A private AI for the whole house

What you need

• One always-on computer with 16GB or more; 24GB-plus for better models. See the hardware guide.
• Docker installed on it (Docker Desktop on Mac or Windows).
• About an hour.

Set it up

1. 1
   Run Open WebUI with Ollama bundled in one container. This is the command from Open WebUI's quick start; replace the secret with a long random phrase of your own and keep it out of anything you share:

   docker run -d -p 3000:8080 \
     -v ollama:/root/.ollama \
     -v open-webui:/app/backend/data \
     -e WEBUI_SECRET_KEY=<a-long-random-phrase> \
     --name open-webui --restart always \
     ghcr.io/open-webui/open-webui:ollama

   Add --gpus=all after -d if the machine has an NVIDIA card.
2. 2
   Open http://localhost:3000 on that machine and create the first account. The first account becomes the administrator.
3. 3
   In the admin settings, download a model (for example gemma4:12b) and set the context to 8,192 or more.
4. 4
   Turn on new sign-ups under the admin authentication settings. New accounts wait as pending until you approve them.
5. 5
   On the other devices, open http://<that-computer's-local-address>:3000. Find the address in the machine's network settings.

Make it the household's

• Give each person a model with their own instructions (see custom models).
• Add shared documents for everyone to ask about (see chat with your documents).
• Turn on memory per account if people want the AI to remember facts about them across chats.