Local AI: open it, first login, settings, fixes
· Updated 18 September 2026 · SparkBox team
Run large language models locally with Ollama, and chat with them through Open WebUI — a private, ChatGPT-like interface. No data ever leaves your server. This is the short card for it — the two minutes after you press Enable, and the page to come back to when something is off.
Before you turn it on
LLMs need significant RAM. 3B models run on 8GB, 7B on 16GB, 13B+ needs 32GB+. The UGREEN DXP2800 with 16GB RAM can comfortably run 7B models.
Open it
Dashboard → Apps → Local AI → Open. The direct address on your network is:
http://<your-box-IP>:8090Create your first account after opening the web UI — that account becomes admin
First login
No preset password. The FIRST account created on the signup screen becomes the admin of your AI chat — create yours straight after enabling, before anyone else on the network does.
Any password the installer generated for you is under Settings → Passwords; the Open button offers to copy it.
Settings that matter
Change these under Apps → Local AI (or when you enable it):
- Default Model — Which model to download by default. Small models (3B-7B) work on limited hardware. Large models (70B) need 64GB+ RAM.
When it breaks
If it will not open, check the app card first — a red dot means it is restarting or unhealthy, and this walkthrough reads the reason from its log. Tom AI on the dashboard can do the same for you.
Where your data lives
/opt/sparkbox/modules/localai/config/Included in the dashboard's Backup by default. Restoring that backup on a fresh install brings Local AI back exactly as it was.
Questions, or did this not match your box?
Every guide here came from a real problem someone hit. If yours behaves differently, say so — that is how these get corrected, and how the fix gets prioritised.
We answer there rather than in a comment box, because that is where the people who have already solved it are.