livekit-graceful-restart.sh decided "no active calls" by counting
established TCP connections on :7881 with `ss ... state established |
grep -c ESTAB`. Two bugs:
- With a `state` filter ss omits the State column, so `grep -c ESTAB`
is always 0: the check never saw anything.
- :7881 is only LiveKit's ICE-TCP media fallback. Signalling is on
:7880 and media is normally UDP.
Reproduced in a two-person call on the local stack: the old check
counted 0 and would have restarted. Now it waits while LiveKit's own
livekit_participant_total is above 0 or any signalling WebSocket on
:7880 is open, and also waits if the metrics can't be read.
The deploy hook now installs the script when it changes (it never
did, so edits to it never reached the server).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Decorations came through Nextcloud's public DAV: a new PHP session (four
cookies) per image and no cache headers, on every decorated avatar. The
set (629 decorations + thumbnails, 573 MB) is now a versioned Gitea
generic package (avatar-decorations 1.0.0, 4 tar parts), pinned by
sha256 in cinny/decorations.sha256. LXC 106's deploy installs it to
/var/www/decorations-<version> (verified against the pin and the
in-package manifest; previous version kept; failures leave the live set
alone), and nginx serves /decorations/ with a one-week public cache.
deploy/publish-decorations.sh publishes a new version and writes the pin.
Fixes#22
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
LXC 106's matrix-deploy hook has not fired since May: Gitea posts to
10.10.10.6:9000, which webhook.service binds to 127.0.0.1. The same hooks
(same secret) are served by webhook-lotus on :9001; the Gitea hook is being
re-pointed there. Before it starts firing again:
- cinny/config.json is merged into the live file instead of copied over it.
The live file carries gifApiKey (injected by lotus_deploy.sh), empty in
git, which a copy would blank. Repo keys win except where the repo value is
empty and the live one isn't. Backup goes to /root/config-backups; the live
file is left untouched on error.
- The script installs its own updates (deploy/lxc106-cinny.sh →
/usr/local/bin/matrix-deploy.sh, after bash -n). Until now repo edits to
it never reached the server.
- A hooks.json change restarts webhook-lotus 15 s after the script exits,
since that listener is the one running this script.
- README: port 9001, the file list, and the history.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
deploy/hooks-lxc*.json in this public repo held the HMAC secrets for
every auto-deploy webhook (and 106's cinny-build token), readable
anonymously. Six values had been committed: the four matrix-deploy
secrets, 106's lotus-deploy secret (publicly reachable through
chat.lotusguild.org/hooks/lotus-deploy) and the cinny-build token.
- All six rotated on the servers and in Gitea (matrix hooks 6-9,
cinny hook 69). The old values are rejected.
- Secrets now live in /etc/webhook/secrets.env (0600) on each LXC, loaded
by a systemd drop-in that also enables webhook's -template mode; the
hooks files use {{ getenv "NAME" | js }} placeholders.
- The repo hooks files are now exact copies of the live ones, so they
also document the *-config-deploy hooks they were missing. Those
three secrets were never committed and are unchanged.
- README: where the secrets live and how to rotate one.
- CI `hooks-no-secrets` fails if a literal secret lands in a hooks
file again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The host's /etc/webhook/hooks.json was replaced on 2026-05-21 with a copy
containing only lotus-deploy, which dropped matrix-deploy and cinny-build;
this repo's copy had the opposite gap. No push to this repo has deployed
to LXC 106 since May. The tracked file is now the union of all three.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Extend voice-limit-guard to enforce a per-room publish-source policy
(io.lotus.room_quality allow_screenshare/allow_camera) for ALL Matrix clients,
alongside the existing participant limit.
- At token issue, re-sign the LiveKit JWT's canPublishSources to drop forbidden
sources (microphone always kept). Verifies our own secret signed the token
first and fails open on mismatch, so a secret drift can never mint a token the
SFU rejects. Limit check and source policy are independent (one's outage can't
skip the other).
- Live (mid-call) enforcement: a background reconcile loop calls LiveKit
UpdateParticipant to revoke a forbidden source from participants who joined
before the policy changed -- which unpublishes their in-progress
screenshare/camera server-side within ~3s and blocks re-publish. Only removes
sources (never grants), preserves other permission flags, fails open, and runs
as a daemon thread that cannot crash or block token issuance.
- Endpoint-specific room-id extraction (/get_token->room_id, /sfu/get->room) so
a client sending both keys can't get a different room's policy applied.
- Auto-deploy the guard on LXC 151 (py_compile-gated, backup + rollback).
- Unit tests: JWT re-sign/verify + tamper, secret-mismatch, source narrowing,
reconcile (never-grant / preserve-flags / disable-on-empty), fail-open.
Numeric bitrate/fps caps are NOT server-enforceable on an SFU (LiveKit forwards,
never transcodes) and remain a Lotus-client-cooperative setting; the
screenshare/camera permission is the hard cross-client lever.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The cinny/lotus-deploy.sh (hyphen) force-deploy variant was an untracked,
redundant duplicate of the CI-gated cinny/lotus_deploy.sh (underscore) added in
c13549f. It's been removed, so its install block here referenced a file that no
longer exists. The CI-gated lotus_deploy.sh is the single source of truth for
the webhook web deploy.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The live /usr/local/bin/lotus_deploy.sh (the `lotus-deploy` webhook target) was
never under version control and had rotted into two deploy-killing bugs that
froze chat.lotusguild.org on an old build:
1. CI gate: it waited on the WHOLE workflow run with a 15-min cap. Web CI shares
the single act_runner with the slow Tauri desktop builds, so a web run could
sit queued >15 min -> "result: timeout" -> deploy aborted. Now it gates only
on the "Build & Quality Checks" commit-status context (build + unit tests),
decoupled from "Trigger Desktop Build", and waits up to 45 min.
2. Dead element-call copy: `cp node_modules/@element-hq/element-call-embedded/...`
under `set -e` aborted every deploy after the widget was forked to
@lotusguild/element-call-embedded. The build already emits dist/public/
element-call; replaced the copy with a presence check.
Also: rsync now excludes config.json so the app deploy stops clobbering the
production runtime config (homeserver list / allowCustomHomeservers) that the
matrix repo owns. lxc106-cinny.sh now installs this script (syntax-checked).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The chat.lotusguild.org nginx config (LXC 106) was edited directly on the box
and never tracked — which is how its CSP drifted (kept a dead Sentry URL and
blocked matrix.org logins). Snapshot it as cinny/nginx.conf (verbatim from prod,
incl. the corrected connect-src that now allows matrix.org/*.matrix.org) and
deploy it via lxc106-cinny.sh: back up the live file, swap, `nginx -t`, and
reload only on success (auto-restore the backup if validation fails, so a bad
config can't take the site down). TLS terminates at the NPM proxy, so this is a
plain HTTP server block with no secrets.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Removed standalone matrixbot/deploy.sh — deploy is handled by the existing
webhook system. Added matrixbot/ block to deploy/lxc151-hookshot.sh: on push,
if any matrixbot/ file changed, source files are synced to /opt/matrixbot and
matrixbot.service is restarted automatically.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Per-LXC deploy scripts (lxc151-hookshot, lxc106-cinny, lxc139-landing, lxc110-draupnir)
- Per-LXC webhook hook configs with unique HMAC-SHA256 secrets
- Livekit graceful restart script + systemd timer (waits for zero active calls)
- Fix hookshot/deploy.sh capitalization bug (Uptime-Kuma, Tinker-Tickets, etc.)
Each LXC independently clones repo and runs its own deploy.sh via adnanh/webhook on port 9000.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Lotus CIandClaude Opus 5.5
jaredandClaude Opus 5