# Installation Ollama und OpenWebUI ## 1. Verzeichnisstruktur vorbereiten Alle Containerdaten liegen unter `/opt/docker`, damit System und Daten sauber getrennt bleiben:
sudo <span class="hljs-built_in">mkdir</span> -p /opt/docker/ollamasudo <span class="hljs-built_in">mkdir</span> -p /opt/docker/open-webui
Optional Zugriffsrechte setzen:
sudo <span class="hljs-built_in">chown</span> -R kiadmin:docker /opt/docker
--- ## 2. Docker-Compose Datei Pfad: `/opt/docker/docker-compose.yml`
<span class="hljs-attr">version:</span> <span class="hljs-string">'3.8'</span><span class="hljs-attr">services:</span> <span class="hljs-attr">ollama:</span> <span class="hljs-attr">image:</span> <span class="hljs-string">ollama/ollama</span> <span class="hljs-attr">container_name:</span> <span class="hljs-string">ollama</span> <span class="hljs-attr">ports:</span> <span class="hljs-bullet">-</span> <span class="hljs-string">"11434:11434"</span> <span class="hljs-attr">restart:</span> <span class="hljs-string">always</span> <span class="hljs-attr">runtime:</span> <span class="hljs-string">nvidia</span> <span class="hljs-attr">volumes:</span> <span class="hljs-bullet">-</span> <span class="hljs-string">/opt/docker/ollama:/root/.ollama</span> <span class="hljs-attr">openwebui:</span> <span class="hljs-attr">image:</span> <span class="hljs-string">ghcr.io/open-webui/open-webui:ollama</span> <span class="hljs-attr">container_name:</span> <span class="hljs-string">open-webui</span> <span class="hljs-attr">restart:</span> <span class="hljs-string">always</span> <span class="hljs-attr">runtime:</span> <span class="hljs-string">nvidia</span> <span class="hljs-attr">ports:</span> <span class="hljs-bullet">-</span> <span class="hljs-string">"3000:8080"</span> <span class="hljs-attr">environment:</span> <span class="hljs-bullet">-</span> <span class="hljs-string">OLLAMA_API_BASE_URL=http://ollama:11434</span> <span class="hljs-attr">volumes:</span> <span class="hljs-bullet">-</span> <span class="hljs-string">/opt/docker/open-webui:/app/backend/data</span> <span class="hljs-comment"># Eigene CSS-Anpassungen (optional)</span> <span class="hljs-comment"># - /home/pleibling/docker/ai/custom.css:/app/build/static/custom.css:ro</span> <span class="hljs-attr">depends_on:</span> <span class="hljs-bullet">-</span> <span class="hljs-string">ollama</span>
--- ## 3. Container starten
<span class="hljs-built_in">cd</span> /opt/dockerdocker compose up -d
Status prüfen:
docker ps
- Ollama-API → `http://:11434` - OpenWebUI → `http://:3000` --- ## 4. Modelle installieren Ollama lädt Modelle nach Bedarf. Installation erfolgt z. B. über:
docker <span class="hljs-built_in">exec</span> -it ollama ollama pull gpt-oss:20b
**Verwendete LLMs in dieser Umgebung:** - **GPT-OSS:20b** → Hauptmodell (groß, sehr leistungsfähig) - **Llama 3.1:12b** - **Mistral:7b** - **Gemma3:12b** - **DeepSeek-R1:8b** - **Qwen3:14b** Optional kannst du auch in OpenWebUI für jedes Modell eigene Presets/Workspaces definieren. --- ## 5. Update der Container
<span class="hljs-built_in">cd</span> /opt/dockerdocker compose pulldocker compose up -d
--- ## 6. Backup-Hinweis Da alle persistenten Daten in `/opt/docker/ollama` und `/opt/docker/open-webui` liegen, reicht ein Backup dieser Ordner. ---