Skip to content

web-ui.md

Latest commit

 

History

History
123 lines (93 loc) · 5.29 KB

web-ui.md

File metadata and controls

123 lines (93 loc) · 5.29 KB
Raw

Web UI

Docksentry includes an optional, lightweight web dashboard. Enable it with WEB_UI=true.

Web UI Status Web UI Logs Web UI Settings

Setup

environment:
  - WEB_UI=true
  - WEB_PASSWORD=your-secret    # optional, recommended
  - WEB_PORT=8080               # default
ports:
  - "9090:8080"

Access at http://your-server:9090 with the configured password (Basic Auth).

Pages

Status

Live overview of all running containers with:

  • Health badges (healthy / starting / running)
  • Pending update badges
  • Pinned container badges
  • Auto-update toggle switches per container
  • Pin/Unpin buttons per container
  • Update buttons for containers with available updates
  • ⚠ on/off toggle to require confirmation for SemVer major bumps (per container)
  • "Check Updates" button to trigger a manual scan
  • Bulk actions bar — multi-select containers via checkboxes, then apply one of: Update / Pin / Unpin / Auto-update on / Auto-update off in a single click
  • Major-update banner — when a major bump is held back by ⚠ on, a yellow banner at the top lists pending containers with Confirm / Skip buttons

With DOCKER_HOSTS set, the table also gets a Host column, a host dropdown next to the search box that narrows the list to one host, and every button acts on the host of the row it sits on — updating nginx on the NAS never touches the local nginx. Deferred major bumps are listed per host too. A host that can't be reached shows up as a single line instead of taking the page down. The per-container detail page (click a container's name) is local-only, so remote rows aren't linked to it. Without DOCKER_HOSTS none of this renders and the page is exactly what it always was.

Logs

View container logs directly in the browser:

  • Dropdown to select any running container
  • Configurable number of lines (10-500)

History

Full update log showing:

  • Timestamp
  • Container name
  • Success/failure status
  • Detail message

Settings

The Settings page is grouped into five tabs:

Tab Contents
General Language, cron schedule, excluded containers, debug mode
Updates Auto self-update toggle, hint about per-container settings on the Status page
Cleanup Auto cleanup, grace hours, backup retention, local-only-images backup
Notifications Disk warning threshold + auto-cleanup-on-warning, quiet hours start/end
Channels Telegram topic ID, Discord webhook, generic webhook

Plus two cards always visible below the tabs:

  • Update Windows (per-container HH:MM ranges + weekdays)
  • Maintenance (one-shot Image Cleanup / Self-Update buttons, both with confirmation dialogs)
  • Info (version, Telegram status, masked credentials)

Hover a ? icon next to any setting label for an inline explanation. Save feedback appears as a brief toast at the top-right.

All settings persist across restarts (saved to /data/settings.json).

Setting Editable in Web UI
Language Yes
Cron Schedule Yes
Debug Mode Yes
Auto Self-Update Yes
Auto Cleanup + grace hours + backup-local-only + backup retention Yes
Disk warning threshold + auto-cleanup-on-warning Yes
Quiet hours (HH:MM start/end) Yes
Update Windows per container (HH:MM range + weekdays) Yes (own section)
Exclude Containers Yes
Telegram Topic ID Yes
Discord Webhook Yes
Webhook URL Yes
Bot Token No (ENV only, masked)
Chat ID No (ENV only, masked)
WEB_PORT, WEB_PASSWORD No (ENV only — would lock you out)

The Update Windows section lets you pick a container, set a HH:MM–HH:MM range, and tick which weekdays the window applies to. Containers without an entry update without restriction.

The Maintenance section provides one-click buttons for Image Cleanup and Self-Update — same actions as Telegram /cleanup and /selfupdate, available headlessly.

Machine-readable endpoints

Two read-only endpoints, guarded by API_TOKENS (see security.md):

  • GET /metrics — Prometheus text format. Reports containers watched, updates pending, one gauge per host, and one per container that has an update — so an alert can name the thing that needs attention rather than only a count.
  • GET /api/status — the same numbers as JSON.

Both read what the status page reads. Neither triggers a check: an endpoint that did would let anyone with a scrape interval drive the update loop.

On a phone

Below 700px the container table is replaced by one card per container — same data, laid out to fit a screen you hold in one hand, with the actions sized for a finger. Above that width the table returns. Both are built from the same values, so they cannot disagree about a container's state.

Security

  • Password protection via Basic Auth (WEB_PASSWORD)
  • Password hashed with SHA-256, never stored in plain text
  • Sensitive values (Bot Token, Chat ID) are masked in the UI
  • For HTTPS, use a reverse proxy — see Security