-
Morphit v1.10.5
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 41smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 17smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 34m7smorphit-release / Build + publish release tarball (push) Successful in 37m34sreleased this
2026-08-08 05:51:01 +00:00 | 86 commits to main since this releaseMorphit v1.10.5
Theme: guided (ansible) installs now advertise the correct relay account, so a new instance is accepted by its peers on the first probe instead of being rejected.
This is a maintenance release. There are no database migrations and no breaking changes.
Fixed
Ansible-installed instances now report their own relay account. Every instance publishes a relay account in
/v1/instance, and peers refuse to trust an instance whose published relay account doesn't match the account that signed its on-chain registration (an anti-impersonation check). The guided ansible installer set the operator tag and account name but never wrote the indexer's relay-account setting, so it silently fell back to the canonical default — meaning a fresh ansible instance advertised the wrong account and every other node rejected its health probe with a "relay account mismatch." The installer now writes the correct value (the same account the relay signs with), matching what the manualmorphit-opswizard already did. A guard was added so no install writer can omit it again.Notes
- No database migrations. No breaking changes.
- Only the guided (ansible) install path was affected; manual
morphit-opsinstalls already set this correctly. - An existing ansible instance can fix this without upgrading by setting
MORPHIT_INDEXER_RELAY_ACCOUNTto its own account and restarting the indexer.
Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads
-
Morphit v1.10.4
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 49smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 39smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 43smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 33m41smorphit-release / Build + publish release tarball (push) Successful in 40m0sreleased this
2026-08-08 03:41:05 +00:00 | 87 commits to main since this releaseMorphit v1.10.4
Theme: instances can finally see each other. This release fixes a bug that stopped every instance from health-probing its peers, so the federation directory shows real, live status for other operators.
This is a maintenance release. There are no database migrations and no breaking changes.
Fixed
Peer health probes now work. The indexer probes other instances over their public address and, as an anti-DNS-rebinding measure, pins the resolved IP for the connection. The code that pinned the IP used an older callback style that the current HTTP library rejects, so the connection failed immediately with an internal "invalid IP address" error — on every peer. The result: other operators' cards in the directory were stuck on "Unreachable" and never picked up their name, tagline, contact link, or Tor/I2P addresses, even when the instance was perfectly reachable. An instance's own card was unaffected because it fills in from local configuration, which is why this stayed hidden until a second instance joined. The IP-pinning now uses the shape the library expects, so probes connect and the directory reflects real peer status. A regression test covers the exact callback contract.
Notes
- No database migrations. No breaking changes.
- After upgrading, previously "Unreachable" peers will flip to their real status (good / syncing / quiet) on the next probe.
Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads
-
Morphit v1.10.3
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 47smorphit-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 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 33m38smorphit-release / Build + publish release tarball (push) Successful in 37m5sreleased this
2026-08-07 22:34:00 +00:00 | 88 commits to main since this releaseMorphit v1.10.3
Theme: smoother upgrades, and a federation that welcomes regional and right-to-left instances. This release fixes a false upgrade error, lets instances brand themselves with the project name, and makes every text field work correctly in Persian, Arabic, and other RTL languages.
This is a maintenance release. There are no database migrations and no breaking changes.
Fixed
Upgrades no longer fail with a false integrity error. Since v1.10.1 every release ships two download files — the small standard tarball and a larger self-contained "offline" bundle — each with its own checksum. The upgrader picked whichever tarball it found first and, separately, whichever checksum it found first, with no guarantee they were the same file. So an online upgrade could fetch the tarball from one place and its checksum from another, decide they didn't match, and stop with a "SHA-256 mismatch" even though nothing was wrong. The upgrader now always uses the small standard tarball for an online upgrade and matches its checksum and signature to that exact file; the offline bundle is used only for genuine offline installs where you supply the file yourself. The integrity check is exactly as strict as before — it was simply comparing the wrong pair.
Instances can brand themselves with the project name. Registering an instance whose display name contained "Morphit" (for example "Morphit Latino") was being rejected as impersonation, which blocked the whole point of running a regional instance. The rule now allows the brand as part of a longer, distinct name while still blocking bare-handle impersonation ("morphit", "@morphit"), look-alike/homograph attacks, and the reserved infrastructure handles ("morphit-fees" and friends). The rightful owner of a reserved name is also exempt for that name, matching how user profiles already worked.
Right-to-left languages now work everywhere text is entered or shown. The half-space (ZWNJ) that Persian and other scripts rely on was being rejected as a forbidden character across instance names, order titles and terms, feedback, and payment-method labels — so a Farsi name or description couldn't be saved at all. That character (and the related joiner) is now accepted everywhere, while the genuinely dangerous zero-width space and bidirectional-override controls stay blocked. On top of that, every field where a user types text and every place user text is displayed now renders right-to-left correctly. Editing an instance's title and description in Farsi through
morphit-opsworks too.Notes
- No database migrations. No breaking changes.
- If a v1.10.1 or v1.10.2 upgrade stopped on a "SHA-256 mismatch," this release resolves it.
- An instance rejected earlier for a "Morphit …" name (or a right-to-left name) can register once the instances processing it are on v1.10.3.
Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads
-
Morphit v1.10.2
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 47smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 38smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 40smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 32m28smorphit-release / Build + publish release tarball (push) Successful in 36m14sreleased this
2026-08-07 07:38:28 +00:00 | 90 commits to main since this releaseMorphit v1.10.2
Theme: a reboot shouldn't take your node down. This release fixes a database-permission defect that could stop the indexer from coming back after a restart, and repairs the gitea.com release mirror.
This is a small but important maintenance release for anyone running a Morphit instance installed the guided (ansible) way. There are no database migrations and no breaking changes.
Fixed
The indexer now always comes back after a restart or reboot. On instances set up with the guided installer, the database's host-permission file (
pg_hba.conf) was left without a rule allowing the indexer and relay to log in over the local network connection they actually use. Everything ran fine until PostgreSQL next re-read that file — typically the first reboot — after which the indexer would fail to connect and restart in a loop, taking/v1/health(and anything that depends on the indexer) down with it. The installer now writes the correct, tightly-scoped permission rule (only the indexer's own database user, only over the loopback address, password-authenticated), so a restart or power cut no longer locks the indexer out. A regression guard was added so this can't silently come back.If you hit this before upgrading, the symptom in the logs was
no pg_hba.conf entry for host "127.0.0.1". After upgrading to v1.10.2 the installer repairs the rule for you; on a node you've already patched by hand, the upgrade simply confirms the rule is present.The gitea.com release mirror now publishes. Forgejo reserves secret names beginning with
GITEA_, so the token for the gitea.com mirror can't be stored under the name the release workflow originally expected. The workflow now reads the token from a name Forgejo allows (GITEACOM_TOKEN), so — with that secret set — each new release is mirrored to gitea.com alongside codeberg.org, keeping upgrades working even if the main forge is unreachable.Notes
- No database migrations. No breaking changes.
- Instances installed the manual (
morphit-ops) way were not affected by the database-permission defect; it was specific to the guided/ansible database setup.
Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads
-
Morphit v1.10.1
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 41smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 42smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 34m26smorphit-release / Build + publish release tarball (push) Successful in 36m8sreleased this
2026-08-07 04:23:29 +00:00 | 91 commits to main since this releaseMorphit v1.10.1
Theme: a node you can look after with less effort. This release is a cluster of practical improvements for the people who run Morphit instances — clearer status while a fresh node catches up, at-a-glance proof that your node is helping host the release, one command that hands your whole health picture to a monitor, and upgrades that keep working when the internet doesn't. Underneath, a brand-new node now writes its catch-up to the database in safer, bounded batches.
This is a maintenance release. There are no database migrations and no breaking changes; an ordinary online instance behaves exactly as before.
Added
See whether your node is seeding the release. Every instance runs a small IPFS node that pins the signed release and rebroadcasts the
ipns://record so releases stay hosted and resolvable as long as any instance is alive. The node-health view now shows an IPFS/IPNS release seeding line —ok,degraded,down, ornot-configured— so you can tell at a glance that your box is doing its share, without digging through logs.One command for monitoring.
morphit-ops health --jsonnow prints the whole node-health view as one machine-readable object — indexer, relay, system (CPU/memory/disk), background services, backups, the warrant canary, and the new IPFS/IPNS seeding state. Point Zabbix (or any monitor) at it — run it on a timer into a file, or through an agent — and alert on any section leaving its healthy state. Host-level operational detail like this stays out of the public health endpoint on purpose, so a passing stranger can't learn that your backups are failing or your disk is full.Upgrade with no internet, from a USB stick or a folder. Copy the signed
-offlinetarball (and its signature) onto the box and runmorphit-ops upgrade --from-file=…, or simply drop it into the offline release folder — the main menu then shows ● update available (offline tarball ready) on its own, and a normal upgrade uses it automatically whenever the network can't be reached. An upgrade that begins online even finishes from that dropped tarball if the connection drops partway through. Unsigned tarballs are refused.Set up your warrant canary during install. The guided home install now offers to set up your warrant canary right then, instead of leaving it as a separate follow-up step.
Clearer instance status. The public instances list now tells apart a node that is still catching up (Syncing) from one that has genuinely fallen behind (stale), and the status dropdown uses plain-language labels for each state.
Changed
Safer initial sync. When a fresh node replays the Blurt chain to build its local view, it now commits each window of blocks in one bounded database transaction rather than one transaction per block. If the machine crashes mid-catch-up, the partial window rolls back cleanly and the node resumes from exactly where it left off — with no half-written state and no risk of a long transaction stalling the database. The same data ends up stored; there is nothing to configure.
Releases mirrored automatically. Each release is now also published to the codeberg.org and gitea.com mirrors (when their tokens are configured), so upgrades keep finding a release even if the main forge is unreachable.
Fixed
IPNS rebroadcast on ansible-managed nodes. An instance installed through the ansible role now also rebroadcasts its IPNS record on a timer. Previously only the hand-run IPFS setup did this, so an ansible-managed node could quietly stop keeping the
ipns://name alive. The rebroadcast needs no key — it simply re-announces the release's own on-chain-signed record — and now runs everywhere.Notes
- No database migrations. No breaking changes.
- The complete node-health picture (including backups, disk, and IPFS seeding) is available through
morphit-ops health --json, locally or over SSH; it is deliberately not exposed on the public/v1/healthendpoint.
Downloads
-
Source code (ZIP)
1 download
-
Source code (TAR.GZ)
1 download
-
Morphit v1.10.0
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 47smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 39smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 39smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 14smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 31m34smorphit-release / Build + publish release tarball (push) Successful in 34m54sreleased this
2026-08-06 22:15:48 +00:00 | 92 commits to main since this releaseMorphit v1.10.0
Theme: Morphit now installs completely offline. From a single self-contained download, with the network cable unplugged, the guided install runs start to finish — and the few steps that genuinely need the internet (your real HTTPS certificate, connecting to Blurt, and, if you opt in, listing your instance on-chain) complete themselves automatically the first time the box sees a connection.
This finishes the offline appliance. An instance can be brought up in a place with no internet at all, on the local network, and it upgrades itself the moment a real link appears — retrying quietly until it succeeds, then stepping out of the way.
Added
Completely offline install. There are now two downloads on each release:
- the usual slim source tarball, which fetches its dependencies while installing (needs internet), and
- a self-contained
-offlinetarball, which bundles everything — the application libraries, the operating-system packages, the Docker images, the Node runtime, and Kubo. Installing from it touches the network for nothing:apt, Docker, and every other step install from the bundle. It is a large download by design.
To install with no internet: download the
-offlinetarball on any machine, copy it to the target, extract it, and run the guided install. When the box is later connected, normal package updates are automatically restored (the offline install temporarily pointsaptat the bundled packages using a reversible setting that leaves your system's own package sources untouched; the finish-when-online helper removes it once you're online).Finish-when-online. After the install, a background helper (
morphit-first-online) watches for a real internet connection — it actually checks that it can reach several Blurt nodes, not just that a cable is plugged in. The first time it succeeds it obtains your real Let's Encrypt certificate (until then the site is served over HTTPS on your local network with a self-signed certificate, so it still works), connects the indexer and relay to Blurt, and — if you opted in — publishes your on-chain instance listing. Each step is done once and remembered; anything that can't finish yet is retried on the next cycle; when everything is done the helper retires itself.Opt-in automatic on-chain listing. The guided install asks whether to list your instance on-chain automatically once online, or leave it for you to run
morphit-ops registeryourself later. The unattended path reuses the same relay key and passphrase your relay already runs with — no key is entered by hand.Fast initial sync. A brand-new instance replays the Blurt chain from the beginning to build its local view. That catch-up used to fetch one block-window at a time; it now fetches several windows at once, each aimed at a different one of your configured Blurt nodes — so no single node is leaned on, and a slow or failing node is stepped around automatically — while still writing every block in strict order, one per transaction. A fresh instance reaches the chain head dramatically faster, and exactly the same data ends up stored. Nothing to configure; a tuning knob (
MORPHIT_INDEXER_BACKFILL_CONCURRENCY) exists for the curious but defaults to one window per node.Changed
The HTTPS step never blocks the install. If Let's Encrypt can't be reached when you install (offline, or your domain isn't resolving yet), the certificate step is deferred to the finish-when-online helper instead of stopping the install. The site comes up immediately and upgrades to the real certificate automatically once it can.
Fixed
Guided-install robustness — from the first real federated deployment. Standing up the first independent Morphit instance surfaced a batch of first-install issues, all now fixed so a fresh install comes up cleanly end to end:
- The real HTTPS certificate is made readable by the proxy on both first issue and on renewal, so the site keeps serving over the renewed certificate without hand-holding.
- The reverse-proxy and firewall containers now start in the right order, with the file permissions and controlled Docker access they need, instead of racing each other on a cold boot.
- The single shared database is set up once and used by both the indexer and the relay (an earlier layout created a second, empty database the relay then couldn't use).
- A multi-word instance name — and your optional Tor and I2P addresses — are now captured and stored intact, so your instance's name and onion/I2P links show up correctly.
Notes
- No database migrations. No breaking changes. An ordinary online install behaves exactly as before — every offline mechanism is dormant unless a bundle is actually present.
- Building the
-offlinetarball (for release maintainers): runbash scripts/build-offline-bundle.shon an Ubuntu 24.04 machine with Docker; it assembles the bundle and writesmorphit-v1.10.0-offline.tar.gz. The release CI attempts this automatically on a best-effort basis, so a build machine without Docker never blocks a release. - If a past install stopped waiting on a certificate or a Blurt connection, nothing needs undoing — re-run the guided install with this version; it finishes offline and the helper takes care of the rest when you're online.
Downloads
-
Source code (ZIP)
1 download
-
Source code (TAR.GZ)
2 downloads
-
Morphit v1.9.22
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 57smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 49smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 46smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 17smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 32m37smorphit-release / Build + publish release tarball (push) Successful in 31m12sreleased this
2026-08-04 05:25:13 +00:00 | 110 commits to main since this releaseMorphit v1.9.22
Theme: the guided install now creates every drop-in configuration directory it writes into, so hardening can't stop on a "directory does not exist" error.
This is the companion hotfix to v1.9.21. That release fixed the SSH-hardening step for machines without an SSH server; this one fixes the same class of problem one step later. On some machines the password-policy step stopped with "/etc/security/pwquality.conf.d does not exist" — the password-quality package installs its main config file but doesn't create the drop-in directory, so writing into it failed.
Fixed
Every hardening drop-in now ensures its directory first. Instead of trusting each system package to have created its own
.ddrop-in directory (some do, some don't), the installer now creates the directory before writing into it — for the password policy, the audit rules, the AIDE integrity rules, and the SSH hardening. The result is the same hardened configuration; it simply no longer depends on a package having pre-made a directory.We also added an internal build check that fails if any future drop-in write forgets to ensure its directory, so this whole class of "directory does not exist" install stoppage can't come back.
Notes
- No database migrations. No breaking changes.
- If your last install stopped at the password-policy step, nothing needs undoing — re-run the guided install with this version and it will pick up cleanly and finish.
Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads
-
Morphit v1.9.21
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 46smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 37smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 39smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 29m50smorphit-release / Build + publish release tarball (push) Successful in 32m0sreleased this
2026-08-04 03:27:15 +00:00 | 111 commits to main since this releaseMorphit v1.9.21
Theme: home-desktop installs no longer stop at the SSH-hardening step when the machine has no SSH server.
This is a small hotfix for the guided install. If you install Morphit on a home computer that doesn't run an SSH server (common for a desktop you sit at and administer locally), the previous version stopped during setup with an error like "Destination directory /etc/ssh/sshd_config.d does not exist." Rented servers (VPS) always have an SSH server, so they were unaffected — but a home node could get stuck here.
Fixed
SSH hardening now applies only when an SSH server is actually installed. The installer checks for an SSH server first:
- If one is present (always the case on a VPS, and on any machine where you added SSH for remote access), it is hardened exactly as before — root login and password login are turned off.
- If there is none (a typical home desktop you run locally), the installer prints a short note explaining there is nothing to harden and moves on, instead of stopping.
We deliberately do not install an SSH server for you. Whether to run one — and expose it — is your decision, not something the marketplace installer should make on your behalf. If you add an SSH server later for remote access, just re-run the installer and it will harden it.
Two small belt-and-braces fixes ride along so the same class of problem can't surface elsewhere: the scheduled security scans (AIDE, rkhunter) are guaranteed their scheduling directories, and the certificate-renewal hook is guaranteed its directory before it's written.
Notes
- No database migrations. No breaking changes.
- If your last install stopped at the SSH step, you don't need to undo anything — just re-run the guided install with this version and it will pick up cleanly and finish.
Downloads
-
Source code (ZIP)
3 downloads
-
Source code (TAR.GZ)
2 downloads
-
Morphit v1.9.20
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 47smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 40smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 39smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 30m24smorphit-release / Build + publish release tarball (push) Successful in 31m23sreleased this
2026-08-03 22:50:48 +00:00 | 112 commits to main since this releaseMorphit v1.9.20
Theme: the end-of-install summary now checks and shows the health of every part of your node — not just "installed and running."
When the guided install finishes, it prints a summary of what came up. In v1.9.20 that summary is far more thorough: instead of a short list, it verifies and displays the status of every subsystem, so you can tell at a glance that your whole node is actually healthy before you announce it.
New
The install summary now covers everything, with a live status for each. After the install, you'll see a checked line for each of:
- Database, signup relay, and marketplace indexer — including a live "is it actually responding?" check against each service's health endpoint, and whether the indexer is caught up or still catching up (which is normal on a fresh node).
- Blurt RPC connectivity and the BTC/XMR price feeds (FX).
- Your verified relay balance, read live from the chain, shown as an amount and an estimate of how many signups it funds.
- The MCP read-only API server.
- The web firewall (BunkerWeb) and your website.
- HTTPS certificate.
- Firewall (UFW) and intrusion protection (fail2ban), now shown as two separate lines.
- Your Tor onion address — the actual
.onionis shown so you can copy it — and your I2P address. - The IPFS node and the hourly release-pinning that keeps the release available over IPFS/IPNS.
- Your warrant canary (with a freshness check) and PGP contact key.
- SEO surfaces (robots.txt and sitemap).
- Your instance settings, nightly backups, and (on a home install) automatic address updates.
- System resources (free disk and memory).
- A final roll-up confirming every background service and timer is active.
Anything still warming up shows a "?", not a failure. A fresh node often needs a minute for the indexer to respond, the price feeds to connect, or RPC to dial in. Those now show a neutral "still starting" marker and no longer hold up the offer to announce your instance — only a genuine failure (a service down, a missing certificate, an inactive firewall, an unfunded relay) does.
Your "Contact this operator" link can now point to a Matrix room, not just an account. The optional Matrix contact in the setup wizard now accepts either your account (
@you:matrix.org) or a room (#support:matrix.org), so you can send traders to a shared support channel instead of a personal account if you prefer. It remains entirely optional — press Enter to skip.Notes
- No database migrations. No breaking changes.
- If any line shows a ✗ or ? right after install, it usually just means that piece is still starting — re-check any time with
sudo morphit-ops status.
Downloads
-
Source code (ZIP)
1 download
-
Source code (TAR.GZ)
2 downloads
-
Morphit v1.9.18
StableAll checks were successfulmorphit-ci / TypeScript typecheck (sweep all workspaces) (push) Successful in 49smorphit-ci / apps/web svelte-check (svelte-kit sync + svelte-aware tsc) (push) Successful in 39smorphit-ci / Integration tests (real Postgres 16) (push) Successful in 42smorphit-ci / ansible-lint (playbook quality gate) (push) Successful in 16smorphit-ci / Smoke suite (run-smokes.sh, triple-pulse) (push) Successful in 30m20smorphit-release / Build + publish release tarball (push) Successful in 31m18sreleased this
2026-08-03 07:34:49 +00:00 | 114 commits to main since this releaseMorphit v1.9.18
Theme: another hotfix on the guided local-install path — the installer now gets through system-account setup and on into deploying the stack.
A small follow-up to v1.9.17. Anyone installing locally (home computer, Raspberry Pi, or an old laptop running Linux) should use v1.9.18.
Fixed
The installer no longer stops while creating its system accounts. On v1.9.17, a "Full guided install" got past the safety checks and began setting up, then stopped with
Group morphit-mcp does not existwhile creating one of Morphit's background-service accounts. One of those accounts was being created before its group existed. It now creates the group first, matching how the other service accounts are handled, so setup continues on into installing the database, relay, indexer, and web server.Changed
Clearer wording about the download. The "run a node" guide now notes that what you download is Morphit's source (a few tens of MB), and that the installer's first step downloads the software libraries (a few hundred MB) — so the
npm installwait at the start is expected, not a hang.Notes
- No database migrations. No breaking changes.
- If a v1.9.17 local install stopped at the
morphit-mcpgroup error, nothing needs undoing — the run stopped before deploying anything. Re-run the installer from a fresh v1.9.18 download.
Downloads
-
Source code (ZIP)
1 download
-
Source code (TAR.GZ)
1 download