← The story behind LLMLifeline
LOCAL AI / DATA SOVEREIGNTYWindows · macOS · Linux

LLMLifeline.

A private AI that runs on your own computer, and that your whole household can share over your home network, without giving your data to anyone.

LLMLifeline puts a chat assistant, image and video generators and a careful computer-use assistant on your own hardware. One computer runs the model. Your other computers, phones, tablets and Chromebooks use it over your home network, and an approved device can even give it tasks from another room.

The models run on your machine, so your conversations never go to an AI company. Sharing answers only devices on your own network, over encrypted connections, and you approve every device with a matching six-digit code.

What you get

  • The installer for Windows (LLMLifeline.exe), macOS (Apple Silicon) and Linux. On its first start it sets up everything else by itself: its Python packages, ComfyUI for images, three uncensored chat models and four image models
  • One chat model per tier: Qwen3.8 27B (high), Gemma 4 12B (middle) and Qwen3.5 4B (low). LLMLifeline starts on the best one your computer runs well, and moves up a tier on its own when you get a bigger graphics card
  • Uncensored image models the same way: Z-Image Turbo (high), RealVisXL 5 (middle) and Realistic Vision 6 (low), next to the SDXL base model
  • Video from a description: Wan video models in large, medium and small. The size your graphics card holds downloads the first time you use video (it needs a card with 6 GB or more)
  • Start a video from a picture or continue one, up to five minutes long, with an estimate of how long it takes and a scene-by-scene sheet for keeping later videos consistent
  • A profile for each person in your home, which learns who they are and what they like, and notices when someone else is typing
  • Private chat with any .gguf model, including ones bigger than your graphics card, with replies as they're written
  • Projects that remember, with your documents indexed so the assistant can quote them
  • Local image generation, and photos the assistant can describe
  • Computer use, off until you turn it on: it asks before each new action and can never upload your data
  • One model shared across your home network, with phones and Chromebooks connecting through the browser
  • Remote control of the assistant from an approved device, under the same permissions
  • Every update, only when you ask: LLMlifeline never updates itself or checks on its own. When you choose to look, it shows what's new, installs only if you say so, and keeps the previous version so you can switch back. Your purchase key also opens the newest downloads whenever you need them

Your key, your computers

  • On its first start, LLMLifeline asks for your purchase key. Activating sends the key and an anonymous ID for that computer to the store, once. After that it checks its license offline, and it only contacts the store again if you ask it to check for updates.
  • A key activates up to 3 computers. A copied LLMLifeline folder won't run on another computer without its own activation, so please don't share your key or the downloads.
  • Moving to a new computer? Reset your computers on your purchase page, once every 30 days.

System requirements

  • Windows. Windows 10 or 11, 64-bit, with Python 3.10 and git installed. An NVIDIA graphics card with a current driver is strongly recommended; without one, LLMLifeline starts on its small model, which runs on the processor. 16 GB of memory at minimum, 32 GB recommended.
  • macOS. A Mac with Apple Silicon (M1 or newer), Apple's Command Line Tools and Python 3.10 to 3.13 from python.org. Intel Macs aren't supported.
  • Linux. 64-bit Linux with Python 3.10 to 3.13, python3-venv, git and a C/C++ compiler (build-essential). An NVIDIA graphics card is optional.
  • Disk space. The first start downloads ComfyUI (about 7 GB) and its models (about 63 GB: three uncensored chat models and four image models; you can skip them and add your own). Video models download the first time you use video, in the size your graphics card holds: 10 to 38 GB. A computer that only uses another computer's model needs under 1 GB.

Honest status

  • Windows is where I build LLMLifeline and use it every day.
  • The macOS and Linux builds pass automated checks on every release, but I haven't used them on a real Mac or Linux PC.
  • Small models follow instructions less reliably; a bigger model works far better, especially for computer use.
  • The chat, image and video models are uncensored: the chat models' built-in refusals have been removed, and the image and video models have no content filter. They answer and draw what you ask. What they make can be wrong, offensive or unsafe, and you decide what to do with it.
  • LLMLifeline is a baseline. It sets itself up with models that fit the computer you have today, and how well it answers, draws and films depends on that computer. As you upgrade, you can move to bigger and more capable models and bring your own: any .gguf chat model, a Hugging Face model folder it converts for you, or any image or video model ComfyUI can run, through your own workflow.
  • If LLMLifeline doesn't run on your computer, email nyctitropist@gmail.com with what happened, and I'll work on a fix in an update. Sales are final, so there are no refunds.