* trek: update install and upgrade workflow
Remove the hardcoded release tag and fetch the latest TREK release. Consolidate builds to npm workspaces (shared, client, server), copy client dist into server/public, and install/prune server deps for production. Add libkitinerary dependency and ensure dependencies during upgrade. Implement backup/restore during upgrades and add migration logic that detects the old systemd ExecStart; when present, rewrite trek.service to run the compiled dist via node --require tsconfig-paths/register, add XDG_CACHE_HOME and QT_QPA_PLATFORM env vars, and switch to network-online.target. Installer changes also create data/uploads directories, set additional .env defaults (TZ, LOG_LEVEL, DEFAULT_LANGUAGE, ALLOWED_ORIGINS), and adjust workspace setup steps.
* add proxy env
Enable Corepack and run Storyteller install/build commands through corepack yarn so the script honors upstream packageManager and avoids Yarn 1 failures.
Use Corepack-managed Yarn when available so installs respect upstream packageManager metadata and avoid Yarn classic mismatch failures during both fresh installs and updates.
* Add runtime script status guard and deleted-script stubs.
Prevent disabled/deleted scripts from running updates with clear user messages, and ensure deleted ct scripts are stubbed automatically so legacy update commands no longer fail with 404.
* Delete ct/ente.sh
* Update booklore.sh
* Add twenty (ct)
* Apply suggestion from @tremor021
* Refactor backup and restore process in twenty.sh
Refactored backup and restore logic to use functions.
* Clean up blank lines in twenty.sh
Removed unnecessary blank lines in the script.
---------
Co-authored-by: push-app-to-main[bot] <203845782+push-app-to-main[bot]@users.noreply.github.com>
Co-authored-by: Slaviša Arežina <58952836+tremor021@users.noreply.github.com>
Co-authored-by: Michel Roegl-Brunner <73236783+michelroegl-brunner@users.noreply.github.com>
Co-authored-by: CanbiZ (MickLesk) <47820557+MickLesk@users.noreply.github.com>
* homelable: preserve MCP server config across updates
The update path runs CLEAN_INSTALL=1 fetch_and_deploy_gh_release, which
wipes /opt/homelable before redeploying. The backup/restore only covers
backend/.env and data/, so an optionally-installed MCP server (set up via
Pouzor/homelable's own scripts/lxc-mcp-install.sh, which targets exactly
this LXC and lives in /opt/homelable/mcp) loses its .env and .venv on
every update. The homelable-mcp service then keeps running on deleted
inodes and dies at the next restart.
Back up mcp/.env when present, and after the deploy restore it, rebuild
the venv (same uv pattern as the backend), restore ownership and restart
the service. Fully conditional: installs without the MCP are unaffected.
* homelable: remove comments per maintainer review
The update path installed into the global interpreter with a global
--ignore-installed flag, which leaves the previous *.dist-info behind
on every dependency bump. Duplicate metadata makes pip resolve against
stale requirements (e.g. downgrading pydantic-core) and the service
crashes at the next restart. Fixes the typing_extensions workaround
(#13548) at the root: in a venv there are no Debian-owned packages to
conflict with, so neither --ignore-installed nor --break-system-packages
is needed.
Follows the existing setup_uv + venv-or-migrate pattern from
prometheus-pve-exporter and esphome. Existing installs are migrated
automatically on the next update; the systemd unit is repointed to the
venv binary.
Fixes#14987
Tom Frenzel