View source for OpenWebUI - 251128-justin

= OpenWebUI Docker Compose Configuration =

'''Reference ID:''' 251128-justin<br>
'''Author:''' Justin

This page contains the <code>docker-compose.yml</code> configuration for Open WebUI, designed to run in front of a local or remote LLM backend (Ollama, llama.cpp, LM Studio, etc.).

== Notes ==

* Typically placed behind an NGINX reverse proxy on a separate "edge" host that terminates HTTPS for your personal domain using Let's Encrypt and forwards traffic to this service on port 3000.
* The named volume <code>open-webui-data</code> is bind-mounted to a specific host path on a backup-managed storage pool (TrueNAS/Synology).
* Application data (users, chat history, uploads, settings, etc.) is expected to grow over time — easily 2–3 GB per year, especially if used for contract and document analysis.

== docker-compose.yml ==

<syntaxhighlight lang="yaml">
# docker-compose.yml for Open WebUI in front of a local/remote LLM backend
# - Designed to work with Ollama, llama.cpp servers, LM Studio, etc. via HTTP API
# - Typically placed behind an NGINX reverse proxy on a separate "edge" host
#   that terminates HTTPS for your personal domain using Let's Encrypt and
#   forwards traffic to this service on port 3000.

services:
  open-webui:
    image: ghcr.io/open-webui/open-webui:${WEBUI_DOCKER_TAG-main}   # Use WEBUI_DOCKER_TAG env var; defaults to 'main' if not set
    container_name: open-webui
    ports:
      - "3000:8080"    # Expose container port 8080 on host port 3000; NGINX reverse proxy points here

    environment:
      # URL of your LLM backend:
      #   - For Ollama: use the HTTP endpoint reachable from this container.
      #     If Ollama is running as a Docker service on the same Docker network,
      #     you can reference it by service name, e.g. http://ollama:11434
      #   - For llama.cpp / LM Studio / other backends: point to their HTTP API
      #     URL (for example, http://llama-server:8080) as long as this
      #     container can reach it.
      - OLLAMA_BASE_URL=http://ollama:11434

      # Optional secret key for Open WebUI.
      # - If **unset/empty**, Open WebUI will auto-generate a key on first start.
      # - If **set**, all session cookies and tokens are signed with this value.
      #   Recommended for multi-user and production deployments.
      # Best practice: set WEBUI_SECRET_KEY via your .env file or environment,
      # not hard-coded here, e.g. WEBUI_SECRET_KEY=$(openssl rand -hex 32)
      - WEBUI_SECRET_KEY=${WEBUI_SECRET_KEY:-}

    volumes:
      # Application data (users, chat history, uploads, settings, etc.).
      # Expect this to grow over time, especially if used for contract and
      # document analysis (easily 2-3 GB per year).
      #
      # We attach a *named* volume here, but map that named volume to a
      # specific host path on a backup-managed storage pool (TrueNAS/Synology).
      # See the `volumes:` section below for the bind configuration.
      - open-webui-data:/app/backend/data

    extra_hosts:
      - host.docker.internal:host-gateway   # Lets the container reach services on the Docker host

    restart: unless-stopped                 # Auto-restart on crash or host reboot

volumes:
  open-webui-data:
    driver: local
    driver_opts:
      # Bind-mount this named volume to a specific host directory instead of
      # using Docker's default volume location. This directory should live on
      # storage that is:
      #   - regularly backed up (e.g., TrueNAS/Synology dataset or share)
      #   - easy to grow/migrate as data usage increases.
      #
      # Example: /mnt/truenas/open-webui-data is a ZFS dataset exported to the
      # Docker host, or a mounted NFS/SMB share from a NAS.
      type: none
      o: bind
      device: /mnt/truenas/open-webui-data
      # To move or expand storage later:
      #   1. Stop the stack: docker compose down
      #   2. Copy data to a new path on your NAS/TrueNAS/Synology
      #   3. Update `device:` to point to the new path
      #   4. Start the stack again: docker compose up -d
</syntaxhighlight>

[[Category:Infrastructure]]
[[Category:Docker]]
[[Category:AI/ML]]
[[Category:Comfac]]