The third shape: a non-mautrix bridge. It confirms the scaffolding is not
mautrix-specific, and it is laid out differently enough to be worth having -
everything sits directly under the base path rather than in config/ and
data/ subdirectories, the owner and homeserver URL are passed on the command
line rather than through a config file, so the unit is where they can be
checked, and identd optionally binds host port 113.
matrix-base's matrix_server_fqn_* family moved into the shared context;
18 of the roles here read one of them.
Three roles now, three shapes - HTTP receiver, mautrix bridge, non-mautrix
bridge - and the scaffolding took no per-shape special-casing beyond each
role's own variables.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The second scenario, chosen to be a bridge rather than another HTTP
component: bridges are roughly 40 of the 70 roles here, and validating the
shape on role two is cheaper than discovering it on role forty. It found
two things the first role could not.
Bridges read a whole family of variables from outside themselves -
matrix_bridges_relay_enabled, _encryption_enabled, _encryption_default,
_msc4190_enabled, _self_sign_enabled, plus matrix_admin - all defined in
matrix-base. Since every bridge reads them, they belong in shared context
rather than in each bridge's scenario: molecule-shared/playbook-context.yml
now carries them along with the identity and path variables the first
scenario had inline. Note it is loaded through vars_files, which outranks
inventory group_vars, so it is authoritative; a scenario that wants to prove
one of these reaches the rendered configuration should say so explicitly.
Appservices call /whoami while starting and refuse to run if the id returned
is not the bot user they were configured as - mautrix-whatsapp exits 17 with
"Unexpected user ID in whoami call". The shared stub now takes the id it
should claim to be. Expect every bridge and bot to need this.
What the scenario proves: the bridge starts and stays up, opens its
appservice port, its rendered config and appservice registration carry the
scenario's tokens and bot user, it created its sqlite database under the
role's data path as the role's uid, and it runs the version the role pins.
It does not bridge anything and never will.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Components here contact a homeserver while starting up and exit if it is
unreachable, so nearly every scenario will need one. Standing up a real
Synapse per role would dominate the run and drag in Postgres, and these
scenarios are not testing Synapse.
The stub answers the handful of endpoints components touch during startup
with the blandest plausible response, and is deliberately permissive: an
unrecognised path returns {} rather than 404, because the goal is to get the
component past its startup checks. It is not an authentication check or a
room state machine, and a scenario should not assert *about* it - if one
starts needing it to behave like a real homeserver, that scenario has
outgrown what these tests are for.
matrix-alertmanager-receiver now includes it instead of carrying its own
inline copy. Verified green afterwards.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Molecule checks for a requirements file at the scenario's default path
before it installs anything, and warns "Missing roles requirements file"
when it is absent - so `requirements-file` pointing elsewhere was silently
ignored and the galaxy roles were never installed. CI failed on a missing
ansible-role-docker; the local run passed only because those roles were
already in ~/.ansible/roles from earlier work, which is exactly the kind of
difference a clean CI run exists to expose.
A symlink satisfies Molecule's check while keeping one copy of the pins.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
molecule-shared/ and the workflow itself are used by every scenario, so a
change to either has to run all of them rather than only the roles whose own
files moved. They were also missing from the paths filter, so such a change
would not have triggered the workflow at all.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The previous commit moved requirements.txt into molecule-shared/ but left
the workflow installing from the old per-role path, so CI failed on a file
that no longer exists while the local run - which reads the shared file
through bin/molecule.sh - passed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three things that would not have scaled to 70 roles:
- The Python and Ansible dependency pins were about to be copied into every
role. They now live once in molecule-shared/, which scenarios reference
relatively, so they cannot drift apart.
- The helper container images used for probing were hardcoded inline. They
are pinned once in molecule-shared/vars.yml, carry `# renovate:`
annotations, and a custom manager in .github/renovate.json keeps them
current - verified with a local Renovate dry run, which offers
curl 8.11.1 -> 8.21.0 and python 3.13 -> 3.14-alpine. Seventy invisible
hardcodes is the blindness class we have been removing elsewhere.
- Running a scenario meant knowing the venv and cd incantation. `just
molecule <role>` does it, and with no argument lists the roles that have
a scenario.
Molecule is deliberately not wired into prek: a run takes minutes, pulls
images and needs Docker, which is fine on request and not fine per commit.
docs/molecule-testing.md covers how to run and write these, including the
four things a role here needs that a standalone role does not. AGENTS.md
points at it rather than carrying the detail.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Not for merging as-is. One role's scenario plus the workflow that would
make per-role testing affordable here, so we can see how it behaves before
deciding whether to do the other 69.
Unlike the MASH repositories, one repository holds every role, so running
everything on every push is not an option. The workflow's first job works
out which roles a push touched and builds the matrix from that; a change to
docs, or to a role with no scenario yet, runs nothing.
Three things this role needed that a MASH role does not:
- The variables matrix-base would supply (matrix_base_data_path,
matrix_domain, matrix_user_name, matrix_group_name and the uid/gid) have
to be provided by the scenario, and the user and group have to exist
before the role's file tasks run.
- The service contacts the homeserver while starting up - it fetches
/_matrix/client/v3/joined_rooms to resolve its room mapping and exits 1
if that fails - so prepare.yml stands up a stub homeserver. Most roles
here are bridges and bots, so this is likely the rule rather than the
exception.
- verify.yml runs as its own play, where role defaults are out of scope, so
the paths it reads are pinned in molecule.yml. The version is deliberately
NOT pinned: it is read from defaults/main.yml so the assertion compares
the running image against what the role ships.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
prek bakes the full path of the currently installed version into the hooks it
generates (var/mise/installs/prek/<version>/...). That path stops existing as
soon as the pinned version changes or old versions are pruned, and the hook's
PATH fallback finds no prek either, so every commit fails until the hook is
regenerated by hand. It also means a hook keeps running the version it was
generated with, long after mise.toml has moved on.
Rewriting PREK to mise's shim makes the hook resolve whatever mise.toml pins at
the time it runs. The accompanying MISE_DATA_DIR / MISE_TRUSTED_CONFIG_PATHS
exports keep that resolution inside this project - without them mise falls back
to the global data directory and silently installs a second copy of the tool.
The patch loop now covers every hook file prek generated rather than just
pre-commit, since default_install_hook_types decides which ones exist.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every dependency bump in the housekeeping list was already automerging,
but as a pull request - which meant an email and an open PR for every
bump even when it merged itself. Branch-push automerge makes green bumps
invisible; a failing one still surfaces as a pull request. The new
Matrix i18n workflow gates the i18n/requirements.txt ones with a real
exercise of the translation toolchain.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BhJbmhyUpt4qSrKD9cbJ5A
Nothing used to test an i18n/requirements.txt bump - the first real
exercise of a bumped Sphinx/Babel stack was the next scheduled
translations run, well after merging. Template extraction runs the same
toolchain the scheduled workflow uses, in a minute or two, so a broken
bump now fails on its branch instead.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BhJbmhyUpt4qSrKD9cbJ5A
The seven pinned com.devture.ansible.role.* roles are maintained by the
same people who review these PRs; their releases are already the review.
Same branch-push automerge (and the same lint gate) as MASH role bumps.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BhJbmhyUpt4qSrKD9cbJ5A
Same rule mash-playbook has been running in production: role bumps from
mother-of-all-self-hosting merge via branch push (no PR) once the branch's
CI checks pass. The trust basis is the role repositories' own gates - each
release is cut by Molecule-verified, autotagged merges over there.
Bumps of devture roles and everything else keep their PRs. Extending the
same treatment to devture roles (here and in mash-playbook) is a discussed
follow-up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BhJbmhyUpt4qSrKD9cbJ5A
The previous image versions set ANSIBLE_WORKER_SESSION_ISOLATION=False in an
attempt to restore SSH prompts on Ansible 2.21. That setting does not bring
the prompts back — it makes Ansible hang silently at the first SSH connection
whenever a prompt would have been needed. See
https://github.com/devture/docker-ansible/issues/6
The new image auto-accepts the SSH host keys of previously unknown hosts
instead. The docs stop recommending ANSIBLE_WORKER_SESSION_ISOLATION=False
for direct (non-Docker) runs for the same reason, and now explain how to use
an ssh-agent for passphrase-protected SSH keys.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Protected rooms, watched policy lists, protection settings and the set of
users allowed to command a bot all belong to a management room rather than
to the bot, and a room can be protected from only one of them. That is what
makes it possible to delegate a single room to another moderator without
handing over power in every other protected room, so it is worth spelling
out, along with what such a split does not cover: the variables naming one
management room, and policy lists shared through subscriptions.
Adding a second management room also means declaring the rooms of a bot
which had been auto-creating its own, which is a step with a sharp edge:
once the bot is no longer auto-creating, any of its rooms missing from the
declaration is unregistered. Document that path and warn about it.
Also drop a note describing "!lists subscribe --insert-before" as a future
release. The role has shipped a version carrying it since v0.2608.0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This reverts commit 421b592690.
The org.matrix.msc4143.rtc_foci property is indeed gone from MSC4143 and
Element Call v0.24.0 no longer reads it, but Element Web does not ship that
version of Element Call yet. v1.12.26 embeds Element Call v0.22.0, which
discovers transports through the well-known property.
That embedded copy cannot consult the homeserver's RTC transports API
instead: it runs as a widget, a widget holds no access token and the API
requires authentication, so its own discovery code skips the API branch
outright. Newer Element Call versions ask their host client over the widget
API (MSC4515) and Element Web already implements that side of it, but the
widget has to speak it too.
Dropping the property therefore left Element Web and Element Desktop users
with MISSING_MATRIX_RTC_TRANSPORT when starting a call.
The restored variables carry a comment about why the property stays, and the
changelog entry is rewritten to announce the revert.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lk-jwt-service v0.6.0 ships a healthcheck which builds its URL as
http://localhost:$LIVEKIT_JWT_BIND/healthz, interpolating the bind
address into the port slot. We set LIVEKIT_JWT_BIND to ":8080", so the
check requests http://localhost::8080/healthz, which does not parse and
can never succeed. The container therefore sits permanently unhealthy,
Traefik skips unhealthy containers, and the service stops being routed,
so Element Call fails to obtain an SFU token.
No value satisfies both sides: the service passes LIVEKIT_JWT_BIND to
ListenAndServe, which needs a full bind address, while the healthcheck
needs a bare port. The check only works when the variable is unset and
its own "8080" fallback applies, which would mean ignoring a configured
port. Correcting the check from here is not possible either, as the
image is built FROM scratch and has no shell for a --health-cmd
override to use.
Turn the healthcheck off, behind a variable so it can be turned back on
once upstream fixes it. This restores the pre-v0.6.0 behavior, where the
image carried no healthcheck at all.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Failing to update the apt cache now reports why, instead of reporting that
the result of the apt module has no cache_updated attribute.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lk-jwt-service v0.6.0 stopped honoring the delay_cs_api_url request parameter
and now locates a user's Client-Server API by fetching the
/.well-known/matrix/client file of that user's base domain itself.
For our own homeserver, this makes the service leave the server over the
public network only to come back to it, which is wasteful at best and does
not work at all where the base domain is not reachable from the host.
Since we know where our homeserver is, we tell the service directly via the
new LIVEKIT_CS_API_URL_OVERRIDES environment variable. Federated homeservers
keep being resolved through discovery.
The service now also joins the container network that the homeserver's
Client-Server API is reachable on, like our other add-on services do.
It used to only be connected to it by coincidence, via the reverse-proxy
network, which left it unable to reach the homeserver when the playbook is
not managing Traefik.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Tuwunel v1.9.0 adds the default `.im.nheko.msc3664.reply` push rule to every
account, but only evaluates its condition when this option is on. Without it,
replies do not affect notification counts or reach pushers.
Kept off, as upstream has it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
