- Keep only the active assistant response in streaming Markdown mode.\n- Remount completed responses in static mode to release incremental AST caches.\n- Preserve final Markdown, code rendering, and theme behavior.
- Show complete user messages with every image immediately after sending
- Analyze up to four bounded images without adding pixels to Design context
- Organize image handling under the extensible attachment domain
- Redact and bound provider error details in copied logs
- Avoid repeat toasts after locale changes
- Show reasoning effort only for supported providers
- Report credit and output-limit failures through localized toasts and debug logs
- Persist optional provider-specific reasoning effort on model profiles
- Classify insufficient-credit connection tests and cover failure diagnostics
- Rename first-party API, RPC, JSON, CORS, SVG, JSX, and related identifiers to preserve acronym casing
- Keep upstream and serialized boundary names unchanged
- Add a lint guardrail and migration notes for exported APIs
- Add one themed placeholder primitive for page, panel, dialog, and setup states
- Vertically center chat, code, assets, variables, and storage placeholders
- Let Vue replace the static boot splash instead of mutating global DOM from views
- Use VueUse for object URLs, timers, and clipboard state
- Replace custom collection and equality helpers with es-toolkit
- Use ofetch and safeDestr at remote API and JSON boundaries
- Add AST-based lint checks for broad unknown object assertions and local JsonObject aliases
- Centralize JsonObject in core and package-local MCP RPC JSON typing
- Replace baseline Record<string, unknown> assertions with named shared/domain types
* fix(acp): show install errors for missing agent CLIs (#172)
* Use npm for install commands, toasts for errors, tighten error matching
- Replace bun add -g with npm i -g (universal baseline)
- Replace initError inline banner with toast.show() (consistent with rest of app)
- Remove overly broad 'not found' / 'no such file' from isMissingCommandError
* Refactor toast API to toast.info/warning/error
- Add scoped command definitions for claude-code-acp, codex-acp, gemini
in Tauri shell plugin capabilities (required for Command.create)
- Show error banner in ChatPanel when transport initialization fails
instead of silently swallowing the error
Implement Agent Client Protocol (ACP) integration so users can use their
existing Claude Code, Codex, or Gemini CLI subscriptions directly in
OpenPencil's AI chat instead of managing separate API keys.
- Add ACPAgentDef/ACPAgentID types and ACP_AGENTS registry to core constants
- Create ACPChatTransport implementing Vercel AI SDK ChatTransport over ACP
- Spawn agent as subprocess via tauri-plugin-shell, communicate over stdio
- Map ACP session/update notifications to UIMessageChunk stream
- Update provider selector to show ACP agents alongside API-key providers
- Hide API key fields for ACP providers in setup/settings UI
- Add tauri-plugin-shell dependency and capability permissions
- ACP agents only appear in Tauri desktop builds (subprocess required)
Engine:
- layoutSizingVertical/Horizontal setters now use child's own layout
direction (not parent's) to map primary/counter axis — matches Yoga
- FILL on cross-axis uses alignSelf:Stretch instead of flexGrow
(flexGrow only works on main axis, caused even distribution bug)
- FILL on main axis auto-sets layoutGrow=1 for leaf nodes
This fixes the 20-step debugging spiral where AI set FIXED heights
on HORIZONTAL children inside VERTICAL parent but Yoga ignored them
(wrote primaryAxisSizing via parent axis, Yoga read counterAxisSizing
via child axis — never saw the FIXED value).
Describe noise reduction:
- Suppress 'absolutely positioned inside flex' — intentional for
decorative overlays (selection handles, badges)
- Suppress 'justify=between with HUG' when node fills parent width
- Suppress 'fixed size + grow' when sizing is FILL (standard pattern)
- 'Inconsistent padding' requires 3+ siblings (not 2)
Prompt:
- Bail-out: delete and re-render after 2 failed set_* attempts
- eval restricted to operations, not layout debugging
Previously the typing indicator only showed during 'submitted' status
(before first token). Once streaming began, gaps between tool calls
had no visual feedback — user saw silence and thought AI stopped.
Now isThinking is true when:
- status is submitted (waiting for first token)
- status is streaming and last part is a completed tool call or
step-start (AI deciding what to do next)
- Track per-run step count, warn AI when 5 steps remain via _warning
field injected into tool results
- Show 'Continue' button in chat when step limit (50) is exhausted
- Add step budget section to system prompt so AI plans ahead
- Reset run step counter on each new agent call via prepareCall
- Mark transport dirty instead of resetting chat on settings change
- Rebuild Chat with new transport while preserving message history
- Watch apiKey and customBaseURL changes in addition to model/provider
- Fix provider select z-index inside settings popover (z-52 + isolate)
- Prevent popover dismiss when interacting with select dropdown
Rendering & layout:
- Fix COUNTER_ALIGN_MAP stretch mapping
- Direction-aware fill sizing based on parent flex direction
- Text without explicit width defaults to WIDTH_AND_HEIGHT auto-resize
- Padding auto-enables vertical auto-layout
- clipsContent propagated as Overflow.Hidden to Yoga
- Text height measurement for HEIGHT auto-resize mode
- Export x/y for absolute children, text w/h per textAutoResize
- Export STRETCH as fill on cross axis
Undo system:
- Undo for auto-layout and layer tree reorder
- Atomic undo for drag + reparent
- Page snapshot undo for AI tool operations
- mouseleave no longer terminates active drag
AI tools & vision:
- Add get_jsx, diff_jsx, and describe tools
- export_image returns image via toModelOutput
- Chunked uint8ArrayToBase64 (stack overflow fix)
- computeAllLayouts after AI tool execution
- onBeforeExecute/onAfterExecute receive ToolDef
Provider & UI:
- Add anthropic-compatible provider
- API type toggle for OpenAI-compatible
- Debug toolbar (dev only)
- Preserve chat on UI toggle
- Enhanced system prompt with JSX reference
Co-authored-by: Anton A S <eddclyde@yandex.ru>
Sort classes using the same algorithm as prettier-plugin-tailwindcss,
configured for Tailwind v4 (reads src/app.css for theme).
Also sorts class strings inside twMerge() and tv() calls.
Eliminates 2 nested ternaries that conflicted with the formatter
(oxfmt strips parentheses that the lint rule requires).
- Move svg-export-{defs,paths}.ts into svg-export/ folder
(defs.ts, paths.ts, index.ts) instead of prefix-based split
- Remove duplicate getChildren lambda in fig-import.ts importPages
- Remove dead dashPattern local in kiwi-convert.ts nodeChangeToProps
- Restore useful explanatory comments in instance-overrides.ts
(DSD propagation, direct vs cloned assignments, seed skipping)
- Rename applyEffectOverrides → applyShapeAndEffectOverrides
(handles polygon props + shadow + blur, not just effects)
- Format src/ with oxfmt
* Add multi-provider AI support (Anthropic, OpenAI, Google AI, OpenAI-compatible)
- Add AI_PROVIDERS registry with per-provider model lists, key placeholders, and URLs
- Refactor use-chat composable: provider factory creates the right AI SDK model
- Per-provider API key storage in localStorage with automatic legacy migration
- New ProviderSetup.vue replaces APIKeySetup.vue with provider selector
- New ProviderSettings.vue popover accessible from gear icon in chat input
- OpenAI-compatible provider with custom base URL and model ID fields
- Install @ai-sdk/anthropic and @ai-sdk/google dependencies
- Update E2E tests for new provider setup flow
* Fix reactivity, key masking, missing model ID field, and chat reset on model change
- Use ref instead of computed for apiKey — computed getter over localStorage
wasn't triggering reactivity when the key was set
- Load stored key for new provider in providerId watcher
- Reset chat on modelId and customModelId change, not just provider switch
- Add custom model ID field to ProviderSetup.vue for OpenAI-compatible
- Simplify ProviderSettings key field — empty input with contextual placeholder
instead of fragile dot-masking that broke on partial edits
- Use data-test-id locator for model selector in E2E test
* Use useLocalStorage from vueuse instead of manual ref + watch + localStorage
Replaces 5 hand-rolled ref/watch/localStorage sync pairs with useLocalStorage.
API key uses a computed storage key that rebinds when the provider changes.
* Use useLocalStorage for collab name persistence
* Review fixes: changelog placement, remove dead class, cursor-pointer
* Extract ProviderSelect component and uiInput helper
- ProviderSelect.vue: shared select dropdown for AI providers
- ui/input.ts: shared input styling (sm/md sizes)
- Remove 7 duplicated input class strings across chat components
* Use uppercase acronyms: keyURL, customBaseURL, supportsCustomBaseURL
* Match select dropdown width to trigger via --reka-select-trigger-width
Move min-w-[var(--reka-select-trigger-width)] into selectContent base
style so all selects get it by default. Remove per-component overrides.
* Uppercase acronyms in variable names, improve setup form layout
- providerID, modelID, customModelID, setAPIKey, AIProviderID
- Keep apiKey lowercase at start (standard JS convention)
- Compact setup form: single column, full-width Connect button
- Fix grammar: 'a OpenRouter' → 'an OpenRouter'
- Shorter promo text
* Allow text selection in chat panel
* Fix chat E2E test: Save → Connect button text
* Disable mermaid in chat markdown renderer
Alias mermaid and beautiful-mermaid to empty shims in Vite config,
preventing vue-stream-markdown from attempting to load them.
* Interleave text and tool calls in chat messages
Render message parts in order instead of grouping all tool calls
first then all text. Use SDK's isToolUIPart/isTextUIPart/getToolName
instead of custom type guards. Fix error state: output-error, not error.
* Implement figma.viewport.scrollAndZoomIntoView()
Figma Plugin API method that centers the viewport on given nodes.
Reuses the same bounding box logic as the viewport_zoom_to_fit tool.
* Catch tool execution errors and return them to the AI
Instead of crashing with an unhandled exception, tool errors are
caught and returned as { error: message } so the AI can retry or
explain the failure. UI detects error outputs and shows them in red
with the error message when expanded.
* Replace esbuild with sucrase for JSX transform
Sucrase is a pure JS transform (201 KB / 46 KB gzip) that works in
both Node/Bun and the browser. Replaces esbuild (13 MB WASM) which
only worked in Node/Bun.
- buildComponent() and renderJSX() are now synchronous
- render tool works in browser AI chat (no more 'esbuild required')
- Handles full JS expressions (map, ternaries, Array.from, etc.)
* Update system prompt to prioritize render tool with JSX
Document available tags, props, layout, text, and sizing options.
Instruct the AI to use full JS expressions in JSX for complex layouts.
* Fix zoom shortcuts to match Figma
Cmd+0: Zoom to 100% (was incorrectly mapped to Zoom to fit)
Cmd+1: Zoom to fit
Cmd+2: Zoom to selection
Shift+1 / Shift+2: same as Cmd+1 / Cmd+2
Add zoomTo100() and zoomToSelection() to editor store.
Refactor zoomToFit() to use shared zoomToBounds() helper.
* Refactor keyboard shortcuts to useMagicKeys
Replace manual keydown handler with VueUse useMagicKeys + whenever
for declarative shortcut registration.
- mod() helper for cross-platform Meta/Control shortcuts
- plain() helper for modifier-free keys
- Proper modifier exclusion (⌘G vs ⌘⇧G no longer conflict)
- Add E2E tests: duplicate, zoom (⌘0/⌘1/⌘2/⇧1/⇧2), auto-layout (⇧A)
- All 26 keyboard shortcut tests pass
* Fix scrollAndZoomIntoView to actually zoom
Match Figma Plugin API behavior (equivalent to Shift-1): compute
zoom level that fits all nodes with padding, capped at 100%.
Previously only set center without adjusting zoom.
Use markRaw to prevent Vue from deep-proxying the Chat class
instance, which broke its internal Vue refs. Add dedent for
multiline prompts.
Tests use mock transport by default. Set TEST_REAL_LLM=1 and
OPENROUTER_API_KEY to run against real OpenRouter.
