Ollama + Docker: Open WebUI setup for your local network

Hugo Paredes demonstrates a self-hosted ChatGPT alternative with Ollama, Docker Desktop and Open WebUI. A Windows 11 desktop handles inference while other devices access the chat interface through a browser. The walkthrough starts with Ollama, Docker Desktop and the models already installed; it focuses on sharing that setup across a local network.

The tutorial deploys Open WebUI in a Docker container on port 3000, finds the host's LAN address and creates a Windows Firewall rule for inbound TCP connections. The demonstrated rule applies to domain, private and public profiles. Paredes then creates the first administrator account, enables registration and approves a pending user. He changes model permissions so that users can select the two Gemma 4 variants he describes as having 2 billion and 4 billion parameters.

RustDesk provides remote access to the Windows laptop and Linux client during configuration. The demonstration also covers creating an account directly from the admin panel and opening the interface on a phone over Wi-Fi. The clients access models running on the desktop rather than performing inference themselves.

The reported host has an Intel i5-14600K, 64 GB of RAM and an RTX 3070 with 8 GB of VRAM. Paredes tests Python script requests and describes the replies as fast, but provides no measured benchmark or minimum hardware requirement.