* feat(ai): render tool calls as summarized, highlighted cards
Every tool call showed only a status and its output as a JSON string,
so render calls hid their JSX, export_image dumped base64, and long
runs filled the transcript with identical rows.
A call now shows a one-line summary read from its input and chips that
select and zoom to the layers it touched, switching to the run's page
when needed. Expanded, it shows the JSX or script it wrote and its
JSON input and output in a read-only CodeMirror view, and exported
images inline. Render calls can be expanded while their input streams,
so the JSX appears alongside the canvas preview. Consecutive calls
beyond three fold into one row that keeps the latest call visible.
CodeMirror loads with the first expanded call. The code theme gains a
monospace fallback because the editor font variable is not always
emitted.
* refactor(ai): drop the unused tool JSON slot and place the JSX summary comment
* fix(ai): keep an opened tool call in place instead of following the output
Opening reasoning already stopped the transcript from following new output; tool calls and tool groups did not, so expanding one near the bottom re-pinned the bottom on every animation frame and slid the card away as it opened. Any disclosure in the transcript now stops following.
* fix(ai): show a pointer over chat tool calls, tool groups, and reasoning
* refactor(app): share CodeMirror setup between the code editor and viewer
CodeViewer repeated CodeEditor's view lifecycle: mounting the EditorView, label, theme, and language compartments, the app-theme watcher, and teardown. useCodeMirror owns that once; each component passes its own fixed and reactive extensions.
* refactor(ai): move tool node lookup and focusing into useToolNodes
ToolNodeChips looked nodes up in the active document and ran the show-on-canvas flow, with its superseded-switch and error handling, inside the component. The composable owns both; the component renders the chips.
* refactor(ai): derive tool call state and input once
ToolCallCard and ToolCallGroup each rebuilt classifyToolState's input from the part, and the card decided inline whether a call had input to show. toolCallState and toolHasInput own those rules beside the other per-call helpers.
* fix(app): use the thin app scrollbar in code editors and viewers
CodeMirror scrolls its own .cm-scroller, which fell back to the platform scrollbar, thick and light in the dark chat. The hosts now give it the shared scrollbar-thin utility.
* feat(code): isolate Design JSX execution
- Transform JSX with the existing Sucrase dependency and execute it in a disposable opaque-origin iframe worker
- Block ambient network capabilities and enforce source, timeout, output, depth, and element limits
- Validate that only bounded plain structured data returns to the application
* feat(code): add editable Design JSX
- Add a lazy CodeMirror editor with JSX syntax support, completion, diagnostics, and bounded scrolling
- Convert validated sandbox output into trusted Design JSX helpers before rendering
- Apply or insert JSX as one undoable graph transaction while preserving dirty drafts
* feat(code): localize JSX editor actions
- Add translated-message fallbacks for editing, applying, inserting, and draft state
- Document the editable JSX workflow in the unreleased changelog
* fix(code): satisfy typed sandbox validation
- Keep the sandbox document terminator literal and preserve optional selection handling under type-aware lint
* test(code): resolve sandbox probes through app aliases
* fix(code): bound sandbox results before cloning
- Enforce string, array, object, depth, element, and byte limits inside the disposable worker
- Retain host-side validation as a second structured-data boundary
* fix(code): keep JSX and HTML editing modes separate
- Switch generated Tailwind output back to OpenPencil JSX before editing
- Close the JSX editor when opening HTML/CSS import and avoid stacking both editors
* feat(code): support authored Design JSX programs
- Allow local constants, function components, arrays, conditionals, fragments, and multiple roots
- Preserve replacement positions for multiple selected roots and reject mixed-parent or locked selections
- Cover multi-root undo, redo, and all-or-nothing rollback
* feat(code): add explicit JSX view mode
- Let authors leave CodeMirror without applying a draft
- Keep the editable surface and HTML/CSS importer mutually exclusive
* feat(code): diagnose unknown JSX vocabulary
- Warn on OpenPencil elements and properties that are absent from the canonical schema
- Surface diagnostics inline through CodeMirror lint markers
* test(code): accept WebKit isolation diagnostics
- Cover the sandbox architecture against Playwright WebKit
- Accept engine-specific wording while preserving the same unavailable-window assertion
* refactor(code): consolidate Design JSX vocabulary
- Drive renderer warnings, completion, and diagnostics from one supported-property schema
- Recognize locally declared components and add focused schema and transform tests
- Replace CodeMirror basicSetup with an explicit feature set and remove the umbrella package
* fix(code): support Design JSX variable helpers
* refactor(code): unify JSX sandbox validation
* fix(code): account for complete sandbox output
* fix(code): preserve locked JSX descendants
* fix(code): preserve JSX sibling order
* fix(code): recompute JSX parent layouts
* feat(code): add live code previews
* test(code): centralize graph assertions
* fix(code): harden live preview sessions
* docs: describe live code editing
* fix(code): update editor accessibility labels
* fix(code): use theme-aware error colors
* fix(dev): stop Vite disconnect error loops
* fix(code): clarify live preview status
* fix(ui): avoid tooltip attribute warnings
* test(code): assert semantic preview status
* refactor(code): remove obsolete editor messages
* fix(code): harden preview concurrency and isolation
* fix(code): cancel stale reset previews
