Shows a separator when the last live assistant answer follows earlier visible content
Keeps render comparisons in sync with the new assistant-text context
Tweaks tool summary spacing and path text styling
Sending used to park the new user message near the top of the viewport by
reserving blank space below it and holding the viewport there while the
reply streamed in. The reserved space showed up on its own at times and the
hold fought other corrections; with a reliable end-follow and pin in place
the effect no longer pays for its machinery.
The anchoring mode, its arm/claim/position/settle lifecycle, the reserved
end space passed to the list and the anchored-turn geometry are removed.
Sending from the live edge is now an instant return to the end, and the
reply is kept in view by ordinary end-follow; sending from mid-history with
auto-follow off still leaves the viewport untouched.
Testing: scroll module tests trimmed to the remaining geometry; ui
type-check, lint and dead-code report; web build.
Opening a panel or resizing the window while pinned to the end of an idle
session bounced the timeline and left it above the end with the pin
released. Every row re-wraps on a width change and the list's total
content length lags a frame behind the rows, so anything scrolling to the
end from it (the list's own maintainScrollAtEnd, scrollHeight) landed on a
blank tail or short of the real end; the idle branch then released the pin
instead of recovering.
An idle session no longer hands end-keeping to the list at all: the hook's
same-frame pin owns it, and during a width resize it holds the measured
bottom of the last real row plus the list footer, for streaming readers
too. The footer size was missing from that measurement (the list does not
expose it through getState) and left the viewport short by the tail
spacer; it now arrives through onMetricsChange. Size compensation applies
only to a reader who left the end.
The follow re-arm band is half a viewport instead of 40px: leaving the end
is decided by a real gesture only, so the band now only decides how close
a reader must come back before follow re-arms.
Testing: scroll module tests updated (footer in the real end, half-viewport
band with a 40px floor); ui type-check and lint; measured in headless
Chrome that scrollTop equals the maximum on every frame of a panel toggle,
a height change and a combined resize, where the previous build ended 122px
and 462px above the end.
Keeps idle readers from snapping back to the end while rows re-wrap
Releases the pin instead of scrolling if a resize moves an idle view off the end
Re-asserts the live edge after resize settle for active streaming sessions
Opening a session (or any relayout that shrinks off-screen size estimates)
could leave the viewport in a phantom tail below the measured content, with
every row out of reach above; a totalSize-change check now detects the
fully blank viewport and returns to the real end, and settling a width
resize re-asserts the end for a reader who was on it. Prompt-rail and
message jumps land on estimated offsets that shift as the target mounts and
measures; a short settle loop now re-aligns the target until layout rests,
backing off on the first user gesture.
The animated maintainScrollAtEnd also ran outside streaming, so opening
a historical session glided visibly through the whole conversation as
late row measurements corrected the end position, and an in-flight glide
could supersede explicit navigation. Corrections are instant unless the
session is actively working.
Returning to the bottom during a stream also lands on the end as of that
moment, and the list's own follow may not have re-armed after the user's
earlier gestures — the queued-send jump then fell behind the growing
reply. goToBottom now re-asserts the edge a few times (150/400/800ms)
until it holds; a new user gesture cancels the window.
Token-by-token streaming mutates the trailing paragraph in place on
every tick: words rewrap, the last line jitters, and the whole reply
reads as flicker. Streamed text now commits only up to the last complete
line — prose arrives a paragraph at a time (a markdown paragraph is one
logical line), code fences reveal line by line, tables row by row — and
a shown block never changes again. A paragraph that runs long without a
newline releases at the last sentence (then word) boundary so the stream
never stalls. Applies to assistant text and reasoning; tool output keeps
its raw tail.
With growth arriving in block steps, the end follow switches to the
list's animated mode so each step is a glide — reveal and scroll read as
one continuous motion. The gesture opt-out now measures at-end from the
live list state instead of the cached flag, which the animated glide
deliberately leaves stale while trailing the edge; without that, a drag
during a glide could leave the scroll-to-bottom pill unshown.
Measured: end-following holds at distance 0 for the whole stream, and
the mobile drag opt-out shows the pill in three of three runs. The
continuous glide costs ~15% more main-thread time per streamed character
than the instant follow — the price of the motion.
Three first-render polish issues:
Code blocks jumped at end-of-stream: streaming defers the per-line
line-number markup, so the finished decorate pass inserted the gutter
column and shifted every code line right. The gutter's horizontal
footprint is now reserved with CSS while the markup is deferred, so the
final pass only fills in numbers and colors.
File references were verified twice per render and re-verified with the
wrong directory: the annotation pass's own DOM writes re-triggered its
MutationObserver, and the pass also ran before the effective directory
resolved — issuing stat probes under an empty directory and a second
time under the real one. The observer now ignores the pass's own
mutations and annotation waits for a resolved directory.
Content replacing the hydration skeleton popped in: it now plays a
one-shot 180ms fade (reduced-motion aware); cached session switches
never carry the class and stay instant. Also removed the dead
disableStaging prop and its unused pendingRevealWork threading.
Verified on a production build over CDP: a streamed code block's text
keeps its exact x-position across end-of-stream, and stat probes for a
message with file mentions are unique per path with the directory header
always present. The hydration fade path could not be exercised in the
harness (the live event stream pre-populates parts in a single-project
environment) and needs an eyeball check on a cold multi-project open.
3.3.x makes maintainScrollAtEnd follow content growth on its own — a
tail row growing in place included — which is exactly what the manual
totalSize correction existed for. Delete that correction (the totalSize
listener now only drives the anchored-turn glide) and pick up 3.3.x's
measurement batching, prepend-flash fixes, and web programmatic-scroll
fixes. Opt the explicit maintainScrollAtEnd config into footerLayout per
the 3.1.1 guidance.
The library's own released-on-user-scroll heuristic proved unreliable
one run in three against synthetic touch, so the gesture state machine
stays authoritative: while a real gesture owns the scroll, the list's
end pinning is switched off through a threaded endPinningReleased prop
and re-engages when the user returns to the end.
Validated with the CDP battery on a production build: stream follow
stays at distance 0, mobile drag releases with the pill shown in three
of three runs, resize oscillation stays at the reduced level, the rail
reaches the last turn, and profiled streaming cost per rendered
character matches the tuned 3.2.0 numbers.
A pinned viewport shook during window resizes while a scrolled-up one
stayed calm: both pinning mechanisms — the growth corrections and the
list's maintain-scroll-at-end — re-assert the end against rows that are
still re-measuring, fighting the size compensation that keeps the free
case stable. While the width is actively changing, stand both down and
hold the reading position the same way the free path does, then re-assert
the live edge once with a single instant write after the resize settles
(only when the viewport was still following).
Both scrollToTurnId and scrollToMessageId hard-returned false when the
target was the trailing (last) turn and its element was not mounted, so
clicking the rail's last item in a long scrolled-up session did nothing.
The trailing entry is a regular list row at the end of the data — scroll
to its index the same way history targets are reached.
A width change re-wraps every row at once, and with size restoration off
the accumulated height delta above the viewport threw the visible content
hundreds of pixels up and down while dragging the window edge. Size
restoration must stay off in steady state — a tool result expanding in
place has to grow downward — so a ResizeObserver on the scroll node
enables maintainVisibleContentPosition size compensation only while the
width is actively changing and releases it 300ms after it settles.
Measured on a continuous 1400-800-1400 drag over a long session pinned to
the live edge: the largest per-step displacement of a visible paragraph
drops from ~680px (median of three runs) to ~200px, and the viewport
never detaches from the end.
The old scroll engine needed message parts to report content growth
(onContentChange) and per-message animation lifecycle callbacks
(AnimationHandlers) so it could re-pin the viewport. The timeline list
measures growth itself now, and the replacement hook had already stubbed
the whole contract with no-ops kept only for source compatibility.
Remove it end to end: the hook exports, the container and list threading,
the ChatMessage/MessageBody signal-only effects, and every part-level
prop and call site. Expand/collapse behavior and reveal animations are
untouched — only the reporting channel goes.
New Streaming section on the Chat settings page with a checkbox that
controls whether the viewport follows new content while a response
streams. Default stays on. With it off, the anchored user message still
parks at the top on send, but no glide or end-follow correction runs and
the list's maintain-scroll-at-end stays disabled; the scroll-to-bottom
pill and session open keep scrolling explicitly.
Persisted through desktop settings like the other chat toggles (auto-save
diff, authoritative apply, sanitize), registered in settings search, and
localized in every locale.
The streaming tail only overlaid the actively streaming message with live
parts from the sync store. When a turn moved to its next step message, the
previous message fell back to its lagging base record, briefly dropping its
completed tool parts — remounting them and replaying their reveal
animation once the record caught up.
Overlay live parts for every message of the streaming tail via a new
useSessionPartsForMessages hook, guarded so an empty live array never
erases parts the record does have. Also key generate-effect glyphs by
index so appended text does not replay earlier characters, and keep tool
row reveal wrappers mounted unconditionally.
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.
* perf: optimize session loading and startup
* fix(chat): stabilize history prepend virtualization
* perf: unblock first session open from startup network contention
Opening the first session after app start waited seconds for its message
fetch. Three independent contributors, each measured via CDP network
capture and Chromium net-log against the packaged desktop app:
- The active-session watchdog fired an uncapped per-directory status poll
and child-session discovery burst at startup, and other subsystems
(git checks, global session pages, command/skill discovery) fanned out
alongside it, saturating the browser's ~6 HTTP/1.1 sockets per origin.
Add a shared background-network gate (concurrency 3) and route the
watchdog, poll-shaped git reads (also priority: low), global session
pages, command/skill loads, and the background update check through it.
- The packaged renderer is cross-origin to the loopback backend, so every
API call needs a CORS preflight; a few slow OpenCode-proxied requests
held the whole pool while preflights and interactive traffic queued
behind them. Lift Chromium's per-host connection cap for loopback via
ignore-connections-limit in the Electron shell.
- OpenCode initializes each directory lazily on its first request, so the
first click paid that cost interactively. Warm the last-used directory
and the three most recently opened projects right after OpenCode
readiness, sequentially and best-effort, overlapping UI startup.
Validation: new background-network tests, lifecycle warmup test, focused
store/sync tests, UI type-check and lint, dead-code report, node --check
plus electron type-check/lint, and CDP first-open measurements on the
packaged app (message fetch socket queue 5.4s -> 0.03s).
* fix(ui): keep interactive git reads out of background queue
---------
Co-authored-by: Bohdan Triapitsyn <artmore@protonmail.com>
Improve OpenChamber responsiveness under large session workloads while fixing
cache, synchronization, and persistence correctness across runtimes, projects,
directories, and worktrees.
- prioritize selected and visible sessions during bootstrap and defer
non-critical enrichment work
- reduce redundant message loading, event processing, store publication, and
hidden sidebar work
- prevent stale session and message requests from overwriting newer
authoritative state
- preserve existing data when authoritative fetches fail instead of treating
failures as successful empty responses
- scope session materialization, messages, drafts, queues, todos, pins,
permissions, folders, tabs, Git state, and pull request data by runtime and
directory identity
- harden runtime switching, reconnect, cleanup, mutation reconciliation, and
persisted-state ordering
- preserve live subagent Task linkage when metadata arrives after an older
message request or while streaming parts are suspended
- coalesce overlapping tail refreshes without losing newer refresh demand
- improve cold-session loading by moving deferrable work out of the critical
bootstrap path
- isolate URL authentication, mobile credentials, native secrets, and other
runtime-owned state across endpoint changes
- bound long-lived caches and remove avoidable allocations from event and
rendering hot paths
- limit virtualization to archive collections where it improves rendering
without disrupting active sidebar layout
- stabilize session folders, pin ordering, expanded state, and persisted
sidebar behavior
- open skill files through the same secure editor and outside-workspace grant
flow used by file navigation, including worktree sessions
- expand regression coverage for stale completions, runtime collisions,
reconnect behavior, persistence races, authoritative empty results, and
subagent refresh ordering
- document the updated synchronization, cache ownership, performance, and
runtime-isolation invariants
- Remove assistant message headers; show provider icon, model, agent,
thinking variant, duration and time in the turn footer (metadata left,
hover-revealed actions right)
- Merge turns started by hidden user messages (subagent nudges) into the
previous turn so Activity, footer and spacing stay continuous
- Treat compaction summary text (info.summary) as justification activity
in sorted mode and skip it when picking the turn summary
- Interleave activity segments with standalone tool rows so Agent Task
sits chronologically between activity sections
* feat(chat): prompt navigator list preview with prompt filtering
The hover preview is now an interactive scrolling mini-list of prompts:
rows render as bordered two-line cards, the highlighted row stays inside
a center dead zone and the list glides only near the window edges, wheel
steps the highlight, and the panel stays open when the pointer moves into
it so a click can be corrected inside the list.
Rail entries are filtered to real prompts: previews are built from
normalized user display parts (synthetic context stripped), fully
synthetic user messages are excluded, and shell-mode messages show their
extracted command via the shared shell bridge helpers.
* fix(chat): render shell command status transitions
The injected /shell text part carries live state in shellAction, which
the render-relevant part comparator ignored — a running→completed update
reached the store but never re-rendered the message row until the next
send. Compare shellAction command/output/status for text parts.
* fix(sync): stream shell bridge part updates while running
Streaming suspension keeps part updates out of the static message records
while an assistant message streams, relying on the live streaming-tail
path to render it. Shell-mode bridge messages are hidden from the
timeline and rendered inside the user row, so they have no live path —
suspension froze their output chunks and left the card without a Show
output action until the run finished. Exempt shell bridges (single bash
tool part parented to a synthetic shell-marker user message) from
suspension; their updates arrive at command-output pace, not delta pace.
* feat(chat): syntax-highlight shell command card
Render the shell-mode command and its output through the shared
WorkerHighlightedCode (Shiki) with bash grammar, matching the bash tool
part presentation, instead of plain pre blocks.
* fix(chat): preserve chronological message order during history pagination
The baseDisplayMessages dedup loop iterated from tail to head (newest
to oldest), keeping the newer occurrence of each message ID. During
history pagination (prepend mode), the server returns older messages
that may overlap with the current view at the boundary. The tail-first
iteration discarded the older (prepended) duplicate in favor of the
newer (existing) one, breaking chronological ordering.
Change the loop to iterate head to tail (oldest to newest) so the
first occurrence of each time-sortable message ID is preserved. Remove
the now-unnecessary .reverse() call.
Fixes#2088
* test(chat): add dedup logic coverage for baseDisplayMessages
Covers message ID deduplication in baseDisplayMessages useMemo:
- First-occurrence preservation during dedup
- Input order maintenance
- Empty input, single-element, all-same-ID edge cases
- History pagination prepend scenario with overlapping IDs
---------
Co-authored-by: bashrusakh <bashrusakh@users.noreply.github.com>
The tanstack rows sit in a wrapper offset with transform: translateY(), and a
transformed ancestor becomes the sticky containing block — turn headers stuck
to the wrapper's overscan-dependent top edge instead of the scroll container,
floating over the previous turn. Offset the wrapper with padding-top instead:
identical geometry, sticky computes against the scroll container again, and
the padding only changes when the virtual window shifts, not per scroll frame.
Adds a Last turn scope to DiffView that renders OpenCode snapshot diffs from the latest user message summary without re-fetching git contents. The view hides Review in that mode and carries the selected diff scope through main and context-panel navigation.
Connects latest-turn changed-file chips in chat to the snapshot diff view on desktop and mobile, while keeping older turn chips static/read-only to avoid misleading affordances and extra subscriptions. Updates localized labels and empty states plus changelog.
Validation: bun run type-check (packages/ui); bun run lint (packages/ui).
- Migrate sidebar session groups, git changes panel, virtualized code
blocks, and JSON tree viewer from virtua to @tanstack/react-virtual;
virtua remains only inside the Pierre diff viewer integration
- Sidebar: preserve scroll position when virtualization enables
mid-session (enable only once the ancestor scroll element is resolved,
seed initial offset from its live scrollTop, render plain rows for the
single pre-paint frame); disable native scroll anchoring on the
sessions scroller; keep row spacing identical between plain and
virtualized modes; absolute row positioning so variable-height rows
cannot drift past the container
- Chat: expand tool/thinking blocks downward by only adjusting scroll
for rows growing above the viewport; raise the desktop history-load
lead to 1.5 viewports so prepends land above the visible area
- Git changes: compute the prefetch window from the first visible row,
skipping overscan rows above the viewport
- Sidebar rows: make the whole highlighted row area clickable, guarded
against double-firing from interactive children
- Replace virtua with @tanstack/react-virtual for chat history on all
surfaces: bottom anchoring (anchorTo: end), key-stable prepend
preservation, and native iOS touch/momentum deferral live in the core
- Patch virtual-core to clamp the render range to real scroll bounds
during transient adjustments
- Rows render in normal flow inside a translated wrapper so sticky user
headers keep working; measurement snapshots cached per session
- Pre-write container height in scrollToFn so the browser cannot clamp
anchor corrections to the stale height; hold the prepend anchor for up
to 180 frames on mobile while fresh rows settle (cancelled by user
input; desktop relies on core anchoring alone)
- Adaptive row-size estimate from per-session measured averages; disable
reveal fade-in for virtualized history rows
- Mobile loads older history only through an explicit localized top
button: no scroll-position trigger and no post-mount background
prepend, so every insert happens from a resting state; a quiet-window
hold defers any stray prepend commit while a touch gesture is active
- Desktop/VS Code keep the seamless scroll-up trigger and progressive
background prepend
- Mobile: defeat iOS momentum scroll when compensating history prepend
(overflow toggle + short rAF watchdog); disable history virtualization
and post-paint background prepends; preload Markdown renderer and use
plain-text Suspense fallback to avoid first-frame geometry shifts
- Desktop: stop double-compensating prepends on the virtualized list -
virtua shift owns the adjustment; remove sticky-anchor heuristics that
misfired as failed restores
- Sync: skip no-op store writes when messages/parts are unchanged
Give touch surfaces a larger, viewport-relative head start for loading older
history so an in-flight fetch completes before the finger reaches the top.
Raise the mobile virtualizer overscan so fast flings stay populated instead of
leaving blank gaps, and drop the fixed itemSize hint so virtua auto-estimates
row heights from measured sizes instead of a flat constant.
Reworks the chat and session-sidebar render paths to cut render cascades, memory
churn, and UI jank on large sessions and big session trees. Behavior is preserved;
the changes are about *when* and *how much* the UI re-renders.
## Chat streaming
- Freeze the streaming message's parts in the bulk turn projection during streaming,
and re-inject live parts only in an isolated tail leaf, so a ~60/sec delta stream
no longer re-runs the whole-session projection or re-renders unrelated rows.
session with referential reuse of unchanged turns.
- Memoize message rows with field-aware comparators instead of reference equality.
- Replace the manual child-session polling in the task tool with the live SSE
stream + a one-shot load, removing a fetch/settle state machine.
## History loading & scroll
- Load an initial page fast, then prepend one older page in the background so the
scroll container has headroom and "load older on scroll-up" fires before the user
hits the absolute top.
- Compensate scroll synchronously (in a layout effect, before paint) for prepends —
including background prepends that don't originate from a user scroll — so the
viewport stays stable instead of judder-correcting on the next frame.
## Markdown rendering
- Render markdown synchronously *styled* on first paint (paragraphs, lists, code
cards, tables, inline code) instead of raw escaped text; the async pass then only
upgrades syntax-highlight colors. Eliminates the flash of full-width raw text.
- Load KaTeX CSS eagerly with the main bundle instead of inside the lazy markdown
chunk, avoiding a late stylesheet injection on first render.
## Sidebar
- Hoist per-row recursive tree walks out of row comparators into per-group
precomputed sets/keys; batch live-session lookups into a single map; add a
group-level memo boundary.
- Isolate rename drafts so per-keystroke typing doesn't repaint the row tree.
## Sync layer
- Add a staleness guard so a slow message fetch can't repopulate a session the user
navigated away from.
- Throw on fetch failure for authoritative loaders so a transient blip can't read as
an empty server response.
## Cleanup
- Remove dead code (unused hooks, params, duplicated inline types) surfaced while
reworking the above.
## Known issue
- A rare, purely cosmetic first-paint width flash can still appear on large sessions;
it has no behavioral or data impact and is tracked for a follow-up runtime trace.
ArrowUp in chat input now moves caret instead of recalling history when the field has text
Removed manual cache-busting workaround and eslint-disable in favor of idiomatic onChange state sync
When switching to a session with long context (especially in Electron
desktop when changing servers), the chat viewport could render blank
until the user scrolled. Two interacting issues caused this:
1. historyVirtualRows memo never recomputed after the first render
because historyVirtualizer (from useVirtualizer's useState) is a
stable reference. Frozen range meant items rendered at the top
while paddingBottom filled the visible viewport after scrolling.
2. pendingInitialRestoreRef replay ran in useEffect (after paint),
showing a frame at scrollTop:0 with the stale virtualizer range.
Fixed by: adding a virtualVersion counter driven by useVirtualizer's
onChange to bust the memo; switching the replay to useLayoutEffect
so scroll position is set before the browser paints.
Compute review transfer state once per chat render
Hide transfer actions when linked review sessions are inactive
Remove session-list scans from individual message rows
Add a packaged-client runtime boundary so the shared UI can talk to local,
desktop, remote, and VS Code runtimes through the right transport instead of
assuming one same-origin web server.
Centralize OpenChamber-owned API access behind RuntimeAPIs, runtimeFetch, and
runtime URL helpers, while keeping official OpenCode traffic on the SDK path.
Support runtime switching, remote host selection, desktop client credentials,
and headless connection links for pairing packaged clients with remote
OpenChamber servers.
Harden the new auth model by moving long-lived client tokens out of browser
URLs, introducing short-lived scoped URL tokens for browser-owned transports,
restricting URL-token access to explicit readable/realtime routes, and making
client-token management session-scoped or self-scoped as appropriate.
Update browser-owned assets and preview proxy flows to work with the split
runtime model, including authenticated project icons, preview token propagation,
CSP-safe preview bridge injection, and preview proxy auth that survives
short-lived URL-token expiry.
Tighten Electron security boundaries for packaged clients by gating privileged
preload state to trusted origins and requiring explicit confirmation before
connect deep-links import or switch remote runtimes.
Also refresh agent guidance and project skills so future runtime/API, auth,
preview, UI, CLI, settings, locale, and drag-to-reorder work follows the new
architecture.
Fix chat history pagination and scroll preservation
Align session history loading with the expected scroll-up pagination UX while
keeping OpenChamber-specific initial message limits for constrained runtimes.
- Separate initial load sizes from older-history pagination size
- Load older messages automatically when scrolling near the top
- Continue fetching history until a visible older turn is available
- Preserve the current viewport synchronously during prepends
- Prevent history loading from fighting pinned-to-bottom follow behavior
- Remove delayed scroll-to-bottom correction that caused jumpbacks
- Fix the virtualizer fallback path that could render a large blank spacer
- Track oldest loaded message per pagination iteration to avoid redundant fetches
Improve chat session switching and history pagination, with most of the aggressive limits scoped to the VS Code webview where the freezes were observed.
Session history loading and pagination:
- Reduce the VS Code message page size to 30 records so switching sessions does not immediately hydrate large histories into the webview.
- Keep manual Load older messages in VS Code fixed at 30 records per request instead of growing the request size over time.
- Add a bounded VS Code initial-tail expansion path from 30 to 50, 80, and 120 records only when the initial page has no user-message turn boundary, preventing large final turns from rendering as an empty chat.
- Lower the normal web message page size from 200 to 150 for a mild shared optimization without adopting the aggressive VS Code limits.
- Make session pagination metadata reactive per session so ChatContainer receives cursor updates from materialization and reconnect paths without requiring a switch away and back.
- Write pagination metadata before publishing newly materialized messages so the first render sees the correct has-more state.
- Store cursor information from direct materialization and reconnect message fetches in the shared session prefetch metadata cache.
VS Code cache and memory pressure reductions:
- Use a shared per-directory session recency map so cache eviction is based on app-level recency instead of whichever useSync instance happened to run.
- Limit VS Code warm session cache retention to 4 sessions and evict heavy inactive message caches after switching away from a large session.
- Disable sidebar session prefetch in VS Code because warming extra sessions was increasing webview memory and GC pressure during navigation.
- Remove dropdown background message prefetch so opening the switcher does not start additional session materialization work.
- Drop cached session-message-record snapshots when evicting session data so stale derived records do not remain after the raw session cache is cleared.
- Add bounded LRU caching for session message record snapshots, with much smaller VS Code limits and a VS Code cap that avoids caching snapshots above 30 messages.
- Bound the turn-window model cache in VS Code and avoid caching turn models for sessions above the VS Code message-page size.
Chat render-path reductions:
- Reuse ChatContainer's already-materialized message records in plan detection instead of adding a second active-session message subscription.
- Add a no-op guard when marking session plan availability so repeated detections do not create new Map references and fan out renders.
- Add no-op guards for session switcher and dropdown open state updates to avoid unnecessary store updates and renders.
- Convert several session-specific hooks to useSyncExternalStore with empty-session no-subscribe behavior so empty IDs do not subscribe to broad store updates.
- Remount the chat viewport when the current session changes, isolating per-session viewport and list state.
- Change the virtualized message-list fallback to render only a tail window when the virtualizer has not produced rows yet, instead of rendering an entire large history.
VS Code layout and header improvements:
- Remove the broad useSessions subscription from the VS Code layout header path and subscribe only to the active session title and initial-session existence.
- Unmount the compact VS Code session sidebar when the user is in chat view instead of keeping the hidden session list mounted and subscribed.
- Compute the latest assistant model and latest context-token usage in a single reverse scan of current-session messages instead of scanning the same list twice.
- Remove switcher git-status warmup work so the switcher reads already-loaded branch labels without starting extra background git status requests.
Markdown and file-reference safeguards:
- Skip expensive syntax highlighting for very large code blocks, with a 200-line cap in VS Code and a softer 1200-line cap in web.
- Add an LRU cap to file-reference stat lookups so the cache cannot grow without bound across many rendered messages.
- Limit the number of file references annotated per render to 40 in VS Code and 200 in web to prevent large assistant outputs from spawning too many stat checks.
- Clear file-link annotations when file-reference mode is disabled so stale attributes and handlers do not remain on previously annotated nodes.
Assistant-message action and preview reductions:
- Skip preview URL scanning on VS Code, mobile, and mini-chat surfaces so assistant text and tool output are not scanned where the preview action is unavailable.
- Skip Save-as-Plan project lookup on VS Code, mini-chat, and mobile surfaces.
- Hide Save-as-Plan and Start MultiRun assistant-message actions on VS Code, mini-chat, and mobile surfaces.
- Resolve the current session directory on demand for assistant actions instead of subscribing each assistant message to the full session list.
Tool and task rendering optimizations:
- Prefer finalized task metadata summaries without fetching child-session messages when the summary is already present.
- Avoid polling or final-fetching task child sessions once a final metadata summary is available.
- Use VS Code-specific task child fetch limits of 30 records for initial, active, and idle fetches.
- Parse diff stats by scanning patch text line-by-line instead of splitting large patches into arrays.
- Count write-tool lines by scanning content instead of allocating a split array for large files.
- Avoid trimming large patch strings just to test whether they contain content.
- Memoize diff and write statistics so unchanged tool parts do not recalculate them on every render.
VS Code bridge improvements:
- Return JSON and text proxy responses through the VS Code bridge as bodyText instead of base64 so the webview avoids synchronous base64 decoding for common API responses.
- Keep binary responses on the base64 path while making bodyBase64 optional in the bridge contract.
- Strip content-length, content-encoding, and transfer-encoding headers from proxied responses because the bridge reconstructs the Response body.
Validation:
- bun run type-check
- bun run lint
- bun run vscode:build
Adds animated reveals for file path tool descriptions
Keeps sorted mode tool animations working without extra chat gaps
Preserves file icons, truncation, and click behavior
- Cache turnWindowModel per sessionId to skip rebuild on re-visit
- Compensate scroll position after virtualizer measurement settles
(RAF-based) so large uncached sessions open at bottom, not top
- Add MessageListHandle.scrollToBottom via virtualizer API
Virtualize message timeline at 5+ turns. Stabilize virtualizer keys,
add prepend-aware scroll compensation for "Load older", and defer
expanded tool bodies by one animation frame. Tighten initial turn
window to 7 turns for faster session open.
The previous system layered three hooks (useScrollEngine, useChatScrollManager,
useChatTimelineController) with overlapping responsibilities, four parallel
ResizeObservers/MutationObservers, and six entry points to "scroll to bottom"
(force-flag combinations, persistent follow loops, materialization recovery).
This produced bugs where users could not break free of auto-follow during
streaming: scrollbar drag, keyboard scrolling and find-in-page were not
detected as user intent, and observers kept restarting the follow loop on
every DOM mutation.
The new architecture replaces the two low-level hooks with a single
useChatAutoFollow that owns scroll behaviour end to end:
- One state: 'following' or 'released'. No follow modes, no pin flags,
no marker pixels.
- One scroll writer: a lerp loop that runs only while the session is
streaming and state is 'following'. Idle sessions never write scrollTop
programmatically.
- One user-intent detector: wheel up, touch drag down, keyboard
(PageUp / Home / ArrowUp), pointerdown on the OverlayScrollbar thumb,
and explicit releaseAutoFollow() calls all flip the state to 'released'.
- A 1.2s grace period after explicit release: re-pin will not auto-engage
inside this window, so a small wheel up cannot snap the user back even
while they remain near the bottom spacer.
- Re-pin and the scroll-to-bottom button share the same threshold: the
height of the empty bottom spacer (10vh on desktop, 40px on mobile).
Released users see the button only after they have scrolled past the
spacer that already exists at the end of the chat.
- Save/restore of scroll position uses ratio mapping, debounced at 150ms
on user-driven scroll events; programmatic writes are masked via a short
window so they never persist as user positions.
- Container reattachment is detected via a useLayoutEffect probe over
scrollRef.current. Listeners and observers re-bind when ChatViewport
mounts after hydration or after the first message promotes a draft
session into a real chat.
- A pending-restore queue replays restoreSnapshot once the scroll
container appears, fixing the case where a hydrating session landed at
the top instead of the bottom.
Removed: useScrollEngine.ts, useChatScrollManager.ts, the persistent
follow loop with its own ResizeObserver+MutationObserver pair, the
materialization-recovery .finally resume that yanked idle users to the
bottom on transient sync gaps, and the openchamber:session-reselected
event (re-select still works through the existing onSessionSelected
callback). The openchamber:chat-force-scroll-bottom event remains for
synthetic-message paths like git-message generation.
Net change: ~1300 lines removed, two hooks replaced with one, one
observer pair instead of four.
* fix: exclude file content from reverted prompt text
Revert and fork now restore only the user's original prompt, not server-injected file content
Uses existing isSyntheticPart helper for type-safe filtering
* fix: keep scrollbar visible when hovering over thumb
* fix: prevent ESC abort from triggering when terminal is focused
* fix: pass directory to permission/question reply calls so approvals actually resolve
* fix: default model selection not responding after Base UI migration
* fix: prevent modal content from shifting and clipping footer buttons
* fix: improve session switching performance and add sub-agent export with prompt collapse
Defer viewport anchor saving to eliminate ~800ms UI freeze when switching sessions
Add export dialog to include sub-agent tasks recursively in markdown export
Add collapse chevron button for expanded user prompts in sticky header
* fix: resolve sidebar scroll and TDZ crash in session sidebar
* perf: reduce CPU overhead and re-renders across chat, layout, and settings
* fix: position collapse button at top of message and prevent ESC abort in terminal
* fix: position collapse button at top and add padding only when expanded
* refactor: extract shared PATH utilities and mobile keyboard hook
* refactor: import shared path-utils in electron, use module-level style constants
- Electron now imports pathLooksUserConfigured/mergePathValues from
shared path-utils.js instead of inline duplication
- ToolPart collapsedCustomStyle moved from useMemo([]) to module const
* fix: resolve remaining merge conflicts and type errors
- Remove duplicate variable declarations in SessionNodeItem
- Remove orphaned export callback body from conflict resolution
- Fix HelpDialog description -> descriptionKey (i18n rename)
* fix: resolve type-check and lint errors in session-actions.test.ts
- Added missing bun:test type declarations (beforeEach, mock, mock.module)
- Removed unused State import
- Replaced 'as any' casts with proper OpencodeClient and ChildStoreManager types
- Added eslint-disable for unused _ parameter in mock function
* fix PR 1028 export and PATH edge cases
* fix startup retry exhaustion state
* remove opencode package lock change
* fix sub-session rename cancellation
---------
Co-authored-by: Bohdan Triapitsyn <artmore@protonmail.com>
* perf: drastically improve cold-start, bundle size, and streaming performance
Cold-start optimizations:
- main.tsx: Remove blocking await on prefs I/O — render immediately with
defaults, hydrate persisted settings asynchronously. Cuts 50-200ms from
time-to-first-paint.
- bootstrap.ts: Split directory bootstrap into 3 phases:
* Phase 1 (blocking): path, config, provider, session status — minimum
data needed to render UI. Mark status complete after this phase.
path.get and session.status must both succeed; they have no fallback.
* Phase 2 (deferred): agents, commands, mcp, lsp, vcs, questions,
permissions — fetched after first paint without blocking.
* Phase 3 (lazy): session messages — loaded without blocking init.
- App.tsx: Keep identical provider tree before/after init to prevent
full subtree remount when isInitialized flips. FireworksProvider and
VoiceProvider are lightweight shells; overlays deferred until init.
Bundle-size optimizations:
- App.tsx + MainLayout.tsx + VSCodeLayout.tsx: Code-split heavy views
(SettingsView, GitView, DiffView, TerminalView, FilesView, PlanView,
OnboardingScreen, SettingsWindow, MultiRunWindow) with React.lazy.
Views load on demand when user switches panels.
- vite.config.ts: Lower chunkSizeWarningLimit from 1200KB to 500KB.
Streaming render optimizations:
- streaming.ts: Throttle streaming store writes ~60Hz → ~1Hz. Busy-session
only scan (Set, O(1)).
- MessageList.tsx: Lower virtualization threshold 40 → 15.
- ChatMessage.tsx: React.memo with areRenderRelevantMessagesEqual.
- MarkdownRenderer.tsx: React.memo with explicit prop comparators.
* fix: address Greptile review feedback on bootstrap and provider tree
- bootstrap.ts: Tighten Phase 1 error guard. path.get and session.status
must both succeed; they have no global fallback.
- bootstrap.ts: Replace dead .catch() on Promise.allSettled() with .then()
that inspects individual results for errors.
- App.tsx: Keep identical provider tree before/after init to prevent full
subtree remount when isInitialized flips.
* perf: lazy-load heavy dependencies (MarkdownRenderer + CodeMirror languages)
MarkdownRenderer dynamic import:
- Move heavy implementation (marked, react-markdown, beautiful-mermaid,
react-syntax-highlighter, ~1500 lines) to MarkdownRendererImpl.tsx
- Replace MarkdownRenderer.tsx with thin lazy wrapper using React.lazy
- All 11 existing imports work unchanged — no consumer code modified
- Full markdown stack loads on first render of markdown content
CodeMirror language lazy loading:
- languageByExtension.ts: remove static imports for 10+ less-common
language packages (@codemirror/lang-go, lang-rust, lang-sql, etc.)
- Keep only 6 most common languages static: javascript, json, css, html,
markdown, python, shell
- Less common languages return null from languageByExtension, causing
callers to fall back to loadLanguageByExtension which dynamically
loads from @codemirror/language-data
- Reduces initial bundle by ~200KB+ of language parsers
---------
Co-authored-by: Shyamalan Kannan <yabuku@Shyamalans-MacBook-Pro.local>
Bohdan Triapitsyn