Merge pull request #1348 from xinnan-tech/custom_tts_api

Custom tts api

Author: xinnan-tech

main/manager-api/src/main/resources/db/changelog/202505151451.sql

@@ -0,0 +1,18 @@
+-- 修改自定义TTS接口请求定义
+update `ai_model_provider` set `fields` =
+'[{"key":"url","label":"服务地址","type":"string"},{"key":"method","label":"请求方式","type":"string"},{"key":"params","label":"请求参数","type":"dict","dict_name":"params"},{"key":"headers","label":"请求头","type":"dict","dict_name":"headers"},{"key":"format","label":"音频格式","type":"string"},{"key":"output_dir","label":"输出目录","type":"string"}]'
+where `id` = 'SYSTEM_TTS_custom';
+
+-- 修改自定义TTS配置说明
+UPDATE `ai_model_config` SET
+`doc_link` = NULL,
+`remark` = '自定义TTS配置说明：
+1. 自定义的TTS接口服务，请求参数可自定义，可接入众多TTS服务
+2. 以本地部署的KokoroTTS为例
+3. 如果只有cpu运行：docker run -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-cpu:latest
+4. 如果只有gpu运行：docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:latest
+配置说明：
+1. 在params中配置请求参数,使用JSON格式
+   例如KokoroTTS：{ "input": "{prompt_text}", "speed": 1, "voice": "zm_yunxi", "stream": true, "download_format": "mp3", "response_format": "mp3", "return_download_link": true }
+2. 在headers中配置请求头
+3. 设置返回音频格式' WHERE `id` = 'TTS_CustomTTS';

main/manager-api/src/main/resources/db/changelog/db.changelog-master.yaml

@@ -155,4 +155,11 @@ databaseChangeLog:
       changes:
         - sqlFile:
             encoding: utf8
-            path: classpath:db/changelog/202505201744.sql
+            path: classpath:db/changelog/202505201744.sql
+  - changeSet:
+      id: 202505151451
+      author: hsoftxl
+      changes:
+        - sqlFile:
+            encoding: utf8
+            path: classpath:db/changelog/202505151451.sql

main/manager-web/src/components/DeviceItem.vue

@@ -28,7 +28,7 @@
       </div>
       <div class="settings-btn" @click="handleChatHistory"
         :class="{ 'disabled-btn': device.memModelId === 'Memory_nomem' }">
-        <el-tooltip v-if="device.memModelId === 'Memory_nomem'" content="未开启记忆" placement="top">
+        <el-tooltip v-if="device.memModelId === 'Memory_nomem'" content="请先在“配置角色”界面开启记忆" placement="top">
           <span>聊天记录</span>
         </el-tooltip>
         <span v-else>聊天记录</span>

main/manager-web/src/components/HeaderBar.vue

@@ -35,10 +35,11 @@
           OTA管理
         </div>
         <el-dropdown v-if="isSuperAdmin" trigger="click" class="equipment-management more-dropdown"
-          :class="{ 'active-tab': $route.path === '/dict-management' || $route.path === '/params-management' || $route.path === '/provider-management' }" @visible-change="handleParamDropdownVisibleChange">
+          :class="{ 'active-tab': $route.path === '/dict-management' || $route.path === '/params-management' || $route.path === '/provider-management' || $route.path === '/server-side-management' }"
+          @visible-change="handleParamDropdownVisibleChange">
           <span class="el-dropdown-link">
             <img loading="lazy" alt="" src="@/assets/header/param_management.png"
-              :style="{ filter: $route.path === '/dict-management' || $route.path === '/params-management' || $route.path === '/provider-management' ? 'brightness(0) invert(1)' : 'None' }" />
+              :style="{ filter: $route.path === '/dict-management' || $route.path === '/params-management' || $route.path === '/provider-management' || $route.path === '/server-side-management' ? 'brightness(0) invert(1)' : 'None' }" />
             参数字典
             <i class="el-icon-arrow-down el-icon--right" :class="{ 'rotate-down': paramDropdownVisible }"></i>
           </span>

main/xiaozhi-server/config.yaml

@@ -675,17 +675,24 @@ TTS:
     speed: 1
     output_dir: tmp/
   CustomTTS:
-    # 自定义的TTS接口服务，请求参数可自定义
-    # 要求接口使用GET方式请求，并返回音频文件
+    # 自定义的TTS接口服务，请求参数可自定义，可接入众多TTS服务
+    # 以本地部署的KokoroTTS为例
+    # 如果只有cpu运行：docker run -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-cpu:latest
+    # 如果只有gpu运行：docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:latest
+    # 要求接口使用POST方式请求，并返回音频文件
     type: custom
-    url: "http://127.0.0.1:9880/tts"
+    method: POST
+    url: "http://127.0.0.1:8880/v1/audio/speech"
     params: # 自定义请求参数
-      # text: "{prompt_text}" # {prompt_text}会被替换为实际的提示词内容
-      # speaker: jok老师
-      # speed: 1
-      # foo: bar
-      # testabc: 123456
+      input: "{prompt_text}"
+      response_format: "mp3"
+      download_format: "mp3"
+      voice: "zf_xiaoxiao"
+      lang_code: "z"
+      return_download_link: true
+      speed: 1
+      stream: false
     headers: # 自定义请求头
       # Authorization: Bearer xxxx
-    format: wav # 接口返回的音频格式
+    format: mp3 # 接口返回的音频格式
     output_dir: tmp/

main/xiaozhi-server/core/providers/tts/custom.py

@@ -1,4 +1,5 @@
 import os
+import json
 import uuid
 import requests
 from config.logger import setup_logging
@@ -12,11 +13,21 @@ class TTSProvider(TTSProviderBase):
     def __init__(self, config, delete_audio_file):
         super().__init__(config, delete_audio_file)
         self.url = config.get("url")
+        self.method = config.get("method", "GET")
         self.headers = config.get("headers", {})
-        self.params = config.get("params")
         self.format = config.get("format", "wav")
         self.output_file = config.get("output_dir", "tmp/")
 
+        self.params = config.get("params")
+
+        if isinstance(self.params, str):
+            try:
+                self.params = json.loads(self.params)
+            except json.JSONDecodeError:
+                raise ValueError("Custom TTS配置参数出错,无法将字符串解析为对象")
+        elif not isinstance(self.params, dict):
+            raise TypeError("Custom TTS配置参数出错, 请参考配置说明")
+
     def generate_filename(self):
         return os.path.join(self.output_file, f"tts-{datetime.now().date()}@{uuid.uuid4().hex}.{self.format}")
 
@@ -27,7 +38,10 @@ async def text_to_speak(self, text, output_file):
                 v = v.replace("{prompt_text}", text)
             request_params[k] = v
 
-        resp = requests.get(self.url, params=request_params, headers=self.headers)
+        if self.method.upper() == "POST":
+            resp = requests.post(self.url, json=request_params, headers=self.headers)
+        else:
+            resp = requests.get(self.url, params=request_params, headers=self.headers)
         if resp.status_code == 200:
             with open(output_file, "wb") as file:
                 file.write(resp.content)