Alexa+ is Amazon’s new, premium voice assistant that costs $20 per month (but comes free with Amazon Prime).Amazon’s top-billed feature is a conversational assistant that’s a lot chattier than the previous version.With Home Assistant, you can achieve something similar for free.
Alexa+ style conversations to Home Assistant Use LLMs to chat and interact with your smart home Amazon’s flagship Alexa+ evolution is a more “conversational” assistant.This has been achieved with the use of a large language model (LLM), essentially fusing the chatbot style of ChatGPT or Gemini with a smart home voice assistant that’s in control of your smart home.There’s more to Alexa+ than just this (and we’ll get to that later).
But if all you want to have chatbot-style interactions using a Home Assistant voice assistant, you’re in luck.You can even choose between using cloud-reliant models (which is what Alexa+ is) or running your own models locally and keeping everything offline.I’ve got two voice assistants on my Home Assistant server right now.
One uses Google’s services (including Gemini) for conversation, text-to-speech (TTS), and speech-to-text (STT); the other uses entirely offline models for all.Gemini is faster and easier to set up, while the offline models took a bit more tinkering but still work consistently.Adding Google Gemini to Home Assistant To add Gemini to Home Assistant, head to Google’s AI Studio, log in, click Get API Key, and copy it to your clipboard.
Now head back to Home Assistant and click Settings > Devices followed by Add Integration on the Integrations tab.Search for “Google Gemini” and then paste the API key you previously retrieved into the box.Hit Submit and then Skip and finish to complete setup.
Alternatively, you can perform a similar process with integrations for OpenAI, Anthropic, and other cloud-based LLMs.You’ll only get so many API calls with a cloud provider per day.Even though this is technically free, you’ll need to pay once you run up against its limitations.
Build a local AI assistant with Ollama and the Wyoming Protocol If you want to roll your own LLM, you can do so.Ollama is a solid starting point, so download and install it on whichever server you’ll be using.From here, head to Settings > Integrations and install the “Ollama” integration.
Point the integration towards the local IP address on which Ollama is running (e.g.http://192.168.0.149:11434).You’ll need to figure out which model you want to use with Ollama and download it.
I personally settled on qwen2, since it’s small and fast enough for my modest M1 Mac mini with 16GB of RAM to handle without choking.I tested this out on my server by running ollama run qwen2 in Terminal (which forced Ollama to download the model), then launching the app and having a mock conversation in a new chat.Now under the Ollama integration, click Add conversation agent and create an agent using your chosen model.
You can edit the prompt, enable Home Assistant control (which you’ll probably want to do), and more.Lastly, you’re going to want some means of handling TTS and STT locally too.The Open Home Foundation’s Wyoming Protocol is perfect for this.
Head to Settings > Devices & services and install “Wyoming Protocol.” Next, head to Settings > Apps and download both “Piper” and “Whisper” then restart your Home Assistant server.Make sure both Piper and Whisper are running, and head back to the Devices & services screen.Home Assistant should automatically detect these services and give you the option to “Add” them both.
Almost done! Building your assistant and controlling your home Everything’s connected From here, it’s a simple case of heading to Settings > Voice Assistant and creating a new one (or editing the one you already have).You can choose a conversation agent from the drop-down menu, as well as STT and TTS models.From here, you’ll be able to use your voice assistant to chat as you would with Alexa+.
You can do this by triggering your Home Assistant smart speaker and asking a question, or by clicking the “Assist” button on the dashboard of the web UI or smartphone and smart watch apps.You can even build or convert your own smart speaker using an ESP32.In addition to chaining commands like “turn off the kitchen lights and lock the front door,” you can ask your assistant all manner of chatbot-related queries.
Just remember that this is an LLM, and it’s subject to the same foibles as any other.It confidently answer incorrectly when you ask how many Rs are in strawberry! Both of these methods allow you to customize the prompt used.This is what is sent to your model of choice before every query.
You can use this to add unnecessary personality, or create an expedient assistant that always answers in short sentences.There’s also a “Think before responding” option for local LLMs, if you’d prefer better quality answers at the cost of speed.Alexa+ wins on personalization and service connectivity But it’s a privacy nightmare There’s a lot more to Alexa+ than just LLM integration, but this is arguably one of the biggest draws for many.
The conversational style means that you can keep chatting to Alexa+ without uttering the wake word again.Home Assistant already has this insofar as it will anticipate a response if it asks you a question, but it’s not quite on the level of Alexa+.The same is true of personalization and learning.
Alexa+ can learn your food allergies and preferences, while remembering which route you took on your last road trip.Home Assistant can reference your calendar, but only if you use or import your own schedule into the platform’s under-utilized calendar element.Alexa+ talks to external services.
You can buy things on Amazon, book a table at a restaurant, or order an Uber just by asking Alexa+.Home Assistant cannot do any of these things, but that’s not necessarily a bad thing.For many, the privacy implications of your smart home assistant knowing where you live, what you eat, and where you’re going will be too much to bear.
Home Assistant Voice Preview Edition Dimensions 84x84x21 mm Weight 96 g Home Assistant Voice Preview Edition is a privacy-first smart speaker built as an alternative to the likes of Amazon Alexa and Google Nest Mini.It adds voice assistant capabilities, including local-only processing, to a Home Assistant-powered smart home. $59 at Ameridroid Expand Collapse reSpeaker Lite Brand Seeed Studio CPU ESP32-S3R8 The reSpeaker Lite Voice Assistant Kit includes a two-mic array, a pre-soldered XIAO ESP32-S3 controller, and an XMOS XU316 audio processor with onboard natural language understanding, interference cancellation, acoustic echo cancellation, noise suppression, and automatic gain control.Hooked up a 5W speaker, you can create your own local voice assistant that you can connect to Home Assistant via ESPHome.
$70 at Amazon $30 at Seeed Studio Expand Collapse Alexa+ will be appealing to those well and truly ensconced in Amazon’s ecosystem.For me, nothing beats Home Assistant—especially not a cloud-dependent service like Alexa that could put your whole smart home at risk.
Read More
