⚠️ THIS REPO HAS MOVED

Textura now lives in the Geometra monorepo.

All future development, issues, and pull requests should go there.

Textura

Live Demo & Interactive Examples

https://github.com/user-attachments/assets/bd67d28e-1841-474e-8d3c-8a7e76f272ba

DOM-free layout engine for the web. Combines Yoga (flexbox) with Pretext (text measurement) to compute complete UI geometry — positions, sizes, and text line breaks — without ever touching the DOM.

Why

The browser's layout engine is a black box that blocks the main thread. Every getBoundingClientRect() or offsetHeight call triggers synchronous layout reflow. When components independently measure text, each measurement triggers a reflow of the entire document.

Yoga solves box layout (flexbox) in pure JS/WASM, but punts on text — it requires you to supply a MeasureFunction callback externally. Pretext solves text measurement with canvas measureText, but doesn't do box layout. Textura joins them: declare a tree of flex containers and text nodes, get back exact pixel geometry for everything.

This unlocks:

• Worker-thread UI layout — compute layout off the main thread, send only coordinates for painting
• Zero-estimation virtualization — know exact heights for 100K items before mounting a single DOM node
• Canvas/WebGL/SVG rendering — full layout engine for non-DOM renderers
• Server-side layout — generate pixel positions server-side (once Pretext gets server canvas)

Installation

npm install textura

Quick Start

import { init, computeLayout } from 'textura'

// Initialize Yoga WASM (call once)
await init()

// Describe your UI as a tree
const result = computeLayout({
  width: 400,
  padding: 16,
  flexDirection: 'column',
  gap: 12,
  children: [
    {
      text: 'Hello World',
      font: '24px Inter',
      lineHeight: 32,
    },
    {
      flexDirection: 'row',
      gap: 8,
      children: [
        { width: 80, height: 80 },                              // avatar
        {
          text: 'This is a message that will wrap to multiple lines based on available width.',
          font: '16px Inter',
          lineHeight: 22,
          flexGrow: 1,
        },
      ],
    },
  ],
})

// result is a tree of computed layouts:
// {
//   x: 0, y: 0, width: 400, height: ...,
//   children: [
//     { x: 16, y: 16, width: 368, height: 32, text: 'Hello World', lineCount: 1 },
//     { x: 16, y: 60, width: 368, height: ...,
//       children: [
//         { x: 0, y: 0, width: 80, height: 80 },
//         { x: 88, y: 0, width: 280, height: ..., text: '...', lineCount: ... },
//       ]
//     },
//   ]
// }

API

init(): Promise<void>

Initialize the Yoga WASM runtime. Must be called once before computeLayout.

computeLayout(tree, options?): ComputedLayout

Compute layout for a declarative UI tree. Returns positions, sizes, and text metadata for every node.

Options:

• width?: number — available width for the root
• height?: number — available height for the root
• direction?: 'ltr' | 'rtl' — text direction (default: 'ltr')

Node types

Box nodes — flex containers with children:

{
  flexDirection: 'row',
  gap: 8,
  padding: 16,
  children: [...]
}

Text nodes — leaf nodes with measured text content:

{
  text: 'Hello world',
  font: '16px Inter',      // canvas font shorthand
  lineHeight: 22,           // line height in px
  whiteSpace: 'pre-wrap',   // optional: preserve spaces/tabs/newlines
}

Both node types accept all flexbox properties: flexDirection, flexWrap, justifyContent, alignItems, alignSelf, alignContent, flexGrow, flexShrink, flexBasis, width, height, minWidth, maxWidth, minHeight, maxHeight, padding*, margin*, border*, gap, position, top/right/bottom/left, aspectRatio, overflow, display.

ComputedLayout

interface ComputedLayout {
  x: number
  y: number
  width: number
  height: number
  children: ComputedLayout[]
  text?: string        // present on text nodes
  lineCount?: number   // present on text nodes
}

clearCache(): void

Clear Pretext's internal measurement caches.

destroy(): void

Release Yoga resources. Mostly useful for tests.

MCP Server — AI Layout Analysis

The Textura MCP server gives AI coding agents (Claude Code, Codex) layout vision — compute geometry, detect issues, validate responsive breakpoints, and auto-fix problems. Works with any framework.

Install

One-line install:

claude mcp add textura npx @razroo/textura-mcp

Uninstall:

claude mcp remove textura

Or manually add to .mcp.json (project-level) or ~/.claude/settings.json (global):

{
  "mcpServers": {
    "textura": {
      "command": "npx",
      "args": ["-y", "@razroo/textura-mcp"]
    }
  }
}

To uninstall manually, remove the textura entry from the config file.

Add to your Codex MCP configuration:

{
  "mcpServers": {
    "textura": {
      "command": "npx",
      "args": ["-y", "@razroo/textura-mcp"]
    }
  }
}

To uninstall, remove the textura entry from the config file.

Open Settings → MCP → Add new MCP server, or add to .cursor/mcp.json:

{
  "mcpServers": {
    "textura": {
      "command": "npx",
      "args": ["-y", "@razroo/textura-mcp"]
    }
  }
}

To uninstall, remove the entry from MCP settings.

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "textura": {
      "command": "npx",
      "args": ["-y", "@razroo/textura-mcp"]
    }
  }
}

To uninstall, remove the entry from the config file.

One-line install:

code --add-mcp '{"name":"textura","command":"npx","args":["-y","@razroo/textura-mcp"]}'

Or add to .vscode/mcp.json:

{
  "servers": {
    "textura": {
      "command": "npx",
      "args": ["-y", "@razroo/textura-mcp"]
    }
  }
}

To uninstall, remove the entry from MCP settings or delete the server from the MCP panel.

Any MCP client that supports stdio transport can use Textura. The server config is:

{
  "command": "npx",
  "args": ["-y", "@razroo/textura-mcp"]
}

To uninstall, remove the server entry from your client's MCP configuration.

Tools

How it works with your code

You ask the agent to check your layout. The agent reads your component code (React, Vue, Svelte, Tailwind, etc.), translates the layout structure into a Textura tree, and calls the MCP:

// Your component:
<div className="flex flex-col gap-4 p-6">
  <h1 className="text-2xl font-bold">Dashboard</h1>
  <div className="flex gap-4">
    <Card>Revenue: $12.4M</Card>
    <Card>Users: 847K</Card>
  </div>
</div>

// Agent translates to Textura tree:
{
  "flexDirection": "column", "gap": 16, "padding": 24,
  "children": [
    { "text": "Dashboard", "font": "700 24px Inter", "lineHeight": 32 },
    { "flexDirection": "row", "gap": 16, "children": [
      { "padding": 16, "children": [{ "text": "Revenue: $12.4M", "font": "16px Inter", "lineHeight": 24 }] },
      { "padding": 16, "children": [{ "text": "Users: 847K", "font": "16px Inter", "lineHeight": 24 }] }
    ]}
  ]
}

The MCP returns exact geometry and issues. The agent applies fixes back to your actual code. No browser needed.

How it works

1. You describe a UI tree using plain objects with CSS-like flex properties
2. computeLayout builds a Yoga node tree from your description
3. For text nodes, it wires Pretext's prepare() + layout() into Yoga's MeasureFunction — when Yoga asks "how tall is this text at width X?", Pretext answers using canvas-based measurement with cached segment widths
4. Yoga runs its flexbox algorithm over the tree
5. The computed positions and sizes are read back into a plain object tree

The text measurement is the key piece: Pretext handles Unicode segmentation, CJK character breaking, Arabic/bidi text, emoji, soft hyphens, and browser-specific quirks — all via Intl.Segmenter and canvas measureText, with 7680/7680 accuracy across Chrome/Safari/Firefox.

Limitations

• Requires a browser environment (or OffscreenCanvas in a worker) for text measurement
• Text nodes use the same CSS target as Pretext: white-space: normal, word-break: normal, overflow-wrap: break-word, line-break: auto
• Use named fonts (Inter, Helvetica) — system-ui can produce inaccurate measurements on macOS

Credits

Built on Yoga by Meta and Pretext by Cheng Lou.