feat(etapa2): transcriber sherpa-onnx-offline em Rust

scripts/setup-fixtures.mjs

@@ -0,0 +1,91 @@
+// Baixa fixtures oficiais do sherpa-onnx para `src-tauri/tests/fixtures/`.
+//
+// O áudio `parakeet-sample.wav` vem dentro do tarball oficial do modelo
+// Parakeet TDT v2 int8. É o áudio usado nos docs do sherpa-onnx para validar
+// a CLI `sherpa-onnx-offline`. Funciona com o v3 (mesma família de modelo).
+//
+// Uso:
+//   node scripts/setup-fixtures.mjs
+
+import { execFileSync } from "node:child_process";
+import {
+	existsSync,
+	mkdirSync,
+	createWriteStream,
+	statSync,
+	renameSync,
+	rmSync,
+} from "node:fs";
+import { dirname, join, resolve, sep as pathSep } from "node:path";
+import { pipeline } from "node:stream/promises";
+import { fileURLToPath } from "node:url";
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = dirname(__filename);
+const repoRoot = resolve(__dirname, "..");
+const fixturesDir = join(repoRoot, "src-tauri", "tests", "fixtures");
+mkdirSync(fixturesDir, { recursive: true });
+
+const FIXTURES = [
+	{
+		name: "parakeet-sample.wav",
+		url: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2",
+		extractFrom: "sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/test_wavs/0.wav",
+		minSize: 10_000,
+	},
+];
+
+async function download(url, dest) {
+	const res = await fetch(url);
+	if (!res.ok) {
+		throw new Error(
+			`download falhou: ${res.status} ${res.statusText} em ${url}`,
+		);
+	}
+	if (!res.body) {
+		throw new Error(`resposta vazia em ${url}`);
+	}
+	await pipeline(res.body, createWriteStream(dest));
+	return statSync(dest).size;
+}
+
+for (const fixture of FIXTURES) {
+	const dest = join(fixturesDir, fixture.name);
+	if (existsSync(dest) && statSync(dest).size > fixture.minSize) {
+		console.log(
+			`✓ fixture já presente em ${dest} (${statSync(dest).size} bytes)`,
+		);
+		continue;
+	}
+	console.log(`↓ baixando ${fixture.url}`);
+	const tmpArchive = join(fixturesDir, "fixture.tar.bz2");
+	const size = await download(fixture.url, tmpArchive);
+	if (size < 1_000_000) {
+		throw new Error(
+			`fixture ${fixture.name} tarball muito pequeno: ${size} bytes`,
+		);
+	}
+	// `tar` no Windows tem problema com paths absolutos em `-C`; usa `cwd`+relativo.
+	const relArchive = tmpArchive.startsWith(fixturesDir + pathSep)
+		? tmpArchive.slice(fixturesDir.length + 1)
+		: tmpArchive;
+	execFileSync("tar", ["-xjf", relArchive], {
+		cwd: fixturesDir,
+		stdio: "inherit",
+	});
+	const extractedPath = join(fixturesDir, fixture.extractFrom);
+	if (!existsSync(extractedPath)) {
+		throw new Error(
+			`caminho extraído ${extractedPath} não existe; verificar nome interno`,
+		);
+	}
+	renameSync(extractedPath, dest);
+	rmSync(tmpArchive, { force: true });
+	const finalSize = statSync(dest).size;
+	if (finalSize < fixture.minSize) {
+		throw new Error(
+			`fixture ${fixture.name} parece corrompida: ${finalSize} bytes (esperado > ${fixture.minSize})`,
+		);
+	}
+	console.log(`✓ fixture pronta em ${dest} (${finalSize} bytes)`);
+}