docker compose for locally hosted llms with tailscale

docker-compose.llms.yaml

services:
  # More info on using tailscale like this can be found in the
  # tailscale blog entry about it: https://tailscale.com/blog/docker-tailscale-guide
  ts-aichat-server:
    image: tailscale/tailscale:latest
    container_name: ts-aichat-server
    hostname: aichat
    environment:
    - TS_AUTH_KEY=tskey-auth-REPLACEME
    - TS_EXTRA_ARGS=--advertise-tags=tag:container --reset
    - TS_STATE_DIR=/var/lib/tailscale
    volumes:
    # you'll probably want to map these to directories rather than volumes
    - ts-aichat-state:/var/lib/tailscale
    - ts-aichat-config:/config
    - /dev/net/tun:/dev/net/tun
    cap_add:
    - net_admin
    - sys_module
    restart: unless-stopped
  ollama:
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            count: all
            capabilities:
            - gpu
    volumes:
    - ollama:/root/.ollama/
    # we don't publish ports when we're using network_mode: service
    #ports:
    #  - 11434:11434
    container_name: ollama
    image: ollama/ollama:latest
    network_mode: service:ts-aichat-server
    depends_on:
    - ts-aichat-server
    restart: unless-stopped
  open-webui:
    environment:
    - OLLAMA_BASE_URL=
    volumes:
    - open-webui:/app/backend/data
    container_name: open-webui
    # we don't publish ports when we're using network_mode: service*
    #ports:
    #  - 3000:8080
    image: ghcr.io/open-webui/open-webui:main
    network_mode: service:ts-aichat-server
    depends_on:
    - ts-aichat-server
    restart: unless-stopped
  litellm:
    image: ghcr.io/berriai/litellm:main-latest
    command: --api_base https://REPLACE_WITH_OLLAMA_API_BASE -m ollama/mistral-openorca
    # we don't publish ports when we're using network_mode: service*
    #ports:
    #  - 4000:4000
    network_mode: service:ts-aichat-server
    depends_on:
    - ts-aichat-server
    restart: unless-stopped
version: "3"