06-voiceManager.js 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400
  1. /**
  2. * 云函数:voiceManager(音色档案 + speaker_id 池)
  3. * 替代:
  4. * POST /api/voice/auto-speaker-id → action=autoSpeakerId
  5. * POST /api/voice-profiles/sync → action=syncProfile
  6. * 新增:
  7. * action=listProfiles 列出所有音色档案
  8. * action=initPool 初始化 speaker_id 池(首次部署用)
  9. *
  10. * 注意:speaker_id 池原本在 docs/音色创建/speaker_id.md 文件里。
  11. * 迁移到云函数后改为存储在 "VoiceSpeakerPool" 表。
  12. * 首次部署后请用 action=initPool 把现有池一次性灌入。
  13. */
  14. const VOICE_TOKEN = readEnv('VOICE_TOKEN') || 'Bearer r:f0333969e312a40e4703e8fe4ed1c600';
  15. const VOICE_TTS_BASE_URL = readEnv('VOICE_TTS_BASE_URL') || 'https://server.fmode.cn/api/volcengine/tts';
  16. const VOICE_MAX_ATTEMPTS = Math.max(1, Number(readEnv('VOICE_MAX_ATTEMPTS') || 3));
  17. async function handler(request, response) {
  18. try {
  19. await Psql.query(`
  20. CREATE TABLE IF NOT EXISTS "VoiceProfile" (
  21. "objectId" VARCHAR(50) PRIMARY KEY,
  22. "timbreId" VARCHAR(255),
  23. "speakerId" VARCHAR(255),
  24. "name" VARCHAR(255) DEFAULT '',
  25. "data" JSONB NOT NULL DEFAULT '{}',
  26. "userId" VARCHAR(255) DEFAULT '',
  27. "occupied" BOOLEAN DEFAULT TRUE,
  28. "createdAt" TIMESTAMPTZ DEFAULT NOW(),
  29. "updatedAt" TIMESTAMPTZ DEFAULT NOW()
  30. )
  31. `);
  32. await Psql.query(`CREATE UNIQUE INDEX IF NOT EXISTS idx_voice_timbre ON "VoiceProfile" ("timbreId") WHERE "timbreId" IS NOT NULL AND "timbreId" <> ''`);
  33. await Psql.query(`CREATE INDEX IF NOT EXISTS idx_voice_speaker ON "VoiceProfile" ("speakerId")`);
  34. await Psql.query(`
  35. CREATE TABLE IF NOT EXISTS "VoiceSpeakerPool" (
  36. "speakerId" VARCHAR(255) PRIMARY KEY,
  37. "trained" BOOLEAN DEFAULT FALSE,
  38. "createdAt" TIMESTAMPTZ DEFAULT NOW()
  39. )
  40. `);
  41. const action = pickParam(request, 'action') || 'listProfiles';
  42. const userId = pickParam(request, 'userId') || '';
  43. if (action === 'autoSpeakerId') {
  44. const used = await Psql.query(`
  45. SELECT s."speakerId" FROM "VoiceSpeakerPool" s
  46. LEFT JOIN "VoiceProfile" v ON v."speakerId" = s."speakerId"
  47. WHERE v."speakerId" IS NULL AND s."trained" = FALSE
  48. ORDER BY s."createdAt" ASC
  49. LIMIT 1
  50. `);
  51. if (!used.length) {
  52. return response.json({ code: 500, success: false, error: 'speaker_id 池已耗尽,请先 initPool 添加' });
  53. }
  54. response.json({ code: 200, success: true, data: { speakerId: used[0].speakerId } });
  55. return;
  56. }
  57. if (action === 'syncProfile') {
  58. const body = pickParam(request, 'profile', 'data') || request.body || {};
  59. const timbreId = body.timbre_id || body.timbreId || '';
  60. const speakerId = body.speaker_id || body.speakerId || '';
  61. if (!timbreId && !speakerId) {
  62. return response.json({ code: 400, success: false, error: '缺少 timbre_id 或 speaker_id' });
  63. }
  64. const now = new Date().toISOString();
  65. const merged = { ...body, timbre_id: timbreId, speaker_id: speakerId, updated_at: now };
  66. let existing = [];
  67. if (timbreId) {
  68. existing = await Psql.query(`SELECT * FROM "VoiceProfile" WHERE "timbreId" = $1 LIMIT 1`, [timbreId]);
  69. }
  70. if (!existing.length && speakerId) {
  71. existing = await Psql.query(`SELECT * FROM "VoiceProfile" WHERE "speakerId" = $1 LIMIT 1`, [speakerId]);
  72. }
  73. if (existing.length) {
  74. const old = typeof existing[0].data === 'string' ? JSON.parse(existing[0].data) : (existing[0].data || {});
  75. const merged2 = { ...old, ...merged };
  76. await Psql.query(
  77. `UPDATE "VoiceProfile" SET "timbreId"=$1, "speakerId"=$2, "name"=$3, "data"=$4, "occupied"=$5, "updatedAt"=NOW() WHERE "objectId"=$6`,
  78. [timbreId, speakerId, body.name || merged2.name || '', JSON.stringify(merged2), body.occupied !== false, existing[0].objectId]
  79. );
  80. response.json({ code: 200, success: true, data: { ...merged2, objectId: existing[0].objectId } });
  81. } else {
  82. merged.created_at = now;
  83. const objectId = generateId();
  84. await Psql.query(
  85. `INSERT INTO "VoiceProfile" ("objectId","timbreId","speakerId","name","data","userId","occupied")
  86. VALUES ($1,$2,$3,$4,$5,$6,$7)`,
  87. [objectId, timbreId, speakerId, body.name || '', JSON.stringify(merged), userId, body.occupied !== false]
  88. );
  89. response.json({ code: 200, success: true, data: { ...merged, objectId } });
  90. }
  91. return;
  92. }
  93. if (action === 'listProfiles') {
  94. const conds = [], params = [];
  95. if (userId) { params.push(userId); conds.push(`"userId" = $${params.length}`); }
  96. const where = conds.length ? `WHERE ${conds.join(' AND ')}` : '';
  97. const rows = await Psql.query(
  98. `SELECT * FROM "VoiceProfile" ${where} ORDER BY "createdAt" DESC LIMIT 1000`,
  99. params
  100. );
  101. response.json({ code: 200, success: true, data: rows.map(rowToObj) });
  102. return;
  103. }
  104. if (action === 'initPool') {
  105. const ids = pickParam(request, 'speakerIds') || [];
  106. const list = Array.isArray(ids) ? ids : String(ids || '').split(/[\s,]+/).filter(Boolean);
  107. if (!list.length) {
  108. return response.json({ code: 400, success: false, error: '缺少 speakerIds(数组或逗号分隔字符串)' });
  109. }
  110. let inserted = 0;
  111. for (const sid of list) {
  112. const r = await Psql.query(
  113. `INSERT INTO "VoiceSpeakerPool" ("speakerId") VALUES ($1) ON CONFLICT DO NOTHING RETURNING "speakerId"`,
  114. [String(sid).trim()]
  115. );
  116. if (r.length) inserted++;
  117. }
  118. response.json({ code: 200, success: true, data: { total: list.length, inserted } });
  119. return;
  120. }
  121. if (action === 'synthesize') {
  122. const payload = pickParam(request, 'payload', 'data') || {};
  123. if (!payload.text && !payload.ssml) {
  124. return response.json({ code: 400, success: false, error: '缺少合成文本' });
  125. }
  126. const upstreamPayload = normalizeSynthesisPayload(payload);
  127. const res = await postJson(`${VOICE_TTS_BASE_URL}/unidirectional`, upstreamPayload);
  128. response.json(res);
  129. return;
  130. }
  131. if (action === 'cloneVoice') {
  132. const payload = pickParam(request, 'payload', 'data') || {};
  133. if (!payload.name || !payload.speaker_id || !payload.audioData) {
  134. return response.json({ code: 400, success: false, error: '缺少声音训练参数' });
  135. }
  136. const upstreamPayload = stripEmpty({
  137. ...payload,
  138. token: VOICE_TOKEN,
  139. });
  140. const res = await postJson(`${VOICE_TTS_BASE_URL}/voice_clone`, upstreamPayload);
  141. response.json(res);
  142. return;
  143. }
  144. if (action === 'deleteProfile') {
  145. const objectId = pickParam(request, 'objectId');
  146. const timbreId = pickParam(request, 'timbreId');
  147. if (!objectId && !timbreId) {
  148. return response.json({ code: 400, success: false, error: '缺少 objectId 或 timbreId' });
  149. }
  150. const r = objectId
  151. ? await Psql.query(`DELETE FROM "VoiceProfile" WHERE "objectId" = $1 RETURNING "objectId"`, [objectId])
  152. : await Psql.query(`DELETE FROM "VoiceProfile" WHERE "timbreId" = $1 RETURNING "objectId"`, [timbreId]);
  153. response.json({ code: 200, success: true, data: { deleted: r.length > 0 } });
  154. return;
  155. }
  156. response.json({ code: 400, success: false, error: `未知 action: ${action}` });
  157. } catch (error) {
  158. console.error('❌ voiceManager 失败:', error.message);
  159. response.json({ code: 500, success: false, error: error.message });
  160. }
  161. }
  162. function normalizeSynthesisPayload(input) {
  163. const payload = { ...(input || {}) };
  164. const officialVoiceType = String(payload.volcengine_voice_type || payload.voiceType || payload.voice_type || '').trim();
  165. if (officialVoiceType && !payload.speaker_id && !payload.speakerId && !payload.timbreId && !payload.timbre_id) {
  166. payload.speaker_id = officialVoiceType;
  167. }
  168. if (payload.timbre_id && !payload.timbreId) {
  169. payload.timbreId = payload.timbre_id;
  170. }
  171. return stripEmpty({
  172. ...payload,
  173. token: VOICE_TOKEN,
  174. isStream: payload.isStream === true,
  175. });
  176. }
  177. function rowToObj(row) {
  178. const data = typeof row.data === 'string' ? JSON.parse(row.data) : (row.data || {});
  179. return { ...data, objectId: row.objectId, occupied: row.occupied, createdAt: row.createdAt, updatedAt: row.updatedAt };
  180. }
  181. function pickParam(request, ...names) {
  182. const sources = [request.params, request.body, request];
  183. for (const src of sources) {
  184. if (!src || typeof src !== 'object') continue;
  185. for (const n of names) {
  186. const v = src[n];
  187. if (v !== undefined && v !== null && v !== '') return v;
  188. }
  189. }
  190. return null;
  191. }
  192. function generateId() {
  193. const chars = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789';
  194. let s = '';
  195. for (let i = 0; i < 10; i++) s += chars.charAt(Math.floor(Math.random() * chars.length));
  196. return s;
  197. }
  198. function stripEmpty(value) {
  199. if (!value || typeof value !== 'object') return value;
  200. const out = Array.isArray(value) ? [] : {};
  201. for (const [key, val] of Object.entries(value)) {
  202. if (val === undefined || val === null || val === '') continue;
  203. if (val && typeof val === 'object' && !Array.isArray(val)) {
  204. const nested = stripEmpty(val);
  205. if (Object.keys(nested).length) out[key] = nested;
  206. } else {
  207. out[key] = val;
  208. }
  209. }
  210. return out;
  211. }
  212. async function postJson(url, body) {
  213. let lastError = null;
  214. for (let attempt = 1; attempt <= VOICE_MAX_ATTEMPTS; attempt += 1) {
  215. try {
  216. const { status, ok, rawText, data, transport } = await requestJson('POST', url, {
  217. headers: { 'Content-Type': 'application/json', 'Accept': 'application/json' },
  218. body: JSON.stringify(body),
  219. });
  220. const code = Number(data && data.code ? data.code : status || 0);
  221. const retryable = !ok || code === 408 || code === 429 || code >= 500;
  222. if (!retryable || attempt >= VOICE_MAX_ATTEMPTS) {
  223. if (!ok) {
  224. return {
  225. code: status || code || 500,
  226. success: false,
  227. error: readError(data, rawText) || `HTTP ${status}`,
  228. upstream: { transport, attempt, maxAttempts: VOICE_MAX_ATTEMPTS },
  229. };
  230. }
  231. return data || { code: 500, success: false, error: '服务返回异常' };
  232. }
  233. } catch (error) {
  234. lastError = error;
  235. if (attempt >= VOICE_MAX_ATTEMPTS) break;
  236. }
  237. await sleep(Math.min(12000, 1200 * attempt * attempt));
  238. }
  239. return {
  240. code: 500,
  241. success: false,
  242. error: `TTS upstream request failed: ${formatFetchError(lastError)}`,
  243. };
  244. }
  245. async function requestJson(method, url, init) {
  246. let lastError = null;
  247. if (typeof fetch === 'function') {
  248. try {
  249. const response = await fetch(url, { method, ...(init || {}) });
  250. return readResponse(response, 'fetch');
  251. } catch (error) {
  252. lastError = error;
  253. }
  254. }
  255. if (typeof require === 'function') {
  256. try {
  257. return await requestJsonWithNodeHttps(method, url, init);
  258. } catch (error) {
  259. lastError = error;
  260. }
  261. }
  262. if (typeof XMLHttpRequest !== 'undefined') {
  263. try {
  264. return await requestJsonWithXhr(method, url, init);
  265. } catch (error) {
  266. lastError = error;
  267. }
  268. }
  269. throw lastError || new Error('No HTTP client is available in this cloud runtime');
  270. }
  271. async function readResponse(response, transport) {
  272. const rawText = await response.text();
  273. let data = null;
  274. try { data = rawText ? JSON.parse(rawText) : null; } catch {}
  275. return {
  276. status: response.status,
  277. ok: !!response.ok,
  278. rawText,
  279. data,
  280. transport,
  281. };
  282. }
  283. function requestJsonWithNodeHttps(method, url, init) {
  284. return new Promise((resolve, reject) => {
  285. try {
  286. const parsed = new URL(url);
  287. const lib = parsed.protocol === 'http:' ? require('http') : require('https');
  288. const headers = init && init.headers ? init.headers : {};
  289. const body = init && init.body ? init.body : '';
  290. const req = lib.request({
  291. method,
  292. protocol: parsed.protocol,
  293. hostname: parsed.hostname,
  294. port: parsed.port,
  295. path: `${parsed.pathname}${parsed.search}`,
  296. headers: body ? { ...headers, 'Content-Length': Buffer.byteLength(body) } : headers,
  297. }, (res) => {
  298. const chunks = [];
  299. res.on('data', (chunk) => chunks.push(chunk));
  300. res.on('end', () => {
  301. const rawText = Buffer.concat(chunks).toString('utf8');
  302. let data = null;
  303. try { data = rawText ? JSON.parse(rawText) : null; } catch {}
  304. resolve({
  305. status: res.statusCode || 0,
  306. ok: res.statusCode >= 200 && res.statusCode < 300,
  307. rawText,
  308. data,
  309. transport: 'node-https',
  310. });
  311. });
  312. });
  313. req.on('error', reject);
  314. if (body) req.write(body);
  315. req.end();
  316. } catch (error) {
  317. reject(error);
  318. }
  319. });
  320. }
  321. function requestJsonWithXhr(method, url, init) {
  322. return new Promise((resolve, reject) => {
  323. const xhr = new XMLHttpRequest();
  324. xhr.open(method, url, true);
  325. const headers = init && init.headers ? init.headers : {};
  326. for (const [key, value] of Object.entries(headers)) {
  327. xhr.setRequestHeader(key, value);
  328. }
  329. xhr.onreadystatechange = function onReadyStateChange() {
  330. if (xhr.readyState !== 4) return;
  331. let data = null;
  332. try { data = xhr.responseText ? JSON.parse(xhr.responseText) : null; } catch {}
  333. resolve({
  334. status: xhr.status,
  335. ok: xhr.status >= 200 && xhr.status < 300,
  336. rawText: xhr.responseText || '',
  337. data,
  338. transport: 'xhr',
  339. });
  340. };
  341. xhr.onerror = function onXhrError() {
  342. reject(new Error('XMLHttpRequest failed'));
  343. };
  344. xhr.send(init && init.body ? init.body : null);
  345. });
  346. }
  347. function readError(data, fallback) {
  348. return data?.error?.message || data?.error || data?.message || data?.msg || fallback || '';
  349. }
  350. function formatFetchError(error) {
  351. if (!error) return 'unknown error';
  352. const message = error.message || String(error);
  353. const cause = error.cause ? `; cause=${error.cause.code || error.cause.message || error.cause}` : '';
  354. return `${message}${cause}`;
  355. }
  356. function sleep(ms) {
  357. return new Promise((resolve) => setTimeout(resolve, ms));
  358. }
  359. function readEnv(name) {
  360. if (typeof process !== 'undefined' && process.env && process.env[name]) {
  361. return process.env[name];
  362. }
  363. return '';
  364. }