-
Morphit v1.11.3
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 50smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 40smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 42smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 17smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 32m48smorphit-release / Build + publish release tarball (push) Successful in 36m29sreleased this
2026-08-13 06:09:04 +00:00 | 25 commits to main since this releaseMorphit v1.11.3
Theme: a calmer, clearer install. The setup wizard reads like a smooth, successful setup — even with the network unplugged — and it can now turn on alerts for you. Plus: you can switch a node between Tor-only and clearnet whenever you like.
Added
Set up alerts right from the install wizard. The guided install now asks — right after the Matrix-contact question — whether you'd like your node to message you on Matrix when something needs attention (low disk, a backup that didn't run, a service down, a TLS certificate nearing expiry). Paste a bot account's access token and the personal
@you:serveraddress to alert, and alerting comes on by default: the Matrix bot plus the disk, SMART, package, and service monitors (and certificate-expiry monitoring on a clearnet node) all start together. Don't have a bot token handy? Press Enter to skip and turn it on any time later withmorphit-ops matrix. Alerts only ever go to a private@user:server, never a#room.A documented way to switch a node between Tor-only and clearnet. A node that started Tor-only can gain a clearnet domain later, and a clearnet node can drop back to Tor-only — either direction, as often as you like. Re-run the guided install and pick the other mode: your database, keys, and existing
.onionaddress are all preserved, the clearnet web front and certificate are added or dropped automatically, and your on-chain registration re-publishes with the new address. See the new "Switching between Tor-only and clearnet" section inRUN-A-MORPHIT-NODE.md.Fixed
The setup wizard no longer looks alarming. On a normal install — and especially an offline or air-gapped one — the wizard used to surface a lot of noise that read like failures even though the install succeeded. That's cleaned up across the board:
- Benign system messages that read like errors ("Permission denied" while setting up temporary directories, an unsandboxed-apt notice, a harmless collection-download message) no longer appear.
- The check that confirms your site serves fresh updates no longer prints a wall of "retrying" lines followed by a warning; it reports a single calm status line, and is skipped with a friendly note when the box is offline.
- A successful container build is no longer labelled as a "warning."
- The warrant-canary and privacy-address (Tor/I2P) steps explain themselves plainly instead of using alarming technical wording — and on an offline box, the canary clearly says it will publish itself automatically once the box is online, rather than looking like a failure.
- The closing "no alerting" notice is now a short, calm, optional suggestion — and doesn't appear at all when you set up alerts during the install.
- Assorted wording that leaned on the word "fail" has been softened where nothing had actually failed.
/v1/healthnow agrees withmorphit-ops healthabout the relay. On a containerized deployment the relay binds to the Docker bridge gateway (so the web container can reach it), not to loopback — and the indexer's public/v1/healthwas probing loopback only, so it reported the relay down even while the local health view (which also tries the bridge gateway) showed it up. The public health check now tries the same addresses, so the two agree.A crash-looping service no longer looks like it's just "starting."
morphit-ops healthread only a unit's top-level state, so a service that was failing and being auto-restarted (for example, a Matrix bot with a bad token or an unreachable homeserver) showed as a calm "starting" indefinitely. It now recognises the restart loop and reports it as failed, with a pointer tojournalctl -u <unit>so you can see why.Notes
- No database migrations. No breaking changes.
- Switching a node's mode, and setting up alerts, are both non-destructive and fully reversible.
Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads