Proxmox LCX + OpenWebUI + DeepSeek R1

Using a Proxmox Virtual Enviroment Linux Container (PVE LCX) I was able to very easily deploy the DeepSeek R1 model for local LLM responses.

There is a great working Proxmox VE Helper-Scripts which deploys the whole thing for you, and then you need to just add the DeepSeek model.

bash -c "$(wget -qLO - https://github.com/community-scripts/ProxmoxVE/raw/main/ct/openwebui.sh)"

The LCX needs to be privileged to be able to access the Nivida Tesla P4 card available in the same server. This means change the setting in the helper script. See here for details on adding the Nvidia card...

./NVIDIA-Linux-x86_64-565.57.01.run --no-kernel-modules
reboot

After the reboot you should be able to connect to the web interface on port 8080 and create the admin user. 

Under admin settings and the models section, you can hit the download button and then add "deepseek-r1" into the "Pull a model from Ollama.com" section