Original post on hessen.social

OpenClaw

Nachdem ich OC in einem Container isoliert & mit Gemini laufen hatte, habe ich nun die nächste Stufe meines OC-Setup genommen:
OC mit eigener Identity auf einem dedizierten Mac mit Ollama & lokalem LLM. 1:1 & Gruppenchat via Telegram & iMessage.

Soeben den Connect via MCP mit limited […]

I have ~60W of load thats showing up as 'untracked consumption' on my #homeassistant energy dashboard. I know I should be happy everything else is known, but that last 60W is bugging me. ;)

5/5 🔄 Y una mejora de estabilidad que se agradece: las automatizaciones largas ya no se detienen si un dispositivo intermedio falla. Si un enchufe no responde, el resto de la secuencia sigue ejecutándose. Parece menor, pero en el día a día se nota.
#HomeAssistant #Domótica #SmartHome #Privacidad

#llm

#AI is making CEOs delusional

#software #homeassistant

https://www.youtube.com/watch?v=Q6nem-F8AG8

Migrated 26 Zigbee and 11 Matter devices to my new #homeassistant .

Now its time to restore the automations 😮‍💨

Claude gère mon blog, mon serveur et mes automatisations Thème WordPress, infrastructure Docker, Home Assistant, RAG : comment Claude Code et les MCPs ont transformé ma gestion de blog solo.

Il y a un an j’étais sur ChatGPT en pensant que Claude c’était que pour les devs. Aujourd’hui il gère mon blog, mon serveur, ma domotique et mes mails. Il a même trouvé un bug que je ne cherchais pas.
www.juliendoclot.com/comment-clau...
#Claude #HomeAssistant

Home Assistant on mycrab.space — your smart home dashboard, accessible from anywhere.

no cloud account. no monthly fee. your automations run local, but the UI is at yourname.mycrab.space.

https://mycrab.space #homeassistant #homeautomation #selfhosted #homelab

CVE-2026-34205: CWE-923: Improper Restriction of Communication Channel to Intend CVE-2026-34205 is a critical security vulnerability classified under CWE-923 (Improper Restriction of Communication Channel to Intended Endpoints) affecting the Home Assistant Operating System up to version 17.1. Home Assistant is an open-s

CRITICAL: Home Assistant OS ≤17.1 exposes unauthenticated endpoints via host network mode. Upgrade to Supervisor 2026.03.02 & segment your IoT network now! 🛡️ radar.offseq.com/threat/cve-2026-34205-cw... #OffSeq #HomeAssistant #SecurityAlert

#HomeAssistant add-ons for #BOINC

Contribute to scientific research projects from your Home Assistant instance.

🚀 New release available!

📦 BOINC v3.3.0

🔗 Release: github.com/hectorespert...

To install it, read the documentation: hectorespert.github.io/boinc-addons

#HomeAssistant add-ons for #BOINC

Contribute to scientific research projects from your Home Assistant instance.

🚀 New release available!

📦 boinctui v2.2.0

🔗 Release: github.com/hectorespert...

To install it, read the documentation: hectorespert.github.io/boinc-addons

Screenshot of my Home Assistant dashboard showing all 4 of my power monitor smart plugs. It shows the fridge using about 60 watts of power, my desktop PC using 180 watts, my room lamps using 20 watts, and my network equipment using 15 watts. Each card is a line graph, so you can see the power usage over time.

Screenshot of a Home Assistant dashboard graph showing computer power draw vs fridge power draw. What's neat is you can see the fridge power going up to 50 watts for 20 minutes and then dropping to near 0 watts, for another 20 minutes. Also shown is my computer power usage which fluctuates between 200 watts and 150 watts.

Photo of the box of ThirdReality smart plugs. This is the 3rd generation, and is a box of 4 plugs. The other box is the Sonoff Dongle Plus MG24, which is a USB plug with an antenna. It's compatible with Home Assistant and Zigbee2MQTT, and I guess it works with the Thread protocol too.

I've been wanting to monitor my energy usage. Some time back I asked Mastodon about favorite smart plugs, and the ones that use #Zigbee seem to be preferred.

After @geerlingguy did a review on the new ThirdReality smart plugs, I finally bought a few, along […]

[Original post on nutmeg.social]

A list of how long until trash is being picked up, ordered by days left

I use auto entities to populate when my trash is being picked up. Now I am about to add ibeacon trackers to all my trash cans and would essentially like to annotate this list with the state of the can - is it by the street or by the house.

Is there a clever way […]

[Original post on mstdn.dk]

A screenshot from a home automation app displays a back patio weather sensor temperature history graph, with 5-minute aggregated data showing a peak of approximately 102°F during the early afternoon hours before dropping to the mid-30s overnight and rising to the current reading of 72.5°F. The blue line graph spans from around 10:00 AM on March 26 through the morning of March 27. (Description assisted by Claude AI.)

Go home, outdoor temperature sensor, you're drunk. It did not get up to 102F yesterday.

(I'm going to check later today if direct sunlight is causing these readings -- or if something is wacky with the sensor.)

#homeAssistant #homeAutomation

Original post on mastodon.social

Then some good news for #HomeAssistant users: 2026.4, coming out on the 1st of April, offers much improved security for backups - independently-audited, to boot.

www.hackster.io/news/home-assistant-2026... […]

Maison connectée : ces produits SonOff en promo changent tout (et coûtent presque rien) - Maison et Domotique Envie d’une maison intelligente sans exploser le budget ? Les SonOff Days offrent des modules ZigBee et WiFi ultra efficaces pour automatiser votre quotidien dès aujourd’hui.

🔥 Bon plan domotique !

Les SonOff Days chez Domadoo ➡️ promos sur capteurs ZigBee, prises connectées, modules WiFi…

Parfait pour démarrer (ou compléter) votre installation Home Assistant 👍

👉 www.maison-et-domotique.com/168910-sonof...

#domotique #smarthome #homeassistant #zigbee #iot #bonplan

ich habe mir noch einen bildschirm (#trmnl) gekauft, obwohl wir wirklich genug davon haben. beim setup **endlich** einen weg gefunden, automatisch screenshots von beleibigen webinhalten oder #homeassistant dashboiards zu machen.
https://wirres.net/articles/trmnl

rooms tiles, funny, like my child's room is mass shooting gallery and the basement has a captive in it

My #homeassistant landing page

Why Home Assistant voice is finally a real self‑hosted replacement for Alexa and Google Assistant, if you keep the AI local and the scope narrow _Home Assistant’s March 4 2026 Android voice launch, combined with its fully local speech pipeline, makes privacy‑first, deterministic voice control a practical reality for home‑lab enthusiasts._ Home Assistant’s new Android voice client turns the long‑standing dream of a truly self‑hosted voice assistant into something you can actually run on a single Raspberry Pi or modest server. The platform now ships a **local‑only speech‑to‑text and intent‑recognition stack** , meaning no audio ever leaves your network. Coupled with the March 4 2026 Android launch, this makes Home Assistant voice a viable, privacy‑preserving alternative to Alexa and Google Assistant— _provided you treat the local voice engine as a deterministic control layer first, and only add an optional local LLM when you deliberately want richer, conversational features_. In the sections that follow I’ll unpack why this architecture works, where it outperforms the big‑cloud rivals, and what misconceptions still need correcting. * * * ## How does the Android voice launch shift the balance toward a fully local assistant? The March 4 2026 rollout introduced an Android client that runs the same on‑device speech pipeline used on Home Assistant’s dedicated hardware boxes. This means **every spoken command is processed locally on the phone** , with the result sent as a plain intent to your Home Assistant core. No remote API calls, no latency spikes caused by internet congestion, and no hidden data collection. Because Android already handles microphone permissions and background services efficiently, the new client eliminates the need for a separate “always‑listening” hub. Your phone becomes the ear of the house, listening only when you invoke the wake word you configure. This mirrors the “always‑on” experience of Echo or Nest speakers but without the cloud dependency. From a home‑lab perspective, the Android client also reduces hardware overhead. You no longer need a dedicated SBC just for voice; the same device that runs your personal VPN or media server can host the voice front‑end. The result is a **simpler, cheaper stack** that aligns with the self‑hosting ethos of minimizing moving parts. ## What privacy and cost advantages make local voice compelling? When you weigh privacy, cost, context handling, reliability, and model quality, the **tipping point often lands on the nature of your workload**. Home Assistant voice scores high on each of those dimensions: * **Privacy** – Audio never leaves your LAN, and intent data stays inside your Home Assistant database. There is no telemetry pipeline feeding a corporate cloud, which is a stark contrast to the “listen‑to‑everything” model of mainstream assistants. * **Cost** – You avoid subscription fees for cloud speech‑to‑text (e.g., Google Cloud Speech) or for premium voice‑skill platforms. The only expense is the hardware you already own, plus the modest CPU overhead of running the on‑device models. * **Context handling** – Because intents are resolved by Home Assistant itself, you can embed any entity or script logic you already have. The assistant can toggle a “movie‑mode” scene, pause a Plex stream, or send a custom notification without needing a separate skill ecosystem. * **Reliability** – With no external API, a local network outage does not cripple voice control. Your lights, locks, and thermostats remain reachable even when your ISP is down. * **Model quality** – Modern on‑device models (e.g., Whisper‑tiny for STT and a lightweight intent parser) have reached a point where everyday commands are recognized with >95 % accuracy in typical home environments. These factors combine to make a **privacy‑first, cost‑effective voice layer** that feels native to your self‑hosted ecosystem, rather than an afterthought bolted onto a cloud service. ## Does local voice deliver the speed and reliability that cloud assistants promise? Home Assistant has repeatedly demonstrated that **keeping the processing local yields a much faster experience for your smart home**. When a command is captured, the on‑device STT model produces text in under a second, and the intent engine resolves the action instantly. By contrast, cloud assistants must transmit audio, wait for server‑side transcription, then receive a response—introducing at least a half‑second of network latency, plus any jitter from ISP congestion. Speed matters not just for convenience but for safety. Imagine a fire alarm scenario where you shout “turn off the kitchen lights” while evacuating; a local assistant will act immediately, whereas a cloud‑based assistant could be delayed by a momentary Wi‑Fi dropout. Reliability is further bolstered by **deterministic intent matching**. Home Assistant’s built‑in intents are defined by you, so you know exactly which phrases trigger which actions. There’s no “skill marketplace” that can be withdrawn or changed without notice, as sometimes happens with third‑party Alexa skills. ## How far can a narrow, deterministic control layer go without a cloud LLM? Even without a large language model, Home Assistant voice can handle a surprisingly wide range of home‑automation tasks. The platform’s **built‑in intent system** lets you define simple commands such as “turn on the living‑room lamp” or more complex scripted actions like “set the thermostat to 68 °F and start the humidifier”. A real‑world demonstration on Reddit showed a user who **replaced all Alexa devices with a fully local stack** (Home Assistant voice + a local LLM + a Jabra 410 microphone) and achieved both simple and complex command handling. The user reported that routine tasks—lighting, media playback, and door locks—worked flawlessly, while more nuanced requests (e.g., “set a movie‑night scene”) were handled by a lightweight local LLM that ran on the same hardware. See the **demonstration of a self‑hosted stack** for details. The key insight is that **deterministic intents cover the majority of day‑to‑day interactions**. Most users only need to toggle devices, launch scenes, or ask for status. By keeping the scope narrow—focusing on these core commands—you avoid the unpredictability and resource demands of a full conversational model. ## When should you add an optional local LLM for richer interactions? While a deterministic layer is sufficient for control, many users eventually crave the **contextual intelligence** that a language model provides. Alexa, for example, forces you to use rigid phrases like “Alexa, set the kitchen light 50 %.” By contrast, Home Assistant’s **local LLM integration (e.g., Ollama)** allows you to say “dim the kitchen lights to half” or “make the living room a bit cozier” and have the system infer the appropriate entity and value. Learn more in the **local LLM integration article**. The sweet spot is to **layer the LLM on top of the deterministic core** , not replace it. The core handles all safety‑critical commands (door locks, alarm disarm, HVAC overrides) with strict intent matching, ensuring that a mis‑interpretation never triggers a dangerous action. The LLM then processes “soft” requests—scene adjustments, natural‑language reminders, or conversational queries about the weather—where a slight ambiguity is acceptable. Because the LLM runs locally, you still retain privacy, but you must consider the **hardware footprint**. A small CPU‑only server can host a 1‑GB model with acceptable latency for home use, while more powerful hardware enables larger models that understand nuanced phrasing. The decision hinges on your **resource budget** and how much conversational richness you truly need. ## What does a fully local hardware replacement look like in practice? The **Home Assistant Voice Preview Edition** is a sleek, Echo‑sized box that runs the entire voice stack locally, offering a plug‑and‑play replacement for Amazon’s devices. It ships with a microphone array, on‑device STT, and the Home Assistant core pre‑installed. Users report that the device feels **instantaneous** and that the privacy guarantees are “rock solid”—there is literally no network traffic for voice processing. See the **Home Assistant Voice Review** for a full rundown. Coupled with the Android client, you can choose between a dedicated hub or a mobile‑first approach, whichever fits your lab’s topology. Both options illustrate that **the hardware barrier to self‑hosted voice has essentially vanished** ; you no longer need to cobble together a Raspberry Pi, a USB mic, and a Docker container. The ecosystem now provides polished, supported devices that integrate seamlessly with the rest of Home Assistant’s automation engine. **Is Home Assistant voice finally the self‑hosted answer to Alexa and Google Assistant?** In my view, the answer is a qualified yes: for anyone willing to keep the AI local, limit the scope to deterministic control, and optionally layer a modest LLM for richer language, Home Assistant now offers a **privacy‑first, fast, and reliable voice interface** that matches—if not exceeds—the core functionality of the big‑cloud assistants. If you’ve already experimented with a local voice stack, or if you’re debating whether to add a local LLM, I’d love to hear about your setup, the hurdles you’ve faced, and the tricks that made your voice assistant feel truly native. Drop a comment below, share your configuration, or challenge the assumptions—let’s keep the conversation (locally) going! ### [ System Audit: Voice Sovereignty ] The “Monthly Bleed” isn’t just about money—it’s about **telemetry**. Moving to a local Voice Pipeline isn’t a downgrade in features; it’s an upgrade in privacy and response time. **The Not-Lame Checklist:** • **Wake Word:** Run via _openWakeWord_ (No cloud listening). • **STT:** Whisper (Local processing on your iron). • **LLM:** Ollama or Home Assistant Assist (Scoped to your entities). More Self-Hosting Guides Explore Hardware Hacks Running a local pipeline on a Pi or an N100? Share your latency specs in the comments—let’s optimize the “Not Lame” stack together.

Learning that there are maturing options for voice control of your home with #HomeAssistant

kindalame.com/2026/03/26/why-home-assi...

'Zon same day delivery... Can now confirm that my new gen 3 doesn't turn off with my pve vm #HomeAssistant restart

Home Assistant 2026.4 is out! 🏠
The biggest new feature: native infrared support. One cheap IR blaster lets you control legacy TVs, ACs, and soundbars from Home Assistant — no hardware replacement needed.

And a lot more!

Full video: youtu.be/GiYTT6v6Apg

#HomeAssistant #SmartHome #HomeAutomation

NEW Video 👉 Home Assistant 2026.4 - Cool New Features & BIG Breaking Changes!

Click the link below to watch the full YouTube video or any other smart home related video on my YouTube channel.

youtu.be/C1e-p-bxQ3U

#HomeAssistant

NEW Video 👇👇👇

Home Assistant 2026.4 April Release Updates 🥳

👉Watch the Full Video : youtu.be/quGiTVndRfQ

#homeassistant #smarthome #release #iot

When #HomeAssistant restarts my gen 2 turns off. I had it monitoring an #Ollama server with rtx 3090 compute.

How disrespectful to my Lenovo P920 to be so aggressively turned off!

Did you happen to ask Third Reality Inc. if this problem could be _fixed_ in a gen 2 firmware update?

www.youtube.com/watch?v=V16M...

Wrote a conversational assistant system for home assistant using local LLM, local TTS and local transcription models

github.com/moimart/conv...

#ai #homeassistant #hassio #claude

Die Lieferung mit #Ikea #smarthome #timmerflotte kam an und ich habe die gleich mal in den #homeassistant eingebunden.

Mal schauen wie die sich schlagen und wie häufig sie die Verbindung zum #Dirigera Hub verlieren.

My Tapo switch died and somehow took my #HomeAssistant HVAC settings with it, I don't remember how to set it up either.

Ooh #HomeAssistant 2026.4 is going to have native Infrared support, so now I can finally automate my bidet!

Home Assistant 2026.3: nuevas automatizaciones, control avanzado de robots aspiradores y comandos por voz en Android | Multimedia | CIBERED Home Assistant 2026.3 llega con mejoras en automatización, limpieza por zonas para robots aspiradores, wake words en Android y nuevas integraciones

🏠 Home Assistant 2026.3: nuevas automatizaciones, control avanzado de robots aspiradores y comandos por voz en Android cibered.com/software/mul...

#HomeAssistant #Domotica #CasaInteligente #Cibered

IKEA just released new Matter over Thread sensors.

So of course I connected them directly into @home-assistant.io

No IKEA hub.
No vendor apps.
Just Matter + Thread.

Are they actually good enough for a real smart home?

youtu.be/joKC9fCGc4w

#homeassistant #smarthome #matter #thread

Home Assistant popup message. "We found compatible devices! These are found on your local network. Some are already added, others may need extra configuration." Listed Devices: Google Cast, Elgato Light, HomeKit Device, Internet Printing Protocol (IPP), LIFX, Google Next, Roborock, Thread, TP-Link Smart Home

I didn't think it would be this easy to integrate my existing "smart home" stuff. This gon' be fun! #HomeAssistant