feat: expand intelligent investigation workspace

server/edge-tts-service.ts

@@ -1,7 +1,10 @@
 import type { IncomingMessage, ServerResponse } from "node:http";
-import { Communicate } from "edge-tts-ts";
+import { EdgeTTS } from "node-edge-tts";
+import { speechVoiceIdSchema, type SpeechVoice } from "../src/shared/tts/voices";
+import { createEdgeTtsVoiceCatalog } from "./edge-tts-voices";
 
 const EDGE_TTS_PATH = "/api/tts/edge";
+const EDGE_TTS_VOICES_PATH = `${EDGE_TTS_PATH}/voices`;
 const DEFAULT_VOICE = "zh-CN-XiaoxiaoNeural";
 const MAX_REQUEST_BYTES = 8 * 1024;
 const MAX_TEXT_LENGTH = 520;
@@ -32,6 +35,7 @@ export type EdgeTtsMiddlewareOptions = {
   synthesize?: EdgeTtsSynthesizer;
   timeoutMs?: number;
   trustProxy?: boolean;
+  listVoices?: () => Promise<SpeechVoice[]>;
 };
 
 function sendJson(response: ServerResponse, status: number, message: string) {
@@ -76,14 +80,15 @@ function getClientKey(request: IncomingMessage, trustProxy: boolean) {
 }
 
 export async function synthesizeEdgeSpeech(text: string, voice: string): Promise<Uint8Array> {
-  const communicate = new Communicate(text, { voice });
+  const communicate = new EdgeTTS({ voice, timeout: DEFAULT_TIMEOUT_MS });
   const chunks: Uint8Array[] = [];
   let byteLength = 0;
 
-  for await (const chunk of communicate.stream()) {
-    if (chunk.type !== "audio") continue;
-    chunks.push(chunk.data);
-    byteLength += chunk.data.byteLength;
+  for await (const chunk of communicate.streamAudio(text)) {
+    if (!Buffer.isBuffer(chunk)) continue;
+    chunks.push(chunk);
+    byteLength += chunk.byteLength;
+    if (byteLength > MAX_AUDIO_BYTES) throw new Error("语音服务返回的音频过大");
   }
 
   if (!byteLength) throw new Error("语音服务返回了空音频");
@@ -99,27 +104,32 @@ export async function synthesizeEdgeSpeech(text: string, voice: string): Promise
 
 export function createEdgeTtsMiddleware(options: EdgeTtsMiddlewareOptions = {}) {
   const synthesize = options.synthesize ?? synthesizeEdgeSpeech;
-  const defaultVoice = options.defaultVoice ?? process.env.EDGE_TTS_VOICE ?? DEFAULT_VOICE;
+  const configuredVoice = options.defaultVoice ?? process.env.EDGE_TTS_VOICE ?? DEFAULT_VOICE;
+  const defaultVoice = speechVoiceIdSchema.safeParse(configuredVoice).success ? configuredVoice : DEFAULT_VOICE;
   const maxConcurrent = options.maxConcurrent ?? DEFAULT_MAX_CONCURRENT;
   const maxRequestsPerWindow = options.maxRequestsPerWindow ?? DEFAULT_RATE_LIMIT;
   const rateWindowMs = options.rateWindowMs ?? DEFAULT_RATE_WINDOW_MS;
   const timeoutMs = options.timeoutMs ?? DEFAULT_TIMEOUT_MS;
   const trustProxy = options.trustProxy ?? false;
   const now = options.now ?? Date.now;
+  const listVoices = createEdgeTtsVoiceCatalog(options.listVoices, now);
   const rateBuckets = new Map<string, RateBucket>();
   let activeRequests = 0;
 
   return async (request: IncomingMessage, response: ServerResponse, next?: () => void) => {
-    if (request.url?.split("?", 1)[0] !== EDGE_TTS_PATH) {
+    const path = request.url?.split("?", 1)[0];
+    const voicesRequest = path === EDGE_TTS_VOICES_PATH;
+    if (path !== EDGE_TTS_PATH && !voicesRequest) {
       next?.();
       return;
     }
-    if (request.method !== "POST") {
-      response.setHeader("Allow", "POST");
-      sendJson(response, 405, "仅支持 POST 请求");
+    const method = voicesRequest ? "GET" : "POST";
+    if (request.method !== method) {
+      response.setHeader("Allow", method);
+      sendJson(response, 405, `仅支持 ${method} 请求`);
       return;
     }
-    if (!request.headers["content-type"]?.toLowerCase().startsWith("application/json")) {
+    if (!voicesRequest && !request.headers["content-type"]?.toLowerCase().startsWith("application/json")) {
       sendJson(response, 415, "请求必须使用 application/json");
       return;
     }
@@ -143,6 +153,19 @@ export function createEdgeTtsMiddleware(options: EdgeTtsMiddlewareOptions = {})
       return;
     }
 
+    if (voicesRequest) {
+      try {
+        const voices = await listVoices();
+        response.statusCode = 200;
+        response.setHeader("Content-Type", "application/json; charset=utf-8");
+        response.setHeader("Cache-Control", "no-store");
+        response.end(JSON.stringify({ voices, defaultVoice }));
+      } catch {
+        sendJson(response, 503, "音色列表暂时不可用，请稍后重试");
+      }
+      return;
+    }
+
     let payload: EdgeTtsPayload;
     try {
       payload = await readJson(request);
@@ -151,10 +174,6 @@ export function createEdgeTtsMiddleware(options: EdgeTtsMiddlewareOptions = {})
       return;
     }
 
-    if (Object.prototype.hasOwnProperty.call(payload, "voice")) {
-      sendJson(response, 400, "语音角色只能由服务端配置");
-      return;
-    }
     const text = typeof payload.text === "string" ? payload.text.trim() : "";
     if (!text) {
       sendJson(response, 400, "缺少待朗读文本");
@@ -164,6 +183,24 @@ export function createEdgeTtsMiddleware(options: EdgeTtsMiddlewareOptions = {})
       sendJson(response, 413, `单次文本不能超过 ${MAX_TEXT_LENGTH} 字`);
       return;
     }
+    let voice = defaultVoice;
+    if (Object.prototype.hasOwnProperty.call(payload, "voice")) {
+      const selectedVoice = speechVoiceIdSchema.safeParse(payload.voice);
+      if (!selectedVoice.success) {
+        sendJson(response, 400, "请选择有效音色");
+        return;
+      }
+      try {
+        if (!(await listVoices()).some((item) => item.id === selectedVoice.data)) {
+          sendJson(response, 400, "所选音色不可用，请重新选择");
+          return;
+        }
+        voice = selectedVoice.data;
+      } catch {
+        sendJson(response, 503, "无法确认音色是否可用，请稍后重试");
+        return;
+      }
+    }
     if (activeRequests >= maxConcurrent) {
       response.setHeader("Retry-After", "1");
       sendJson(response, 429, "语音服务繁忙，请稍后重试");
@@ -171,7 +208,7 @@ export function createEdgeTtsMiddleware(options: EdgeTtsMiddlewareOptions = {})
     }
 
     activeRequests += 1;
-    const synthesisTask = Promise.resolve().then(() => synthesize(text, defaultVoice));
+    const synthesisTask = Promise.resolve().then(() => synthesize(text, voice));
     void synthesisTask.finally(() => {
       activeRequests -= 1;
     }).catch(() => undefined);