- Update tool count from 29 to 75 across all docs and locales - Expand all locale pages (de/es/fr/it/pl) to full translations - Add missing German sections: Advanced Rendering, Z-Order, ScrubInput, CI/CD, JSX Renderer, Code Panel, full figma-comparison matrix - Write complete figma-comparison (152-row matrix) for all locales - Expand comparison, features, tech-stack, contributing for es/fr/it/pl - Fix garbled text in locale mcp-tools reference pages - Architecture diagrams use '75+ tools' to avoid staleness
64 lines
2.8 KiB
Markdown
64 lines
2.8 KiB
Markdown
# Tech Stack
|
|
|
|
## Core Technologies
|
|
|
|
| Layer | Technology | Why |
|
|
|-------|-----------|-----|
|
|
| **Rendering** | Skia CanvasKit WASM | Same engine as Figma — proven performance, GPU-accelerated, pixel-perfect |
|
|
| **UI Framework** | Vue 3 + VueUse | Reactive composition API, excellent TypeScript support |
|
|
| **Components** | Reka UI | Headless, accessible UI primitives (tree, slider, etc.) |
|
|
| **Styling** | Tailwind CSS 4 | Utility-first, fast iteration, dark theme |
|
|
| **Layout** | Yoga WASM | CSS flexbox engine from Meta, battle-tested in React Native |
|
|
| **File Format** | Kiwi binary + Zstd | Figma's own format — compact, fast parsing, .fig compatible |
|
|
| **Collaboration** | Trystero + Yjs | P2P WebRTC via MQTT signaling, CRDT sync, y-indexeddb persistence |
|
|
| **Color** | culori | Color space conversions (HSV, RGB, hex) |
|
|
| **AI/MCP** | MCP SDK + Hono | 75 tools for AI coding tools, stdio + HTTP transports |
|
|
| **Desktop** | Tauri v2 | ~5MB native app (vs Electron's ~100MB), Rust backend |
|
|
| **Build** | Vite 7 | Fast HMR, native ES modules |
|
|
| **Testing** | Playwright + bun:test | Visual regression (E2E) + fast unit tests |
|
|
| **Linting** | oxlint | Rust-based, orders of magnitude faster than ESLint |
|
|
| **Formatting** | oxfmt | Rust-based formatter |
|
|
| **Type Checking** | typescript-go (tsgo) | Native Go implementation of TypeScript type checker |
|
|
|
|
## Key Dependencies
|
|
|
|
```json
|
|
{
|
|
"canvaskit-wasm": "^0.40.0",
|
|
"vue": "^3.5.29",
|
|
"yoga-layout": "^3.2.1",
|
|
"reka-ui": "^2.8.2",
|
|
"tailwindcss": "^4.2.1",
|
|
"culori": "^4.0.2",
|
|
"fzstd": "^0.1.1",
|
|
"fflate": "^0.8.2",
|
|
"trystero": "^0.20.0",
|
|
"yjs": "^13.6.24",
|
|
"y-indexeddb": "^9.0.12"
|
|
}
|
|
```
|
|
|
|
## Why Not...
|
|
|
|
### Why not SVG rendering?
|
|
|
|
SVG is slow for complex documents. Every node is a DOM element — 10,000 nodes means 10,000 DOM nodes with layout, paint, and compositing overhead. CanvasKit draws everything to a single GPU surface. (Penpot still defaults to SVG rendering but has an opt-in Rust/Skia WASM renderer in development.)
|
|
|
|
### Why not Electron (like Figma desktop)?
|
|
|
|
Tauri v2 uses the system webview (~5MB) instead of bundling Chromium (~100MB). The Rust backend provides native performance for file I/O and system integration.
|
|
|
|
### Why not React (like the original plan)?
|
|
|
|
The project migrated from React to Vue 3 early in development. Vue's reactivity system and VueUse composables proved more ergonomic for the editor's state management needs.
|
|
|
|
### Why not custom layout engine?
|
|
|
|
Yoga is maintained by Meta, battle-tested across billions of React Native devices, and implements the CSS flexbox spec. Building a custom engine would be months of work to reach the same correctness level.
|
|
|
|
## Planned Technologies
|
|
|
|
| Technology | Purpose | Phase |
|
|
|-----------|---------|-------|
|
|
| CSS Grid in Yoga | Grid-based auto layout | Blocked on upstream (facebook/yoga#1893) |
|