The barely-visible shift at end-of-stream was the reserved status-row
inset being released when the row unmounted. The measured overlay
height now only grows during a session and resets on session switch, so
the finished message's metadata footer appears exactly where the status
row was, with zero layout movement.
The floating status/working row let streaming text show through it and,
because the scroll math did not know the overlay existed, the reveal
correction targeted the very bottom edge — the live line ended up
hidden beneath the status row ('catches up to almost-the-end'). The
overlay now has a solid background, its height is measured with a
ResizeObserver and fed to the timeline as the occluding overlay height,
so the live streaming line glides to rest just above the status row and
the list reserves matching end inset at rest.
The phase-2 reveal correction now scrolls animated; successive
corrections restart the smooth scroll from the current position, so
streaming follows as one continuous motion. User gestures interrupt the
native smooth scroll, so free-scrolling is unaffected.
Phase 2 of the anchored turn (following the live edge once the reply
outgrows the viewport) never ran: the reveal correction was triggered
only by entries-array changes, and the streaming tail grows inside a
single row without touching the array. The hook now also subscribes to
LegendList's totalSize and re-runs the same guarded correction on every
content growth.
Six reported defects, one root theme — our port diverged from the
reference semantics:
- a user gesture no longer collapses the reserved anchored end space
(that snap-to-bottom was the 'anchoring vanishes on a micro-scroll');
it only disarms the follow/anchor machinery, and anchor remeasures
after the user takes over can no longer re-position the viewport
- the anchor arms synchronously BEFORE the send reaches the store, and
the claim compares against the arm-time baseline — arming a frame
after the optimistic row committed is why anchoring 'rarely worked'
and sending from mid-history did nothing
- gestures are selective (wheel up, touch/pointer away from the end,
PageUp/Home/ArrowUp), so scrolling toward the end or clicking a row
at the live edge no longer kills follow
- 'at end' is a tight 40px band against the full content length instead
of the list's half-viewport isNearEnd, so the scroll-to-bottom pill
appears when the viewport actually leaves the end
- reaching the end re-arms follow without knocking an anchored turn out
of its mode, and the overlay scrollbar suppression follows scroll
ownership rather than the anchor's mere existence
- the status/working row moved out of the list footer to a fixed spot
above the composer: inside the list it walked down with every
streamed line, and its unmount was part of the end-of-stream jerk
Sending a message now parks that message near the top of the viewport
and streams the reply into reserved end space below it, instead of
jumping to the bottom and chasing it.
- swap @tanstack/react-virtual for @legendapp/list in the chat timeline; the
streaming tail becomes a normal list row rather than a separately rendered
block, so one component owns the scroll position
- add timelineScrollAnchoring: pure anchored-turn geometry plus the three
scroll modes (following-end / anchoring-new-turn / free-scrolling)
- replace useChatAutoFollow with useChatTimelineScroll, which opts out of
automatic movement on real gestures via a generation counter instead of the
timer windows the old implementation needed to recognise its own writes
- move the load-older button, question/permission cards, recap, status row and
bottom spacer into the list header/footer, since the list owns its container
- extract useScrollShadow so the shadows can attach to that container
maintainScrollAtEnd and maintainVisibleContentPosition replace the manual
prepend anchor-hold and the mobile quiet-window prepend deferral.
Validated: workspace type-check, lint, web build, ui tests per file.
Scroll behaviour itself is unverified and needs manual testing on web, desktop
and iOS.
The hover info card (project/branch/PR/time) added noise without pull —
the tab title plus the sidebar already cover it. The tooltip body
component, its per-hover subscriptions, and the native title attribute
are gone; right-click/menu and the status dot are untouched.
The 'Always show editor toolbar' preference is gone — the docked
toolbar under the file tabs is now the only mode, and the floating
hover toolbar branch in the files editor (with its open-state and
outside-click machinery) is deleted. The stored preference is dropped
by a store migration and removed from desktop settings persistence,
settings search and every locale.
Navigation section order now reads: file editor keymap, auto-save,
terminal shell + login shell, Terminal Quick Keys, and the Session
tabs group last.
The General section renders OpenChamberVisualSettings with an explicit
visibleSettings allowlist, so the new control was searchable but never
rendered.
A new Session tabs group (web/desktop only) turns the header session
tabs off; disabled, the header renders the exact pre-tabs view — plain
session title with meta row and the always-visible session menu (the
same block VS Code uses). The Alt+W close-tab shortcut no-ops while
tabs are off. Registered in settings search with a matching anchor;
labels translated in all locales.
The close-with-neighbour-activation logic moved into a shared helper
(the strip and the shortcut use the same path). Alt+W is the default
because the browser owns Cmd/Ctrl+W on web; desktop users can rebind it
— the action is registered as customizable, so it appears in the
Settings shortcuts section and the shortcuts help automatically. VS
Code is excluded (it has no session tabs).
Auto-add means the strip only grows, so past 20 tabs the oldest one
leaves the working set (the newly opened tab is always the survivor).
The branch/worktree/PR/project subscriptions that feed the hover
tooltip moved into the tooltip body component, which mounts only while
the tooltip is open — a resting tab now subscribes only to its status
dot's session status and unread count.
Rename from another tab's menu stored only a boolean that the
session-switch reset effect wiped before the menu finished closing, so
the tab activated but rename never began. The pending rename now
carries the target session id and begins exactly when that session
becomes active (whether the menu closes before or after the switch),
without the reset effect cancelling it — and without an in-flight
rename being cancelled by unrelated re-renders. Tab background/text
state changes animate at 75ms so activation reads immediate.
The title fade no longer sits on controls: it lifts while the tab is
hovered or its controls/menu are open (hard clip instead) and is not
applied at all while the active tab renames. The active tab clips its
title instead of showing an ellipsis, matching inactive tabs. Tab
separators drop the extra transparency that made them nearly
invisible.
Keeps enough space for the chat panel when the sessions sidebar stays visible
Bases the expanded layout breakpoint on the sidebar width instead of a fixed value
Right-click now opens the session menu under the cursor (the sidebar's
context-menu pattern; the same header-supplied items back both the
"..." dropdown and the context menu via injected menu primitives) and
still never changes the active tab. Hovering a tab shows the sidebar's
session tooltip — title, last-activity time, project, branch and PR
status — after a delay, suppressed while a menu is open or a drag is in
flight. Each tab carries the sidebar's status dot at its end (accent
while the session runs, info-blue for unread), hidden while the hover
controls overlay it. Tab titles fade out instead of ending in "...",
and while the active tab is renaming its hover controls stay hidden so
only the rename controls show.
The strip now owns a single dropdown per tab, fed by the header with
items bound to that tab's session — rename (activates the tab first),
copy id, share/copy link/unshare, export and move-to-worktree (active
tab only, they need the loaded directory), close other tabs, archive
and delete with the confirm dialog targeting the right session. The
separate inactive-tab menu is gone, and there is no Close item — the
tab's close button covers it, now placed after the menu button.
Right-click opens that menu without activating the tab. The menu's
anchor overlay stays mounted until the close animation finishes, which
removes the popup flashing in the top-left corner on close. The
scroller hides its scrollbar via a dedicated CSS class (the bar was
shifting the header content vertically).
The old flex spacer split the header in half, boxing the strip into the
middle while tabs shrank to slivers; the spacer now renders only for
the VS Code and surface-title layouts, tabs keep a fixed width (w-44)
and the strip scrolls behind the right-side buttons. Every tab gains a
hover-revealed close button next to the menu; the active tab's menu is
hover-revealed too (it was always visible) — the header passes it into
the strip so both overlays behave identically. The active tab's inner
padding matches inactive tabs.
Uniform compact tabs (h-7, 6px radius, 13px medium) that share the strip
width and shrink before scrolling, thin separators between inactive
neighbours that disappear around the active or hovered tab, a
hover-revealed menu button anchored to the tab's end, a soft selection
pill for the active tab, and hidden scrollbars with fade edges. The
active tab is single-line: the project/branch meta row stays only in the
VS Code header, which keeps the plain title.
Web/desktop header replaces the single session title with a strip of
soft pill tabs, one per session the user has opened (sidebar, palette
or deep link — opening anywhere adds a tab once). The active tab is the
familiar title block — rename, meta row and the full session menu —
inside a gently selected pill; a brand-new draft shows as a transient
pill until its session exists. Inactive tabs show the title with a
hover-revealed "..." menu (close tab, close other tabs, copy id) that
nudges the text like sidebar rows, and close by middle-click too.
Tabs drag to reorder, scroll behind the right-side header buttons with
soft fade edges, respect the reserved window-controls inset, and
persist across reloads. Closing the active tab activates its neighbour
(or opens a new draft when it was the last). Tab ids whose session is
not in the loaded list stay stored but hidden, so a partial session
list never destroys the working set. VS Code keeps the plain title;
mobile is untouched.
activeSurface was permanently 'chat' after the legacy mobile layout
removal, so the whole concept is gone: the store field, surfaceGuard,
setActiveSurface/setSurfaceGuard, the per-runtime surface memory in
prepare/restoreForRuntimeSwitch, and WorkspaceSurface itself. All ~30
setActiveSurface('chat') call sites were no-ops and are deleted;
always-true 'is the chat active' checks in keyboard shortcuts, Header
and ChatContainer are unconditional now. FilesView's dirty-file guard
kept its file-switch and close protection but drops the surface-switch
branch nothing could trigger. TerminalView visibility comes only from
its callers. The router keeps parsing legacy ?tab= links (they open the
matching context-panel surface) via its own RouteTab type and no longer
serializes a tab or diff file into URLs — desktop URLs never carried
them anyway.
MainTab/activeMainTab/setActiveMainTab/setMainTabGuard were deprecated
mirrors of the surface names — every call site now uses
activeSurface/setActiveSurface/setSurfaceGuard directly and the aliases
are gone, including the persisted mirror field.
The 'diagram' surface had no way to open it (navigateToDiagram had no
callers except a .drawio attachment click that navigated to a surface
nothing rendered); the surface, DiagramView, and its store plumbing are
removed, and a .drawio attachment now opens in the file panel.
?tab= deep links map to the matching context-panel surface instead of
setting a main-area surface nothing renders, and a persisted non-chat
surface can no longer rehydrate into a blank main area.
Phone viewports run the separate MobileApp shell (and a viewport crossing
now reloads into it), so the desktop layout's mobile branch was
unreachable: the drawer machinery, the full-screen secondaryView surface
switch (including the terminal/diagram desktop carve-out nothing could
trigger), the mobile header with its tab bar, the Cmd+number tab
shortcuts, the mobile quota panel, and the surface guard that reset
non-chat tabs. DrawerContext had no consumers left and is deleted.
Header drops from 2630 to 1853 lines; the desktop render is unchanged.
The mobile-vs-desktop surface is stamped once at boot, so a browser
window narrowed past the phone threshold kept the desktop shell (and
its legacy squeezed layout) until a manual reload. A viewport watcher
now reloads into the other shell once the resize settles — the same
mechanism the old Settings toggle used. Fixed shells (Capacitor,
desktop, VS Code) and ?surface= overrides never switch.
With the new mobile app reachable this way, the old/new mobile layout
preference is gone: phones always get the mobile app.
Sidebar session/folder/group search and the Todos, Memory, Plans and
Notes filters required the whole query as one literal substring; they
now match tokens in any order and ignore punctuation via
matchesRankQuery, keeping their own list order and tree structure.
Branch, project, agent, model, provider, stash, SSH-host, skill-catalog
and archive filters each had their own toLowerCase().includes (or no
ordering at all); the git branch and gitmoji pickers also let cmdk
re-filter and reorder on top of the manual filter, silently dropping
rows. All of them now go through rankByQuery/matchesRankQuery: results
are relevance-ordered, multi-word queries match in any order, matching
ignores punctuation, and cmdk filtering is disabled where the ranked
list is already final. rankBranchesForQuery keeps relevance order
instead of re-sorting matches alphabetically; the model picker now also
matches model ids.
One matcher for every dropdown filter: multi-token queries match in any
order, matching is punctuation-insensitive, single-token queries
tolerate typos, and results come back relevance-ordered (prefix, then
word-boundary, then substring, original order on ties). matchesRankQuery
is the boolean companion for lists that keep their own grouping.
ghostty-web marks the terminal container contenteditable but only its
hidden textarea opts out of IME text mangling, so iOS and Android
keyboards uppercased the first letter of every command. Set
autocapitalize/autocorrect/spellcheck off on the container after open.
The tab list lived only in per-tab sessionStorage, so a new browser
tab, another device, or cleared storage showed an empty terminal
sidebar while PTYs kept running server-side, and orphans leaked until
the idle sweep. Add GET /api/terminal/sessions and adopt unknown
server sessions into the local tab projection (additive only; a failed
listing changes nothing).
The idle sweep also reaped terminals in background tabs because only
the active tab holds a WebSocket attachment. Add POST
/api/terminal/touch and have open clients periodically refresh
activity for every session their tabs reference.
Directories and files were rendered as fixed category blocks, so an
exact file match sat below unrelated directories. Merge both result
sets and rank them with the shared fuzzy scorer against the full
relative path. Multi-word queries now match tokens in any order
(longest token queries the server, the rest filter client-side), and
path truncation keeps the parent segments next to the file name so
index.md-heavy trees stay distinguishable.
Cmd+P scored server file hits by basename only, so a query like
"solo-is-a" returned nothing for solo-is-a-team-size/index.md.
Score by relativePath and widen the server candidate limit to 40
so client-side reranking has enough to work with.
Two mitigations for local multi-instance contention over the shared relay
identity:
- A standby instance now waits a 2-minute grace period after the host claim
frees before taking over, so a cleanly restarting host (app update or
relaunch) — which reclaims at boot with no wait — always wins the restart
window instead of stranding paired devices on another process.
- Dev instances never host the relay passively: dev scripts set
OPENCHAMBER_RELAY_HOST=off and the Electron dev shell is detected via
OPENCHAMBER_ELECTRON_DEV. Explicit enable/pairing on such an instance still
force-claims; OPENCHAMBER_RELAY_HOST=on overrides.
A dedupe-keyed client re-mint (QR rescan, password/passkey re-login) replaces
the stored record, which also reset the operator-visible name to the app's
hardcoded default ('OpenChamber Mobile'). The app-reported name is now only a
fallback: an explicit pairing label wins, otherwise the replaced record's
label is kept, and the default applies only to a first-ever pairing.
The chat selection menu (desktop pill and mobile sheet), the chat
comment input, the file-preview Comment pill, the shared inline comment
editor, and the composer's context chip preview popover now use the
same translucent blurred glass surface as the btw panel, keeping their
borders and soft shadow.
The project knowledge panel's section sidebar becomes a horizontal pill
strip on mobile, where a half-width side column left no room for
content. The autocomplete popups get their 120px minimum height back —
browser keyboard panning could put the anchor above the measured
boundary and collapse them to nothing. The status row's 24px footer-swap
reservation now applies only to the assistant-status instance; the
accessory-only row above the composer (pending changes) takes the
normal 8px instead of floating a stray gap.
Bohdan Triapitsyn