ops
Security model
What DeepLM Ops can and cannot do, and how the container is locked down.
Read-only by design
DeepLM Ops is a read-only assistant. It answers questions about your GPUs and Slurm jobs, and the model runs locally on llama.cpp, on your own machine.
Local only
The run command publishes the chat page on 127.0.0.1:8090 only, so it is reachable from your machine and not from the network.
The sign-in link is a password
The link printed in docker logs dlm-ops contains a token. Anyone with the link can use your instance, so keep it out of chats, issues and screen shares. docker restart dlm-ops makes a new one.
Container hardening
The run command starts the container with a locked-down configuration. See the Docker run flags reference for what each flag does.