Containers Panel¶

The Containers panel (press 1) shows all Docker containers — running and stopped. Press a to toggle between showing all containers and only running/paused containers.

List Columns¶

Each container row displays:

• State icon (color-coded: green for running, red for stopped, yellow for other)
• Container name
• Port mappings (first exposed port)
• Health status badge (green=healthy, red=unhealthy, yellow=starting) when the container has a health check configured
• Uptime / exit status

Detail Tabs¶

Select a container and press Enter or Tab to view its detail tabs:

Logs¶

Live-streamed container logs with token-level syntax highlighting. Individual tokens — HTTP methods, status codes, URLs, IPs, timestamps, JSON keys, state keywords, and file paths — are each colored distinctly, giving you immediate visual parsing of log output. The log buffer holds the most recent 1000 lines.

Logs are selection-driven — streaming starts when you select a container and stops when you navigate away.

Follow pause — scrolling up in the Logs tab (keyboard or mouse wheel) pauses auto-follow so you can read scrollback; a ⏸ follow paused — G to resume indicator appears in the detail tab bar. Press G or scroll back to the bottom to resume following.

Severity badges appear in the Logs tab header showing running counts of ERROR, WARN, INFO, and DEBUG lines, each colored by severity level.

Log filter — press f to open the filter overlay. Type a search query and press Enter to filter. Two modes are available (toggle with Tab):

• Exact — case-insensitive substring match
• Fuzzy — all words must appear in the line (AND logic), in any order

Matched text is highlighted and a match count is displayed. Press Esc to clear the filter.

Log compare — press m on a container to pin it as a compare target. When a compare target is pinned and you're viewing the Logs tab, both log streams are displayed side by side in left/right columns. Each column scrolls independently (j/k for primary, Shift+J/Shift+K for secondary). Press m again on the pinned item to unpin. Each panel remembers its compare target independently.

Stats¶

Real-time CPU and memory usage rendered as sparkline charts. The stats collector maintains a 60-sample ring buffer, giving you a rolling view of resource usage.

Network I/O — cumulative RX/TX bytes with rate sparklines (bytes/sec) showing transfer activity over time.

Block I/O — cumulative read/write bytes with rate sparklines showing disk activity.

PIDs — current process count inside the container.

A log severity sparkline appears below the resource charts, showing the distribution of log severity over time. Each position is colored by the dominant severity in that time bucket (red for errors, yellow for warnings, blue for info).

Stats are also selection-driven to avoid unnecessary API overhead.

Env¶

Environment variables set in the container, displayed as key-value pairs.

Config¶

Container configuration details, grouped into sections:

• Identity — ID, name, image
• Status — state, status, health badge, creation time; for an exited or dead container, the exit code with OOM killed and any daemon error; for a container with a healthcheck, the failing streak and the last probe's exit code and output (usually the actual reason it is unhealthy)
• Network — published ports
• Runtime — restart policy (e.g. on-failure (max 5)) and how many times the container has restarted
• Mounts — each volume or bind mount as source → destination (type, ro)
• Compose — project and service, for Compose-managed containers

The Env and Config tabs come from a single docker inspect call, refreshed when the container starts, dies or changes health.

Labels¶

Every label set on the container, sorted by key. Labels are how compose projects, orchestrators and image build metadata identify themselves.

Files¶

Shows all filesystem changes made inside the container compared to its base image. Each change is marked with a color-coded indicator:

• A (green) — file added
• C (yellow) — file changed
• D (red) — file deleted

Uses Docker's container.changes() API, which works on both running and stopped containers. Changes are fetched once on selection and cached.

Patterns¶

Groups similar log lines into templates by replacing variable parts (IDs, IPs, timestamps) with <*> wildcards. Templates are ranked by frequency, helping you quickly identify the most common log patterns and spot anomalies.

For example, three lines like User alice logged in from 192.168.1.1, User bob logged in from 10.0.0.1, and User charlie logged in from 172.16.0.1 would be grouped into: User <*> logged in from <*> (3 occurrences).

Sorting¶

Press o to open the sort overlay. Sort containers by:

• State (default) — running first, then stopped
• Name — alphabetical
• CPU % — highest CPU usage first
• Memory % — highest memory usage first
• Network I/O — highest total network bytes first
• Block I/O — highest total disk bytes first
• PIDs — highest process count first

Use j/k to navigate, Enter to apply, R to reverse sort direction. You can also press R globally to toggle reverse sort. The sort menu and the toast describe the current order in words — "A→Z" / "Z→A" for names, "running first" / "stopped first" for state, "highest first" / "lowest first" for stats.

Actions¶

Press x to open the context menu, or press an action's key directly. Only actions that apply to the container's current state are offered:

Open in Browser¶

b is offered for running containers with at least one published TCP port. It opens http://<host>:<port> (https for ports 443/8443) with the platform opener — xdg-open, open, or rundll32 on Windows — without going through a shell. The host is localhost for a local socket, and the daemon's host when the endpoint is tcp:// or ssh://. When several ports are published, the first is opened and the toast lists the others.

Interactive Exec¶

The exec action opens a shell inside the container, /bin/sh by default. Choose another with --shell <path> or the SIDEKICK_DOCKER_SHELL environment variable (see CLI Commands). If the shell does not exist in the image — common for distroless and scratch images — a toast says so instead of the session silently closing.

How the session runs depends on whether the optional node-pty native module can be loaded:

• Without node-pty (the default for npm install -g sidekick-docker, which does not ship it) — the dashboard suspends its rendering and hands the terminal to docker exec -it. End the shell with Ctrl+D / exit to return to the same selection and view state.
• With node-pty (e.g. a source checkout, where it is a dev dependency) — the shell runs embedded in a dashboard overlay. Press Ctrl+] to detach back to the dashboard, or end the shell with Ctrl+D / exit.