* feat(tui): treat paste markers as atomic segments in editor
Paste markers like `[paste #1 +123 lines]` are now treated as single
atomic units for cursor movement, word navigation, deletion, and
wrapping. Only markers with valid paste IDs (present in the editor's
pastes Map) are treated atomically.
* fix(tui): word-wrap oversized atomic segments in editor
* feat(coding-agent): add fold/unfold to tree branch navigation
Remap left/right arrow keys in the session tree to jump between branch
segment starts (first child of each branching point). Left on a foldable
node (root or segment start) folds it, hiding its descendants. Right on
a folded node unfolds it. Nested fold state is preserved. Search and
filter changes reset all folds.
Folded branches show ⊞ in the connector (├⊞/└⊞) or as a prefix for
roots. Active path marker • appears after the fold indicator.
Add Page Up/Down support for paging through the tree list, consistent
with session selector and config selector.
* feat(coding-agent): improve tree branch folding closes#1724
---------
Co-authored-by: Sviatoslav Abakumov <dust.harvesting@gmail.com>
When Kitty keyboard protocol is not available (e.g. inside tmux), fall
back to xterm modifyOtherKeys mode 2 so that Shift+Enter, Ctrl+Enter,
and other modified keys are distinguishable from plain Enter.
tmux users need to add to ~/.tmux.conf:
set -g extended-keys on
set -g extended-keys-format csi-u
closes#1872
The previous lazy-loading change switched to require("koffi") inside
enableWindowsVTInput(). In ESM, require is undefined, so the call
threw and VT input mode was silently not enabled on Windows.
Use createRequire(import.meta.url) at module scope and
cjsRequire("koffi") at the call site. This keeps koffi externalized
for Bun binaries while restoring Windows VT input behavior, including
multiline paste handling.
Changed koffi import from top-level to dynamic require in
enableWindowsVTInput() and added --external koffi to bun build.
This prevents embedding all 18 platform .node files (~74MB) into
every compiled binary. For Windows builds, only the win32_x64
koffi.node is shipped alongside the binary.
Binary size reduction: darwin-arm64 142MB -> 67MB, archive 43MB -> 28MB.
Add pasteToEditor(text) method that pastes text into the editor via
bracketed paste sequences, triggering paste handling (including collapse
for large content). Unlike setEditorText which directly replaces content,
pasteToEditor routes through handleInput on the active editor component.
- Add pasteToEditor to ExtensionUIContext interface
- Add handleInput to EditorComponent interface (was missing, all
concrete implementations already had it)
- Implement in interactive mode via bracketed paste sequence
- Add fallback in RPC mode (delegates to setEditorText)
- Document in extensions.md
* feat(tui): extract KillRing and UndoStack, add to Input
Extract kill ring and undo logic from Editor into reusable classes:
- KillRing: ring buffer with accumulation for consecutive kills
- UndoStack<S>: generic stack with clone-on-push semantics
Refactor Editor to use both classes. Add kill ring (kill/yank/
yank-pop), undo with coalescing, and deleteWordForward to Input.
* feat(tui): extract handleBackspace() and handleForwardDelete()
Adds Terminal.prepareForExit() to disable Kitty protocol and wait for
in-flight release events before fully stopping. This prevents escape
sequences from leaking to the parent shell over slow SSH connections.
Fixes#1204
- Remove height change detection (only width changes trigger full redraw)
- Change clearOnShrink default to false (use PI_CLEAR_ON_SHRINK=1 to enable)
- Fix viewport check to use previousLines.length instead of maxLinesRendered
(prevents false positive redraws when appending lines after content shrunk)
- Add clearOnShrink setting to /settings in coding-agent
- Remove line truncation in custom message component (always show full content)
With the Kitty keyboard protocol (flag 4), terminals report both the
logical key codepoint and the physical base layout key (PC-101 QWERTY
position). The key matching logic in matchesKittySequence matched
against both unconditionally, which caused a single keypress to match
multiple shortcuts on remapped keyboard layouts.
For example, on a Dvorak layout with xremap, pressing Ctrl+K sends
CSI 107::118;5u (codepoint=107 'k', base layout=118 'v'). The
unconditional base layout match made this also match Ctrl+V, which is
bound to pasteImage. Since CustomEditor checks app-level keybindings
(pasteImage) before editor keybindings (deleteToLineEnd), Ctrl+K was
silently intercepted by the paste image handler instead of deleting
to end of line. The same issue affects symbol keys, as Dvorak also
remaps symbols to different physical positions (e.g., '/' sits where
'[' is on QWERTY).
The fix restricts base layout key matching to cases where the codepoint
is not a recognized Latin letter (a-z) or symbol (/, -, [, ;, etc.).
When the codepoint is already a recognized key, it is authoritative and
the base layout key is ignored. This preserves non-Latin layout support
(Ctrl+К on a Russian layout still matches Ctrl+K via base layout key
107) while preventing false matches from differing physical key
positions on remapped layouts.
Both matchesKittySequence and parseKey are updated with the same logic.
When moving up/down through lines of varying lengths, the editor now
remembers the original column position and restores it when reaching a
line long enough to accommodate it.
Example: cursor at column 10, move up to a shorter line (cursor clamps
to end), move up again to a longer line - cursor returns to column 10.
Implementation:
- Add preferredVisualCol instance property (nullable)
- Set it when moving to a shorter line during vertical navigation
- Clear it when arriving at a line that fits the preferred column
- Clear it on any horizontal movement or editing via setCursorCol()
- Detect line rewrap by checking if cursor is in middle of line
- When pressing right at end of prompt, set preferredVisualCol so
subsequent up/down navigation uses that column position
- Add height change detection (analogous to existing width detection)
- Auto-detect when content shrinks below maxLinesRendered and trigger
full re-render to clear leftover empty rows
- Only triggers when no overlays are active (overlays need padding)
Fixes empty rows appearing below footer when:
- Closing /tree or other selectors
- Clearing multi-line editor content
- Any component shrinking
Also adds regression tests for resize handling and content shrinkage.
Changed isImageLine() from using startsWith() to includes() to detect
Kitty and iTerm2 image escape sequences anywhere in a line, not just
at the start. This prevents TUI width checks from failing on lines
containing image data, which could cause crashes when rendering tool
results with images (e.g., when reading image files).
Also added comprehensive test coverage for isImageLine() including:
- Both iTerm2 and Kitty protocols
- Regression tests for long lines and terminals without image support
- Negative cases to ensure no false positives
Fixes crash: 'Rendered line exceeds terminal width' when image
escape sequences appear in output.
* fix(tui): blockquote multiline rendering and wrapping
Fix two bugs in blockquote rendering:
1. ANSI codes split incorrectly on newlines - when text tokens contain
newlines, applyDefaultStyle() wrapped the entire string including the
newline in ANSI codes. Any caller splitting by \n would break the codes.
Fixed by splitting text by newlines before styling in renderInlineTokens().
2. Wrapped blockquote lines lost the │ border - long blockquote lines
that wrapped to multiple lines only had the border on the first line.
Fixed by wrapping within the blockquote case and adding border to each line.
* test(tui): add test for inline formatting inside blockquotes
When cursor is on the first visual line and up is pressed, jump to
start of line instead of doing nothing. When cursor is on the last
visual line and down is pressed, jump to end of line instead of doing
nothing. This matches standard behavior in most native text inputs.
