audio_codebuddy/deploy-package/scripts/build-lang-matrix-json.js

#!/usr/bin/env node
/**
 * build-lang-matrix-json.js — 把 alicloud-voices-matrix.json 转成 LANG_ROUTE 路由表
 */
const fs = require('fs');
const os = require('os');
const path = require('path');

// ── 1. 规范化语言字符串 ────────────────────────────────────
function parseSimpleLang(p) {
  if (!p) return null;
  p = p.trim();
  if (!p) return null;
  if (p === '中文' || p === '普通话') return { base: '中文', dialect: '普通话' };
  if (p === '日语') return { base: '日语', dialect: '通用' };
  if (p === '韩语') return { base: '韩语', dialect: '通用' };
  if (p === '印尼语') return { base: '印尼语', dialect: '通用' };
  if (p === '美语' || p === '美式英语' || p === '美式英文' || p === '英语(美式)') return { base: '英语', dialect: '美式' };
  if (p === '英式英语' || p === '英式英文' || p === '英语(英式)') return { base: '英语', dialect: '英式' };
  if (p === '英语') return { base: '英语', dialect: '通用' };
  // 中文方言裸词
  const DIALECT = {
    '广东话': '粤语', '粤语': '粤语',
    '闽南话': '闽南', '闽南': '闽南',
    '东北话': '东北', '东北口音': '东北', '东北': '东北',
    '四川话': '四川', '四川': '四川',
    '山东话': '山东', '山东': '山东',
    '河南话': '河南', '河南': '河南',
    '湖南话': '湖南', '湖南': '湖南',
    '陕西话': '陕西', '陕西': '陕西',
    '安徽话': '安徽', '安徽': '安徽',
  };
  if (DIALECT[p]) return { base: '中文', dialect: DIALECT[p] };
  return null;
}

// 用 regex 全局匹配 "{base}(dialect1、dialect2)" 或 "{base}" 这两种 base
// 支持的全角/半角括号。注意 raw 里用的是 "中文" "英文" 不是 "英语"
const LANGBASE_GROUP_RE = /(中文|英文|日语|韩语|印尼语)[\(（]([^\)）]+)[\)）]/g;
// 裸 base(不在 group 里的:纯 "英文"、"美式英语" 等)
const LANGBASE_BARE_RE = /(中文|英文|日语|韩语|印尼语|普通话|中文东北口音|美式英语|英式英语|美式英文|英式英文)(?=[\s,，、]|$)/g;
const DIALECT_SEP_GLOBAL = /[、，,]/g;

function normalizeLangString(s) {
  if (!s) return [];
  const cleaned = s.replace(/\s+/g, ' ').trim();
  const result = [];
  const seen = new Set();
  // 1. 先匹配 "base(dialect、dialect)" 这种 group
  const groupMatches = [...cleaned.matchAll(LANGBASE_GROUP_RE)];
  for (const m of groupMatches) {
    const base = m[1];
    const dialectText = m[2];
    for (const d of dialectText.split(DIALECT_SEP_GLOBAL)) {
      const trimmed = d.trim();
      if (!trimmed) continue;
      const lang = dialectToLang(base, trimmed);
      const key = `${lang.base}-${lang.dialect}`;
      if (!seen.has(key)) { seen.add(key); result.push(lang); }
    }
  }
  // 2. 再匹配剩余的纯 base(可能紧跟 group)
  let stripped = cleaned;
  for (const m of groupMatches) stripped = stripped.replace(m[0], ' ');
  const bareMatches = [...stripped.matchAll(LANGBASE_BARE_RE)];
  for (const m of bareMatches) {
    const base = m[1];
    const lang = baseToLang(base);
    const key = `${lang.base}-${lang.dialect}`;
    if (!seen.has(key)) { seen.add(key); result.push(lang); }
  }
  return result;
}

// 方言标准名(支持 raw 里的所有写法)
const DIALECT_ALIAS = {
  '广东话': '粤语',
  '广东': '粤语',
  '东北话': '东北',
  '东北': '东北',
  '东北口音': '东北',
  '四川话': '四川',
  '四川': '四川',
  '山东话': '山东',
  '山东': '山东',
  '河南话': '河南',
  '河南': '河南',
  '湖南话': '湖南',
  '湖南': '湖南',
  '陕西话': '陕西',
  '陕西': '陕西',
  '安徽话': '安徽',
  '安徽': '安徽',
  '闽南话': '闽南',
  '闽南': '闽南',
  '普通话': '普通话',
};

function dialectToLang(base, dialect) {
  if (base === '中文') {
    const d = DIALECT_ALIAS[dialect] || dialect;
    return { base: '中文', dialect: d };
  }
  if (base === '英语') {
    const d = DIALECT_ALIAS[dialect] || dialect;
    return { base: '英语', dialect: d || '通用' };
  }
  return { base, dialect: dialect || '通用' };
}

function baseToLang(base) {
  if (base === '普通话') return { base: '中文', dialect: '普通话' };
  if (base === '中文东北口音') return { base: '中文', dialect: '东北' };
  if (['美式英语', '美式英文'].includes(base)) return { base: '英语', dialect: '美式' };
  if (['英式英语', '英式英文'].includes(base)) return { base: '英语', dialect: '英式' };
  if (base === '中文') return { base: '中文', dialect: '普通话' };
  if (base === '英文') return { base: '英语', dialect: '通用' };
  return { base, dialect: '通用' };
}

// ── 2. 标准化 key: 中文(粤) → "zh-YUE" ────────────────────────
const ZH_DIALECT_TO_ISO = {
  '普通话': 'CMN', '粤语': 'YUE', '东北': 'NEN', '四川': 'SCN',
  '山东': 'SHD', '河南': 'HEN', '湖南': 'HNN', '陕西': 'SHX',
  '安徽': 'AHM', '闽南': 'MNN',
};

function langToKey(lang) {
  const { base, dialect } = lang;
  if (base === '中文') {
    const k = ZH_DIALECT_TO_ISO[dialect];
    return k ? `zh-${k}` : 'zh-MISC';
  }
  if (base === '英语') {
    if (dialect === '美式') return 'en-US';
    if (dialect === '英式') return 'en-GB';
    return 'en-MISC';
  }
  if (base === '日语') return 'ja-JP';
  if (base === '韩语') return 'ko-KR';
  if (base === '印尼语') return 'id-ID';
  // raw 字段作为兜底
  return `${base.replace(/[^a-z0-9]/gi, '').toLowerCase() || 'misc'}-MISC`;
}

function langLabel(key, base, dialect) {
  if (base === '中文') return '中文(' + (dialect || '普通话') + ')';
  if (base === '英语') return '英语' + (dialect && dialect !== '通用' ? '(' + dialect + ')' : '');
  return base || key;
}

// ── 3. 推荐音色(从 docs/alicloud-tts-lang-voice-matrix.md §三 手动权威) ───────
const RECOMMENDED = {
  'zh-CMN':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '欢脱元气女,支持 9 种方言' },
  'zh-YUE':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanyue',     desc: '粤语标准女声' },
  'zh-NEN':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '东北话 → longanhuan_v3 兜底' },
  'zh-SCN':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '四川话兜底' },
  'zh-SHD':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '山东话兜底' },
  'zh-HEN':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '河南话兜底' },
  'zh-HNN':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '湖南话兜底' },
  'zh-SHX':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longshange',    desc: '陕西话专用户(陕哥)' },
  'zh-AHM':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanhuan_v3', desc: '安徽话 → longanhuan_v3 兜底' },
  'zh-MNN':  { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longanmin',     desc: '闽南话专用户' },
  'en-US':   { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'longava',       desc: '美式英语女声' },
  'en-GB':   { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'loongluna',     desc: '英语英伦少女' },
  'ja-JP':   { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'loongtomoka',   desc: '日语女声' },
  'ko-KR':   { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'sohee',         desc: '韩语女声' },
  'id-ID':   { vendor: 'bailian', model: 'cosyvoice-v3-flash', voice: 'loongindah',    desc: '印尼女声' },
};

// ── 4. 主流程: 读 raw → 生成 routes ─────────────────────────────
// 注: cosyvoice-v2 / v1 已淘汰,只保留 v3-flash / v3-plus / Qwen-TTS
const DEPRECATED_MODELS = new Set(['cosyvoice-v2', 'cosyvoice-v1']);

const rawPath = process.argv[2] || path.join(os.tmpdir(), 'alicloud-voices-matrix.json');
const raw = JSON.parse(fs.readFileSync(rawPath, 'utf8'));

const langSupported = new Map();
const langVoices = new Map();

for (const [model, voices] of Object.entries(raw)) {
  if (DEPRECATED_MODELS.has(model)) continue;  // v2/v1 已淘汰,跳过
  for (const v of voices) {
    const langs = normalizeLangString(v.lang);
    for (const lang of langs) {
      const key = langToKey(lang);
      if (!langSupported.has(key)) langSupported.set(key, new Set());
      langSupported.get(key).add(model);
      if (!langVoices.has(key)) langVoices.set(key, []);
      langVoices.get(key).push({
        model, voice_id: v.voice_id, name: v.name, desc: v.desc, lang,
      });
    }
  }
}

// ── 5. 输出 JSON ─────────────────────────────────────────────
const out = {
  version: '2026-07-12',
  note: '由 scripts/build-lang-matrix-json.js 从 alicloud 官方页面抓取数据后生成。不要手编。',
  recommended: Object.keys(RECOMMENDED),
  languages: {},
};

let totalKeys = 0;
for (const [key, modelSet] of [...langSupported.entries()].sort()) {
  totalKeys++;
  // 取代表性 lang 用于 label
  const sample = langVoices.get(key)[0];
  const base = sample?.lang?.base || '中文';
  const dialect = sample?.lang?.dialect || '普通话';
  out.languages[key] = {
    label: langLabel(key, base, dialect),
    base,
    dialect,
    supportedModels: [...modelSet].sort(),
    voiceCount: (langVoices.get(key) || []).length,
    recommended: RECOMMENDED[key] || null,
    fallbackChain: RECOMMENDED[key]
      ? [RECOMMENDED[key].voice, 'longanhuan_v3']
      : ['longanhuan_v3'],
    sampleVoices: (langVoices.get(key) || []).slice(0, 30).map(v => ({
      voice_id: v.voice_id, name: v.name, model: v.model,
    })),
  };
}

const outPath = process.argv[3] || path.join(__dirname, '..', '..', 'server', 'src', 'config', 'alicloud-tts-lang-voice-matrix.json');
fs.mkdirSync(path.dirname(outPath), { recursive: true });
fs.writeFileSync(outPath, JSON.stringify(out, null, 2), 'utf8');
console.error(`写入 ${outPath}`);
console.error(`共 ${totalKeys} 个 lang_key,${Object.keys(RECOMMENDED).length} 个推荐`);
console.error();
console.error('推荐列表 (UI 默认显示):');
for (const k of out.recommended) {
  const l = out.languages[k];
  if (!l) { console.error(`  ${k}: (not in matrix, will fallback)`); continue; }
  console.error(`  ${k}: ${l.label} (${l.voiceCount} voices, models: ${l.supportedModels.join(',')})`);
}