Skip to main content

Configuring Ollama

Ollama is the engine that lets you run LLM models locally on your machine. DM Loremind offers three modes depending on your setup.

If you chose the desktop mode​

In this case, you will need to install Ollama separately and configure it so that it is reachable by the DM Loremind Docker container. At the time of writing this documentation, the official Ollama page is the following: Ollama link

Which mode should you choose? (Docker part)​

During installation, the installer asks two successive questions.

1. "Do you already have Ollama installed on this machine?"​

Answer: Yes → secure host mode​

The installer automatically calls the secure-host-ollama.{sh,ps1} helper, which configures your existing Ollama so that it is reachable by the DM Loremind Docker container without being exposed to the local network or the Internet.

  • Linux: Ollama listens on the Docker gateway IP (172.17.0.1 by default). This IP is never routed outside the machine. The systemd override is written to /etc/systemd/system/ollama.service.d/loremind-host.conf.
  • Windows: Ollama listens on 0.0.0.0 (technically required with Docker Desktop), but the Windows firewall is configured to only allow the loopback and the Docker Desktop subnets. The added rules are prefixed LoreMind-Ollama-*.

The URL configured in .env is OLLAMA_BASE_URL=http://host.docker.internal:11434.

2. "Do you want to install Ollama via Docker now?" (if the first answer was No)​

Answer: Yes (default) → embedded mode​

An ollama service is added to the stack via the local-ollama Docker profile. Ollama runs in a dedicated container, on the internal Docker network, never exposed to the LAN or the Internet. Models are stored in the ollama-data Docker volume (persistent across restarts and updates).

  • URL: OLLAMA_BASE_URL=http://ollama:11434 (internal Docker DNS).
  • No network or firewall configuration required.
  • Automatic NVIDIA GPU support if available.

To download a model:

docker exec -it loremind-ollama ollama pull gemma4:26b
docker exec -it loremind-ollama ollama list

Answer: No → deferred mode​

No Ollama configuration is applied. The installer finishes without Ollama. You configure Ollama later via DM Loremind's Settings page by entering the URL of your Ollama server.

Running the hardening helper manually​

If you chose deferred mode and then installed Ollama later on your machine, or if you want to switch from embedded mode to host mode:

Linux​

bash secure-host-ollama.sh
# Then in the .env of the installation folder:
#   OLLAMA_BASE_URL=http://host.docker.internal:11434
# And: docker compose up -d

Windows (PowerShell admin)​

.\secure-host-ollama.ps1
# Then edit .env (in %LOCALAPPDATA%\DM Loremind\):
#   OLLAMA_BASE_URL=http://host.docker.internal:11434
# And: docker compose up -d
Safe to re-run

The helpers remove their old rules before recreating them. This is useful, for example, if you have reset Docker Desktop and the subnets have changed.

Undoing the hardening configuration​

Linux​

sudo rm /etc/systemd/system/ollama.service.d/loremind-host.conf
sudo systemctl daemon-reload && sudo systemctl restart ollama

Windows (PowerShell admin)​

Get-NetFirewallRule -DisplayName "LoreMind-Ollama-*" | Remove-NetFirewallRule
[Environment]::SetEnvironmentVariable("OLLAMA_HOST", $null, "User")