misc: add stream answer for AI response

src/services/speaker/ai.ts

@@ -4,10 +4,11 @@ import {
   SpeakerCommand,
   SpeakerConfig,
   QueryMessage,
+  SpeakerAnswer,
 } from "./speaker";
 
 export type AISpeakerConfig = SpeakerConfig & {
-  askAI?: (msg: QueryMessage) => Promise<string>;
+  askAI?: (msg: QueryMessage) => Promise<SpeakerAnswer>;
   /**
    * 切换音色前缀
    *
@@ -200,7 +201,7 @@ export class AISpeaker extends Speaker {
   ];
 
   async askAIForAnswer(msg: QueryMessage) {
-    let data: any = {};
+    let data: { answer?: SpeakerAnswer } = {};
     const { hasNewMsg } = this.checkIfHasNewMsg(msg);
     for (const action of this._askAIForAnswerSteps) {
       const res = await action(msg, data);

src/services/speaker/base.ts

@@ -8,7 +8,7 @@ import {
 } from "mi-service-lite";
 import { sleep } from "../../utils/base";
 import { Http } from "../http";
-import { ResponseStream } from "./stream";
+import { StreamResponse } from "./stream";
 
 export type TTSProvider = "xiaoai" | "doubao";
 
@@ -59,7 +59,7 @@ export class BaseSpeaker {
   async response(options: {
     tts?: TTSProvider;
     text?: string;
-    stream?: ResponseStream;
+    stream?: StreamResponse;
     audio?: string;
     speaker?: string;
     keepAlive?: boolean;
@@ -79,7 +79,7 @@ export class BaseSpeaker {
 
     if (ttsNotXiaoai && !stream) {
       // 长文本 TTS 转化成 stream 分段模式
-      stream = ResponseStream.createResponseStream(text!);
+      stream = StreamResponse.createStreamResponse(text!);
     }
 
     let res;
@@ -139,7 +139,7 @@ export class BaseSpeaker {
   private async _response(options: {
     tts?: TTSProvider;
     text?: string;
-    stream?: ResponseStream;
+    stream?: StreamResponse;
     audio?: string;
     speaker?: string;
     keepAlive?: boolean;

src/services/speaker/speaker.ts

@@ -1,5 +1,6 @@
 import { firstOf, lastOf, sleep } from "../../utils/base";
 import { BaseSpeaker, BaseSpeakerConfig } from "./base";
+import { StreamResponse } from "./stream";
 
 export interface QueryMessage {
   text: string;
@@ -10,12 +11,18 @@ export interface QueryMessage {
   timestamp: number;
 }
 
+export interface SpeakerAnswer {
+  text?: string;
+  url?: string;
+  steam?: StreamResponse;
+}
+
 export interface SpeakerCommand {
   match: (msg: QueryMessage) => boolean;
   /**
    * 命中后执行的操作，返回值非空时会自动回复给用户
    */
-  run: (msg: QueryMessage) => Promise<string | undefined | void>;
+  run: (msg: QueryMessage) => Promise<SpeakerAnswer | undefined | void>;
 }
 
 export type SpeakerConfig = BaseSpeakerConfig & {
@@ -105,7 +112,7 @@ export class Speaker extends BaseSpeaker {
         if (answer) {
           if (noNewMsg()) {
             await this.response({
-              text: answer,
+              ...answer,
               keepAlive: this.keepAlive,
             });
           }

src/services/speaker/stream.ts

@@ -1,6 +1,6 @@
 type ResponseStatus = "idle" | "responding" | "finished" | "canceled";
 
-interface ResponseStreamOptions {
+interface StreamResponseOptions {
   /**
    * 单次响应句子的最大长度
    */
@@ -23,12 +23,12 @@ interface ResponseStreamOptions {
   batchSubmitTimeout?: number;
 }
 
-export class ResponseStream {
+export class StreamResponse {
   // 将已有的大篇文字回复 chuck 成 stream 回复
-  static createResponseStream(text: string, options?: ResponseStreamOptions) {
+  static createStreamResponse(text: string, options?: StreamResponseOptions) {
     const { maxSentenceLength = 100 } = options ?? {};
     if (text.length > maxSentenceLength) {
-      const stream = new ResponseStream(options);
+      const stream = new StreamResponse(options);
       stream.addResponse(text);
       stream.finish();
       return stream;
@@ -38,7 +38,7 @@ export class ResponseStream {
   maxSentenceLength: number;
   firstSubmitTimeout: number;
   batchSubmitTimeout: number;
-  constructor(options?: ResponseStreamOptions) {
+  constructor(options?: StreamResponseOptions) {
     const {
       maxSentenceLength = 100,
       firstSubmitTimeout = 200,
@@ -173,7 +173,7 @@ export class ResponseStream {
   }
 }
 
-const stream = new ResponseStream();
+const stream = new StreamResponse();
 
 // ai onNewText
 // {

tests/speaker.ts

@@ -1,5 +1,5 @@
 import { AISpeaker } from "../src/services/speaker/ai";
-import { ResponseStream } from "../src/services/speaker/stream";
+import { StreamResponse } from "../src/services/speaker/stream";
 import { sleep } from "../src/utils/base";
 
 export async function testSpeaker() {
@@ -13,16 +13,16 @@ export async function testSpeaker() {
   const speaker = new AISpeaker(config);
   await speaker.initMiServices();
   // await testSpeakerResponse(speaker);
-  await testSpeakerStreamResponse(speaker);
+  // await testSpeakerStreamResponse(speaker);
   // await testSpeakerGetMessages(speaker);
   // await testSwitchSpeaker(speaker);
   // await testSpeakerUnWakeUp(speaker);
-  // await testAISpeaker(speaker);
+  await testAISpeaker(speaker);
 }
 
 async function testAISpeaker(speaker: AISpeaker) {
   speaker.askAI = async (msg) => {
-    return "你说：" + msg.text;
+    return { text: "你说：" + msg.text };
   };
   await speaker.run();
   console.log("finished");
@@ -60,7 +60,7 @@ async function testSpeakerResponse(speaker: AISpeaker) {
 }
 
 async function testSpeakerStreamResponse(speaker: AISpeaker) {
-  const stream = new ResponseStream();
+  const stream = new StreamResponse();
   const add = async (text: string) => {
     stream.addResponse(text);
     await sleep(100);