The easiest way to run LLM locally with Docker

This article will be as short as it is easy to set up, provided you already have Docker installed on your machine.
This solution works on any machine because it only uses the CPU, which is available on every machine that works :)

Docker Compose:

services:
  webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: llm-webui
    ports:
      - 4444:8080/tcp
    volumes:
      - /mnt/Docker/llm/open-webui:/app/backend/data
    extra_hosts:
      - "host.docker.internal:host-gateway"
    depends_on:
      - ollama
    environment:
      - OLLAMA_BASE_URL=http://192.168.1.30:11434
      - USER_AGENT=Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36
    cpus: 1
    mem_limit: 2G
    pids_limit: 100

  ollama:
    image: ollama/ollama
    container_name: llm-ollama
    ports:
      - 11434:11434/tcp
    healthcheck:
      test: ollama --version || exit 1
    volumes:
      - /mnt/Docker/llm/ollama:/root/.ollama
    cpus: 3
    mem_limit: 10G
    pids_limit: 400

Change the OLLAMA_BASE_URL variable to the IP address corresponding to your machine, and add the path corresponding to your machine in the volumes section.

When you're ready, start the stack.

After that you can reach the webUI on this URL (modify the IP):
192.168.1.30:4444

Download libaries

This is also very simple. Select an LLM from this link, then modify the command below accordingly, e.g.:
docker exec -it llm-ollama ollama pull llama3.2-vision:11b

List currently installed libaries:
docker exec -it llm-ollama ollama list

Remove one of them:
docker exec -it llm-ollama ollama rm llama3.2-vision:11b

Note: I have observed that roughly the same amount of memory is required as the size of the library.

That's it. It wasn't that hard, was it? :)