Open WebUI — Setup
Browser-based AI frontend. Connects to Ollama on LXC 107 for local model inference. Runs as a Docker container on LXC 122 (ai).
Links
- GitHub: https://github.com/open-webui/open-webui
- Website: https://openwebui.com
Infrastructure
| Host | Internal | URL |
|---|---|---|
| Docker (LXC 122) | 192.168.1.122:8088 | https://ai.eva-00.network |
Observability
Logs
Open WebUI logs are collected via Grafana Alloy Docker discovery and shipped to Loki.
| Query | Purpose |
|---|---|
{container="open-webui"} |
All container output |
{container="open-webui"} \|= "error" |
Errors only |
{container="open-webui"} \|= "ollama" |
Ollama connection issues |
Access: Grafana → Explore → Loki → Enter query
Metrics
Open WebUI does not export Prometheus metrics by default. Use Loki logs to diagnose issues.
IaC
| Artifact | Path |
|---|---|
| Playbook | ansible/playbooks/ai.yml |
| Workflow | .forgejo/workflows/ai.yml |
| Compose | services/open-webui/docker-compose.yml |
Ollama connection
Open WebUI connects to Ollama via:
OLLAMA_BASE_URL=http://192.168.1.107:11434
Ollama must be running on LXC 107 for model inference to work. See the Ollama setup docs.
First-run
On first deploy, Open WebUI prompts to create an admin account via the web UI. Subsequent sign-ins use PocketID SSO.