- Replace regex-based list and error parsing with xmldom
- Decode entities through the parser and support namespaced pagination responses
- Treat malformed listing documents as empty pages
- Store desktop secrets in native OS credential services through a narrow Tauri bridge
- Provide session-only and encrypted remembered browser stores with explicit switching
- Migrate legacy plaintext keys only after verified writes and cover failure recovery
- Pin websocket-driver 0.7.5 across the Trystero Firebase dependency chain\n- Restore a clean critical dependency audit\n- Reuse the shared font family/style shape exposed by scene-graph
- Introduce Nuxt UI-style Tailwind Variants themes and 26px panel primitives\n- Align section and field action rails across labeled property rows\n- Add Storybook with dark/light states, accessibility tooling, and CI build coverage\n- Standardize UI acronym casing across app and Vue SDK types
- Add a repo-wide TypeScript AST duplicate shape check to the quality gate
- Extract shared named types for existing cross-file duplicates
- Keep same-file oxlint coverage for editor feedback
- shapeTextForClipboard() now processes all lines, not just the first
- buildFigmaClipboardHTML() passes CanvasKit-shaped positions to
derivedTextData when available (browser with CanvasKit loaded)
- Add @chenglou/pretext for proper Unicode-aware word wrapping as
fallback when CanvasKit isn't available (CLI, headless, .fig export)
- Falls back to basic word-boundary wrapping if pretext fails
- Configure oxfmt custom import groups for workspace, app, package, and test aliases
- Keep type imports grouped with their matching source category instead of one global tail group
- Expand the format script to cover formatter config, Vite files, and scripts
- Merge structural lint rules into oxlint.json and remove the second config file
- Add #tests/* path aliases and rewrite deep test helper imports
- Ban new deep parent-relative imports through the shared lint config
Use @ai-sdk/deepseek instead of openai-compatible, which properly
handles reasoning_content in conversation history. DeepSeek V4
requires reasoning_content on every assistant turn — the dedicated
provider backfills it automatically.
Adds DeepSeek V3 and R1 models to the provider list.
Vector PDF output via jsPDF + svg2pdf.js (lazy-loaded, 124KB gzipped).
Reuses existing SVG export pipeline — text stays selectable, paths
stay sharp at any zoom.
- packages/core/src/io/formats/pdf/ — SVG→PDF conversion module
- export_pdf tool for MCP/AI
- PDF format in IO registry and app export panel
- CLI: bun open-pencil export file.fig --format pdf