Your infrastructure. Your database. Your network.

The free tier is a complete product. It is sized for one person.

One operator, one org, one box, three days of history. That is a real tool and it stays free. It stops being enough the moment a second person needs their own login, or a second team needs their own org, or an incident review asks what happened last month.

Telemetry — metrics, logs and HEP — stays in SQLite on the container's volume in both tiers, and rebuilds itself from your controllers if you lose it. Only the control plane (orgs, people, servers, encrypted tokens) moves to Postgres, because that is the part you cannot re-derive.

Written for the person who has to sign off on it.

Buying an observability tool usually means handing a vendor a path into production. This one is the opposite shape, and the reasons are structural rather than promises — they are consequences of how the licence and the transport are built.

One thing this page will not claim: traffic to your controllers on port 8687 is plain HTTP. The credential is protected by the request signature; the response bodies are not. That is fine inside a private network and is not fine across the public internet — put the plane behind your VPN or mesh, the same as you would any other admin API.

Paste it in. Nothing restarts.

You already have the installation. Enterprise is a token you paste into it — there is no separate build, no migration, and no window where the dashboard is down. The screen below is the free tier: the licence goes in the box on the left, and the panel on the right disappears the moment it verifies.

Pasting is the fastest route and the wrong one for a fleet you rebuild from scratch. For that, ship the licence as configuration — below.

Or ship it as configuration.

A licence supplied by environment is read at boot, so a container that is destroyed and recreated comes back licensed with no manual step. Whichever source is newer wins, so a token pasted into the UI later still takes precedence over a stale one in your environment.

docker run -d --name voodu-webui \
  -p 3000:3000 \
  -v voodu:/rails/storage \
  -e VOODU_LICENSE="eyJhbGciOiJSUzI1NiIsInR5cCI6IkpXVCJ9..." \
  ghcr.io/thadeu/voodu-webui

docker run -d --name voodu-webui \
  -p 3000:3000 \
  -v voodu:/rails/storage \
  -v /etc/voodu/license.jws:/run/secrets/voodu-license:ro \
  -e VOODU_LICENSE_FILE=/run/secrets/voodu-license \
  ghcr.io/thadeu/voodu-webui

The compose file below is the shape most Enterprise installations run: the licence as a file secret, and the control plane in Postgres. Telemetry stays on the volume in both cases.

services:
  webui:
    image: ghcr.io/thadeu/voodu-webui
    restart: unless-stopped
    ports:
      - "3000:3000"
    volumes:
      - voodu:/rails/storage
      - ./license.jws:/run/secrets/voodu-license:ro
    environment:
      VOODU_LICENSE_FILE: /run/secrets/voodu-license

      # Control plane in your own Postgres. Omit it and everything
      # runs on SQLite in the volume — still a supported Enterprise
      # deployment, just one you back up differently.
      DATABASE_URL: postgres://voodu:CHANGE_ME@postgres:5432/voodu

      # How long telemetry is KEPT. The licence caps how far back the
      # UI will search; this caps what is on disk at all.
      VOODU_RETENTION_DAYS: "90"
    depends_on:
      postgres:
        condition: service_healthy

  postgres:
    image: postgres:18-alpine
    restart: unless-stopped
    environment:
      POSTGRES_USER: voodu
      POSTGRES_PASSWORD: CHANGE_ME
      POSTGRES_DB: voodu
    volumes:
      - pgdata:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U voodu"]
      interval: 5s
      timeout: 3s
      retries: 10

volumes:
  voodu:
  pgdata:

There is no secret key to generate: the container creates and persists its own on first boot, in the volume. Migrations run on boot and are idempotent, so upgrading is pull and up -d — keep the volume and nothing else is needed.

What happens when it runs out.