add voice

leochen-g · Oct 9, 2024 · a0d7bd0 · a0d7bd0
1 parent 4c837b8
commit a0d7bd0
Show file tree

Hide file tree

Showing 8 changed files with 90 additions and 3 deletions.
diff --git a/src/botInstance/coze.js b/src/botInstance/coze.js
@@ -6,6 +6,7 @@ import { getPuppetEol, isWindowsPlatform } from '../const/puppet-type.js'
 import dayjs from 'dayjs'
 import { extractImageLinks } from '../lib/index.js'
 import nodeFetch from 'node-fetch'
+import {getText2Speech} from "../proxy/multimodal.js";
 
 
 class CozeAi {
@@ -127,6 +128,12 @@ class CozeAi {
         parentMessageId: id
       }
       let replys = []
+      if(this.config?.openTTS) {
+        replys = await getText2Speech(text, this.config.ttsConfig)
+        if(replys.length) {
+          return replys
+        }
+      }
       let message
       if (this.config.showQuestion) {
         message = `${content}${this.eol}-----------${this.eol}` + (this.iswindows ? text.replaceAll('\n', this.eol) : text)

diff --git a/src/botInstance/cozev3.js b/src/botInstance/cozev3.js
@@ -5,6 +5,7 @@ import { ContentCensor } from "../lib/contentCensor.js";
 import { getPuppetEol, isWindowsPlatform } from '../const/puppet-type.js'
 import dayjs from "dayjs";
 import { extractImageLinks } from '../lib/index.js'
+import {getText2Speech} from "../proxy/multimodal.js";
 
 
 class CozeV3Ai {
@@ -101,6 +102,12 @@ class CozeV3Ai {
         };
       }
       let replys = []
+      if(this.config?.openTTS) {
+        replys = await getText2Speech(text, this.config.ttsConfig)
+        if(replys.length) {
+          return replys
+        }
+      }
       let message;
       if(this.config.showQuestion) {
         message = `${content}${this.eol}-----------${this.eol}` +  (this.iswindows ? text.replaceAll('\n', this.eol) : text);

diff --git a/src/botInstance/dify.js b/src/botInstance/dify.js
@@ -5,6 +5,7 @@ import { ContentCensor } from "../lib/contentCensor.js";
 import { getPuppetEol, isWindowsPlatform } from '../const/puppet-type.js'
 import dayjs from "dayjs";
 import { extractImageLinks } from '../lib/index.js'
+import {getText2Speech} from "../proxy/multimodal.js";
 
 
 class DifyAi {
@@ -99,6 +100,13 @@ class DifyAi {
         };
       }
       let replys = []
+      console.log('是否开启语音', this.config?.openTTS)
+      if(this.config?.openTTS) {
+        replys = await getText2Speech(text, this.config.ttsConfig)
+        if(replys.length) {
+          return replys
+        }
+      }
       let message;
       if(this.config.showQuestion) {
         message = `${content}${this.eol}-----------${this.eol}` +  (this.iswindows ? text.replaceAll('\n', this.eol) : text);

diff --git a/src/botInstance/officialOpenAi.js b/src/botInstance/officialOpenAi.js
@@ -8,6 +8,7 @@ import { getPuppetEol, isWindowsPlatform } from '../const/puppet-type.js'
 import { v4 as uuidv4 } from "uuid";
 import dayjs from "dayjs";
 import { extractImageLinks } from '../lib/index.js'
+import {getText2Speech} from "../proxy/multimodal.js";
 let chatGPT = null
 
 
@@ -198,6 +199,12 @@ class OfficialOpenAi {
       }
 
       let replys = []
+      if(this.config?.openTTS) {
+        replys = await getText2Speech(text, this.config.ttsConfig)
+        if(replys.length) {
+          return replys
+        }
+      }
       let message;
       if(this.config.showQuestion) {
         message = `${content}${this.eol}-----------${this.eol}` + (this.iswindows ? text.replaceAll('\n', this.eol) : text);

diff --git a/src/botInstance/qany.js b/src/botInstance/qany.js
@@ -4,6 +4,7 @@ import { ContentCensor } from '../lib/contentCensor.js'
 import { getPuppetEol, isWindowsPlatform } from '../const/puppet-type.js'
 import dayjs from 'dayjs'
 import { extractImageLinks } from '../lib/index.js'
+import {getText2Speech} from "../proxy/multimodal.js";
 
 class QAnyAi {
   constructor(config = {
@@ -102,6 +103,12 @@ class QAnyAi {
        needHistory: true
       }
       let replys = []
+      if(this.config?.openTTS) {
+        replys = await getText2Speech(text, this.config.ttsConfig)
+        if(replys.length) {
+          return replys
+        }
+      }
       let message
       if (this.config.showQuestion) {
         message = `${content}${this.eol}-----------${this.eol}` + (this.iswindows ? text.replaceAll('\n', this.eol) : text)

diff --git a/src/handlers/on-scan.js b/src/handlers/on-scan.js
@@ -12,13 +12,25 @@ export function resetScanTime () {
   scanTime = 0
 }
 
+function extractLastPathWithCheck(url) {
+  const parts = url.split('/');
+  if (parts[parts.length - 2] === 'x') {
+    return parts[parts.length - 1];
+  } else {
+    return null; // 或者可以返回一个错误消息
+  }
+}
+
 function getQrcodeKey(qrcode) {
   if(!qrcode || !qrcode.startsWith('http')) return
   let url = new URL(qrcode);
   let searchParams = new URLSearchParams(url.search.slice(1));
   if(searchParams.get('key')) {
     console.log('获取到二维码信息中的key')
     globalConfig.setQrKey(searchParams.get('key'))
+  } else if(extractLastPathWithCheck(qrcode)) {
+    console.log('获取到二维码信息中的key');
+    globalConfig.setQrKey(extractLastPathWithCheck(qrcode));
   } else {
     globalConfig.setQrKey('')
   }

diff --git a/src/proxy/config.js b/src/proxy/config.js
@@ -1,4 +1,5 @@
-export const AIBOTK = 'https://api-bot.aibotk.com/open/v1/api'
+// export const AIBOTK = 'https://api-bot.aibotk.com/open/v1/api'
+export const AIBOTK = 'http://127.0.0.1:7002/open/v1/api'
 export const AIBOTK_OUTAPI = 'https://outapi.aibotk.com/open/api/v1'
 export const TULING = 'http://openapi.tuling123.com/openapi/api/v2'
 export const TXHOST = 'http://api.tianapi.com'

diff --git a/src/proxy/multimodal.js b/src/proxy/multimodal.js
@@ -21,7 +21,7 @@ export async function getVoiceText(file, aiConfig) {
     let config = {
       method: 'post',
       maxBodyLength: Infinity,
-      timeout: 30000,
+      timeout: 60000,
       url: AIBOTK_OUTAPI + '/voice/text',
       headers: {
         ...formData.getHeaders(),
@@ -43,6 +43,44 @@ export async function getVoiceText(file, aiConfig) {
   }
 }
 
+/**
+ * 文字转语音
+ * @param text
+ * @param aiConfig
+ * @returns {Promise<*|string>}
+ */
+export async function getText2Speech(text, aiConfig) {
+  try {
+    const env = await getAibotConfig()
+    const { apiKey } = env
+    let config = {
+      method: 'post',
+      maxBodyLength: Infinity,
+      timeout: 60000,
+      url: AIBOTK_OUTAPI + '/text/speech',
+      headers: {
+        'Content-Type': 'application/json',
+        Authorization: `Bearer ${apiKey}`,
+      },
+      data : {
+        text,
+        aiConfig
+      }
+    };
+
+    const result = await axios.request(config)
+    if(result.data.code === 200) {
+      return result.data.data
+    } else {
+      console.log('语音生成出错', result.data.message)
+      return '';
+    }
+  } catch (e) {
+    console.log(`语音生成出错: ${e}`)
+    return '';
+  }
+}
+
 /**
  * 识别图像
  * @param images
@@ -57,7 +95,7 @@ export async function getImageVision(images, question, config) {
     const reqConfig = {
       method: 'post',
       maxBodyLength: Infinity,
-      timeout: 30000,
+      timeout: 60000,
       url: AIBOTK_OUTAPI + '/image/vision',
       headers: {
         Authorization: `Bearer ${apiKey}`,