![\"Cubing](\"docs/img/logo.png\"){kind=logo}
\n\n \n
\n Random-state scrambles for WCA events, with a Rust core compiled to WebAssembly.
\n It generates the scrambles on cubing.center.\n
\n ![\"npm\"](\"https://img.shields.io/npm/v/@cubingcenter/scrambler?logo=npm&color=CB3837\")
![\"TypeScript\"](\"https://img.shields.io/badge/TypeScript-ESM_%2B_CJS-3178C6?logo=typescript&logoColor=white\"){kind=icon}
![\"Rust](\"https://img.shields.io/badge/Rust-WebAssembly-654FF0?logo=webassembly&logoColor=white\"){kind=icon}
![\"Node.js](\"https://img.shields.io/badge/Node.js_%2B_browser-339933?logo=nodedotjs&logoColor=white\"){kind=icon}
![\"MIT](\"https://img.shields.io/badge/license-MIT-blue\"){kind=icon}
![\"Flow](\"docs/img/flow-dark.svg\")
\ndocs/img/flow.mmd (Mermaid).\n\n- **State first, scramble second.** A uniform legal state is sampled (the\n permutation and orientation of every piece), the solver finds a short\n solution and the scramble is that solution inverted. No state is more\n likely than another, which is what the WCA requires.\n- **Rust/WASM core.** The 2x2, 3x3, 4x4, Pyraminx, Skewb, Square-1 and FTO\n solvers run in Rust compiled to WebAssembly. The module is\n embedded in the package, so there is nothing extra to download or configure.\n- **Precomputed tables.** Each solver uses pruning tables generated once,\n versioned with checksums and loaded on demand.\n- **Every scramble is checked.** Before returning it, the scramble is applied\n and must reach the sampled state; on the 4x4 a failed search retries the\n same state with larger budgets, never a different state.\n\n### Performance\n\nMean per scramble, warm, Node.js (1.0.0):\n\n| 222 | 333 | 444 | pymx | skwb | sq1 | fto |\n| ------- | ------- | ------ | ------- | ------- | ------- | ------- |\n| 0.03 ms | 6.09 ms | 473 ms | 0.03 ms | 0.04 ms | 7.04 ms | 62.5 ms |\n\nCold times, p95, memory, table sizes and the comparison with cubing.js are in\n[bench/README.md](bench/README.md).\n\n## Install\n\n```sh\npnpm add @cubingcenter/scrambler\n# or\nnpm install @cubingcenter/scrambler\n```\n\nIt ships as ESM and CommonJS, with TypeScript types.\n\n## Usage\n\n```ts\nimport { generateScramble, generateScrambles, getSupportedEvents } from \"@cubingcenter/scrambler\";\n\nconst events = getSupportedEvents(); // [{ eventId: \"222\", name: \"2x2x2 Cube\", ... }, ...]\n\nconst result = await generateScramble(\"333\", { seed: \"demo\" });\nconsole.log(result.scramble); // \"D B2 F2 D B2 L2 F2 L2 D R2 ...\"\nconsole.log(result.metadata?.faceletString); // final state, 54 characters in U R F D L B order\n\nconst batch = await generateScrambles(\"333\", 5, { seed: \"session-1\" }); // seeds session-1:0 … session-1:4\n\nconst pll = await generateScramble(\"333\", { subset: \"pll\" }); // training subsets for 3x3\n```\n\nEach result has `scramble` (text), `moves` (list), `quality` (`random-state`\nor `random-move`), `generator` and `metadata`. For random-state events,\n`metadata` carries the target state, the `facelets` ready to draw and solver\ndata. Options, events, metadata and errors are documented in\n[docs/api.md](docs/api.md).\n\n### In the browser\n\nImporting the root loads every puzzle. In a web app, import only the puzzle\nyou need; each subpath registers its own generator:\n\n```ts\nimport { generateScramble, initializeScrambler } from \"@cubingcenter/scrambler/puzzles/pyraminx\";\n\nawait initializeScrambler(\"/scrambler-tables\", \"pymx\");\nconst result = await generateScramble(\"pymx\");\n```\n\n`initializeScrambler` downloads the solver tables from the given URL, where\nthe `puzzles/