setHtml() resets the document layout. sb->setValue(sb->maximum()) was
reading a stale maximum before relayout completed. Use moveCursor(End)
+ ensureCursorVisible() which forces layout and scrolls reliably.
Also: fix session list state updates destroying delegates (rename fix)
already included in previous commit.
- Add configurable close-to-tray (Settings > System Tray)
- QSystemTrayIcon with show/hide toggle and quit action
- closeToTray property persisted in theme.conf
- Window hides instead of quitting when enabled
- Fix session rename being interrupted by state changes
- onStateChanged no longer calls refreshSessions()
- State updates in-place via sessionStateUpdated signal
- Delegates track state locally, preserving rename TextField
- Replace Qt.callLater scroll debounce with Timer (16ms) to coalesce
rapid contentHeight changes and eliminate scroll flickering
- Add execute button (play icon) on code blocks for scripted languages
(bash, python, perl, ruby, etc.) — launches in terminal
- Fix assistant bubble width: always use 95% of available width
- Add minimum width (280px) to code fence blocks
- Move emoji popup into prompt RowLayout so it appears near input
- Disable wheel scroll inertia (direct contentY) while keeping flick
inertia intact
- Apply all fixes to both KF5 and KF6 QML files
- Tool blocks: expanded via sub-blocks (same as assistant) so embedded
diffs and code fences render properly with coloring
- Call blocks: always JSON, use formattedContent (pretty-print + highlight)
- Prose sub-blocks use Text.MarkdownText for native backtick/bold/link support
Tool results often contain markdown with code fences and diffs.
Using formattedContent treated the whole block as raw text.
Now tool blocks expand into parsed sub-blocks (prose, code, diff)
just like assistant blocks, so embedded diffs get colored.
HTML RichText collapses multiple spaces, destroying JSON indentation.
Wrap all highlighted output in <pre> tags and use real newlines instead
of <br>. pre-wrap allows wrapping at container width.
QSyntaxHighlighter applies colors to the layout only (not accessible
via QTextBlock iteration). Replace with a custom AbstractHighlighter
subclass (HtmlBuilder) that builds HTML directly in applyFormat().
This actually produces colored output.
QML scope issues made JS-based formatting unreliable. New approach:
- FormattedContentRole in the model computes pretty-printed HTML in C++
- Call blocks: always prettyJson()
- Tool blocks: detect JSON/diff and format, otherwise escaped HTML
- QML just binds text: block.formattedContent || ""
The onVisibleChanged approach failed because block.blockContent could
not resolve through the Loader scope. Now using direct binding:
- call blocks: text bound to chatModel.prettyJson(block.blockContent)
- tool blocks: detect JSON/diff, format accordingly
- prettyJson() uses QJsonDocument for proper indentation + KSyntaxHighlighting
- Only evaluated when visible (lazy)
ListView.positionViewAtEnd() causes freezes by forcing instantiation
of all delegates to measure total height. Flickable+Column avoids this
- contentHeight is computed directly from Column.implicitHeight
- Scroll to bottom is just contentY = contentHeight - height
- No virtualization needed with only 20 blocks
- Eliminates all positionViewAtEnd and _scrolling guard complexity
- Code blocks show plain text immediately, highlight via Timer after
150ms (prevents blocking the render thread during initial load)
- Reduced initial blocks from 40 to 20
- Detect append vs full reset in setRawText for better session switch
- Model shows last 40 blocks initially, positioned at end
- Scrolling near top triggers loadEarlier() which prepends 30 blocks
- Scroll position maintained after prepend via contentY adjustment
- All blocks still parsed (for searching later) but only tail rendered
- Reduced initial render cost significantly
With 300KB+ chat logs producing hundreds of blocks, instantiating all
delegates was causing unresponsiveness. Cap the model at the last 80
blocks and set cacheBuffer: 200 to limit off-screen rendering.
- Cache parseSubBlocks per block (only reparse when content changes)
- Cache highlightedHtml results (code+language key)
- Prevents expensive KSyntaxHighlighting and regex on every render
- Also fixes scroll loop with _scrolling guard
- ChatBlockModel::reparse() now does incremental diffs instead of
beginResetModel which destroys all delegates and resets scroll
- ListView scroll uses _scrolling guard to prevent infinite loop
from positionViewAtEnd triggering contentHeightChanged
- Loader delegate: set explicit height from item.implicitHeight so
ListView can calculate correct contentHeight for scrolling
- Prose sub-blocks: escape HTML, convert newlines to <br> before
linkifying paths (RichText ignores bare newlines)
Replace flat TextArea chat view with structured ListView:
- ChatBlockModel parses raw chat text into typed blocks (user,
assistant, reasoning, call, tool, error, info)
- Sub-blocks parsed from markdown: prose, fenced code, diffs
- Code blocks rendered with KSyntaxHighlighting (auto light/dark theme)
- Diff blocks with red/green line coloring
- Copy button on code and diff blocks
- File paths linkified and clickable (opens with xdg-open)
- Reasoning/call/tool blocks collapsible (click to expand)
- User blocks highlighted with subtle accent background
- Error blocks in red
- Scroll pinning preserved (auto-scroll when at bottom)
Replace single-line TextField/QLineEdit with multi-line TextArea/QPlainTextEdit.
Enter sends, Shift+Enter and Ctrl+J insert newlines. Max height capped so the
input area does not take over the panel.
Scan ~/.local/share/color-schemes/ in addition to /usr/share/color-schemes/
so user-installed themes (e.g. Catppuccin) are discovered. Mode is now any
scheme name rather than just system/light/dark. Settings dialog uses a
ComboBox listing all available schemes plus "System".
- Kate: add always-visible stop button next to prompt input
- All front-ends: sort sessions alphabetically by ID
- All front-ends: only auto-scroll chat if at bottom, show
floating go-to-bottom button when scrolled away
- CMakeLists: set KDE_INSTALL_PLUGINDIR to QT6_INSTALL_PLUGINS
so plugins install to lib64/qt6/plugins/ (where Kate looks)
m_ghost was parented to the view, so Qt parent-child cleanup could
destroy it when a view was replaced. The raw pointer remained non-null,
causing a double-free on the next viewChanged signal.
Fix by using QPointer<OllieGhostProvider> which auto-nulls on destroy.
Also replace qobject_cast with dynamic_cast for KF5 KTextEditor
interfaces to avoid crashes when qt_metacast accesses uninitialized
KateViewConfig during early view activation.
Replace KirigamiDelegates.RoundedItemDelegate (not available in all
Kirigami versions) with PlasmaComponents.ItemDelegate. Add explicit
highlightedTextColor bindings to labels and icon in the session list
delegate for both KF5 and KF6 builds.
The GhostOverlay is parented to the view widget and gets destroyed by
Qt child cleanup before ~OllieGhostProvider runs. Stop calling
hideOverlay() in destructor, connect destroyed() signal to null the
pointer.
- Add Default Backend / Default Model config for chat sessions
(overrides OLLIE_BACKEND / OLLIE_MODEL env vars)
- Pass configured backend/model to CreateSession D-Bus call
- Move "Completion enabled on startup" checkbox below text fields
- Deduplicate status bar indicator when multiple views exist
- Make install target depend on build
- Remove Escape-to-dismiss (conflicts with Kate popups)
- Fix parent not passed to InlineNoteProvider causing provider recreation
on every viewChanged signal (killed debounce timer, no completions fired)
- Add view() accessor, use it in setupGhostProvider check
- Null-check focusProxy() to prevent crash during early viewChanged
- Multi-line completions: first line as inline note, continuation lines
rendered via GhostOverlay floating widget
- Guard event filter and signal handlers against active completion popups
- Add suggestion cycling (Meta+O,N / Meta+O,P) with history
- Add Meta+O,C to toggle completions on/off with status bar indicator
- Add Meta+O,D to dismiss current suggestion
- Add Meta+O,Tab to manually trigger a completion
- Add config page for completion backend and model override
- Pass backend/model config via D-Bus Complete context parameter
Adds OllieGhostProvider that renders AI completions as translucent
gray text after the cursor. Uses the new Complete D-Bus method with
500ms debounce. Tab accepts, Escape dismisses, any edit clears.
Works on both KF5 and KF6 builds.
Levi Neely
Levi Neely