dsh-plugin-safe-upgrade

Configuration

Everything is auto-detected: the unit from the process cgroup, the install directory from the running dsh binary, the health URL from --host/--port, and the profiles from $DSH_HOME/profiles. To override, add a row to your profile's cordis.patch.yml. A patch replaces the whole config, so list every key you want to keep.

- id: safe-upgrade
  config:
    channel: latest              # dist-tag that /upgrade offers first
    failOnWarnings: false
    telegram:                    # optional; credentials are read at send time
      envFile: /etc/dsh/alerts.env
      tokenVar: TELEGRAM_BOT_TOKEN
      chatVar: TELEGRAM_CHAT_ID

Key	Default	Meaning
unit / scope	detected	systemd unit and system/user scope
installDir	detected	folder whose node_modules holds @deepseek-ai/dsh
profiles	all	profiles validated with --dump-config
healthUrl	from --port	any HTTP answer except 5xx counts as up (401 is normal)
healthTimeoutMs	150000	how long a new boot may take
bootTimeoutMs	120000	how long plugins may stay loading before the boot counts as unhealthy
idleWaitMs	300000	how long an upgrade or rollback waits for running turns before it gives up without stopping dsh
channel	latest	dist-tag offered first; tags behind it are hidden
checkIntervalHours	6	npm dist-tag check interval
keepPrev	2	previous install copies kept for rollback
keepGoodTags	10	good-* tags kept
snapshots	true	auto-commit config edits
failOnWarnings	false	treat warnings as failures (dry runs, upgrades, manual rollbacks)
uiCheck	auto	auto: roll back on dsh's failure screen, warn if the check can't run; true: also roll back when it can't run; false: skip
chromium	detected	path to a Chromium/Chrome binary for the UI check
uiTimeoutMs	45000	how long the UI may take to show the composer
telegram	off	{envFile?, tokenVar?, chatVar?}