Open WebUI + Ollama for a Local Chat Front Door

Ollama is great for model management. Many people still want a ChatGPT-like browser UI for family, team demos, or non-developer users. Open WebUI is a common front door.

• Browser chat against local models
• Model switching without teaching everyone the CLI
• A stack that stays on your LAN if you bind carefully

1. Install and start Ollama. Pull at least one model you actually use.
2. Confirm the Ollama API responds on the default port (typically 11434).
3. Run Open WebUI via Docker, pointing it at the Ollama base URL. On Linux, host networking or host.docker.internal patterns may be required so the container can reach Ollama.
4. Create an admin user on first visit and disable open signup if this is more than a personal toy.
5. Select the model in the UI and send a test prompt.

• UI lists your pulled models
• A short chat returns tokens
• Stop Docker/Ollama and confirm you know the restart order

• Empty model list: Open WebUI cannot reach Ollama — fix networking first
• Slow answers: model too large for hardware; try a smaller quant
• Exposed to the internet by accident: bind to localhost or put behind auth and reverse proxy

A chat UI is not an agent harness. For coding agents, keep using CLI/IDE tools with verification commands. This recipe is for accessible local chat, not autonomous tool use.