Sessions of this account not seen for 90+ days (never the current one; a
session with no last-seen time is left alone) get a nudge in the existing
security strip, after the device-security ones and with the same rules
("Not now" snoozes 7 days, gone after 3 dismissals; a snoozed security
nudge no longer hides it): "You have 3 sessions you haven't used in 3
months. Signing them out keeps your account safe if an old device is
lost." [Review]
Review opens Settings -> Devices with those sessions ticked for sign-out
(one confirm, the usual re-auth incl. SSO); they are listed first and
marked "Inactive for 3+ months". OIDC accounts sign out on the
provider's dashboard, so nothing is ticked there.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Landscape, at most 500px tall, touch input (a sideways phone is ~840px
wide, the Tablet size class, so the trigger is the screen shape), in this
room's joined call, chat closed: the call takes the whole screen. No room
header, sidebar or call-status bar; safe-area insets respected.
- Taps inside the call frame never reach the app, so while the controls
are hidden a transparent layer covers the call: the first tap shows the
call bar plus Chat / Exit full screen, which hide again after 4 s
without a touch; with them up the call underneath stays usable.
- In an Android browser tab the first tap also enters the browser's own
full screen (hides the address bar); left again on rotate back, hang
up or exit. No effect on iPhone or the installed app.
- Exit full screen lasts until the phone is upright again.
- The frame is transparent (room wallpaper), so in full screen it gets an
opaque base colour under the wallpaper.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The fork now checks, before joining (lotusDeviceCheck=1), that the mic /
output the user last picked are plugged in. When one is missing it asks
us (io.lotus.device_missing): "Your usual microphone isn't connected.
Arctis 7 isn't connected. Join with Laptop Mic instead?" [Join anyway]
[Cancel]. Join anyway isn't asked again this session for the same
device(s); Cancel closes the call before it starts. A device lost during
a call is a toast ("Arctis 7 disconnected — now using Laptop Mic").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The login page looked for an OAuth issuer in .well-known
(m.authentication), which matrix.org no longer publishes, so its users
got the password / SSO compatibility forms. Detect native OAuth the way
the SDK finds the auth server for refresh: load /auth_metadata next to
the login flows. When present:
- Login shows "Continue with <server>" only (as Element does); a legacy
SSO loginToken coming back from a redirect still completes.
- Register opens the auth server's sign-up page (prompt=create, where
prompt_values_supported says so) instead of the legacy forms.
Servers without it (Lotus) are unchanged.
Checked against the live servers: matrix.org login and register show
"Continue with matrix.org" and no password form; matrix.lotusguild.org
keeps its password form.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The wheel ran as soon as the composer appeared; when it landed during the
room's initial jump-to-bottom it was snapped back and "Jump to Latest"
never showed. Seen ~1 in 5 runs since matrix-js-sdk 44 shifted the load
timing (failed the post-merge e2e run of #447 on lotus). Wait for the
newest message, then retry the scroll until the button appears.
10/10 locally.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The fork synced to upstream v0.28.0 (element-call #61): matrix-js-sdk
develop with the SDK 44 MatrixRTC changes, delayed-leave delegation.
v0.28 renamed the delayed-leave config keys and its embedded build ships
no config.json (defaults: 18 s). Ship the Lotus values with the build
(build/element-call/config.json -> public/element-call/config.json) so
the desktop app keeps the 60 s grace from element-call lotus #46. The web
deploy keeps the server's own config.json (rsync excludes it); that file
already carries both the old and new keys.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Picks up three releases of MatrixRTC / delayed-event fixes and
matrix-sdk-crypto-wasm 18.3.1 -> 18.9.0.
SDK 43 moved OAuth (MSC3861) into the SDK and dropped oidc-client-ts, so
the native-OAuth sign-in is ported:
- login discovers the auth server via /auth_metadata, registers with
OAuth2.registerClient, and keeps the PKCE verifier + device in its own
sessionStorage entry keyed by `state` (single use; CSRF check);
- refresh is done by the SDK (oauthClientId + onTokenRefresh); the #16
cross-tab protection is kept by a narrow guard on refresh-token grants
(the SDK calls the global fetch for OAuth, not fetchFn), which
serialises them with a Web Lock and adopts a sibling tab's tokens;
- logout and forced logout revoke via mx.logout() (the SDK revokes at
the auth server for OAuth sessions);
- the callback page parsed its params on every render, re-running the
single-use code exchange in a loop; parsed once now.
Also: notification_type is deprecated in MSC4075 (ring in DMs, notify
elsewhere when omitted), _onRTCSessionMemberUpdate ->
ensureRecalculateSessionMembers, precache limit 6 -> 10 MB for the
7.8 MB crypto wasm.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
When the server keeps rejecting this device's /keys/upload with 400 "One
time key ... already exists", the device's encryption store has diverged
and the SDK retries forever (matrix-rust-sdk#5200, still open). Synapse
saw 474k of these from May 30 to Jul 8 2026. The fix is a fresh login, so:
- keyUploadWatch observes /keys/upload responses through the client's
fetchFn (the response the SDK gets is untouched; a clone is read).
Three conflicts within 15 min = stuck; any successful upload clears it.
- KeyUploadBanner (critical strip, same slot as the sync/clock banners):
"Encryption on this device is out of sync" with Details and
"Fix: sign out…", which opens the normal logout dialog (it already
warns when the device isn't verified / backed up).
Server side, Grafana now alerts on the same signal (E2EE Key Upload
Conflicts) and on delayed-event 5xx (Call Delayed-Event Errors).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
On the desktop app with a working keychain (Windows Credential Manager),
the tokens leave localStorage: it keeps a token-less shell blob
(tokensIn: keychain) and the full session lives in memory, read from the
keychain at boot before anything reads the session, so getFallbackSession
stays synchronous. Web, Linux and older desktop builds are unchanged.
- A new session (login, token refresh) is written to localStorage in full
first; the mirror verifies the keychain copy and only then strips the
tokens, so one durable store always holds the newest tokens.
- Existing installs migrate on first boot the same way; a failing keychain
write keeps the tokens in localStorage.
- If a shell is there but the keychain entry is missing or unreadable, the
app shows Retry / Sign in again instead of logging out, and the mirror
is not started (it would clear the entry).
Fixes#105
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
vimeo.com/event/{id} (and /event/{id}/{hash} for unlisted events) now get
the same click-to-play card as Vimeo videos, embedding Vimeo's live-event
player (vimeo.com/event/{id}/embed[/{hash}], dnt=1). The issue assumed
this needed vimeo.com in the CSP, but frame-src is already https: on web
and desktop.
Refs #200
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
MSC2931: a widget with m.always_on_screen (auto-granted) can send
set_always_on_screen; the host now answers it, and a widget that asked
keeps running in a small floating window (title, Back to room, close)
when the Widgets panel closes or you switch rooms. Back to room returns
to it with the panel open on that widget, without reloading it.
Moving an iframe reloads it, so widget frames now live in an app-level
layer from the start (widgetSessions.ts) and are positioned over the
panel's placeholder while docked. A widget that did not ask is torn
down as before. One floating widget at a time; it also goes when the
widget withdraws, is removed from the room, or on logout. A permission
request while floating grants nothing extra.
Refs #205
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The call-ended toast gets a Details button (and stays 10 s) that opens a
card with who was in the call and the peak, connection verdict and
reconnects, how long your / each person's / everyone's connection was
poor (likely the server when everyone), RTT, packet loss and jitter, and
your talk, mute, deafen and screenshare time, in technical wording.
Deafened time is tracked by CallEmbed; the rest comes from the fork's new
io.lotus.call_summary `details` (lotus.29). Settings → Calls gets
"Summary After Calls" (on by default).
Fixes#253
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Each widget in the Widgets panel (and the open widget's header) gets an
"Open in a new window" button, so a pad or board can sit beside the chat.
It opens the same filled-in URL the frame uses (widgetCompleteUrl, now
shared) with noopener,noreferrer, standalone as in Element. Only
cross-origin https URLs get the button: room state can carry anything,
e.g. a script URL from another client.
Refs #205
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Long-press opens the message action sheet (#166), but on iOS the system
takes a long-press for itself: the link/image callout or the text-selection
loupe races the 450 ms press. Message rows on touch screens now carry
data-touch-actions (no -webkit-touch-callout), and the pressed row carries
data-pressing from touch-down until the finger lifts, which turns text
selection off for that moment only. Copy Text stays in the sheet.
Refs #199
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
- Developer tools (send event, state event editor, room account data) and
setMarkedUnread go through the typed helpers in utils/room and
utils/accountData (new setRoomAccountData) instead of casting at the
call site; sendStateEvent accepts a developer-chosen type string.
- SecretStorage passes the key description as the SDK's
SecretStorageKeyDescription, CallEmbed replies with an IWidgetApiRequest,
and fillMissingPowers writes through a Record instead of `as any`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
#265: in the compact layout (phones, narrow windows) the control bar
stacked into four rows (~330px, 40% of a portrait phone). It is now one
row: mic, deafen, camera, chat, soundboard, More and an icon-only End.
Screenshare, its audio mute and fullscreen move into More; Share Screen
is only offered where getDisplayMedia exists. (The self-view comes from
the fork, lotus.27.)
#277: "Mute Screenshare Audio" showed whenever screensharing was allowed,
as a speaker icon that did nothing. It now shows while the fork reports
remote screenshare audio (io.lotus.screenshare_audio_state), or while it
is muted, so a hidden mute cannot silence the next share.
Fixes#265Fixes#277
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
- Modal500 takes a label: Room/Space Settings ("<name> settings") and
Settings dialogs have accessible names.
- aria-selected on plain buttons (inbox filter chips, date/time picker
chips, verification/usage/notification menu items) -> aria-pressed; the
Permissions power chips open a menu -> aria-haspopup + aria-expanded.
- Labels tied to the room pack name input and the Export From/To dates.
- UserAvatar defaults to alt="" (decorative next to a name / in a
labelled button) unless a caller passes one.
- Composer placeholder: Slate's inline opacity 0.333 overrode the #222
AA-contrast class (2.66:1); it is dropped so the class (0.75) applies.
- e2e/a11y.spec.ts also audits Room Settings → Permissions, Emojis &
Stickers and Export, checks the dialog name, and audits the inbox.
Fixes#278
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
#279: Room Settings → Soundboard said clips are "inherited by every room
under this space" for plain rooms too; the wording now follows
room.isSpaceRoom(). The Push to Deafen description only mentions Escape
while rebinding. "can pin a messages" → "can pin a message".
#280: the tracking stripper removed Reddit's context=, which is what
makes a comment permalink show its parent comments. Only share_id and
rdt are stripped now.
Fixes#279Fixes#280
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Settings → About only showed cinny's package version, the same on web
and desktop and unchanged across many deploys a day. vite.config.js now
stamps the short commit (git, or LOTUS_BUILD_COMMIT; "unknown" without
git) and build time into the bundle, and About shows "Web build <commit>
· built <date>", linking the commit. On desktop it also shows "Desktop
<version>" from Tauri's built-in plugin:app|version (allowed by
core:default).
Fixes#276
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Production Synapse caps delayed events at 24h, but the dialog let any
later time be picked and then showed the raw server error. The date
input now has min/max (today .. now+24h), a too-late time shows "You can
schedule up to 24 hours ahead." as soon as it is picked (and blocks
Schedule), and ORG.MATRIX.MSC4140_DELAY_TOO_LARGE maps to the same text.
Fixes#275
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The strip showed for PREPARED and for the first SYNCING after it (or after
RECONNECTING), i.e. until the next /sync returned: ~30 s of long-poll on a
quiet account while the app already worked. A SYNCING means a sync just
succeeded, so the strip now shows only during CATCHUP; failures still show
as Reconnecting / Connection lost.
Fixes#274
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
#269 read-receipt pill sits at the message row's right edge, not where
the text ends (outside the shrink-to-fit content column; grows in
Compact). Bubble layout unchanged.
#270 Most Active Members: the bar's % is of its own track, so every row
shows its count.
#271 Noise Suppression description at T200 like the rest.
#272 decoration preview shows your own avatar under the decoration; the
empty state is a visible dashed circle.
#273 grouped settings get a divider between tiles (SettingTileGroup, via
globalStyle; SettingTile roots carry data-setting-tile), and on phones
the Sync tile's buttons sit under its description.
Fixes#269Fixes#270Fixes#271Fixes#272Fixes#273
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
OutboxFeature loaded lotus_outbox_v1 once and then saved its own
in-memory list over the key on every change, so a second tab's save
dropped the first tab's unsent messages. persist() now re-reads the
stored list and applies the change (add/remove by txnId) to it.
Fixes#268
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
#266: remembered widget approvals (lotus.widgetConsent.*) survived logout,
so they applied silently to the next account on the device and listed the
previous account's rooms; clearPlaintextCaches now sweeps them. Space
structure, call membership and Lotus room-policy state join
PROTECTED_STATE_TYPES, so no widget can be granted them.
#267: any website could auto-submit a form to /share-target and land the
user in the share picker with attacker-chosen text and files. The worker
now refuses a share whose Origin is another site. (Sec-Fetch-Site, as the
issue suggested, never reaches a service worker: checked in Chromium.)
Origin null/absent, as the OS share sheet sends, stays accepted, and
/share now says the content came from another app or website. Both logout
paths await clearSharedPayload(); the server-forced one also clears the
soundboard clip cache, as #57 intended.
Fixes#266Fixes#267
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Members and bookmarks were 266px, the gallery 320, the threads list 340,
a thread and widgets 360. Swapping panels resized the room view, so the
header icons slid sideways (74px for members -> threads) and a second
click on the same spot opened Widgets instead of closing the panel. All
six now use SIDE_PANEL_WIDTH (340px, styles/sidePanel.ts).
Fixes#263
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The topic editor's hand-rolled markdown treated any single _ or * as
emphasis, inside URLs and #alias:server / @user:server too, and built the
plain topic by stripping tags, so the underscores were lost for every
client. It also left " unescaped in link targets.
buildTopicContent() now uses the composer's inline markdown parser (links,
bare URLs and HTML escaping as for messages) and treats _ between two
letters/digits as literal (CommonMark), except inside code, link targets
and URLs. The plain topic drops the markers but keeps links as
"text (url)" and decodes entities.
Fixes#262
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The Server ACL editor treated an absent allow_ip_literals as false, so the
first save in a room with no ACL (or one omitting the field) silently banned
IP-literal servers. The spec default is true. normalizeServerAcl() fills the
spec defaults once and feeds the initial state, the dirty check and the save.
It also stops showing an existing ACL with no allow list as allow-all: per
spec that allows no servers, and the /acl command already keeps it as-is.
The checkbox is driven by onClick (folds has no onChange), now readOnly so
React stops warning about a checked input without onChange.
Fixes#261
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
DND was sent as presence unavailable with the literal status text
"dnd": it replaced your status, and other clients showed "dnd". Where
the server supports MSC3026 (busy_presence), DND now sends presence
org.matrix.msc3026.busy and leaves the status alone (Synapse keeps
busy across /sync set_presence=online). Busy renders as red "Do Not
Disturb" with the real status line; the legacy unavailable + "dnd"
encoding is still recognised and still used on servers without it.
Status saves / clears no longer downgrade busy. The SDK rejects
non-spec presences, so busy goes through a direct authed PUT.
Fixes#283
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
A passive read (just looking at the room) sent a threaded receipt for
every unread thread you don't follow, so the Threads panel's Unread
filter and Mark all read only ever worked for followed threads.
Decision (2026-10-19): leave them unread until opened, but stop
counting unfollowed threads toward the room's unread dot
(getUnreadInfo subtracts them like muted threads; a mention makes a
thread followed, so highlights are never hidden). isThreadFollowed
moves to utils/room. Adds a Playwright regression.
Fixes#314
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The pacman / apt update command installed the downloaded package with
no integrity check. Download it with the release's SHA256SUMS.txt and
.minisig into a fresh temp dir, verify the signature with the updater
key (minisign), check the hash, then install.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
"Notes only visible to you" overstated it: notes live in plain account
data that homeserver admins can read. Say so under the note box (and
link it as the box's description): "Only you can see this in Lotus Chat.
It's stored on your homeserver and isn't end-to-end encrypted."
Fixes#329
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Settings (every page, pickers, the denoise tester) was in the entry
chunk though it only renders when opened. Load it lazily behind a
focusable spinner (it opens inside a FocusTrap), and warm the chunk
when the app goes idle. SettingsPages and the status localStorage keys
move to tiny modules so callers don't pull the dialog back in.
sanitize-html (+postcss, htmlparser2) and chroma-js get their own vendor
chunks; with Settings gone the chunk graph had dropped them into the
entry chunk.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The entry chunk on lotus (d7bc6d94) is 360238 bytes gzip, past the
358317 budget seeded in September, so every PR now fails Build &
Quality Checks. The growth is ~40 merged fixes (call-state, status,
notification, search and settings changes), not one regression.
Re-seed per the file's rule: measured sizes +10% (total 1666894 →
1833583, largest 360238 → 396262).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
On desktops with no tray host (stock GNOME), "Keep running in the tray"
hid the window with no icon to bring it back. Ask the desktop app
(has_tray_host); without a tray, say so, focus "Quit when I close the
window", and offer "Keep running in the background" as the secondary
choice. Older desktop builds keep the current dialog.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The fork now refuses a clip it cannot publish (audio-only rooms grant
only the microphone source) with reason "not_permitted". Say "The
soundboard isn't available in this voice channel" instead of the
generic error. Older forks are unaffected.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The PiP had no way to get out of the way short of dragging it. Add a
minimise / expand button: minimised it is a 220×40 bar with the call
room's name (still draggable, tap to return to the call). The choice is
remembered; with no saved choice, phones start minimised.
Fixes#323
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
After a homeserver stall Element Call showed "Connection lost" while the
call bar still looked joined. Handle the fork's io.lotus.call_error:
show "Call disconnected" with a Reconnect button (io.lotus.reconnect),
and reconnect automatically once sync is back.
Fixes#318
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The fork reports what actually runs as io.lotus.denoise_state, but the
host never listened ("unsupported action"), so a failed ML denoiser
silently published the raw mic. Handle it: toast when ML did not start
or fell back to another model. Remove LotusDenoiseFeature, which waited
for a postMessage nothing sends any more and accepted it from any frame.
Refs #260
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
LOTUS_FEATURES.md had not been updated since September and was missing
55 shipped features. Add a "June–October 2026 Additions" section,
grouped by area, one line per feature with its commit and issue.
LOTUS_REFERENCE.md's blocked-features note is re-checked against
production Synapse 1.162.0.
Fixes#294
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Custom Window Chrome, Launch on login / When I close the window and the
keychain status sat inside Appearance, between Monochrome Mode and
Twitter Emoji. Move them to a Desktop section (above App Updates),
rendered only in the desktop app.
Fixes#290
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The denoise worklets/wasm/models were only copied into dist/ by the
lotus-denoise closeBundle hook, so on the dev server the call frame got
the SPA fallback for them and silently used the raw mic. The asset list
is now shared (denoiseSources) between that copy and a dev-server
middleware serving the same paths. Also send lotusTheme=true with
lotusTransparent, which the fork warned about in every call.
Fixes#293
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Where the on-device Translator/LanguageDetector APIs are missing,
Settings said to use Chrome/Edge "or the Lotus desktop app", but the
Linux and macOS apps (WebKit) lack them too, and showed the advice to
people already in the app. Word it by platform. (The per-message
Translate action was already hidden where unsupported.)
Fixes#349
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
manifest.json locked the installed app to portrait (no landscape video,
screenshare or call view on Android / tablets) and its "New Message"
shortcut pointed at /, the home page. Drop the orientation lock and
point the shortcut at ./direct/create/ (Create Chat).
Fixes#292
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
When a message was already in the target language, or its language
could not be detected, Translate switched itself back off with no chip,
no "Translating…" and no error. Show a short toast instead: "This
message is already in English." or "Couldn't tell what language this
message is in, so it wasn't translated."
Fixes#287
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Closing the dialog via "I've accepted — retry" also deactivated its focus
trap, whose onDeactivate is "Later": every close set the 10 s snooze, so
a retry made before the terms were really accepted failed silently.
Retry now closes without snoozing (and clears any earlier snooze).
Fixes#311
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Custom emoji and sticker images, pack avatars and power-level badge
images were uploaded with their EXIF intact (GPS location, timestamp),
although they are published to every member of the room or space. Route
those three pickers through a new useStrippedFilePicker, which applies
the same stripImageMetadata as attachments and avatars.
Fixes#332
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Previews are built from the plain body, clicks follow formatted_body's
hrefs, and a sender controls both: text "https://paypal.com" linked to
evil.example showed a real PayPal card. Drop body URLs whose site is
only claimed by the visible text of a mismatched link (linkSafety).
Fixes#286
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The embed fallback used the preview's og:url whatever site was posted.
og:url is set by the page, so evil.example could render as a playable
YouTube card (attacker's video) whose links go to evil.example. Only use
an og:url on the same provider as the posted link, including the
provider's own short-link domains (youtu.be, spotify.link, redd.it, …).
Fixes#326
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
SettingTile rendered its description in a <p>, and many tiles pass
blocks (badges, Box, Text), so Settings had <div>/<p> inside <p> and
React warned on every page. Render it as a <div>. UploadCardRenderer
called startUpload() during render, which now updates the desktop
taskbar-progress hook mid-render; start it from an effect instead.
Refs #291
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The ? dialog showed 7 shortcuts and left out Ctrl+K, ↑ to edit, all
composer formatting keys and the push-to-talk / deafen keys. Add them:
formatting rows are generated from the editor's INLINE_HOTKEYS /
BLOCK_HOTKEYS so they can't drift, and the call keys show the user's
own bindings.
Fixes#334
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Forwarding from an encrypted room into an unencrypted one posts the text
(and a re-uploaded plaintext attachment) where the server can read it,
with the original author attached, and the dialog said nothing. Show a
note above Send naming the unencrypted destination(s).
Refs #330
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Compression re-encodes through a canvas to JPEG. Only PNG was skipped,
so a transparent WebP came out on a black background, and an animated
GIF or WebP became one still frame. GIF and APNG are no longer offered;
an animated WebP (VP8X animation flag) or any image with a transparent
pixel is kept as the original.
Fixes#331
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
When a room admin turned off Allow Screen Sharing, the sharer got two
toasts: the correct "turned off by a room admin" and the fork's
"ended" notice, "the shared window was closed", which was wrong. Skip
the ended notice while the room policy forbids sharing.
Fixes#297
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The service worker posts inline notification replies in the clear (it
cannot encrypt). The Reply action was only withheld for rooms that were
encrypted when the notification was shown, so a reply typed hours later
into a room that has since turned on encryption went out unencrypted.
It also posted with the newest session the worker knew, not the
notification's account.
The worker now GETs the room's m.room.encryption state first and only
sends on a definite 404; notifications carry the user id and the worker
uses that account's session (or refuses). "Reply not sent" says why.
Fixes#299
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The first-use PiP position was bottom-right, 72px from the bottom: on
desktop over Send, on phones over Send, Emoji and the message box, so
the first thing you typed or tapped in the text room hit the call.
Default to top-right under the room header instead. A dragged position
is still remembered.
Refs #323
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
notify() shows an in-page toast while the window is focused and only
falls back to an OS notification otherwise, but both message and invite
notifications were gated on Notification.permission === granted. Anyone
who declined (or was never asked) heard the ping with nothing on screen.
Gate only the OS path on the permission.
Fixes#324
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
"Clear after 1 hour" lived in that device's localStorage, so a status
set on a laptop that was then closed stayed on the server until the
laptop opened Lotus again. The expiry is now also kept in account data
(io.lotus.status_expiry: { status_msg, expires_at }). Every device's
StatusExpiryMonitor clears the status when due, but only if it is still
the one that was set to expire (otherwise it just forgets the expiry).
Fixes#300
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
A timed mute writes a server-side push rule (muted everywhere), but the
unmute timer lived in one browser's localStorage: close that laptop and
the room stayed muted on the phone until it was reopened, and forever
after a logout there. Timers now live in account data
(io.lotus.mute_timers: { timers: { [roomId]: unmuteAt } }). Every
client schedules from it, reschedules when it changes, and the first one
online after expiry unmutes. An indefinite mute clears a pending timer.
Local timers from before are migrated once.
Fixes#333
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
dedupePolicyEntries keyed on the raw recommendation, so m.ban and
org.matrix.mjolnir.ban for the same entity never matched and the rule
showed twice (both labelled "Ban"). Key on the normalised value; move
the pure helpers to policyEntries.ts with unit tests.
Fixes#321
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
SecurityBanner asked pickBanner whether another strip would show without
the dismissed set, so a dismissed Kuma announcement (often pinned for
weeks) still counted and suppressed the verify / backup nudge on every
device that had dismissed it. The dismissed set is now a shared jotai
atom (still persisted to localStorage) used by both banners. Adds a
Playwright regression.
Fixes#307
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
plain() deleted those characters everywhere, so "join #lotus-general"
lost its #, snake_case and @some_user lost underscores, "->" became
"-", "5 * 3" became "5 3", and "<b>" became "<b". Strip only syntax:
heading / quote / list markers at line start, paired **x** *x* _x_ `x`,
and HTML tags.
Fixes#308
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
After a reload during a DM call the toast said "You were in the call in
Bob Builder"; it now says "You were in a call with Bob Builder". DM
calls are never auto-rejoined (it could ring the other person), and if
they have left by the time you tap, the toast says "Call ended" instead
of starting a new, ringing call.
Fixes#302
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Server search also returns the m.replace events, so one message edited
twice came back as three rows. Fold edits into their message: one row
at the best-ranked position, showing the latest m.new_content and
pointing at the original event so Open jumps to the real message.
Fixes#288
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
A filter matching nothing left Favorites / Projects / Rooms / Low
Priority headers with nothing under them. While filtering, sections
without a match are hidden, and with no match at all the list shows
"No rooms match “…”" with a Clear filter button.
Fixes#289
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The board opens with focus in its search field, and its FocusTrap used
the shared stopPropagation helper, which ignores Escape while an input
is focused, so Escape never closed it (Chromium, Firefox, WebKit).
Always close on Escape here, still stopping the key from closing the
room or modal underneath. The GIF picker already uses the default.
Fixes#304
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
"The call failed to load… try again" only offered Back, so the user had
to back out and press Join again. Try again starts a fresh embed for
the same room with the same mic/camera/sound (which disposes the hung
one); Back stays as the secondary action.
Fixes#301
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
matrix.to can't send anyone to this deployment (its Cinny adapter is
app.cinny.in, web-instance[] is Element-only), so a friend scanning a
room QR or clicking an invite link was offered Element or app.cinny.in.
Share Room (link + QR) and the Invite dialog (link + QR) now use the
Lotus permalink (correct on desktop via webAppUrl). The space menus
keep both "Copy Link" (matrix.to, for other clients) and "Copy Lotus Link".
Fixes#285
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
PollContent read mEvent.getId() once at render. For your own new poll
that is the local echo id (~!room:txn), and nothing re-rendered it when
the SDK swapped in the real $id, so the vote related to the local id
and the SDK threw synchronously (getPendingEvents under chronological
ordering): no vote was ever sent. Track the id via LocalEventIdReplaced
and Status, block voting while it is still local, and catch synchronous
throws. Adds the Playwright regression the issue asks for.
Fixes#284
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
deviceIds are hashed per frame/origin, so the ids the menu got from the
chat page's enumerateDevices() were unknown inside the call iframe: the
fork fell back to the default output while the menu showed a tick on
the chosen device. Handle io.lotus.audio_output_state (also silencing
the fork's "host did not ack" log) and build the menu from the fork's
list and effective selection; older forks fall back to the old list.
Fixes#310
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The tray only knew what this browser had scheduled (localStorage), but
Synapse lists every pending one via MSC4140 GET /delayed_events. Load it
when the tray mounts and on window focus (localStorage stays as a first-
paint cache), keeping only non-state m.room.message events. Logout now
says "N scheduled messages still send after you log out" and offers to
cancel them.
Fixes#322
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
During a homeserver stall Element Call retries update_delayed_event every
2 s, and the host opened a new HTTP request each time: 12 pending in a
60 s stall, all landing together on recovery (404 x5, 429 x7). Share the
in-flight restart per delay id and abort it after 5 s.
Refs #318
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
A hung Synapse (accepts connections, never answers) kept sync SYNCING
for up to 110 s (pollTimeout 30 s + the SDK's 80 s buffer), so no strip
showed and Kuma was never asked. A watchdog now notices a /sync overdue
by 45 s, confirms with a 5 s GET /_matrix/client/versions, then shows
"Lotus Chat's server isn't responding…" and polls Kuma at once.
Fixes#317
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
More than 10 call features subscribe to CallControl's state_update at once
(call bar, controls, hotkeys, AFK, muted-talk, join/leave sounds,
announcements, rejoin, policy toast, glance, desktop thumbbar/SMTC). At
the default limit Node's events logged a MaxListenersExceededWarning, and
the console kept the warning's emitter (this CallControl, and through it
the whole Element Call document) alive: ~11 MB per call per the audit.
setMaxListeners(100); these are real, unsubscribed-on-cleanup listeners.
Not reproduced in this environment (headless web logs no warning: fewer
subscribers without the desktop hooks); this is the fix the issue
measured (swallowing the warning freed all CallControls).
Fixes#316
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Two bugs in the one-row "+N" collapse:
- measure() used the border box, which grows by 4 px while collapsed
(margin -2 / padding 2), so a row within 4 px of fitting flipped
collapsed↔expanded every frame (flicker, layout shifts, ~11 % of a core
per looping message). It now measures the content box.
- The collapsed style used the `margin` shorthand, which overrode the
timeline's marginTop and, when removed, cleared it; the row then stuck
to the message text. Longhand margins keep marginTop (minus 2 px).
Verified headless (12 reactions, viewport sweep 1300–1400 px): before,
90 style flips in 1.5 s at 1344 px and margin-top 0px after a cycle;
now 0 flips at every width and margin-top stays 8px.
Fixes#305
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
createRoom granted org.matrix.msc3401.call.member at user level only for
voice rooms. Every other room fell back to Synapse's state_default (50),
so in a private group room (where Lotus shows the call button and rings)
only admins could start a call; members saw "1 Live · Join" and then
"You don't have permission to join!". All new non-space rooms now get
call.member: 0 (spaces keep their own override).
Verified headless through Create Room → Private: before, no call.member
entry (state_default 50); now 0. Existing rooms are unchanged; an admin
can set it in room permissions.
Fixes#328
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
handleReject related the MSC4310 decline to the caller's call.member event
(the notification's reference), but MSC4310 and Element Call match a
decline against the rtc.notification event itself, so the caller kept
"Calling…" for the full 90 s after being told "Declined", and declines
from other clients (correctly referencing the notification) didn't stop
Lotus ringing on other devices.
- The incoming call keeps the notification's own event id; Reject relates
the decline to it.
- The "declined on another device" dismissal and declinedRing accept
either id (older Lotus builds referenced the membership).
- The missed-call line is skipped for a call we declined.
Verified headless (DM, Bob rings, Reject): the decline used to reference
the membership; it now references the notification. Unit test added.
Fixes#313
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
DeliveryStatus showed whatever mEvent.status was at the last timeline
re-render, and a local echo's status changes (SENDING → SENT → confirmed,
NOT_SENT → retried) only emit event-level events, so the last message in
a quiet room kept "Sending..." (or a stale "Failed to send. Click to
retry" after an outbox retry) indefinitely. It now re-renders on
MatrixEventEvent.Status / LocalEventIdReplaced.
Not browser-verified in this change (checks: tsc, unit tests, eslint).
Fixes#306
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
POST /rooms/{id}/upgrade copies only Synapse's fixed list of state, so an
upgrade silently dropped everything Lotus keeps in room state: an
audio-only voice channel became camera/screenshare-allowed, user limits
and bitrate caps vanished, the soundboard pack was lost, and a room set
to delete messages after 7 days kept new ones forever.
RoomUpgrade now reads the client-owned state first (utils/roomUpgradeState,
unit-tested: io.lotus.*, m.room.retention, role tags, room emoji packs)
and re-creates it in the replacement room; a partial copy is reported in
the dialog instead of closing silently.
Verified headless through the UI (v10 room with voice limit, quality
policy, a soundboard pack and retention → upgrade): before, none
survived; now all four are in the new room.
Follow-up (not here): for spaces, the children's m.space.parent still
point at the old space.
Fixes#327
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
useNotificationsQuiet evaluated the quiet-hours window (and the
"Pause notifications" expiry) only at render time, and the long-lived
message/invite notifiers rarely re-render, so someone leaving Lotus open
overnight with quiet hours 22:00–07:00 got every sound all night (and a
30-minute pause could outlast itself). The hook now re-evaluates every
30 s.
Verified headless (notifications granted, window unfocused): quiet
hours set to start ~75 s after the app opened; a DM before the window
notified; one inside the window, with the app left alone: before it
still notified, now it's silent.
Fixes#325
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
- #319: logout wiped the drafts, then reloaded, and the reload's pagehide
made RoomInput flush the open composer's (decrypted) draft straight back
to localStorage. clearPlaintextCaches now sets a module flag and
persistDraft stops writing once it's set (resets with the new page).
- #296: the device-wide encrypted-search opt-in (searchCacheEnabled)
survived logout, so the next account started indexing decrypted text
without opting in. It's now cleared, along with the SDK's per-user
filter key.
Verified headless with a throwaway account (search opt-in on, "SECRET
half-typed reply" in the composer, Settings → Log out): before, the draft
and searchCacheEnabled=true remained; now neither does. Unit test added.
Fixes#319Fixes#296
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
CallEmbed asked Element Call to send a call-start notification for every
non-DM call. In persistent voice channels and space rooms it carries an
@room mention: every member got a highlighted ping (and phone push)
whenever a moderator or admin joined first, and other clients showed it
as a missed call. Now only rooms that can ring (isRingableRoom: DMs and
invite-only groups outside a space) get ring/notification; everything
else passes sendNotificationType=none (element-call-embedded 0.25.0-lotus.23).
Verified with a local build of the fork: an admin joining the Voice Lounge
sends none, and Bob gets no new notification (before: +1 highlighted).
Fixes#295
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Since the switch to @fontsource-variable/inter 5 (May), the package
registers the family 'Inter Variable', but the app still asked for the
4.x name 'InterVariable', so every user got their platform's default
sans-serif (Arial / Helvetica / DejaVu) and Settings → Appearance →
font "Inter" did nothing.
--font-secondary (index.css) and the App font map now ask for
'Inter Variable' first, keeping 'InterVariable' as a fallback for a
locally installed old version.
Verified headless: before, 0 Inter faces loaded and
document.fonts.check('16px "Inter Variable"') false; now the face loads
and the check is true.
Fixes#320
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Several dialogs are <Box as="form"> with the header ✕ (folds IconButton,
no type attribute, so a submit button) as the form's first button. On
implicit submission, Enter in a text field "clicked" ✕, closing the dialog
and throwing away what was typed (a poll's question and answers, a
schedule, a report reason).
The header close/cancel IconButtons now have type="button": Create Poll,
Schedule Message, Report User, Report Room, Account Data editor, State
Event editor, device verification setup, and the message delete/report
dialogs.
Verified headless: in Create Poll, typing a question and pressing Enter
used to close the dialog; now it stays open with the question kept.
Fixes#315
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The media gallery and activity log register filtered timeline sets on the
room (getOrCreateFilteredTimelineSet); the room re-emits their events as
live RoomEvent.Timeline, so every matching event reached "new event in
the room" listeners twice for the rest of the session. The timeline
window shifted by an extra message per event (loaded messages vanished),
and notifications, unread counts, widget feeds and ring handling saw
duplicates.
isFromFilteredTimelineSet() (unit-tested) identifies emits from a filtered
set; the room timeline, roomToUnread, message notifications, clock-skew
sampling, call ring/decline handling and widget event feeds ignore them.
Verified headless (40 messages, every 5th a file; gallery opened; Bob
sends 3 files + 1 text): before, the first rendered message drifted from
the room start to "file15"; now it stays put.
Fixes#259
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
m.room.retention can be set to anything by anyone with state power. A
fat-fingered `max_lifetime: 1000` blanked the timeline for every Lotus
user and, for members with "Enforce Message Retention" on, the sweeper
redacted all their messages within a minute. The RETENTION_MIN_MS floor
existed but was never applied.
- effectiveRetentionMs(): anything under 10 min counts as off; used by
getRoomRetentionMs (sweeper) and the timeline's hiding.
- The sweeper never deletes under a policy less than 24 h old, so a
mistake or a malicious change can be reverted before anything is lost.
(Local hiding of expired messages still applies at once.)
Verified headless (the issue's repro, enforcement on): after
max_lifetime 1000, all 4 messages stay visible and none are redacted
after 45 s (before: timeline blank, all 4 redacted). Unit tests for the
floor and the grace window.
Fixes#298
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
initClient awaited navigator.storage.persist() before creating the client.
Chromium and WebKit answer at once, but Firefox shows a permission prompt
and leaves the promise pending until it's answered, so the app sat on
"Heating up" (no sync, no crypto, no IndexedDB) for every Firefox user who
didn't notice the prompt.
The request is still made (KE-1 only needs that), but startup waits at
most 1.5 s for the answer. The comment claiming persist() never prompts
is corrected.
Verified in Playwright Firefox with the prompt left unanswered
(persist() still pending): before, still "Heating up" after 20 s; now the
room list after 12 s.
Fixes#303
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
After a reload/restart within a ring's lifetime, the catch-up sync replays
the ring together with the caller's hang-up and our own decline, and the
MatrixRTC session can keep the stale join. The client rang again for up to
90 s and showed "1 Live" (call screen instead of chat) for minutes.
- Before ringing, check the room STATE (utils/callRing.ts, unit-tested):
no ring if we already declined that ring or are in the call; if the
caller has no active membership yet, re-check after 4 s (a real call's
membership can land just after its ring) and ring only if they joined.
- The "caller gone" dismissal also consults room state, not only the
possibly stale session.
- When sync becomes live (startup or reconnect), recompute every active
RTC session from the current state, clearing a stale "Live".
Verified headless (Bob's membership + ring in an invite-only room):
live ring rings; hang-up stops it; reload within the window: before
rang again + "1 Live", now no ring and no "Live"; a ring that arrives
just before the membership still rings.
Fixes#312
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Follow-up to #359: the personal room name now also shows in message-search
result groups and the search room filter (whose room search matches it
too), Inbox → Notifications, the threads list and the thread panel header.
Verified headless with general renamed to "Zebra Lounge": a global message
search and Inbox → Notifications both show "Zebra Lounge".
Fixes#336
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Joining a voice channel alone sends a call notification (m.rtc.notification)
to the room, and the timeline rendered every unanswered one as "Missed call
from X", so everyone in the channel saw a missed call from someone who had
just joined it. Ringing already excluded these rooms; the missed-call line
didn't.
utils/callRing.ts (unit-tested) holds the one rule for both: only DMs and
invite-only groups outside any space ring or show "Missed call"; voice
rooms, space channels, restricted and public rooms don't.
Verified headless with the same two notifications from Bob: the Voice
Lounge chat showed 2 "Missed call" lines before, 0 after; a small
invite-only group still shows its missed call.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
matrix.lotusguild.org turned off open registration (2026-10-16; accounts
come from SSO/LLDAP), so the login page's registration probe
(POST /_matrix/client/v3/register) now answers 403 "Registration has been
disabled" instead of 401 + UIA flows. The boot smoke test treated that
console line as unexpected and failed on every branch (seen on #360).
collectConsole now skips a 403 only when its source URL is the register
endpoint; any other 403 still fails the test.
Verified: boot.spec passes locally in chromium, webkit and iphone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
background-position can't run on the compositor, so every frame of the
animated chat backgrounds repainted the whole chat area: 51 to 72% main-thread
busy while idle in a room at 60 fps. The drifts are slow (3 to 22 px/s,
under a pixel per frame), so each animation now uses steps(): about 12
updates/s for fireflies, stars and aurora and about 24/s for rain and
pulse.
Measured headless (20 s idle in a room, CDP TaskDuration):
none 0.7% → 0.9%; rain 50.7 → 5.9; stars 71.2 → 4.5;
pulse 64.9 → 5.8; aurora 58.9 → 4.2; fireflies 71.7 → 3.8.
Aurora and pulse used ease-in-out; with steps() their motion is now
even-paced. Pause animations / reduced motion still strip the animation.
Fixes#337
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Two easy bypasses of the "this link goes somewhere else" check:
- Any whitespace in the link text disabled it, so "Log in at
chat.lotusguild.org" → evil.example wasn't flagged. hostsFromText now
scans each word for host-shaped tokens; a link is flagged if any named
host is a different site. Bare file names (main.py, report.pdf) and
version numbers are skipped inside sentences so ordinary links aren't
flagged.
- Shared-hosting suffixes counted as one site (alice.github.io vs
evil.github.io, *.pages.dev, netlify.app, vercel.app, …).
registrableDomain now treats a curated list of them as public suffixes
(longest match, incl. s3.amazonaws.com). Curated rather than the full
PSL to stay inside the bundle budget.
Verified with unit tests and headless on real messages: the four bypass
cases are flagged; the honest sentence, file/version text and "click
here" are not.
Fixes#335
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The personal room name (io.lotus.room_names) was only used by the
sidebar, header, intro and call bar; the quick switcher couldn't find a
room by it ("No match found") and showed the server name, as did the
forward/share pickers, toasts, OS notifications and reminders.
useRoomMeta gains useLocalRoomNames() / getLocalRoomName(); the quick
switcher matches both names and shows the personal one, the forward and
share pickers show and filter by it, and message/reminder notifications
use it for the room title.
Verified headless with general renamed to "Zebra Lounge": Ctrl+K "Zebra"
and "general" both list "Zebra Lounge"; the forward picker filtered by
"zebra" lists "Zebra Lounge".
Not covered here (follow-up): message-search result headers, the inbox
notifications page and the threads list still use the server name.
Refs #336
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
- Tab into a link in the hidden part used to scroll the overflow:hidden box
itself (top cut off, and it stayed that way); focus inside now expands
the message, and collapsing resets the scroll.
- "Read more" / "Show less" is announced as a toggle (aria-expanded,
aria-controls); the arrows are aria-hidden.
- The fade is a mask on the content instead of a gradient in a fixed
surface colour, so there's no band in bubble layout, on highlights or
over chat backgrounds.
- The cap is 20rem instead of 320px, so a larger font shows as many lines.
Verified headless on a 41-line message with a link on the last line:
aria-expanded false → focusing the link expands it (aria-expanded true,
no clipping); bubble layout fades into the bubble with no band.
Fixes#343
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Logout left lotus-per-user-volume (Element Call's per-person volume map:
whom the previous user called and turned up/down; same origin on the web)
and settings-sync-meta:<userId> (a key naming the account) in
localStorage. clearPlaintextCaches now removes both; device preferences
(`settings`) are kept as before.
Verified: after joining/leaving a call and logging out via Settings, both
keys are gone (unit test added).
Fixes#347
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Settings sync mirrored enterForNewline to io.lotus.settings, so turning it
on for a phone keyboard switched the desktop to Ctrl+Enter-to-send (and
vice versa). It is now in DEVICE_LOCAL_KEYS.
Verified with one account on a phone and a desktop profile: toggling it
on the phone leaves the desktop at Enter-to-send, and it is no longer in
the synced account data.
Fixes#348
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
Lotus encodes DND as status_msg "dnd". After switching back to Auto, the
heartbeat read the server's status, adopted "dnd" as the user's own and
kept re-sending it, so the custom status was lost and other clients kept
showing "dnd"; the Settings status field showed "dnd" too.
pickHeartbeatStatus (unit-tested) never adopts the "dnd" sentinel and
falls back to the device's real status; Profile neither mirrors it into
the local status nor shows it in the field.
Verified on the local Synapse: "🎮 Gaming" → DND (unavailable/dnd) → Auto
→ online/"🎮 Gaming" (before: online/"dnd").
Still open in #283: while in DND other clients show the literal "dnd";
that needs MSC3026 busy presence on the homeserver (matrix#19).
Refs #283
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
usePresenceUpdater sent a one-off PUT /presence, but the SDK kept
long-polling /sync with the default set_presence=online and Synapse treats
every sync as "online", so Invisible / Hide Online Status / Idle / DND
flipped back within seconds — a privacy setting that silently didn't work.
The presence helpers now also call mx.setSyncPresence(): offline for
Invisible / Hide Online Status, unavailable for Idle / DND (and auto-mode
idle / hidden tab), online otherwise. The PUT stays for the immediate
change.
Verified on the local Synapse, another account polling every 5 s for 60 s
after a reload in each mode: invisible → offline ×12, hidePresence →
offline ×12, idle → unavailable ×12, dnd → unavailable ×12, online →
online (currently_active) ×12; /sync carries set_presence=offline /
unavailable / online accordingly (before: online* throughout).
Fixes#282
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The composer's location button sent the exact position (6 decimals) from a
single click once the browser permission had been granted, discarding the
fix's accuracy, so a ±3.5 km Wi-Fi/IP fix was shown as exact.
- ShareLocationDialog: "Share your location?" with the coordinates and
±accuracy, a "Check on OpenStreetMap" link, and Cancel / Send
approximate (~1 km) / Send. Focus starts on Cancel; Esc cancels.
- utils/geoShare.ts (unit-tested): RFC 5870 geo URI with `;u=` accuracy;
approximate shares round to 2 decimals and never claim better than 1 km.
- Location cards show the ± when the event carries `u=`.
- The toolbar icon is a pin instead of the globe.
Verified headless (fix 40.748817,-73.985428 ±3500 m): the click opens the
dialog and sends nothing; Cancel/Esc send nothing; Send approximate →
geo:40.75,-73.99;u=3500; Send → geo:40.748817,-73.985428;u=3500; the card
reads "40.74882, -73.98543 (±3.5 km)".
Fixes#340
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
ReminderMonitor pushed an ordinary toast (auto-dismissed after 4 s) and
deleted the reminder at once, with no OS notification — one that came due
while the app was in the background or the user was away was lost.
Now the reminder toast is sticky; clicking it opens the message and
dismissing it closes it, and only then is the reminder deleted (an unseen
one returns on the next start). When the window isn't focused it also
raises an OS notification (respecting "show notifications" and quiet
hours; no decrypted E2EE text, per N106).
Verified headless (notifications granted, window unfocused): due at
t+24 s, the toast stayed up for the remaining 51 s, one OS notification
"Reminder :: call the dentist before 5", the reminder stayed in
io.lotus.reminders until dismissed, then was removed.
Fixes#338
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
useVoiceChannelFull compared the number of MatrixRTC memberships (one per
device) with io.lotus.voice_limit, so one person on desktop + phone, or a
crashed tab whose membership hadn't expired, filled two slots and locked
others out, while the server-side guard (which counts distinct users)
would have admitted them. utils/voiceChannelFull.ts (unit-tested) counts
distinct senders; the header/sidebar counter shows people too.
Verified on the local stack: limit 2, @aud_main in the call on two
devices; @aud_bob's prescreen now shows 1/2 with Join enabled (was
"Channel Full (2/2)"), matching the guard's 200.
Fixes#342
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
The pagehide keepalive write sent { presence: 'offline' } with no
status_msg, which Synapse treats as clearing it, so closing or reloading
any Lotus tab wiped the user's status for everyone (intermittently, as
the 10 s presence rate limit sometimes 429'd the write). It now sends the
current status along with the offline state.
Verified against the local Synapse: with a status set from another
device, two reloads and a tab close keep it (server shows
offline/"Gone fishing" after the close; before, the first reload wiped it).
Fixes#339
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5
`npm audit --audit-level=high --omit=dev` (Build & Quality Checks) started
failing on every branch after new advisories were published:
- source-map-js 1.2.1 → 1.2.2 (high, GHSA-68fv-2mgg-jv7q; transitive via
sanitize-html → postcss, in range — lockfile only)
- i18next-http-backend 4.0.0 → 4.0.2 (low, GHSA-xvq9-wjp8-hwqf)
- katex 0.16.47 → 0.18.11 (low, GHSA-238p-pmpm-9mq7; fixed in 0.18.2).
0.17/0.18's breaking changes are an internal __defineFunction API and
prefixed internal CSS classes; Lotus uses neither (KaTeX.tsx calls
renderToString and imports the package's own stylesheet).
Only these four lockfile entries change (plus katex's own CLI dependency
commander 8 → 15). npm audit --omit=dev: 0 vulnerabilities.
Verified headless against Vite with a fresh dep cache (katex 0.18.11
served): inline and display math render (\frac, \sqrt, \sum with limits),
KaTeX fonts load, no page errors; translations still load (en strings,
no raw keys). tsc clean, 1319 unit tests pass, eslint 0 errors (36
warnings, unchanged), prettier clean, production build OK.
Fixes#345
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Giphy and Tenor link previews forced every GIF into the same box — full
card width, max 200px tall, object-fit: cover — whatever its shape: a
square GIF lost half its height, a portrait one three quarters (only the
middle band showed, so captions and faces at the top/bottom were cut), and
a 100×80 GIF was blown up 4×.
utils/gifPreviewSize.ts (unit-tested) fits the GIF inside 400×320 from
og:image:width/height, keeping its aspect ratio and enlarging small GIFs
at most 2×. The card centres the GIF on the surface colour with
object-fit: contain, reserves the box before it loads, and the GIF badge
sits on the GIF's corner. Without dimensions it shows at natural size,
capped by the card. The >10 MB thumbnail fallback is requested at 2× the
box (800×640) so it isn't upscaled.
Verified headless (fixture previews, desktop / Pixel 7 / 320px phone):
480×200 → 398×166, 300×300 → 320×320 (292×292 on 320px), 240×480 →
160×320, 100×80 → 200×160, all 100% visible; the no-dimension fallback
and narrow cards keep the ratio. tsc clean, 1325 unit tests pass, eslint
0 errors (36 warnings, unchanged), prettier clean.
Fixes#341
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
One "security" strip, in the same top slot and style as the status banner
(#124), for this device, in priority order:
- "Verify this device" — cross-signing exists but this device isn't
verified (can't unlock backed-up history, untrusted to others). First,
because verifying with the recovery key also connects the backup.
- "Connect this device to your key backup" — a backup exists, this device
isn't using it.
- "Set up key backup" — no backup at all (includes accounts without
cross-signing; the setup flow does both).
The button opens Settings → Devices (existing flows); "Not now" snoozes.
Rules: nothing in a device's first 24 h; "Not now" snoozes that nudge 7
days; 3 dismissals stop it; never on a healthy device; waits until sync
has settled (never under "Connecting…"); outage/maintenance/connection
strips win. Per-device record in localStorage, wiped on logout. Mounted
inside the Matrix client context and an error boundary (an early version
outside the context crashed the app on load — caught before pushing).
Also: the status strip wraps its text on phones instead of truncating it
when there's no Details button (benefits #124's strips too).
Design approved on #123 (real-client screenshots there). Tests: 4 unit
(the #123 state table, loading, timing, stored record); e2e: nothing during
the grace period, "Set up key backup" → Settings → Devices, "Not now"
holds across a reload. Unit 1319, Playwright 29 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
`sudo pacman -U <url>` also fetches `<url>.sig` and failed (404) on
CachyOS: remote packages fall under RemoteFileSigLevel (signature
required) and we don't sign the package. A downloaded file installs under
LocalFileSigLevel (optional): `curl -LO <url> && sudo pacman -U ./…`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
npm audit --omit=dev started failing every PR: brace-expansion <=1.1.20
(via @eslint/eslintrc → minimatch 3) has three high-severity DoS
advisories. Lockfile-only patch bump; nothing else changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Reported on CachyOS: "Check for Updates → The update downloaded but
couldn't be installed … Permission denied (os error 13) at path
/usr/bin/tauri_current_app…". The app was installed from the Arch package;
Tauri's Linux updater can only replace an AppImage.
With cinny-desktop's new update_install_kind command:
- pacman / deb installs: the toast says the update is available and opens
Settings → General → App Updates, which shows the package-manager command
(`sudo pacman -U …pkg.tar.zst`, or the .deb + `sudo apt install`) with
Copy command and Download package — no Install & Restart that can't work.
"Copied" only when the clipboard write actually succeeded.
- other distros: a link to the downloads page.
- Windows, AppImage, and desktop builds without the command: unchanged
in-app update.
- a native "package-managed" refusal shows the same help.
Tests: unit (kinds, commands, refusal detection); in the real client with a
simulated desktop bridge: pacman → toast + Settings command/buttons,
install never attempted; older desktop → in-app flow as before. Unit 1321,
Playwright 26 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
When the homeserver, voice calls or sign-in break, or maintenance is under
way, say so in the app — driven by the Kuma status page
(isitup.lotusguild.org/status/matrix), managed from Kuma's UI. The client
asks Kuma directly (not via our servers) so it still hears "the server is
down" when our servers can't tell it.
- config.json `statusPages`, keyed by homeserver: users of other servers
never contact Kuma.
- utils/kumaStatus.ts (pure, unit-tested): parse Kuma 2.x's public JSON;
a group is down when any monitor fails two checks in a row (down+down or
pending+down); maintenance = windows under way; announcements = incidents.
One strip at a time: server down (connection lost AND Kuma confirms) >
server having problems > maintenance > calls down > announcement;
sign-in problems on the login screen only.
- Wording about the user's own connection: "Connection lost … our status
checks say the server is up, so it may be your internet connection" ONLY
when Kuma checked the server after this client's connection dropped and
it passed; a stale "up" (Kuma needs a minute or two to notice an outage)
keeps the plain "Connection Lost!".
- useServerStatus: polls only while visible; 5 min, 60 s while something is
wrong or the connection is lost, at once when it drops; backoff; any
failure = no banner (Kuma being unreachable never looks like Matrix
being down); GET only, no cookies.
- UI in the existing banner slot and style (ContainerColor/Line like the
sync and clock banners); Details expands; dismiss for calls-down and
announcements (an edited announcement comes back); calls-down note above
Join; phone: one line + Details.
Needs the CSP connect-src to allow https://isitup.lotusguild.org (matrix
repo) before it can fetch in production; until then it fails quiet.
Tests: 15 unit tests (live page layout, two-check rule, any-monitor rule,
unknown/garbage, UTC beat times, stale-vs-fresh "up", priorities, login vs
client, maintenance, announcements + dismiss/edit); e2e (fixtures for Kuma):
calls-down strip + dismiss across reload, other homeservers make no
requests, Kuma 500 → nothing, lost connection + Kuma down → critical strip
instead of "Connection Lost", + fresh "up" → "may be your connection",
+ stale "up" → plain "Connection Lost". Unit 1315, Playwright 26 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Step 1 of moving the desktop app's login out of the webview's plaintext
localStorage: keep a verified copy of the tokens in the OS keychain
(Windows Credential Manager, via cinny-desktop's new secure_session_*
commands). The session is still read from localStorage exactly as before,
so nothing about login changes and a keychain problem can't log anyone out.
Step 2 (a later release, once this has run on real installs) switches reads
to the keychain and drops the tokens from localStorage.
- sessions.ts: onSessionPersisted — listeners told about every session
write (login, token rotation) and removal (logout); a throwing listener
can't break the write.
- keychainMirror.ts: desktop only. Mirrors userId/deviceId/accessToken/
refreshToken (not the rest of the session); reads first and writes only
when the copy differs, then verifies by reading back; serialized, 5 s
timeouts; no session → clear (also covers a logout whose reload beat the
clear). Every failure is a status, never an exception. A desktop build
without the commands reads as "unsupported", so this can ship before the
desktop side.
- Settings → General (desktop): "Login in the system keychain" status.
Tested: unit tests (fake keychain: store, no rewrite when current, rotation,
clear, unsupported, missing commands, denied write, read-back mismatch,
timeout); a simulated desktop app with a fake keychain, 14/14 (login
mirrors only the secrets, Settings status, reload verifies without
rewriting, logout clears, an existing session is mirrored after upgrade,
Linux/denied show an honest status and stay logged in); the real Linux
desktop binary (commands answer "unsupported", login unaffected, Settings
says so). Unit 1295 pass, Playwright 20 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The desktop app loads the bundled Element Call page from its own origin
(http://localhost:<port>), so the call frame can read the app's storage
(login token) and DOM — the hole #43 closed on the web by moving the page
to call.chat.lotusguild.org.
The desktop's local server can also answer on http://127.0.0.1:<port>: the
same server and bundle, but a different origin (and still a secure
context). resolveDesktopCallPageUrl loads the bundled page from there when
the desktop config sets `desktopCallOrigin`:
- only a loopback http origin on the SAME port as the app, no path, query
or credentials;
- only when the app itself runs on http://localhost (release builds; debug
builds on tauri:// keep the same-origin page);
- unset (every desktop build until cinny-desktop opts in, together with the
server bind, CSP and permission changes it needs): unchanged.
The web app is unchanged (elementCallUrl as before).
Tested in a simulated desktop app (Tauri bridge stub + the desktop
config.json, served on localhost and 127.0.0.1) against a local Synapse +
LiveKit, two users: call page from http://127.0.0.1:<port>, parentUrl =
the app origin; the frame gets SecurityError on parent.localStorage and
parent.document (same-origin control: readable); join, speaking indicator,
mic off/on, screenshare start/stop, layout switch and hang-up all work, no
page errors — 12/12 in 5 of 6 runs, like the same-origin control (3 of 4;
the misses on both sides were the local LiveKit connection).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
"Ahead" is now reported only once it has held for a minute of fresh
samples (a stalled server delivers late and reads as ahead). The test sends
its ticks, checks nothing is shown yet, fast-forwards the page clock past a
minute and sends two more.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Incident 2026-09-29: the homeserver's host ran out of memory and stalled for
~2 minutes. The /sync that finally went out carried events whose `age` was
computed ~30 s before it arrived, so every client showed "Your computer's
clock is 30 seconds ahead of the server" while the real problem was the
server (all host clocks were within 0.25 s the whole evening).
The skew estimate was the median of the last 5 samples, and a sample is
local skew + delivery delay, so one late /sync with a handful of events
tripped it.
- Estimate = the LOWEST sample of the last 5 minutes: delay only ever adds,
so the fastest-delivered event is the truest.
- "Behind" (which a delay can't cause) is reported as soon as there are 3
samples, like before. "Ahead" must hold across samples received at least
a minute apart, so a single late burst never trips it.
- Samples are aged on the monotonic clock, and a change of the local clock
(someone fixing it) resets the measurement, so the warning clears at once.
- Only events stamped by our own homeserver are sampled: a federated event's
origin_server_ts is the other server's clock.
- Wording: "This device's clock is … Voice calls and encrypted messages can
fail until it's corrected." / call bar "Device clock … : calls may fail"
(was "will fail").
Unit tests: the incident (late burst after normal traffic, and a fresh
client whose first samples are all late), mixed slow/fast deliveries,
ahead only after a minute, behind at once, hysteresis, clock fixed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Until now a send that failed (offline, homeserver down, a blip) went
straight to "Failed to send": nothing retried it, and a reload dropped it
without trace (chronological pending ordering keeps local echoes in memory
only).
- Outbox (utils/outbox.ts + features/outbox/OutboxFeature): own message
sends (text, stickers, reactions, polls; not call signalling or
redactions) are mirrored to localStorage from their first local echo until
the server confirms them or the user cancels.
- After a reload they come back as local echoes via room.addPendingEvent,
same shape as the SDK's own. Recent ones (< 1 h) are sent again with the
same txnId; older ones come back as "Failed to send" for the user to
retry or cancel. Ones the server already has (transaction id seen in
/sync) are dropped, so no duplicates.
- Retries: network failures (ConnectionError, 408/429/5xx) are re-sent when
the connection returns (sync recovers or the browser goes back online),
and after a blip while online (5 s, backing off, max 10 per message).
Oldest first, in order per room. 4xx / consent / encryption failures are
left to the user.
- UI: a network failure while offline shows a clock, "Queued. Will send
when you're back online" (thread view too), not the red ✕. The ✕ is now
a button: click to retry.
- Logout wipes the outbox with the other plaintext caches (the content is
decrypted, like drafts).
Tested end to end against a local Synapse (Chromium): offline → queued →
sent once on reconnect; homeserver unreachable → queued → sent once; failed
send → reload → sent once and shown once; server accepted but response lost
→ reload → no duplicate; 2 h old entry → failed, not sent, click ✕ → sent;
cancel → gone after reload; encrypted room → restored message goes out as
m.room.encrypted with no plaintext and decrypts; one-off failure retried by
itself in ~5 s; no page errors. Unit tests for the pure parts; Playwright
20 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Lotus permalinks (#130) were built from window.location.origin. In the
desktop app that's the local tauri-plugin-localhost server (hash-routed), so
"Copy Lotus Link" copied e.g. http://localhost:…/#/home/!room…, which works
for nobody else. And the link recogniser only knew that local base, so a real
https://chat.lotusguild.org/home/… link in a message opened the browser
instead of the room.
- useLotusShareBase: in the desktop app, links for other people use config
`webAppUrl` (https only, set by cinny-desktop #23) in web path-routing
form; otherwise the origin, as before. Used by "Copy Lotus Link" on
messages, the space menu and space tabs.
- The recogniser accepts several bases: the origin, plus `webAppUrl` in the
desktop app.
- The web app is unchanged.
Verified with a simulated desktop (Tauri bridge + webAppUrl) against a local
Synapse. Copied links are https://chat.lotusguild.org/home/<room>/<event> and
https://chat.lotusguild.org/<space>. A public link in a message renders as
the room pill and clicking it opens the room in-app, with nothing sent to the
system browser. The web app still copies origin links. Unit tests for the
base selection; Playwright 20 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Rendering only; the vote logic (tally, optimistic votes, end poll, keyboard
radiogroup, voters) is unchanged.
- Every row shows "N votes · P%". Multiple choice looked broken as
"100% / 100%" with no counts; the footer now says "voters" there.
- Results are a thin progress bar under each answer: accent for your pick,
success for the winner, neutral otherwise. The old full-row fill was the
same grey as the row, so a 100% answer just looked disabled.
- Radio and checkbox indicators are 18px with a 2px border in a colour mixed
from the theme's text colour. Primary.ContainerLine was nearly invisible,
especially in dark themes.
- The winner shows a star and "Winner" instead of a second check mark.
- Bordered card, 460px wide (max 100%, same width for every poll). The header
is now "Poll · Single choice | Pick up to N | Results hidden until the end |
Final results", replacing the letter-spaced "◉ POLL · …" line.
- Footer: plain muted text that wraps, and chips ("Who voted", "End poll")
that never wrap. Previously the chip label broke onto two lines inside a
one-line chip and spilled out of it. Reproduced before/after in Lotus
Terminal at 150% zoom, a 360px phone, and dark at 125%.
- Lotus Terminal theme: data-winner gets its own green border rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The Linux desktop app runs on WebKitGTK, which ships without WebRTC (2.52
has no RTCPeerConnection; 2.54 disables it outright pending a libwebrtc
backend around 2.56), so calls can't work there. Until now the call button
just disappeared, the call room said "Your browser does not support WebRTC"
with Join disabled, and an incoming call couldn't be answered.
In the desktop app (isTauri) without WebRTC:
- call rooms: "Calls aren't available in the desktop app on Linux yet: its
web engine has no WebRTC" + an "Open in browser" button;
- incoming-call overlay: the same, with "Answer in browser";
- room header: the call button stays, and opens the room in the browser.
The link is the room in the web app (config.json `webAppUrl`, https only,
new key); the user presses Join there. Deliberately not an auto-join link:
a crafted URL must not be able to join a call and open someone's mic. It
opens through the desktop's new-window handler (web/mail schemes only → the
system browser). Without `webAppUrl` the explanation shows with no button;
browsers without WebRTC keep the old message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
lotus.22 draws the "Share your screen?" prompt inside the call frame on
request and drops the corner button. Verified against the published
package: no-delegation (Firefox path) cross- and same-origin — bar button,
in-frame prompt, Cancel, Share → screenshare tracks, bar Stop; Chromium
unchanged; picture-in-picture prompt fits and Share works; room policy
hides the button and refuses the prompt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
With a fork that reports `screensharePrompt` (element-call lotus-screenshare-
prompt), the call bar and status bar keep their screenshare button on every
engine. Where the click can't be delegated, starting asks the fork to show
"Share your screen?" inside the call frame (io.lotus.prompt_screenshare)
instead of our own confirm; its Share click starts the share. Stopping works
from the bar as before (no click needed in the frame). Chromium is unchanged.
- useScreenshareMode: hidden (older fork's corner button) | prompt | direct.
- The room's call policy is also pushed to the fork in prompt mode, so the
prompt never opens where sharing is forbidden.
- Picture-in-picture: the "Return to call" overlay covers the frame; while
the fork reports the prompt open it lets clicks through to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
lotus.21 checks widget messages against the host's origin (works same- and
cross-origin) and accepts soundboard clip bytes. Same-origin behaviour is
unchanged; verified against the published package: join, both screenshare
paths, PTT/deafen in the frame, layout/reactions/settings, speaking and mic
level, soundboard, avatars, muted-speech warning, per-person volume, and the
foreign-frame spoof stays blocked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Groundwork for serving the call page from its own origin
(call.chat.lotusguild.org). Inert until config.json sets `elementCallUrl`:
without it the bundled same-origin page is used exactly as today.
- callPageUrl: resolves `elementCallUrl` — absolute https only (http only on
localhost for development); anything else, and the desktop app, fall back
to the bundled page so a bad value can't break calls. Set once from the
loaded client config.
- CallEmbed builds the widget URL from it; the widget origin (used by the
message guard and Capability Delegation) follows automatically.
- Soundboard: a host blob: URL can't be fetched from another origin, so
io.lotus.inject_audio now also carries the clip's bytes (`audio`). Forks
that predate it ignore the field and use `url`, so this is safe on the
released fork.
Needs element-call's lotus-call-origin branch (host-origin message check +
inject_audio bytes) released and pinned before `elementCallUrl` is set.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
matrix-widget-api's host transport handled a message from ANY window on the
page as long as it carried the widget's id; its strictOriginCheck only
compares with the host's own origin and is off by default. The call's id is
the fixed 'call-embed', so any other frame (a room widget, a URL-preview
embed) could post fromWidget actions as the call. Reproduced locally: an
opaque-origin frame posting one io.lotus.hotkey keydown for the PTT key
turned a push-to-talk user's mic on ("● Live").
restrictWidgetMessages() swaps each ClientWidgetApi transport's listener
for one that requires ev.source === the widget iframe's window and
ev.origin === the widget's origin. Applied to the call and to room widgets
(so one widget can't impersonate another). Verified: the spoof no longer
opens the mic; PTT/deafen from inside the call, screenshare, speaking
indicator and room widgets (capability prompt, send, live events) unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Firefox, Safari and the WebKitGTK desktop app can't hand the user's click to
the call frame (no Capability Delegation), and getDisplayMedia needs it. The
host used to click EC's hidden footer button through the DOM instead, which
dies with same-origin. Now (pins element-call-embedded 0.25.0-lotus.20):
- those engines get `lotusFrameScreenshare`, the fork shows EC's own
screenshare button in the frame, and the call bar and status bar hide
theirs once controls_state reports `frameScreenshare`; the
screenshare-audio mute stays;
- the room's call policy is pushed with io.lotus.set_frame_screenshare, so
the frame button hides where the server would refuse a share, like ours;
- Chromium keeps the delegated io.lotus.set_screenshare from the host bar.
Removed the fallbacks for forks older than lotus.14, which read or clicked
EC's DOM: the screenshare/layout/settings/reactions/leave button lookups
and their MutationObservers, the frame-window hotkey binding, and the
speaking/muted tile scrape in useCallSpeakers (io.lotus.call_state is the
only source now). getCallDocument is gone; the host's only handle on the
frame is postMessage.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The SDK routes thread replies out of every room timeline set, the gallery's
detached one included, into the room's Thread objects, so photos posted in
a thread never reached the gallery. The gallery now merges media from the
loaded threads (deduped, newest first) and refreshes on ThreadEvent.NewReply,
waiting for decryption in encrypted rooms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
markAsRead ran on every room visit (timeline at the bottom and focused) and
sent a threaded receipt for every unread thread, so a reply in a thread you
started or replied in lost its unread badge the moment you glanced at the
room, without opening the thread.
Reads from just viewing the timeline are now "passive":
- threads you follow (started, replied in, or were mentioned in) stay unread
until their panel is opened; other threads are still cleared so they don't
keep the room dot lit forever;
- while a followed thread has an unread reply, the main receipt is scoped to
the main timeline instead of unthreaded, because an unthreaded receipt
also reads every older thread reply (the next main message would clear the
thread anyway). The check also asks whether the latest reply is read, since
the thread's count lags when the reply and a main message share a sync;
- the thread open in the panel is skipped, as the panel sends its own
receipt (was two identical receipts per reply).
Explicit "mark as read" (room menu, Escape, bulk actions) still clears
everything. Unit tests for each rule plus a local-homeserver e2e that checks
the server's per-thread count survives a reply + newer main message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Thread images opened the room media lightbox, which looks the event up in
the room's detached media timeline. Thread replies never reach that
timeline, so a thread image always showed "1 / 1" with no prev/next, after
paging the room's media up to six times. The thread panel now builds the
viewer's items from its own root + loaded replies, and "Go to message"
scrolls the thread panel instead of the room.
The viewer gains a "Copy image" button: fetches the displayed media (blob
URL for E2EE, authenticated URL otherwise), re-encodes to PNG when needed,
and writes it via ClipboardItem with a promise so Safari keeps the click's
user activation. Hidden where ClipboardItem is missing. No "open in new
tab": an E2EE blob URL is revoked when the viewer closes and authenticated
media 401s in a bare tab.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Pins @lotusguild/element-call-embedded 0.25.0-lotus.19.
- Styles: the fork now hides its own footer (`lotusHostControls=1`) and
sets its root color-scheme from the theme, so the host no longer
injects `#lotus-ec-styles` or sets inline styles on EC's DOM. The two
other injected rules matched nothing in EC 0.25 (dead). The
transparent background was already the fork's (`lotusTransparent`).
- Hotkeys: PTT / deafen keys pressed with focus inside the call frame now
arrive as `io.lotus.hotkey` (the host sends the codes via
`io.lotus.set_hotkeys`), instead of listeners on the frame's window.
The window binding stays only for a fork that doesn't report `hotkeys`.
- Fixes (with lotus.19): pressing the deafen key M with focus in the call
also hit EC's own "M = toggle mic" shortcut, so the first press turned
the mic ON instead of deafening — even in push-to-talk mode.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The call bar's Start/Stop Screenshare no longer clicks EC's hidden button
on Chromium (incl. WebView2): it sends io.lotus.set_screenshare with
postMessage `{ delegate: "display-capture" }` inside the user's click, so
the frame can call getDisplayMedia on engines that require the click.
matrix-widget-api has no postMessage options, so its sendInternal is
swapped for that one synchronous send; delegation needs the frame's real
origin, not `*`.
Firefox, Safari and WebKitGTK (Linux desktop) have no delegation and
still click EC's button (needs same-origin, which is still on). Gated on
the fork reporting `screenshareAction` in controls_state.
Pins @lotusguild/element-call-embedded 0.25.0-lotus.17.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Probed how real providers fail before picking signals:
- X renders an EMPTY frame for a deleted/private/suspended post and says
so only via postMessage `twttr.private.no_results`. The post embed now
swaps to "This post isn't available…" with an "Open on X" link.
- A hung frame never fires `load`. After 20 s every player (media, rich
posts, TikTok, Steam widget, X) overlays "This embed is taking too long
to load" with Retry (remounts the iframe) and "Open on <site>". A late
`load` clears it.
- Instagram and Bluesky show their own "removed / not found" page, and a
refused request still fires `load` (browser error page), so neither needs
or can use a guess. A missing height message is not treated as failure.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The reaction button's aria-label used getShortcodeFor(), which returns
undefined until the lazily loaded emoji data arrives. The same button read
"🎉 reaction, 1 person" on first render and "tada reaction, 2 people" after
any later re-render. It now always uses the emoji itself (screen readers
speak it by its proper name, e.g. "party popper").
Custom (mxc) emoji were labelled just "custom emoji"; they now use the
shortcode carried on the reaction event (":lotus_blob: reaction").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Adds 530 decorations in 19 new categories (631 total). Decorations that
contain Discord branding (the Clyde-visor helmets) are excluded.
The decorations are ~1 MB animated PNGs, so a picker that rendered every
one would pull ~590 MB while scrolling. The picker now:
- shows static 144px WebP thumbnails (~9.5 KB each, `thumbs/` on the CDN)
and loads the animated file only on hover, focus or selection, with a
fallback to the full file if a thumbnail is missing;
- mounts one category at a time behind tabs, plus a name search across
all categories.
scripts/makeDecorationThumbs.py builds the thumbnails from the busiest
frame of each animation (many start on an empty frame).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
d2503332 made Enter pick the mention suggestion, which also caught
"/kick @alice:server" + Enter (the e2e "/kick failure is reported" test):
typing a complete ID opens the mention list too, and Enter picked it instead
of running the command. Enter now picks only while a partial name is being
typed (no ':' in the query); a complete @user:server or #alias:server sends
as typed, like before.
Chromium e2e 19 passed; "@bo" + Enter still picks, ":smi" + Enter still
sends.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
With the mention list open, pressing Enter sent the half-typed text
("hey @bo") instead of inserting the highlighted person — found while
investigating the "composer hit a snag" report; only Tab picked. Enter now
picks the top suggestion for people, rooms and commands, as in Discord,
Slack and Element.
useAutocompleteEnter listens in the window capture phase (so it runs before
the composer's submit handler) and only while the list has suggestions —
with no command match Enter still sends. Emoji suggestions are deliberately
left on Tab only, so "lol :p" + Enter still sends rather than inserting an
emoji. IME composition and Shift/Ctrl+Enter are untouched. Tab now inserts
the member's display name, same as a click.
Verified: "@bo" + Enter inserts the mention and sends nothing; the next
Enter sends "hey @bob hi" with m.mentions; ":smi" + Enter still sends.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Reported on chat.lotusguild.org: picking a name from the @-mention list
showed "The message composer hit a snag." — clicking OK continued with the
draft intact. The composer's known failure mode is a transient render error
right after an autocomplete insert (slate-react's DOM selection sync racing
the model; 477df4ae fixed one such path). It did not reproduce locally in 60
attempts (click / Enter / Tab, fast and slow, display names with spaces and
emoji, Chromium and WebKit), so this makes it harmless and diagnosable:
- ComposerErrorBoundary: the first failure clears the selection and remounts
the composer immediately, with no notice (the draft is intact). A second
failure within 5 s shows the notice, now saying the draft is safe, with
"Reload composer" and "Copy details" (time, error, stack, component stack,
browser). Every failure is logged to the console as "[composer] render
error".
- Used for the room composer and, newly, the thread panel composer (which had
no boundary, so the same error took the whole panel down).
Verified with a temporary injected crash (not committed): one crash recovers
with the draft kept and no notice; two within 5 s show the notice, Copy
details copies the report, Reload composer restores the full draft.
Chromium e2e 19 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The GIF button needs a Giphy key in the client config. Without one (the
desktop app until now), switching "GIF Picker" on did nothing and nothing
said why. With no key the tile now explains it and the switch is disabled;
with a key it is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Jared felt the first redesign was too sparse for a full theme: most of the
time only two webs and a spider were on screen.
- Will-o'-the-wisps: six soft green/orange glows wandering and pulsing,
always present while animating (not in the still scene, where static glows
would sit on text).
- A witch on a broom crosses just below the header about once a minute
(redrawn so she reads as a seated figure, not a blob).
- Bats every 20 s (was 30), plus a faster swarm of eight every 60 s.
- The big spider now climbs up its thread and drops back with a bounce.
- A second, small spider on the left web; a tiny jack-o'-lantern with a
flickering candle hangs by the right web (desktop).
Rendered and inspected at the witch/swarm moments in light, dark and phone,
plus reduced; still 60 fps on vs off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Draft for human review (#239 must be approved and closed by a human).
- Removed the full-screen purple wash, fog and the blurred green "moon" that
sat over the member list and made it unreadable.
- Corner cobwebs (generated radial threads + sagging rings), thin and
readable in both themes, smaller on phones.
- A spider on a thread in the empty space under the member list (near the
top on phones), bobbing and swaying.
- Every 30 s a group of four bats flaps across the top of the screen (a real
bat silhouette; wing beat without flattening them into dashes).
- Every 50 s a small, see-through ghost rises and fades (outlined so it
shows on light themes; text stays readable through it).
- Theme-aware ink via --hw-ink/--hw-web: dark silhouettes on light themes,
pale dusk-violet on dark ones.
- Reduced motion: webs and the spider at rest. Settings swatch: night sky,
crescent moon with a bat across it, a corner web.
Rendered and inspected: light/dark/phone at the bat and ghost moments,
reduced on desktop and phone, the Settings swatch; 60 fps on vs off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The main chunk had crept to 356.9 kB gzip against a 349.9 kB budget (a
hard gate on pull requests, only a warning on pushes, so it went unnoticed
until PR #241). jsQR (camera scanning) and qrcode (drawing the QR) were
statically imported by the device-verification components, so every
startup loaded them. Both are now dynamic imports: jsQR when the scanner
opens (fetched in parallel with the camera permission), qrcode when a QR is
drawn.
Main chunk 356.9 → 302.5 kB gzip (largest chunk is now matrix-sdk at
304.7 kB, within budget); chromium e2e 19 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Draft for human review (#237 must be approved and closed by a human).
- Removed the full-screen amber wash, sun-shaft bands and vignette: they
turned the user's theme beige (muddy brown on dark) and lowered contrast
everywhere. The leaves carry the season on their own. Also removed the
dust motes (invisible specks) and the now-unused keyframes.
- Gust: every 42 s (first ~8 s after load) five leaves blow across the screen
together, rising and spinning, in ~5 s. Linear timing — an eased curve is
applied per keyframe segment and made them hang mid-air for a second.
- Softer flutter (leaves flattened into slivers mid-flip), stronger leaf
colour (opacity 0.58–0.73, less pale gold tip), brighter on dark themes.
- Reduced motion on phones: no icon rail there, so the resting leaves tuck
further off-edge and smaller instead of landing on the timeline icons.
- Settings swatch: new optional `preview` prop on seasonal overlays; Autumn
draws a warm backdrop with arranged leaves (the reduced scene rendered in
the tile as a black square with two edge fragments).
Rendered and inspected: light/dark/phone at 10/15/20 s, reduced on desktop
and phone, gust tracked frame by frame, the Settings swatch; 60 fps with the
theme on vs off (p95 frame 16.7 ms both).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Draft for human review (#237 must be approved and closed by a human).
- Bug: every falling leaf sat in a leaf-sized wrapper pinned at the top of
the screen with `contain: layout paint style`. Paint containment clips
children to that box, so a leaf stopped being drawn as soon as it fell out
of its first ~40 px: the animated theme showed almost nothing while the DOM
reported 10–16 leaves on screen. Now `contain: layout style`.
- Leaf shapes: a flag-style maple (bold lobes, stem) that stays recognisable
at 30 px (the old outline read as a star/splat), an oak with a stem, and an
aspen; gradient body with darker veins.
- Palette: crimson, pumpkin orange, gold, russet, burnt orange (was muted
tans that read as beige on the light theme).
- 12 leaves at 28–52 px, opacity 0.5–0.65 so text under a passing leaf stays
readable; brightened on the dark themes, where they sank into the wash.
- Reduced motion: resting leaves moved into empty layout areas (middle of
the left rail, under the member list), partly off-edge, instead of on the
sidebar, members, composer and corner controls.
Rendered and inspected at 10/15/20 s in light, dark, phone and reduced
motion (screenshots on #237).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Typed operators alongside the existing from:, each one setting the filter the
buttons already set (so typing and clicking end in the same state, shown in
the filter bar):
in:general in:"voice lounge" room by name / alias / id (unique partial
name ok); a room outside this page's scope
switches to all rooms
after:2026-09-01 before:7d date range; relative h/d/w/m/y
has:link | image | video | file contains-link / media filters
is:pinned pinned-only filter
- searchOperators.ts: pure parser + room resolver (unit-tested); quoted
values; keys case-insensitive; only at a word start, so URLs and times
aren't mistaken for operators; anything it doesn't understand stays in the
searched text with a notice listing the valid operators.
- Notices: no matching room; operators with no words or from: ("Add a word to
search for, or a from:@user.") — then the text is kept in the box.
- from:bob now resolves to a known @bob:server (this homeserver first); the
old path sent "@bob", which matched nobody unless picked from autocomplete.
Verified in Chromium on a local Synapse: `raid in:"Thread Lab" from:bob`
sets rooms=!…Thread Lab, senders=@bob:localhost and finds bob's message;
dates, an unknown room, operators-only and has:gif give the expected params
and notices.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
While unmuted, the microphone button (call view and the bottom call bar)
shows three small green bars in the icon's corner driven by the fork's
io.lotus.mic_level (element-call v0.25.0-lotus.16), so you can see the mic is
actually picking you up. Nothing is drawn while muted or silent; the button's
label is unchanged (the bars are decorative).
Your own name no longer appears in the "… is speaking" line: that is now for
everyone else, and your voice is on the meter. Avatar speaking rings still
include you, and the "You're muted" talking-while-muted notice is unchanged.
Verified in a local call with a fake-tone mic: data-mic-level cycles 3/2/1
with the tone on both buttons, disappears on mute and returns on unmute; the
status line reads "bob is speaking..." only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
When a multi-line paste looks like code, a chip above the composer asks
"That looks like code. Format it as a code block (js)?" with Format as code /
No thanks. It never converts on its own; typing, No, or 8 s dismiss it, and
Settings → Editor → "Offer to Format Pasted Code" turns it off.
- utils/looksLikeCode.ts (pure, tested): ≥ 3 lines and ≥ 2 of indentation
with depth changes, statement terminators, operators + brackets, keywords,
monospace/<pre> clipboard HTML, or SQL clause lines. Prose guards: quoted
replies, URL lists, long sentence-punctuated lines, plain-word lines,
markdown lists. Language guess only when fairly sure (js/ts/python/sql/
rust/c/php).
- Accept rebuilds the pasted paragraphs as one code block from their plain
text (code lines hold text only) and leaves the caret after it.
- Code blocks carry an optional lang → <code class="language-js"> (whitelisted
identifier only).
Verified in Chromium: pasting a JS function shows the chip; Format as code →
sent formatted_body is <pre><code class="language-js"> with indentation intact;
a three-line prose paste shows no chip; typing after a paste dismisses it.
Unit tests: 14 detector fixtures + 2 output tests; chromium e2e 19 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Timeline images were focusable (tabIndex 0) but only opened on click, so a
keyboard user could Tab to one and do nothing with it, and screen readers
heard a plain image. They are now role="button" labelled "View image: <name>"
and open the viewer on Enter or Space. Focus already returns to the image when
the viewer closes (checked for mouse and keyboard).
Verified in Chromium: Tab → Enter opens the viewer, Escape returns focus to
"View image: red.png", Space opens it again; chromium e2e suite 19 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The SDK gives a thread root its own id as threadRootId, so the
`!isThreadedMessage` guard hid the thread action on exactly the message that
has a thread; the "N replies" chip was the only way in. Roots now get the
action in both the hover bar and the ⋮ menu, labelled "View thread" /
"View Thread (14 replies)" with the thread icon; other messages keep
"Reply in Thread". Same handler, which opens the thread panel.
Verified in Chromium: on a 14-reply root both entries show and open the panel;
an ordinary message still shows "Reply in thread".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Windows' emoji font has no flag glyphs, so 🇺🇸 (:flag_us:) showed as the
letters "US". Two fixes:
- A TwemojiFlags @font-face over the bundled Twemoji file, limited by
unicode-range to the regional-indicator letters (U+1F1E6–1F1FF), is put in
the font stack on Windows only. Every other emoji stays native, other
platforms keep their own flags, and the file is fetched only when a flag is
on screen.
- The Appearance font setting overwrote --font-secondary with a stack that
had no emoji families at all, so neither this nor the existing "Twitter
emoji" switch reached message text. The font map now keeps
var(--font-flags), var(--font-emoji) before the generic family.
Verified in Chromium via CDP platform fonts: with a Windows user agent 🇺🇸 is
drawn by "Twemoji Mozilla" while 😀 stays on the system emoji font; with a
Linux user agent both stay native.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The thread header subtitle now says who started the thread and when
(64af736c); the snapshot's room-name line is a pattern for that.
CI run 2104's Playwright job failed on this; the chromium suite passes
locally (19 passed, 3 skipped).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Once the fork reports io.lotus.controls_state (element-call
v0.25.0-lotus.14), the call bar's Spotlight/Grid View, Reactions and Settings
send io.lotus.set_layout / toggle_reactions / open_settings instead of
clicking Element Call's hidden buttons, and the call bar's screenshare and
layout state come from that report instead of a MutationObserver on EC's DOM.
Older forks never send the report and keep the DOM path.
Screensharing itself is still started/stopped via EC's button: it needs the
user's click to reach the frame (Capability Delegation, step 3 of #43).
Verified in a two-person local call (Synapse + LiveKit) against a locally
built fork: controls_state arrives on join; Spotlight View / Grid View flip
EC's layout and the reported layout; Reactions opens EC's reactions menu;
Settings opens EC's settings modal; the four actions go over postMessage;
starting and stopping a screenshare flips the host's button via the report.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A row in the Threads list only opened the thread panel; there was no way to
get to the root message in the room's timeline. Each row now has a small
"Go to message" button (a sibling of the row button, laid over its corner —
no nested buttons) that navigates the room to the root event. On a phone,
where the list covers the timeline, it also closes the list, and the button
is the bigger touch size.
Verified in Chromium: desktop jumps to the root with the list still open;
Pixel 7 jumps and closes the list (34 px target).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The subtitle was only the room name. It now reads e.g. "Thread Lab · started
by alice, Fri 09:18 PM" (the user's clock/date preferences via the #139
formatter), which is what you want to know when arriving from the Threads
list. Verified in the 360 px desktop panel: fits without clipping.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A captioned image/video (MSC2530: `filename` set and `body` differs) put the
caption in the header, truncated to one line, so a long caption was
unreadable. The header now shows the file name; the caption is shown in full
under the media (wrapped, scrolls past 25vh). The image alt text uses the
caption, and Download now saves under the real file name instead of the
caption.
Verified in Chromium: header "lake-sunset.png", the 150-character caption
fully visible below the image.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The viewer's arrows were 32 px at the screen edges on a phone, under the
44 px target the rest of the mobile work uses. On a coarse pointer they use
the next IconButton size (48 px measured on a Pixel 7); desktop is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
- sendRoomMessage (composer's fire-and-forget sends: text, location, voice,
files, GIFs) and sendRoomEvent (polls, poll responses/ends, forwards,
reactions, edits) in utils/room.ts carry the one cast each needs
(`keyof TimelineEvents` / `RoomMessageEventContent`, no `any`).
- sendRoomMessage also swallows the rejected promise: a failed send already
shows on the local echo (Failed to send + Retry, or the consent prompt), so
it no longer surfaces as an unhandled error in the console.
- getAccountData narrows to `keyof AccountDataEvents`; ForwardMessageDialog's
guard now narrows `contentToSend` itself (same behaviour).
- `as any` 39 → 23; eslint warnings 46 → 36, ratchet tightened to 36.
Verified in Chromium on a local Synapse: a text message, a quick reaction and
an edit all reach the server with the right content; a consent-blocked send
no longer logs an unhandled MatrixError. 1219 unit tests pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Zoom on a phone was only the +/- buttons; the pinch people try first did
nothing. usePinchZoom tracks a two-finger gesture on the viewer's media area
and scales from the zoom it started at (clamped 1×–5×, snapping back to
exactly 1× when released near it, so one-finger swipe navigation re-arms).
The area gets touch-action: none so the browser doesn't zoom the page
instead. One-finger swipe and pan already ignore multi-touch.
Verified on an emulated Pixel 7 with CDP two-point touch: spread 80→200 px
gives 250 %, closing to 120 px gives 150 %, closing fully returns to 100 %,
and a one-finger swipe afterwards still moves 2/2 → 1/2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Room widgets were limited to display-only capabilities because the driver
couldn't serve anything else. Now:
- classifyWidgetCapabilities: display caps are still granted silently; reading
or sending events/state in the widget's OWN room is offered to the user;
everything else (other rooms' timelines, to-device, account data, uploads,
user directory, delayed events) stays denied. Writing protected state
(power levels, join rules, encryption, membership, ACLs, widgets, …) is
never offered, and the driver refuses it again at send time.
- WidgetPermissionPrompt names the widget, the host that runs it and who added
it; each request in plain words ("Send messages of type m.text in this
room · as you"); reading is pre-ticked, sending is not; Deny / Escape grant
nothing extra. "Remember my choice" stores it per viewer (localStorage),
tied to the widget URL, so a changed URL asks again.
- GeneralWidgetDriver implements sendEvent / readRoomTimeline / readRoomState
/ readEventRelations, each refusing any room but the widget's own;
RoomWidgetView feeds the room's live (decrypted) events and state updates,
which ClientWidgetApi forwards only if the widget holds the capability.
Verified in Chromium with a cross-origin test widget against a local Synapse:
power-levels and timeline:* requests are never shown; after allowing
send+read the widget's message lands on the server, its power-levels write
is rejected, it receives live messages, and after a reload the remembered
choice skips the prompt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
A homeserver with a consent requirement (Synapse user_consent) rejects sends
with 403 M_CONSENT_NOT_GIVEN until the user accepts its current terms. The
message just showed "Failed to send" with no reason.
Listen for the SDK's HttpApiEvent.NoConsent and show a dialog naming the
user's own homeserver (the client works with any server, so no Lotus-specific
wording), with "Review and accept" opening the server's consent_uri (http(s)
only; anything else is dropped) and "I've accepted — retry sending" resending
every event that failed for this reason. "Later" snoozes it for 10 s so
background retries don't re-open it immediately.
Verified in Chromium against a local Synapse with the send endpoint answering
M_CONSENT_NOT_GIVEN: dialog shows, link opens, retry delivers the message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
16 menus (room header, Home/Direct/Space sidebar tabs, create tab, lobby,
room-nav item, history visibility, client root) set
returnFocusOnDeactivate: false, so pressing Escape dropped focus to
<body> and a keyboard user had to start over from the top of the page.
That option was there so focus wouldn't be pulled back to the menu button
when an item opens a dialog; since ce8ed89f dialogs move focus into
themselves (their initial focus runs after the menu's return), so it's no
longer needed. The autocomplete menu (focus must stay in the editor) and
the two menus whose opener disappears are left as they were.
Checked: room header menu → Escape → focus on "More options" (was
<body>); menu → Leave Room → focus inside the Leave Room dialog on
Cancel, same as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Follow-up to ce8ed89f for the modals that have no visible heading to
point at: the three "Seen by" reader lists (read-receipt pill, "is
following", message menu), View source, the text/PDF file viewers, the
room-card join error and the user-profile modal get role="dialog",
aria-modal and an aria-label; their traps move focus in (fallbackFocus
on the dialog) where the trap is local.
Verified: the receipt pill opens a "Seen by" dialog with focus on
Close; Escape closes it and focus returns to the pill.
The remaining unnamed Modal/Dialog uses are startup/loading and error
screens (config, feature check, spec versions, client root, password
reset) and wrappers around components that carry their own role (image
viewer).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Most modals rendered a folds Dialog/Modal with no role and no name, and
their focus traps used `initialFocus: false`, so focus stayed behind the
modal and a screen reader never announced it.
- 32 dialogs with a visible heading: role="dialog", aria-modal,
aria-labelledby → the heading (given an id), tabIndex=-1.
- 4 dialogs that already had a name (Leave Room, room topic viewer,
server ACL, room-nav prompt): role + aria-modal.
- Their focus traps drop `initialFocus: false` for focus-trap's default
(keep an already-focused autoFocus field, else the first tabbable
element) with the dialog itself as fallbackFocus, so a dialog without
a tabbable node can't crash the trap. Traps that live in a parent
(UIA stages, Logout, Forward, Invite) get the semantics only.
- The file drop overlay is deliberately left alone (not a dialog).
Checked at runtime: Join with Address, Delete Message, Report Message,
Leave Room and Logout open as named dialogs with focus inside and close
with Escape (Tab first when a text field has focus — the shared
stopPropagation keeps Escape from discarding typed text, by design).
The axe e2e spec (6 tests) passes; eslint warnings unchanged (46).
17 modals with no heading (image/file viewers, loading screens) remain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The first time the window is closed, a dialog asks "Keep Lotus Chat
running?" — Keep running in the tray (default, focused) / Quit when I
close the window — with an opt-in "Start Lotus Chat when I sign in"
checkbox in the same moment (per the approved design: one dialog, no
wizard). The choice is saved natively; Settings → General → "When I
close the window" changes it later (tray / quit / ask me).
The dialog is role="dialog" aria-modal, labelled and described, with
focus on the default button. Web-side flow verified with a stubbed
native side: event → dialog → checkbox + Quit sends autostart enable +
resolve_close_request("quit"); the Settings select saves "tray".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The "N replies · <time>" chip under a thread root was named only by its
visible text, so a screen reader never said it opens the thread, and the
unread dot / muted bell were visual-only. Its accessible name is now the
visible text first (WCAG 2.5.3) plus "view thread", "unread replies" and
"muted" as applicable, e.g. "1 reply · 12:16 PM, view thread".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The logo sat on a pure-black (#000) square inside the icon's #0A0A0A
background, so the adaptive (circle/squircle) home-screen icon showed a
faint square around the logo — the "floating in an odd box" the
checklist warns about. Recoloured that backing square to the background.
Checked: both icons load in production (200, image/png); the logo's
furthest pixel is 29 % of the width from centre (Android safe zone: 40 %),
corners opaque — so no clipping in any launcher mask.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
- Skip link: following #main-content left focus on <body> (the <main>
target wasn't focusable), so nothing was announced and the URL got a
stray fragment. <main> is now tabIndex=-1 and the link focuses it.
- Keyboard-shortcuts dialog (?): no role, and focus stayed in the
timeline, so it opened silently. Now role="dialog" aria-modal, and focus
moves into it (Escape still returns focus to where you were).
- Reaction viewer (both the reaction context-menu path and "View
Reactions"): same — role="dialog" aria-modal aria-label="Reactions",
focus moves in.
Keyboard-only checks (Playwright): Tab → skip link → Enter focuses
<main>; Tab reaches the room list, Enter opens a room, typing lands in the
composer, Enter sends (verified on the server); focus ring visible. "?"
opens the dialog with focus inside, Escape returns to the same element,
"?" in the composer stays text. Topic viewer, Search dialog and reaction
viewer all return focus to their opener.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The call-bar and status-bar toggles change their label with the state
("Deafen"/"Undeafen", "Turn off/on microphone", "Start/Stop Video", …)
AND set aria-pressed, which made screen readers say "Deafen, pressed"
while you were not deafened, and gave mic/screenshare a pressed state in
one bar but not the other. Per the ARIA practice, a toggle uses either a
changing action label or aria-pressed with a fixed label; these use the
former, so aria-pressed is removed from all nine.
Checked in a live call: every control now announces only its action, in
both states (deafen → "Undeafen", mic → "Turn on microphone", …).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The full-screen incoming-call overlay (shown when you're not already in a
call — the common case) had no role or accessible name on any ancestor of
its buttons, so a screen reader landed on an unlabelled button with no
hint that a call was ringing. The in-call banner variant already had
role="alert" + a label. The overlay's Dialog is now role="alertdialog",
aria-modal, labelled "Incoming voice|video call from <caller>".
Found while verifying #187 DP6 (decline still dismisses when the decline
send fails — it does).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Status messages are saved per device and re-sent with every presence
heartbeat (a presence write without status_msg clears it on Synapse). A
device never receives its own user's presence changes made on other
devices, so the DP3 fix in db864326 — mirroring remote changes from the
Profile page — could never fire: device B kept a status that device A had
cleared and re-published it on its next state change.
Heartbeats now reconcile with the server first: GET our own presence,
send the server's current status_msg and bring the local copy in line.
Falls back to the local copy when the read fails, when the server shows
us offline (invisible mode clears the status by design), and for 15 s
after this device saved/cleared its own status (a server read that
hasn't caught up yet can't override a fresh save).
Verified with two sessions of the same user against local Synapse:
B sets "dp3 old status" → A clears it → B goes hidden→visible → server
stays "" and B's local copy is removed (before: back to "dp3 old status").
A sets "dp3 new from A" → B heartbeat keeps it and adopts it locally.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Steam double-encodes its meta tags (`&quot;`), so after Synapse's
single decode the card showed "The "Perpetual Testing
Initiative" …". The preview's og:title / og:description /
og:site_name are now decoded once where the preview is fetched, so every
card (about 20 read those fields directly) gets clean text. Rendered as
React text only, so decoding can't inject markup; exactly one level is
decoded.
Verified on the Portal 2 store link: "The \"Perpetual Testing
Initiative\" has been expanded…", no literal " left on the page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
RoomView focuses the composer on any "typing" key pressed while nothing
editable has focus, and Escape counted as typing. Closing the GIF picker
with Esc (its focus trap returns focus to the GIF button) therefore
landed in the composer instead, and so did Esc on any other room control.
Escape, CapsLock, Insert, ContextMenu, PrintScreen and Pause are now
excluded like the other non-typing keys.
Verified at 1300 px and 320 px: Esc closes the picker and focus is back
on "Insert GIF"; the 320 px picker fits with no page overflow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Removed every `as any` in src/, then restored only the ones tsc still
needs: 163 → 36. 112 were in lotus-terminal.css.ts (`'… !important' as
any` etc. — vanilla-extract's style types accept these strings as-is),
so its file-wide eslint-disable goes too. The casts were type-only, so
emitted code and generated CSS are unchanged.
eslint warnings 49 → 46; the ratchet is tightened to match.
tsc clean; 1202 unit tests pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
Shown only while Launch on login is on, and only when the native side
answers `get_start_minimized` (older desktop builds just don't show it).
Toggling calls `set_start_minimized`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
register() rejects on non-http(s) origins, and the rejection was
unhandled. The desktop app's debug build loads from tauri://localhost,
which surfaced as a Sentry "serviceWorker.register() must be called with
a script URL whose protocol is either HTTP or HTTPS". Skip registration
there, and catch any other failure (e.g. SWs disabled) with a warning.
The app works without a SW.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
The comment said on.soundcloud.com links need an oEmbed round-trip. They
don't: Synapse follows the redirect for the preview and the og:url
fallback re-parses the canonical track URL. Checked with a real short
link (the play facade loads the w.soundcloud player).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA
2026-09-23 22:09:34 -04:00
400 changed files with 18999 additions and 3524 deletions
@@ -232,7 +233,7 @@ All decoration files are 256×256 APNGs. They animate natively in all modern bro
Decoration preference is stored in the public Matrix profile field `io.lotus.avatar_decoration` (a slug string, e.g. `lotus_flower`). Any Lotus Chat user viewing your profile sees your current decoration.
**CDN:**
Files are self-hosted on the Lotus Nextcloud instance. Direct access: `https://drive.lotusguild.org/public.php/dav/files/{token}/cinny-decorations/{slug}.png`. `<img>` elements load cross-origin freely — no CORS headers needed.
Static files at `https://chat.lotusguild.org/decorations/{slug}.png` (thumbnails: `thumbs/{slug}.webp`), cached for a week. They come from the versioned `avatar-decorations` package in Gitea's package registry, installed by the matrix repo's LXC 106 deploy (see its README, "Avatar decorations"). To add or replace decorations, publish a new package version there, then update the catalog in `avatarDecorations.ts`. They used to be served from Nextcloud (matrix #22). `<img>` elements load cross-origin freely, so no CORS headers are needed.
**Module-level cache with in-flight deduplication:**
`useAvatarDecoration(userId)` fetches the profile field once per user per session. A `Map<userId, slug|null>` cache prevents redundant requests; a second `pending` waiters map ensures multiple components requesting the same userId simultaneously share one HTTP request rather than firing duplicates.
@@ -1522,6 +1523,98 @@ Checks for a new desktop release every 12h and offers a one-click update. `Tauri
---
## June–October 2026 Additions
Features shipped since the sections above were last updated (Gitea #294). Each
line names the commit that introduced it.
### Calls
- **Selectable ringtone + volume:** pick the incoming-call ringtone (`4a875884`).
- **In-call incoming-call banner:** a small banner for a second call instead of a full ring (`c67aed01`).
- **Denoise tester:** Settings → Calls lets you audition the ML models and calibrate the noise gate (`14cfa021`).
- **Mic level meter** on the mute button (`3449338b`, #146).
- **"You're muted" nudge** when you talk into a muted mic (`73b1d1e3`, #117).
- **Undeafen catch-up toast:** what you missed while deafened (`33e16e85`, #128).
- **Admin turned off screen share / camera:** a toast explains why your share or camera stopped (`34574178`, #223).
- **Voice-limit / audio-only chip** in the room header (`7f17940d`, #134).
- **"N Live" pill names** who is in the call on hover / focus (`a475531b`, #157).
- **Live-call dot on the space rail** (`4fe9c870`, #148).
- **Missed-call line** in the timeline for DMs and small groups (`c921f115`, #162).
- **Answer with camera or audio-only:** answering says when it turns the camera on (`d2f56817`, #135).
- **Audio output quick-switch** in the call bar (`98c80f36`, #119).
- **Call-ended summary toast:** "Call ended · 41 min · connection was good" (`af1c0ee1`, #143).
- **No WebRTC on desktop:** offers "Open in browser" instead of a dead end (`fff583fc`).
- **Call page on its own origin** (desktop, opt-in) (`7d7a379c`, #43).
### Messaging & composer
- **Reactions collapse** to one row with a +N chip (`bf05751e`, #138).
- **Consecutive photos / videos** render as one gallery grid (`bd8c79e0`, #137).
- **Inline audio playback** for `m.file` audio attachments (`eb34b047`).
- **Recently used GIFs** in the GIF picker (`5b58e5fe`).
- **Pasted image → caption field:** a single pasted or dropped image lands in its caption (`0c45bde8`, #129).
- **Offer to format pasted code** as a code block (`75d55e86`, #107).
- **Tap to preview stickers / GIFs** on touch, tap again to send (`8d11a62e`, #147).
- **Link-mismatch confirmation** before opening a link whose text names another site (`74d8e311`, #122).
- **Embed fallback** for hung embeds and deleted X posts (`ba5b1ffe`, #200).
- **"Back to where you were"** after jumping in the timeline (`7da91bd8`, #127).
- **Offline outbox:** unsent messages survive a reload and retry (`02d86cae`, #112).
- **Terms-consent explanation** when "Failed to send" is the server wanting its terms accepted (`c96c47dd`).
### Search & navigation
- **Quick switcher (Ctrl/⌘+K)**, with a configurable key (`9d4679d2`).
- **Search filters:** `has:image/file/video`, recent searches, pinned only (`1ee0f0b5`).
@@ -34,7 +34,7 @@ This file keeps only what a contributor needs to have open while working: the tw
## 🚫 Blocked Features (server / upstream gated)
Re-run `/_matrix/client/versions` + `unstable_features` after each Synapse upgrade. **Re-checked on 1.157.1 (2026-07-23): no change — all four below are still `false`.** The 1.156.0→1.157.1 delta unblocked nothing (it's a bugfix release; the only feature-bearing release in the gap was 1.156.0, which we were already running).
Re-run `/_matrix/client/versions` + `unstable_features` after each Synapse upgrade. **Re-checked on 1.162.0 (2026-10-18, Gitea #294):** live location (`msc3489`/`msc3672`) and relation redaction (`msc3892`) are still unadvertised; thread subscriptions (`msc4306`), sticky events (`msc4354`) and QR login (`msc4108`) are still `false`. New: `org.matrix.msc3026.busy_presence` is advertised as `false`, and enabling it would give Do Not Disturb a real presence state (#283). `msc3266` and `msc4222` are no longer listed, but room preview still works (`im.nheko.summary` and `v1/room_summary` both answer 200). (Previous check: 1.157.1 on 2026-07-23, no change then.)
- **[BLOCKED] Live Location Sharing** (MSC3489 + MSC3672 both `false`) — real-time GPS beacons over the existing static share.
- **[BLOCKED] Reaction/Relation Redaction** (MSC3892 `false`) — remove a reaction without redacting the parent; current full-redaction fallback is acceptable.
"$comment":"Gitea #96. Budgets are seeded from the dist/assets gzip sizes on the tree at seed time, +10% headroom. Regenerate deliberately (not just to silence a failure) when a real feature addition grows the bundle: measure the new gzip sizes and bump these with the same +10% margin.",
"totalGzipBytes":1731120,
"largestChunkGzipBytes":358317
"$comment":"Gitea #96. Budgets are seeded from the dist/assets gzip sizes on the tree at seed time, +10% headroom. Regenerate deliberately (not just to silence a failure) when a real feature addition grows the bundle: measure the new gzip sizes and bump these with the same +10% margin. Re-seeded 2026-10-18 from lotus d7bc6d94 (total 1666894, largest 360238 bytes gzip) after the October audit fix batch (~40 PRs) grew the entry chunk past the old budget.",
?'Lotus Chat can keep running in the system tray when you close this window, so messages, notifications and calls still reach you. Open it again from the tray icon.'
:"Your desktop doesn't show tray icons, so if Lotus Chat keeps running after you close this window, the only way back is to launch it again."}
Loaded 100 of 400 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.