* fix(install): generate /etc/machine-id on systems without /proc
The install script read the fingerprint UUID from
/proc/sys/kernel/random/uuid, which FreeBSD does not have. On pfSense
the redirect still created /etc/machine-id, cat failed, and the agent
was left with an empty machine-id file. The [ ! -f ] guard then skipped
regeneration on every later run.
Fall back to uuidgen when /proc is unavailable, mirroring how the
script already picks between sha256sum and FreeBSD's sha256, and leave
no file behind when neither source exists.
* fix(install): regenerate empty machine id
The final message always echoed $PORT, which falls back to the default when -p
is not passed. Existing service files are kept as they are, so a plain upgrade
on a host with a custom port reported that the agent runs on 45876 regardless
of the actual configuration.
Read the address from the active service file instead. LISTEN is checked before
PORT to match the agent's own precedence in GetAddress, and the value is read as
text so host:port and unix socket paths are reported as configured.
- Keep chart documentation aligned with the application version
- Update app versions automatically after application releases
- Restore the main application release as latest after chart publishing
- Serialize release runs to prevent GitHub Pages index races
- Publish drafts using the release ID from Release Please
- Keep Helm chart releases out of the latest release channel
- Publish releases before Chart Releaser indexes them
- Match Chart Releaser to v-prefixed release tags
- Create draft releases before uploading chart packages
- Update the chart index before making releases immutable
- Keep Helm releases out of the latest release channel
- Serialize chart publishing to avoid index races
- Run the workflow only when Helm-related files change
Add curl retries/timeouts, archive integrity checks, binary existence
checks, and temp dir cleanup on all failure paths. Unify --mirror flag
handling in hub script to match agent. Use cat instead of tee for
systemd service file, quiet systemctl output.
* feat: add outbound heartbeat monitoring to external endpoints
Allow Beszel hub to periodically ping an external monitoring service
(e.g. BetterStack, Uptime Kuma, Healthchecks.io) with system status
summaries, enabling monitoring without exposing Beszel to the internet.
Configuration via environment variables:
- BESZEL_HUB_HEARTBEAT_URL: endpoint to ping (required to enable)
- BESZEL_HUB_HEARTBEAT_INTERVAL: seconds between pings (default: 60)
- BESZEL_HUB_HEARTBEAT_METHOD: HTTP method - POST/GET/HEAD (default: POST)
Allows users to explicitly choose Scoop or WinGet for installation
instead of relying on auto-detection. Useful when both package
managers are installed but the user prefers one over the other.
On Alpine Linux, the correct command to add a user to an existing group is addgroup <username> <groupname> rather than usermod -aG. The usermod command is part of the shadow package which is not installed by default on Alpine.
henrygd
Aditya Raj SinghandGitHub
MartinandGitHub
hank
Daniel NikoloskiandGitHub
mhlas7andGitHub
Anikesh ThakurandGitHub
Iriome SantanaandGitHub
Amir MoradiandGitHub
Loïc TosserandGitHub