Let's look at a few sol­id choic­es. Some are for talk­ing to your smart home, oth­ers are for help­ing you write code, and some are gen­er­al-pur­­pose assis­tants you can run on your own serv­er.

For Your Smart Home: Home Assistant's Assist

If you're into home automa­tion, you prob­a­bly know about Home Assis­tant. It’s an open-source plat­form that lets you con­trol all your smart devices from one place. They also have a voice assis­tant called Assist.

The main idea behind Assist is local con­trol and pri­va­cy. Unlike the big-name assis­tants that send your voice com­mands to the cloud, Assist can run entire­ly on your own hard­ware at home. This means what you say stays in your house.

To get start­ed, you'll need a device run­ning Home Assis­tant. This could be a Rasp­ber­ry Pi or a small serv­er. Then, you need a way to talk to it. Home Assis­tant offers ded­i­cat­ed hard­ware called the Voice Pre­view Edi­tion, which is a small speak­er with a micro­phone designed to work direct­ly with Assist. You can also build your own voice assis­tant device for around $13 using hard­ware like an ESP32.

Once you have the hard­ware set up, you con­nect it to your Home Assis­tant instance. The sys­tem guides you through the process. You can choose a wake word, like "Okay Nabu," "Hey Jarvis," or "Hey Mycroft." After that, you can start giv­ing it com­mands to con­trol your lights, check the tem­per­a­ture, or run any automa­tion you've set up.

For speech pro­cess­ing, you have two main choic­es. You can process every­thing local­ly if your hard­ware is pow­er­ful enough (an Intel N100 proces­sor or bet­ter is rec­om­mend­ed for full local pro­cess­ing). Or, you can use the Home Assis­tant Cloud ser­vice, which helps with pro­cess­ing and can be faster on less pow­er­ful devices.

A Flexible Voice Platform: OpenVoiceOS (OVOS)

Anoth­er strong con­tender in the voice assis­tant space is Open­VoiceOS, or OVOS. Think of it as the spir­i­tu­al suc­ces­sor to Mycroft AI. It's a com­­mu­ni­­ty-dri­ven plat­form designed to be a flex­i­ble foun­da­tion for voice inter­faces on all sorts of devices, from smart speak­ers to robots.

OVOS is built to be mod­u­lar. It has a core sys­tem that man­ages dif­fer­ent ser­vices like skills, audio, and speech pro­cess­ing. This lets you swap out com­po­nents. For exam­ple, you can choose dif­fer­ent speech-to-text (STT) and text-to-speech (TTS) engines. While the default STT requires an inter­net con­nec­tion, you can set up self-host­ed, offline options.

Get­ting OVOS run­ning is straight­for­ward. You can use their installer on a Lin­ux sys­tem or run it in a Dock­er con­tain­er for a sta­ble set­up. For Rasp­ber­ry Pi users, there's a ded­i­cat­ed image called RaspOVOS. The plat­form is com­pat­i­ble with many of the skills orig­i­nal­ly devel­oped for Mycroft, giv­ing you a good base of abil­i­ties to start with.

Because it's so flex­i­ble, peo­ple have used OVOS for cre­ative projects, like build­ing a smart mir­ror that can switch between dif­fer­ent AI per­sonas. The com­mu­ni­ty is active, and you can find help and share ideas through their chat and dis­cus­sion forums.

For a General-Purpose Assistant: Leon

If you're look­ing for an assis­tant that's less focused on smart homes and more on gen­er­al tasks, check out Leon. Leon is an open-source per­son­al assis­tant that you can run on your own serv­er. It’s built with Node.js and Python and uses AI con­cepts to under­stand and respond to you.

Leon's goal is to be your vir­tu­al brain. You can ask it to do things, and it will exe­cute them. It’s designed to be self-host­ed, so your data stays with you. It sup­ports var­i­ous speech-to-text and text-to-speech ser­vices, includ­ing some that can be self-host­ed, giv­ing you more con­trol over pri­va­cy.

The set­up involves cloning its code from GitHub and run­ning it on your serv­er. It’s a bit more hands-on than some­thing like Home Assis­tant Assist, but it gives you a per­son­al assis­tant that you con­trol com­plete­ly.

For Coding: Open Source AI Coding Assistants

For devel­op­ers, there are sev­er­al open-source AI assis­tants designed to help write code. These are alter­na­tives to tools like GitHub Copi­lot. The big advan­tage is that you can run them local­ly, which is impor­tant if you work with sen­si­tive code.

One pop­u­lar option is Con­tin­ue. It's an exten­sion for VS Code and Jet­Brains that lets you con­nect to any AI mod­el, whether it's a cloud ser­vice or a mod­el you're run­ning on your own machine. You can use it for auto­com­ple­tion, chat­ting with your code to under­stand it bet­ter, and even automat­ing edits across mul­ti­ple files.

Anoth­er one is Tab­by, a self-host­ed AI cod­ing assis­tant. You can run it on your own infra­struc­ture, giv­ing you full con­trol over your data. Zed is anoth­er tool that stands out; it's a code edi­tor built from the ground up with AI as a core fea­ture.

These tools offer trans­paren­cy and cus­tomiza­tion that pro­pri­etary solu­tions can't match. You can see exact­ly how the AI is han­dling your code and tai­lor it to your spe­cif­ic needs.

Putting It All Together With a Web UI

No mat­ter which back­end AI mod­el you choose to run, you'll like­ly want a good user inter­face to inter­act with it. Open WebUI is a pop­u­lar choice. It's a self-hostable web inter­face that looks and feels a lot like Chat­G­PT, but you con­trol every­thing.

You can deploy it eas­i­ly using Dock­er. Once it's run­ning, you can con­nect it to var­i­ous AI mod­els, includ­ing local ones you run with a tool like Olla­ma. A key fea­ture of Open WebUI is its built-in Retrieval-Aug­­men­t­ed Gen­er­a­tion (RAG) capa­bil­i­ty. This lets you upload your own doc­u­ments, like PDFs or text files, and the AI can answer ques­tions based on that pri­vate infor­ma­tion. This is use­ful for cre­at­ing a per­son­al knowl­edge base or a ques­­tion-answer­ing sys­tem for a busi­ness.

Choos­ing an open-source AI assis­tant comes down to what you want to do. For smart home con­trol, Home Assistant's Assist is a nat­ur­al fit. For a flex­i­ble, hack­able voice plat­form, Open­VoiceOS is a great option. For cod­ing, tools like Con­tin­ue and Tab­by give you con­trol over your devel­op­ment work­flow. By com­bin­ing these with a self-host­ed inter­face like Open WebUI, you can build a pow­er­ful, pri­vate AI assis­tant tai­lored just for you.