Seed a default admin account and harden the installer; update symlink behavior and LXC hook argument handling.
- Install: downgrade Node setup to 22, write ADMIN_EMAIL/ADMIN_PASSWORD into /opt/trek/server/.env for initial boot, chmod the file, wait for app health, then remove plaintext creds from the env and print the default admin credentials. Remove previous DB patching script and credentials file generation. Add health-check failure handling.
- ct/trek.sh: check for /opt/trek instead of ~/.trek, run npm ci without --production, and recreate server data/uploads by removing any existing dirs and creating explicit symlinks.
- Installer: mirror symlink strategy used in the container (rm then ln -s) and ensure generated ENCRYPTION_KEY note; add ADMIN_EMAIL default.
- json: set default username to admin@trek.local, update notes about seeded admin, ENCRYPTION_KEY storage, and APP_URL recommendation.
- tools/pve/lxc-prehook.sh: fix append_unique_line_in_ct to pass positional arguments into the bash -c snippet safely (avoid parent-shell expansion).
These changes ensure a reproducible default admin creation flow without leaving plaintext credentials, improve symlink handling, and fix a bug in the LXC prehook.
Change Foldergram installation to store media and config under /opt/foldergram_media (create media dir, write env file there, update systemd EnvironmentFile). Update ct/foldergram.sh to remove temporary backup/restore during updates. Update json/foldergram.json config_path and add an info note about where to place media. Add tools/pve/ct-batch-create.sh — a new Proxmox CT batch-creator script (interactive/unattended modes, caching, storage selection) to download and deploy multiple community-scripts CTs.
Add optional Shlink Web Client installation and update workflow, switch init to use shlink-installer, and expose the web UI on port 3000. Changes include: use vendor/bin/shlink-installer init (sourcing /opt/shlink/.env) instead of direct db:migrate/create; create data directories (cache, locks, logs, proxies, temp-geolite); add interactive prompt to install the web client, fetch and deploy shlink-web-client GitHub releases, generate /opt/shlink-web-client/servers.json with the initial API key, and configure an nginx site serving the client on port 3000. The ct update script now also checks for and updates the web client release. Update metadata (json/shlink.json) to set interface_port to 3000 and add an info note about API (8080) vs Web Client (3000). Also print the 3000 URL on completion.
Updated the PR script to handle cases where the application requirements are missing or unchecked more effectively. Added checks for the 'keep-open' label and improved messaging for maintainers.
- Replace custom git clone + version tracking with fetch_and_deploy_gh_release tarball mode
- Replace manual version comparison with check_for_gh_release in update script
- Remove git from dependencies (no longer needed for tarball download)
- Version file ~/.clickstack now managed by tools.func
- Switch tag format from hyperdx@* to @hyperdx/app@* (v2.x releases)
- Remove @types/hyperdx__lucene workaround (fixed in v2.x)
- Fix API build path: dist/index.js -> build/index.js
- Use next start instead of standalone server for app
- Add IS_LOCAL_APP_MODE for auth-free local deployment
- Add DEFAULT_SOURCES/DEFAULT_CONNECTIONS for auto-provisioning
- Update script: fresh clone approach for major version changes
- Add corepack setup to update script for yarn version changes
Yarn Berry caches resolution descriptors in yarn.lock in a multi-line
block format that sed line-deletion cannot properly handle. The
@types/hyperdx__lucene -> npm:@types/lucene alias persists across all
partial cleanup attempts. Deleting yarn.lock forces a fresh resolution
from package.json only, which is safe for a build-from-source install.
- Use sed to remove all lines referencing the alias from yarn.lock
(both block headers and inline workspace dependency refs)
- Also strip @types/lucene resolution target (package doesn't exist)
- Remove .yarn/install-state.gz to clear Yarn Berry cached state
- Keep node -e for package.json cleanup (devDeps + resolutions)
Removing from packages/app/package.json devDependencies alone is not
enough — Yarn Berry still resolves the alias from yarn.lock entries.
sed removes all yarn.lock lines referencing the broken package.
- Event store migrations use PostgreSQL-specific SQL (SERIAL, UUID) that
fails on SQLite — replaced with doctrine:schema:update --force
- Run messenger:setup-transports BEFORE event_store schema to avoid
messenger_messages table conflict
- Read model migrations (--em=read_model) include role seeding (ROLE_USER,
ROLE_ADMIN) which doctrine:schema:create skips
- Fix default_credentials in json to match script (admin@slink.local)
