I'm the author. I wanted a phone OS purpose-built for talking to an LLM framework running on my own server. It's easier to tell my LLM assistant to "write a text to my wife with the time of the last meeting on my calendar for the day, and set a 5 minute timer for eggs" than it is to navigate through the calendar, contacts, messages, then timer apps, finding the info, writing the message, etc. I couldn't find a mobile operating system that strips all non-essential apps, including the app store, and that is built around "type/hold to talk" commands sent to your own server. I imagine this is how we will be interacting with technology in the future, rather than downloading apps, so I started an open source project for it.
RistOS is the frontend OS for that. It's a de-Googled OS for the Pixel 10a where a push-to-talk assistant replaces the launcher, and the endpoint it posts to is a blank field you fill in. Your server does the bulk of the work. You set up speech-to-text and speech synthesis on your server, and whatever tools your server exposes are the tools the phone has. You change what the phone can do by editing a server rather than by writing an Android app.
I've built a backend for myself and I'm working on a version I can share. I intend to publish it within about a month. Until then you can build your own backend using the docs in the repo.
Right now, this is similar to connecting a dumb phone to your own LLM, so there are some limitations. There is no app store or web browser. This works for me because I'm trying to declutter my devices, remove social media and other distractions from my phone and lead by example in my home by using the kind of device I'd be comfortable with my kids using. My calendar, contacts, notes, etc. are stored off-device. This means I can't access them without service or wifi. I'm never really in a position where I don't have service but still need phone functionality, so this isn't a problem for me. Also, I travel a lot, so I'd rather my contacts and history sit on a server I control than on a handset that can be stolen or handed over and searched at a border.
Aside from fellow tech minimalists and people who want more control over the way they integrate LLMs into their personal devices, I imagine this could also be useful if you're managing a fleet of devices and want to give people access to your company's internal LLM without having other apps on the device.
Disclosures:
- I am also working on commercial tools that will use this OS, but nothing is for sale today, there's no signup, and the OS neither requires it nor ships its address.
- Emergency calling is not yet validated. Ordinary calls work so far, but emergency calling on a VoLTE-only network is a separate IMS registration I haven't exercised, and nobody has confirmed on a live network that a 911 call from this build connects.
- I've only tested on Pixel 10a.
- What's on it: calling, texting, contacts, camera, gallery, maps, flashlight, settings, and your backend if you connect it. Nothing else, and no way to add anything.
Apache-2.0 for the layer I wrote, kernel and GPL userspace source per release. It's a distinct OS based on the GrapheneOS tree. It's not affiliated or endorsed.
I'm the author. I wanted a phone OS purpose-built for talking to an LLM framework running on my own server. It's easier to tell my LLM assistant to "write a text to my wife with the time of the last meeting on my calendar for the day, and set a 5 minute timer for eggs" than it is to navigate through the calendar, contacts, messages, then timer apps, finding the info, writing the message, etc. I couldn't find a mobile operating system that strips all non-essential apps, including the app store, and that is built around "type/hold to talk" commands sent to your own server. I imagine this is how we will be interacting with technology in the future, rather than downloading apps, so I started an open source project for it.
RistOS is the frontend OS for that. It's a de-Googled OS for the Pixel 10a where a push-to-talk assistant replaces the launcher, and the endpoint it posts to is a blank field you fill in. Your server does the bulk of the work. You set up speech-to-text and speech synthesis on your server, and whatever tools your server exposes are the tools the phone has. You change what the phone can do by editing a server rather than by writing an Android app.
I've built a backend for myself and I'm working on a version I can share. I intend to publish it within about a month. Until then you can build your own backend using the docs in the repo.
Right now, this is similar to connecting a dumb phone to your own LLM, so there are some limitations. There is no app store or web browser. This works for me because I'm trying to declutter my devices, remove social media and other distractions from my phone and lead by example in my home by using the kind of device I'd be comfortable with my kids using. My calendar, contacts, notes, etc. are stored off-device. This means I can't access them without service or wifi. I'm never really in a position where I don't have service but still need phone functionality, so this isn't a problem for me. Also, I travel a lot, so I'd rather my contacts and history sit on a server I control than on a handset that can be stolen or handed over and searched at a border.
Aside from fellow tech minimalists and people who want more control over the way they integrate LLMs into their personal devices, I imagine this could also be useful if you're managing a fleet of devices and want to give people access to your company's internal LLM without having other apps on the device.
Disclosures:
- I am also working on commercial tools that will use this OS, but nothing is for sale today, there's no signup, and the OS neither requires it nor ships its address.
- Emergency calling is not yet validated. Ordinary calls work so far, but emergency calling on a VoLTE-only network is a separate IMS registration I haven't exercised, and nobody has confirmed on a live network that a 911 call from this build connects.
- I've only tested on Pixel 10a.
- What's on it: calling, texting, contacts, camera, gallery, maps, flashlight, settings, and your backend if you connect it. Nothing else, and no way to add anything.
Apache-2.0 for the layer I wrote, kernel and GPL userspace source per release. It's a distinct OS based on the GrapheneOS tree. It's not affiliated or endorsed.