- 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
Replace the react-markdown/Prism component tree with an HTML-string
pipeline (marked -> KaTeX -> Shiki -> DOMPurify -> DOM decorators) patched
into the DOM via morphdom, with per-block reconciliation and a paced
streaming reveal. Cuts streaming CPU versus the previous renderer while
preserving file-reference links, mermaid diagrams, table export, and
agent/skill/favicon link handling. Public MarkdownRenderer/
SimpleMarkdownRenderer props are unchanged (drop-in).
* 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>
- Drop characterData from scroll-follow and file-link MutationObservers;
ResizeObserver already catches content growth.
- extractMermaidBlocks: bail out when markdown has no 'mermaid' at all.
- MarkdownCodeBlock: skip Prism highlighting while code is actively
changing, render plain pre until settled for 300ms.
* feat: implement LaTeX math rendering via KaTeX
- Add katex, remark-math, and rehype-katex dependencies
- Integrate KaTeX CSS with light/dark theme color overrides
- Wire remark-math and rehype-katex plugins into MarkdownRenderer
- Support inline ($...$) and display ($$...$$) math delimiters
- Add plan and task documentation
* fix(katex): display inline error messages for invalid LaTeX
- Configure rehypeKatex with throwOnError: false to render errors inline
- Use destructive color var for error text
- Update feature documentation with verification results
* chore: remove latex PR artifacts and hardcoded color fallback
---------
Co-authored-by: Bohdan Triapitsyn <artmore@protonmail.com>
- Unify send and session resumes around the latest chat tail
- Keep smooth follow active during assistant message growth
Remove staged chat rendering from the main scroll path
* feat(chat): add arrow key navigation for thinking mode in model selector
Adds keyboard arrow key navigation for thinking mode selection within the model picker, enabling a fully keyboard-driven workflow for model and variant selection.
Changes:
- Model Selector: Added ←/→ arrow key handling to cycle thinking variants while picker is open
- Introduced transient state to track per-model variant selections during picker lifecycle
- Updated metadata display logic to show thinking mode only after adjustment with arrow keys
- Added contextual footer hints that appear only when highlighted model supports thinking variants
- Modified Enter key behavior to apply both selected model and pending thinking variant together
- Help Dialog: Added in-picker navigation shortcuts (↑↓ navigate, ←→ adjust thinking)
- Documentation: Updated CHANGELOG.md files with feature entry
Behavior:
- Navigate models with ↑↓, adjust thinking with ←→, press Enter to apply both
- Footer hints update dynamically based on highlighted model capabilities
- Thinking display persists only for models adjusted with arrow keys
- No-op for models without variants (clean, no feedback)
Testing:
- Type-check, lint, and build all pass
- Manual testing verified on dev server
* fix: allow native context menu on links in chat messages
The desktop app was blocking the context menu on all elements except
specific allowlisted ones (terminal, input, textarea, etc.). This
prevented users from right-clicking on HTTP links in chat messages
to access the 'Open Link' option.
Added 'a' (anchor) tag to the allowlist to restore native context
menu functionality for links.
Fixes#708
* fix: use tauri.shell.open for external links in desktop app
- Add global window.open override to init_script that routes HTTP/HTTPS
URLs through tauri.shell.open() instead of window.open()
- Add openExternalUrl utility that prefers tauri.shell.open with window.open fallback
- Add openExternalUrl to MarkdownRenderer for link safety.onLinkCheck
- Replace window.open with openExternalUrl in ProvidersPage for OAuth URLs
Fixes#708
* fix: unify external link opening across desktop and UI
Added a shared URL opener that only allows http/https links.
Replaced duplicated Tauri/window link-open logic in key UI sections.
Removed fragile desktop window.open override and markdown external-link modal behavior.
---------
Co-authored-by: Bohdan Triapitsyn <artmore@protonmail.com>
* refactor: add canonical turn projection pipeline for chat
Centralizes turn/activity/summary derivation, migrates render consumers to shared projection sources, and adds projector regression coverage for retries, malformed parent fallbacks, and activity segmentation.
* refactor: stream assistant text live in turn-first render path
* refactor: complete phase-3 turn timeline history/navigation
* refactor: complete phase-4 turn pipeline cutover
Remove legacy grouping adapters and lock key edge cases with executable regression tests so turn-centric rendering stays stable across retry/history/navigation flows.
* fix: align chat history loading path with opencode parity
* fix: stabilize chat streaming and reduce render churn
* feat: add blurIn streaming text animation and smooth lerp-based auto-scroll
Streaming markdown now uses Streamdown blurIn animation (150ms, ease-out) for per-word reveal
Auto-scroll during streaming uses continuous exponential smoothing (lerp) instead of restarting spring animations
One-shot scroll-to-bottom (button click) uses motion spring for natural deceleration
* fix: respect user scroll-up during streaming and re-pin on scroll back to bottom
* fix: stabilize activity stream and tool progress rendering
Activity now stays chronologically consistent and avoids duplicate reasoning/justification.
Tool rows show running state and duration earlier with smoother pending behavior.
Removed legacy trim/timestamp paths to prevent message drops and simplify Activity UI.
* fix: reverted regression of revert functionality
* fix: stabilize chat activity layout and tool progress behavior
Stopped user messages from re-animating when new user messages appear.
Made Activity and aggregated tool rows wrap inline consistently without misaligned headers/icons.
Fixed tool part matching so duration and shine no longer reset across pending/in-progress status updates.
* fix: restore activity mode behavior and stabilize tool rendering
Re-enabled Collapsed/Summary/Detailed/Changes defaults for Activity and tool expansion.
Fixed tool row wrapping/alignment, removed user fade re-animations, and disabled FadeInOnReveal animations.
Improved tool status handling to reduce timer/shine resets and made reasoning bold render as normal text.
* refactor: simplify chat tool and message rendering
Replace activity-group rendering with flat inline part rendering
Remove obsolete tool output and justification activity settings
Keep user message animation and improve edit/apply-patch file path headers
* style: narrow chat and composer columns equally
* fix: Tool status checks and abort flow regression
* fix: improve multi-file apply patch header rendering
* fix: make timeline message jump scroll to selected chat entry
Conversation Timeline clicks now reliably scroll to the selected message in chat
Timeline dialog waits for navigation result before closing to avoid silent failures
Chat message scrolling now resolves the active chat scroller consistently
* chore: remove branch-added chat tests and clean lint leftovers
Removes newly added chat and message test files from this branch
Fixes lint blockers in ProgressiveGroup, ToolPart, and messageStore
Keeps type-check, lint, and build passing after cleanup
* fix: keep composer status row closest to the input
Status/todo row now renders below queued messages and attachments.
Linked GitHub issue/PR chips stay above status so input context order is consistent.
Improves chat composer readability while assistant is working.
* fix: disable assistant header animations and polish chat status behavior
Assistant message header now renders without transition effects
Removed a lint issue from the working placeholder props
Chat status and message flow updates improve consistency during active turns
* feat: increase bottom spacer in mobile chat for better scrolling
* feat: add sorted chat mode with progressive activity rendering
Adds global chat render mode settings for sorted and live behavior
Renders non-stop assistant output in Activity as messages complete
Disables streaming text animation in sorted mode and restores classic reasoning/justification blocks
* fix: stabilize chat timeline navigation and older history loading
Prevents auto-loading older history from scroll and keeps loading behind the button
Fixes timeline jump-to-message behavior that snapped back to the bottom
Reduces chat list flicker and empty-state glitches on large session pagination
* fix: stabilize sorted activity text and hide reasoning timers
Prevented sorted justification from appearing before message completion by requiring explicit non-stop finish.
Removed duration values for sorted Thinking and Justification blocks.
Kept smooth sorted auto-follow behavior tuned to reduce bottom-jump jitter.
* refactor: simplify chat tool rendering paths and add parts docs
Centralized tool icon mapping into a shared presentation helper used by grouped and expandable rows.
Removed static-tool-specific branches from ToolPart to keep it focused on expandable tools.
Added a new DOCUMENTATION.md explaining where to change tool descriptions and rendering behavior.
* feat: enhance event stream handling for message parts
- Refactor event key building functions to improve clarity and reduce redundancy.
- Introduce `applyPartDelta` method in message store for handling delta updates to message parts.
- Update `useEventStream` to utilize the new `applyPartDelta` method for processing incoming deltas.
- Enhance text extraction logic to accommodate new part types and improve merging behavior.
- Implement logic to skip delta-driven updates for ongoing text/reasoning parts to prevent overwriting.
- Adjust server-side text merging logic to handle trimmed text comparisons and improve consistency.
* feat: enhance OpencodeService with delta handling and stale delta management
- Introduced globalSseStaleDeltas to track stale deltas.
- Added methods to generate keys for delta and updated part events.
- Updated global SSE loop to skip processing stale deltas.
- Refactored global SSE event handling to improve performance and reliability.
refactor: remove unused streaming part handling in messageStore
- Eliminated streaming part queue and associated logic.
- Simplified message merging logic by removing redundant checks.
- Streamlined message handling to improve clarity and maintainability.
chore: clean up unused streaming text normalization functions
- Removed deprecated functions related to streaming text identity and merging.
- Simplified normalizeStreamingTextPayload to focus on current functionality.
* refactor: simplify streaming event passthrough
Use SDK-native global event streaming in the UI client
Remove dead server-side SSE normalization helpers
Keep streaming payloads closer to upstream behavior
* revert: remove message memory settings UI
Drop the restored message memory settings screen
Keep session settings focused on current defaults
Preserve chat history limits as internal behavior
* feat: add compact collapsed Activity preview in sorted mode
Collapsed Activity now shows a live preview of the latest 7 entries
Older activity rows roll up behind a +N more indicator
The +N more indicator is clickable to expand the full Activity
* fix: adjusted activity changed files text prominence
* fix: make sorted mode snap to bottom when not streaming
* perf: show chat scrollbar only on user-driven scrolling
Chat scrollbar now stays hidden during automatic follow-to-bottom
Scrollbar visibility is gated by real user scroll intent
Reduced scroll overhead by removing timer-based intent tracking
* refactor: align web and vscode session activity streaming
Route VS Code SSE through SDK-based proxy flow
Restore controlled web session activity tracking and cooldowns
Keep chat activity behavior consistent across runtimes
* fix: adjusted muted foreground themes color
* fix: stabilize session activity and attention updates
Keep streamed chat state in sync as sessions complete
Restore reliable attention indicators for inactive sessions
Tighten server and UI event handling for chat updates
* feat: add chat render previews and file icon visibility setting
Added animated chat render mode previews in Chat settings.
Added a new toggle to show or hide file icons across tool rows.
Persisted chat render, activity, and mermaid modes to shared settings file.
* fix: update default GitHub OAuth client ID
Use the new OpenChamber org GitHub OAuth app ID by default.
Keep web server and VS Code auth defaults in sync.
* fix: sync chat settings across VS Code chat views
Broadcast settings changes to both sidebar and editor chat webviews.
Apply synced settings in each webview via shared settings refresh.
Ensures chat render mode updates appear immediately in editor tabs.
* fix: stop writing session folders file in VS Code workspaces
Disable disk persistence for session folders in VS Code webview runtime
Keep session folder state in VS Code/webview storage instead of project files
Preserve existing behavior for non-VS Code runtimes
* style: polish expanded tool panels and bash output readability
Removed extra borders/labels in expanded tool cards for a cleaner UI
Added softer scroll shadows and full-height decorative activity line
Switched bash command/output to code-style typography with tighter spacing
* feat: add default-open tool controls for chat activity
Added Chat settings toggles to open Bash and Edit tools by default
Default-open behavior now applies across both sorted and live activity rendering
Users can still manually collapse default-open tools, plus bash output height was reduced
* fix: align chat loading skeleton with message layout
Match skeleton width and horizontal alignment to chat message column
Add one-line tool-style skeleton rows with subtle circular placeholders
Increase top spacing so loading state no longer touches chat header
* fix: prevent blank screen when adding inline diff comments
Sanitize persisted inline comment drafts to ignore invalid data
Harden diff annotations against malformed line numbers
Fail safely when annotation updates error instead of crashing the view
* fix: streamline VS Code archived sessions sidebar
Show archived sessions in VS Code only for the active workspace directory.
Render archived sessions as a flat list in VS Code without project-root subfolders.
Hide the "Open in Side Panel" session action in VS Code.
* docs: clarify named and quick Cloudflare tunnel usage
* feat: make chat file paths openable from rendered responses
* perf: speed up chat file-path links and open behavior
* fix: open chat file references at mentioned lines
* fix: prevent context panel flicker on blocked file opens
* refactor(settings): new IA shell + projects section + skills catalog discoverability
* chore(settings): split providers list by scope; show user before project
* fix: navigation flow in mobile Settings
* feat: redesign settings pages to use modern elevated surface patterns
* feat: replace helper text with tooltips in settings
* ui: redesign update dialog and fix external link routing
- Restructures UpdateDialog to focus on changelog readability with a wider max-w-4xl canvas
- Highlights @username contributor mentions with theme primary color
- Strips excessive vertical padding and right-aligns compact action buttons
- Disables streamdown's internal link safety dialog in favor of direct Tauri shell routing
* feat: refactor Git identities into dedicated Git settings page
* feat: unify sidebar background styling across VS Code and web/mobile
* fix: adjust button styling and layout for mobile settings pages
* feat: add MCP settings page and sidebar
* feat: hide models in provider view (thanks to @nguyenngothuong)
* feat: add "Add new provider" option to model selector dropdown
* fix: local evroc logo + provider dropdown icons
* fix: increase width of provider menu
* fix: dark theme background color for better contrast
* feat: update @opencode-ai/sdk dependency to v1.2.10
* fix: restore session sorting to only use updated time
* fix: added settings for sessions deletion dialog
* fix: adjust padding on settings pages for better layout
* fix: standardize select dropdown height across UI
* fix: agent selector UI and notification settings
* fix: remove redundant helper text from settings pages
* fix: update UI layout for description fields
* fix: remove border-none and shadow-none from textarea classes
* fix: enable context menu on sidebar items
* feat: refactor UI controls and layout patterns across settings pages
* fix: use headerless blocks when page title already provides context
* fix: remove subtask option from command settings
* fix: refactor mcp page settings
* fix: reduce spacing in skills configuration pages
* feat: refactor voice settings
* feat: refactor settings sidebar sections
* feat(chat): add mermaid preview popups and fullscreen diagram viewer
Enable opening Mermaid diagrams from markdown and file attachments with a dedicated fullscreen dialog, while tightening preview loading behavior and sizing for more reliable interaction.
* refactor(chat): extract shared preview overlay hooks
Centralize fullscreen preview transition and viewport lifecycle logic so image and Mermaid dialogs stay behaviorally aligned while reducing maintenance overhead.
* fix(security): enforce workspace boundaries for fs read endpoints
Validate /api/fs/read and /api/fs/raw paths against active workspace roots and canonical realpaths to block traversal and symlink escapes before serving file contents.
* fix(chat): remove dead Mermaid copy component
Drop an unused merge-leftover component in MarkdownRenderer to keep lint clean without changing Mermaid preview behavior.
* feat: show current branch in empty chat state
* fix: display current git branch for worktrees and update them with branch change
* refactor: improve read tool output parsing with structured data
* feat: add support for message part delta events
* fix(chat): improve streaming rendering, scroll behavior, and assistant action visibility
* feat: Add mermaid diagram support to chat markdown rendering
* fix: update table download functionality to include success notification and remove unused MarkdownRenderer import
* refactor: streamline Streamdown component props for improved readability
* fix: preserve Streamdown code-block markers and use native Tauri cache clearing
* feat: add context overview panel to view conversation details
* feat: added themes system
* feat: smart sidebar auto-hide for files/diff tabs + lower files sidebar threshold
* feat: added Checkbox component and update theming
- Add reusable Checkbox component for toggles across UI
- Replace several inputs with Checkbox in settings and commands panels
- Add DiffIcon and apply surface/border theming to key UI areas
* feat: Add convert-vscode-theme.cjs to convert VS Code themes to OpenChamber format
* refactor: remove unused permission logic from ChatInput
- Remove unused permission rules parsing logic from ChatInput
- Memoize renderTheme in DiffWorkerProvider to avoid unnecessary recalculations
- Remove forceOpaque helper in vscode theme adapter
* fix: guard VSCode theme loading in MarkdownRenderer
* feat: add custom user themes loading and reload
- Load user themes from ~/.config/openchamber/themes at runtime
- Expose /api/config/themes to fetch custom themes
- Allow theme reloading from Settings → Theme → Reload themes in the UI
Update all import statements to use @opencode-ai/sdk/v2
Modify Vite configurations to alias new SDK path
Update API client method signatures for SDK v2 compatibility
Bohdan Triapitsyn