GET
/v1/health返回本地页面服务状态和当前 TTS capability base URL。
定位
Voice Assistant 现在提供教学 persona、prompt 编排和前端体验；浏览器通过同源 /api/tts/jobs 创建正式 TTS job。
快速开始
- 启动本地页面：
python -m voice_assistant serve --host "127.0.0.1" --port 8000 - 选择教学 persona、语速和语气。
- 前端创建
POST /api/tts/jobs并轮询GET /api/tts/jobs/{id}。 - 成功后播放 capability 返回的
audioUrl。
最小 curl 请求
curl -X POST "http://127.0.0.1:8000/api/tts/jobs" \
-H "Content-Type: application/json" \
-d '{
"text": "Please read this clearly.",
"voice": "Dean",
"stylePrompt": "Clear, focused, teacher-like reading voice...",
"formattedText": "Please read this clearly.",
"audioFormat": "wav",
"metadata": {
"source": "voice-assistant",
"personaId": "persona_dictation_coach",
"speed": "standard",
"toneMode": "coach"
}
}'本地接口
GET
/v1/personas返回当前可用 persona、语速选项和语气选项。该接口用于前端 bootstrap 和调试，不负责合成。
POST
/v1/tts/normalize将教学输入编排为 TTS consumer 可用的 normalized payload。
请求字段
text: 待合成文本，1-5000 字符personaId: persona 标识speed:gentle/standard/brighttoneMode:companion/coach/encourage/livelyaudioFormat:wav/mp3
成功响应示例
{
"text": "学习不能只停留在记住知识。",
"voice": "冰糖",
"stylePrompt": "角色：你是“陈老师”...",
"formattedText": "学习不能只停留在记住知识。",
"audioFormat": "wav",
"metadata": {
"source": "voice-assistant",
"personaId": "persona_morning_recital",
"speed": "standard",
"toneMode": "coach",
"paceProfile": "medium"
}
}前端接入示例
Node.js / Browser
const job = await fetch("/api/tts/jobs", {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify(normalizedPayload)
}).then((response) => response.json());
const detail = await fetch(`/api/tts/jobs/${job.id}`).then((response) => response.json());
if (detail.status === "success") {
audio.src = detail.audioUrl;
}