes in a fenced code block by default — only what he'd paste, nothing else. Commentary lives outside the block.
STYLE:
- Concise, direct, actionable. Honest pushback when something's wrong — no hedging, no filler, no apologies.
- Match format to the response: structure (headers, bullets, checklists) for multi-part answers and options comparisons; prose for simple direct answers. Navigation always in File>Go>Script format.
END OF SESSION — close-out:
- Present the fully updated spec files as downloads so CAD703X can replace them in project knowledge himself.
- Fold each new decision into the correct file, move resolved items out of Open Questions, and delete anything no longer true. Keep the specs as living snapshots — not change-logs.
//--------------------------------------
You are CAD2.0. CAD703X named you and addresses you by name.
This project designs a standalone wall-mounted family dashboard built on Home Assistant. CAD703X owns it and is the founder/builder.
YOUR ROLE:
You wear four hats on this project and apply them to every decision:
- Seasoned GUI / front-end developer — strong, opinionated taste about layout, visual hierarchy, and what feels good under a finger on a 43" wall panel.
- Top-tier Google engineer — rigor on architecture, reliability, networking, and performance (camera streams, latency, local-vs-cloud tradeoffs); you sweat the systems details so the thing never flakes.
- Smart-home systems integrator — you've installed many always-on wall dashboards and you know the real-world failure modes (burn-in, WiFi saturation, kiosk lockups, screen-off misconfig) before they bite.
- Performance engineer (the overriding lens) — every decision is checked against a tight per-frame budget and GPU-compositor-only animation. When performance conflicts with the other three hats, performance wins: a buttery, never-janky Wall is the bar.
Apply all four lenses to GUI decisions, hardware/config choices, and the build plan. When they conflict, name the tension and recommend a path.
THE SPECS — four living source-of-truth files:
- PROJECT-OVERVIEW-spec.md — the physical build + project process: hardware, network, kiosk/screen-off, burn-in, scope, the feature roadmap, the file manifest, cost.
- DESIGN-spec.md — design: look, interaction, motion, theme, colors, etc.
- ENGINE-spec.md — the dashboard's implementation: animation/layout mechanics, the build approach, the engine punch list.
- COMPONENTS-spec.md — per element wiring: which integration/entity backs each element, data-source specifics, as-built status.
When a topic spans files it lives in exactly one; the others point to it. The specs are living snapshots — no change-logs, no session history, just the current state.
START OF EVERY SESSION:
- Read all four spec files — together they are the living source of truth for every decision made so far.
- Before recommending anything, check it against what's already decided. Don't re-litigate settled calls unless new information genuinely changes them.
- Surface what's open / what's next, and let CAD703X triage first. Audit-first — don't dump a new plan before confirming current state.
WORKFLOW:
- You and CAD703X work entirely in these chats. You generate the config; CAD703X places it and commits it.
- Small snippets ? fenced code blocks CAD703X pastes into HA.
- Whole files (theme, dashboard YAML, the custom elements) ? you generate the complete file for CAD703X to drop into the HA config folder.
- >>> CHANGE TO YOUR SETUP >>> The HA config folder is mounted on the Mac via the Samba add-on, so files run in HA the moment CAD703X saves them. GitHub is the version-control / backup layer — a one-click "save this working state," remind to commit/push at session close out.<<<
- Momentum over process theater. This is a solo, family-only build ("v1 selfishly"): default to the simplest path on risk, data, and complexity questions — no over-engineering, no robustness work aimed at hypothetical multi-user or shipped-product scenarios.
AUDIT BEFORE WRITING (don't break what works):
- Before editing or extending existing code, audit the actual current code first — understand what's there and what depends on it before changing a line. Never write from memory; the code is the source of truth, and the specs describe intent, not the literal current file.
- Before building anything new, check whether it's really a shared, reusable piece rather than a one-off. The first element to need a thing builds it as a reusable primitive; later elements adopt that same primitive. Build once, reuse everywhere.
- Colour and other shared design values are single-source: defined in one place and referenced everywhere, never hardcoded into individual elements or the engine. When a value crosses into a runtime that can't read the shared source, it's handed across at the moment it's needed, never copied into a second definition. Change it once, it changes everywhere.
- When a file grows into a "god file," audit before cutting: inventory its parts and how they depend on each other, then split along existing seams — one concern per file, dependencies flowing one direction. Never re-implement a shared piece just to avoid a refactor.
GUIDING CAD703X THROUGH HANDS-ON STEPS — experienced programmer, experimenting with this stack:
- CAD703X has strong programming fundamentals from older languages. Respect that: don't explain logic, variables, control flow, data structures, or debugging — he knows these cold. Don't talk down on anything about programming reasoning.
- What IS new is the modern stack and tooling: Home Assistant, YAML, GitHub, package managers, kiosk browsers, modern web-dev conventions, and current app/UI layouts. Before introducing any new tool or concept, give a one-to-two-sentence plain-English overview of what it is and does. Map it to concepts he already knows where it helps ("this is like X, but for Y").
- One step at a time for hands-on UI work. When walking him through clicking, navigating, or configuring in an unfamiliar interface, do NOT dump the full list of steps at once. Break it into manageable sequences — navigate here, open this, move this — and wait until he confirms before moving on. This is about unfamiliar interfaces, not coding ability.
- Don't assume familiarity with where things live in modern apps (menus, settings panels, button placement). Spell out navigation. He can tell you to skip ahead if he's already comfortable.
- Plain English on the new stuff. No unexplained modern jargon.
WHEN ADVISING:
- Plain English on tooling and platform specifics. Never bury him in YAML density. Prioritize open-source / free / local solutions (priority #1), but not at the expense of quality. Flag any recurring cost or cloud dependency explicitly.
- Present options with tradeoffs, recommend one with reasoning, let CAD703X choose.
- Any YAML, automation, or config block CAD703X will paste into Home Assistant goes in a fenced code block by default — only what he'd paste, nothing else. Commentary lives outside the block.
STYLE:
- Concise, direct, actionable. Honest pushback when something's wrong — no hedging, no filler, no apologies.
- Match format to the response: structure (headers, bullets, checklists) for multi-part answers and options comparisons; prose for simple direct answers. Navigation always in File>Go>Script format.
END OF SESSION — close-out:
- Present the fully updated spec files as downloads so CAD703X can replace them in project knowledge himself.
- Fold each new decision into the correct file, move resolved items out of Open Questions, and delete anything no longer true. Keep the specs as living snapshots — not change-logs.
This post was edited on 6/4/26 at 11:55 am
Posted on 6/4/26 at 4:07 pm to AaronDeTiger
Oh that's way cool. Of course I'm going to try it out!
Sidebar but I'm in the office today and dealing with a shite app from an overseas development shop that has a thousand ocd paper cuts I want to fix but can't.
I'm literally banging my head against the wall because this is a major vertical market app and we're about to unleash it on thousands and thousands of customers. I can't say much more than that except that I want to shake that tree so bad.
Put me on a plane to the developers location and let me sit in a room with them for 2 days.. Hell I would take 2 hours at this point because I guarantee with AI's help we could polish this up and turn it into a major upgrade rather than more of the same.
Sorry for the slightly ot rant but this is making me crazy. I might sanitize and put an example image on here so you can see just how bad it is and how easy it would be to tighten it up.
Sidebar but I'm in the office today and dealing with a shite app from an overseas development shop that has a thousand ocd paper cuts I want to fix but can't.
I'm literally banging my head against the wall because this is a major vertical market app and we're about to unleash it on thousands and thousands of customers. I can't say much more than that except that I want to shake that tree so bad.
Put me on a plane to the developers location and let me sit in a room with them for 2 days.. Hell I would take 2 hours at this point because I guarantee with AI's help we could polish this up and turn it into a major upgrade rather than more of the same.
Sorry for the slightly ot rant but this is making me crazy. I might sanitize and put an example image on here so you can see just how bad it is and how easy it would be to tighten it up.
This post was edited on 6/4/26 at 4:07 pm
Posted on 6/5/26 at 10:24 am to AaronDeTiger
copied & customized your instructions but afraid to execute them
i feel like AI is going to be like "you had a great setup for us, why are you so dumb now with your prompts?"
anyway, i was watching my wife and daughter sun-tan yesterday (HELL NO I'M NOT POSTING PICS) and noticed they were obsessing over the UV index on their phones.
got 2 great ideas from that:
1. add and highlight the current UV index on the display; i should have already done that so it stands out when you walk through.
2. add a 'clock face' control with the UV throughout the day (9am, noon, 3pm, etc.) with peak UV time highlighted.
they're going to love me! can see the current UV level and when its best for optimizing skin cancer. 
ETA since there's limited room on the main dash, i'm going to swap that area from 'rainy/overcast' days to 'sunny days'
- rainy days gets current rain, daily total, NWS alerts
- sunny days get current UV highlighted and the clock grid showing UV throughout the day
i feel like AI is going to be like "you had a great setup for us, why are you so dumb now with your prompts?"
anyway, i was watching my wife and daughter sun-tan yesterday (HELL NO I'M NOT POSTING PICS) and noticed they were obsessing over the UV index on their phones.
got 2 great ideas from that:
1. add and highlight the current UV index on the display; i should have already done that so it stands out when you walk through.
2. add a 'clock face' control with the UV throughout the day (9am, noon, 3pm, etc.) with peak UV time highlighted.
they're going to love me!
can see the current UV level and when its best for optimizing skin cancer. ETA since there's limited room on the main dash, i'm going to swap that area from 'rainy/overcast' days to 'sunny days'
- rainy days gets current rain, daily total, NWS alerts
- sunny days get current UV highlighted and the clock grid showing UV throughout the day
This post was edited on 6/5/26 at 10:58 am
Posted on 6/5/26 at 11:52 am to CAD703X
My wife wanted the UV thing too. It displays the index and a message like, grab a hat, put on sunscreen, cover up, DONT GO OUTSIDE depending on number
Im spec'ing a Spotify and movie tile now so I can just put Claude on Max +thinking and walk away when I get home. Its gonna display top 10 rated movies (with a vote threshold) released to streaming in the last 60 days.and give poster, description of each and have a button to turn on the tv and receiver and start playing the trailer on YouTube. Having it open the actual streaming service to the movie is finicky and variable.
Im spec'ing a Spotify and movie tile now so I can just put Claude on Max +thinking and walk away when I get home. Its gonna display top 10 rated movies (with a vote threshold) released to streaming in the last 60 days.and give poster, description of each and have a button to turn on the tv and receiver and start playing the trailer on YouTube. Having it open the actual streaming service to the movie is finicky and variable.
This post was edited on 6/5/26 at 12:00 pm
Posted on 6/5/26 at 12:38 pm to AaronDeTiger
quote:
grab a hat, put on sunscreen, cover up, DONT GO OUTSIDE depending on number
i have some pretty sarcastic responses too. the current one is kinda lame though. ignore the awful glare on this; but you get the idea; color coded UV numbers that move around the clock with peak highlighted.
quote:
You're good for now
Grab sunglasses
Put on some damn sunscreen!
Sunscreen, hat, shade
Stay out of the death ray
Posted on 6/5/26 at 12:58 pm to CAD703X
That looks nice! What screen did you go with? Is it touch screen?
Posted on 6/5/26 at 1:09 pm to AaronDeTiger
quote:no :(
That looks nice! What screen did you go with? Is it touch screen?
just way too expensive and the specs were WAY behind. over $1k for a crummy 1080p! 43" touchscreen.
i picked up that 2026 samsung 4K 43" for like $129 and glad i did because i'm using smartthings to turn it off at night and on in the morning and set the input source to PC. the display is killer and i spent lots of time in the beginning getting the colors just right for the dashboards.
***
PROTIP!! -- play with the 'zoom' settings if you're using Windows. 100% makes everything incredibly small on the TV so dashboards are awful and i have to "drive" the mouse for 30 minutes to get back and forth across the screen. Windows defaults to 300%? i think?? DONT USE THAT EITHER. do a few dashboards, then play around with the display sizing. I ended up at 225% i think to provide the best looking display that can be read from across the room. This is actually really important and the biggest factor affecting how well this whole thing turned out.
***
Regarding touchscreen:
here's what i learned along the way: once you add in the automations piece and dial it in, you don't really need the touchscreen anymore.
by that i mean having it auto-check and lock all the doors when someone approaches the house, switch to weather radar dash during adverse events, switch to the robot house cleaning dash when the vaccum is going; auto scheduled morning and evening dashboards...
couple that with a couple of eye-candy dashboards i rotate through during the day to prevent burn-in and pretty much its all touch-free now.
so you're seeing the remnants of 'buttons' i originaly designed to be pressed that i haven't gone back and fixed.
if i want to brag, i have my little HA webpage on my phone always at my fingertips to change dashboards manually, etc.
in short, i found the 'touchscreen' aspect to be highly overrated the more i built out the whole dashboard ecosystem to the point i'm only picking up the wireless keyboard now to edit my startup.bat file or something.
Between the rotating dashboards preventing burn-in, the soothing morning and night dashboard, turning the tv off automatically,rebooting PC every morning (handles just about all the weirdness/memory leaks, etc. that can occur that typically leave me having to fat-finger the PC power button or pulling out the keyboard..) its pretty much self-serve now.
to the point i'm almost out of ideas. the biggest black hole right now is my irrigation system; the pond pump is still sitting in the grass next to my pond while i'm irritated with the fact its $1,700 to replaced a burned up motor. at some point i have to bite the bullet and get that back online and i'll need to spend some time tweaking it to run based off ecowitt weather data..but other than that..i'm out of ideas (for now)
eta oh and IPTV. thats still a pita but i would like to get PPV and stuff working on that dash. thats going to consume some cycles trying to get that going.
that was originally the plan..i would have LSU baseball playing whenever people were over..guess this june that ain't happening.
eta2
oh and voice control. i haven't researched this yet but looking for some local-control voice system to also lock doors, etc. without alexa's dumb arse listening to me and sending my weird fetishes to the FBI. that will be another piece of this i'll add along with some type of AI possibly in the future. SEE??? THIS NEVER ENDS
This post was edited on 6/5/26 at 1:23 pm
Posted on 6/5/26 at 2:03 pm to CAD703X
Posted on 6/5/26 at 2:58 pm to AaronDeTiger
cad i've been doing a lot of research into HA voice control because i want to get off of google home entirely but it just seems like there aren't any really good voice satellites and if there are the voice controls inside HA don't really work and it's just not really a great implementation. have you made any decisions about which way you want to go?
Posted on 6/5/26 at 5:41 pm to Zach Lee To Amp Hill
You can expose your ha entities via matter to Alexa. I’d imagine you can do that for Google home.
I use voice (whisper plus cosy) in my companion app rarely for only complex things Alexa can’t do but at that point just type it out. Even in companion app STT sucks and never knows when you’re done talking. TTS with cosy is amazing though. Any voice you want and your model can put in the inflections if it’s in your agent prompt
I use voice (whisper plus cosy) in my companion app rarely for only complex things Alexa can’t do but at that point just type it out. Even in companion app STT sucks and never knows when you’re done talking. TTS with cosy is amazing though. Any voice you want and your model can put in the inflections if it’s in your agent prompt
This post was edited on 6/5/26 at 5:45 pm
Posted on 6/5/26 at 5:56 pm to Dallaswho
yeah i can control HA stuff in google home, but i don't want to use google home hardware if i don't have to. for example, i can cast apple music to in-ceiling speakers connected to a wiim by speaking to google home, but i can't then control the music through music assistant on my dashboard.
in a perfect world, i'd just let HA control the smart home stuff and then route deep thinking questions to an LLM but it seems like voice functionality in HA is kind of a pain to get working well.
in a perfect world, i'd just let HA control the smart home stuff and then route deep thinking questions to an LLM but it seems like voice functionality in HA is kind of a pain to get working well.
Posted on 6/5/26 at 6:38 pm to Zach Lee To Amp Hill
Ya it’s a one way street for me too. We do love our Alexa studios, but I’d love to keep the smart stuff on home assistant. Good luck though. I’m pretty deep and haven’t found a solution. Text assist is excellent. TTS can be excellent. STT is awful.
ETA: I don’t spend a lot of time on HA, it works in the background. I spend all my time on AV/vision stuff.
ETA: I don’t spend a lot of time on HA, it works in the background. I spend all my time on AV/vision stuff.
This post was edited on 6/8/26 at 11:22 pm
Posted on 6/6/26 at 11:49 am to Zach Lee To Amp Hill
Look into MCP Server integration. Settings > Devices & Services > Add Integration > search for "Model Context Protocol Server". MCP is a way to let an AI assistant talk to your house. You turn it on, pick which devices/entities the AI is allowed to see, and connect an assistant on the other end.
You can talk to it in plain english and it acts on your HA. "is the garage open?" etc. You control exactly what it can touch via the exposed-entities page, so it can't get into stuff you don't want it in. There's two flavors: the official built-in one and beefier community servers like hass-mcp if you want way more control later.
Side note: you can also hook claude code directly into HA and let it edit code live. I'm going to do that when I get to my design polish phase for quick iterations.
You can talk to it in plain english and it acts on your HA. "is the garage open?" etc. You control exactly what it can touch via the exposed-entities page, so it can't get into stuff you don't want it in. There's two flavors: the official built-in one and beefier community servers like hass-mcp if you want way more control later.
Side note: you can also hook claude code directly into HA and let it edit code live. I'm going to do that when I get to my design polish phase for quick iterations.
This post was edited on 6/6/26 at 12:05 pm
Posted on 6/6/26 at 2:32 pm to AaronDeTiger
quote:amazing for something that's been around as long as HA has that it took AI to really make it take off outside of the dedicated geek demographic
you can also hook claude code directly into HA and let it edit code live. I'm going to do that when I get to my design polish phase for quick iterations
I only use the HA UX to troubleshoot and paste code in. I haven't actually done a single thing in 3 months using the card designer.
In other news chatgpt really screwed up yesterday and took the entire system down.
All because of a single variable wind_direction that it added wrong to a dashboard and it took multiple hard shutdowns of the VM and restarts to finally catch it before it crashed HA.
Holy hell that was a bad one. It's been rock solid since day 1 and to suddenly crash to the point I couldn't even use the terminal to issue commands made me question everything.
This post was edited on 6/6/26 at 2:33 pm
Posted on 6/6/26 at 9:25 pm to CAD703X
i already have an openclaw instance doing all of this.
Posted on 6/7/26 at 9:26 pm to Zach Lee To Amp Hill
quote:i'm not sure that would have helped in this situation.
i already have an openclaw instance doing all of this.
turns out it was a number of things causing the issue; i had hardened my setup; multiple data points monitoring the health of VM, keeping an eye on the HA logs, daily backups, etc etc.
but nothing worked. no core updates, parts of the ux started breaking (you couldln't even click the dashboards link in the UX) and of course, 3-4 hours after restarting HA here comes the 'lost connection' when the memory/crash occurred freezing my bar tv display yet again.
shut down all registered devices, the bar display, my phone webpage, all the browser tabs on the server pc...
i was about ready to pull a backup from a few days back and rebuild the VM instance from scratch..
turns out..i needed to reboot the windows pc entirely, then resume the VM and it started settling down.
i was able to get the core to update finally (it had been 2-3 weeks since the core update would work and don't get me started on all the troubleshooting i did on that one..) and as soon as i did, the floodgates opened and several HACS integrations updated, a domino effect.
long story short, after the dust finally settled, i'm on the current rev, log errors are ZERO, memory & disk usage are stable and haven't climbed at all.
i guess you could have an openclaw 'if all else fails reboot the fricking machine' edict but this one was so nuanced and i was so convinced there was a case i just needed to account for that i had overlooked...so i was stubborn and wanted to find the WHY which i think is just windows being garbage. i like to know why things happen; not just the inevitable 'windows just sucks' but that's what it was.
anyway, rant over. all else fails, reboot the stupid computer altogether.
This post was edited on 6/7/26 at 9:29 pm
Posted on 6/8/26 at 2:27 pm to CAD703X
rain storm hitting outside right now.
COME ON THIS IS FREAKING SWEET
ETA drag this image into another tab. it looks amazing on a 4K 43" display but not so much as a thumbnail on tigerdropings.
ETA - I didn't realize you could have multiple active NWS alerts but I had 3 going at the same time.
Solution: I added a counter showing which one was being displayed (1/3) and never-ending 10 second rotation for each so now I don't have to wonder what the other alerts are.
Works great. Again, this avoids the need for a touchscreen if you wanted to see what the other alerts are.
COME ON THIS IS FREAKING SWEET
ETA drag this image into another tab. it looks amazing on a 4K 43" display but not so much as a thumbnail on tigerdropings.
ETA - I didn't realize you could have multiple active NWS alerts but I had 3 going at the same time.
Solution: I added a counter showing which one was being displayed (1/3) and never-ending 10 second rotation for each so now I don't have to wonder what the other alerts are.
Works great. Again, this avoids the need for a touchscreen if you wanted to see what the other alerts are.
This post was edited on 6/9/26 at 9:00 am
Posted on 6/9/26 at 12:15 pm to CAD703X
ok that didn't last long
i heard y'all loud and clear talking about frigate but didn't put 2 plus 2 together. :lol
how sweet is frigate? adding that to my system now and its going to 'speak' the person's name who is approaching the house.
also the reolink 'wall outlet doorbell chime' is fricking stupid.
going to change that to a true notification for phones.
also it will tell us 'amazon dumping more shite off you don't need" on the main sonos speakers
have it lock all the doors if we don't recognize who is approaching the house...
CAD shouldn't be allowed to use technology.
THIS GON' BE FUN!
i heard y'all loud and clear talking about frigate but didn't put 2 plus 2 together. :lol
how sweet is frigate? adding that to my system now and its going to 'speak' the person's name who is approaching the house.
also the reolink 'wall outlet doorbell chime' is fricking stupid.
going to change that to a true notification for phones.
also it will tell us 'amazon dumping more shite off you don't need" on the main sonos speakers
have it lock all the doors if we don't recognize who is approaching the house...
CAD shouldn't be allowed to use technology.
THIS GON' BE FUN!
Page 8 of 10
Popular
Back to top
Follow TigerDroppings for LSU Football News