# server.py
import sys
#print("当前Python路径：", sys.executable)
import os
from mcp.server.fastmcp import FastMCP
from volcengine_mcp.volc_tts import VolcTTS
from dotenv import load_dotenv
from mcp.types import TextContent
from volcengine_mcp.config import get_voice_type, VOICE_LIST
from typing import Optional
import uuid
load_dotenv()
VOLC_APPID = os.getenv("VOLC_APPID")
VOLC_TOKEN = os.getenv("VOLC_TOKEN")
#print("VOLC_APPID:", VOLC_APPID)
#print("VOLC_TOKEN:", VOLC_TOKEN)
tts_client = VolcTTS(VOLC_APPID, VOLC_TOKEN)
mcp = FastMCP("VolcEngineTTS")
#print("准备注册 volcengine_tts 工具")
@mcp.tool(
description="""
将文本转为语音并保存为音频文件，支持所有火山引擎TTS参数。
支持自然语言描述音色，如"青年男声""甜美女声"等。
参数说明：
text (str): 要合成的文本内容。
voice_type (str, 可选): 使用的音色名称，默认为官方女声。
encoding (str, 可选): 音频编码格式，默认为mp3。
speed_ratio (float, 可选): 语速，1.0为正常语速。
uid (str, 可选): 用户ID，默认uid123。
cluster (str, 可选): 集群，默认volcano_tts。
emotion (str, 可选): 情感标签。
enable_emotion (bool, 可选): 是否启用情感。
emotion_scale (float, 可选): 情感强度。
rate (int, 可选): 采样率，默认24000。
bitrate (int, 可选): 比特率，默认160。
loudness_ratio (float, 可选): 响度，默认1.0。
explicit_language (str, 可选): 明确语言。
context_language (str, 可选): 上下文语言。
silence_duration (float, 可选): 静音时长。
with_timestamp (bool, 可选): 是否返回时间戳。
extra_param (dict, 可选): 额外参数。
output_dir (str, 可选): 音频文件保存目录，优先级高于环境变量 OUTPUT_DIR，默认当前目录。
output_filename (str, 可选): 自定义音频文件名（无需扩展名），如不指定则自动生成唯一文件名。
返回：
TextContent，包含合成成功提示和音频文件路径。
English:
Convert text to speech with a given voice and save the output audio file to a given directory.
Directory is optional, if not provided, the output file will be saved to the current working directory.
Args:
text (str): The text to convert to speech.
voice_type (str, optional): The name of the voice to use.
encoding (str, optional): Output audio encoding, default is mp3.
speed_ratio (float, optional): Speed of the generated audio. 1.0 is normal speed.
uid (str, optional): User ID, default is uid123.
cluster (str, optional): Cluster, default is volcano_tts.
emotion (str, optional): Emotion tag.
enable_emotion (bool, optional): Whether to enable emotion.
emotion_scale (float, optional): Emotion intensity.
rate (int, optional): Sampling rate, default is 24000.
bitrate (int, optional): Bitrate, default is 160.
loudness_ratio (float, optional): Loudness, default is 1.0.
explicit_language (str, optional): Explicit language.
context_language (str, optional): Context language.
silence_duration (float, optional): Silence duration.
with_timestamp (bool, optional): Whether to return timestamp.
extra_param (dict, optional): Extra parameters.
output_dir (str, optional): Directory to save the output audio file. Priority: argument > env OUTPUT_DIR > current dir.
output_filename (str, optional): Custom audio file name (no extension needed). If not specified, a unique file name will be generated.
Returns:
Text content with the path to the output file and name of the voice used.
"""
)
def volcengine_tts(
text: str,
voice_type: Optional[str] = "zh_female_wanqudashu_moon_bigtts",
encoding: Optional[str] = "mp3",
speed_ratio: Optional[float] = 1.0,
uid: Optional[str] = None,
cluster: Optional[str] = "volcano_tts",
emotion: Optional[str] = None,
enable_emotion: Optional[bool] = None,
emotion_scale: Optional[float] = None,
rate: Optional[int] = 24000,
bitrate: Optional[int] = 160,
loudness_ratio: Optional[float] = 1.0,
explicit_language: Optional[str] = None,
context_language: Optional[str] = None,
silence_duration: Optional[float] = None,
with_timestamp: Optional[bool] = None,
extra_param: Optional[dict] = None,
output_dir: Optional[str] = None,
output_filename: Optional[str] = None
) -> TextContent:
"""调用火山引擎语音合成，将文本转为语音mp3，返回音频文件路径"""
# 类型安全：所有 int 类型参数都强制转为 int
if rate is not None:
rate = int(rate)
if bitrate is not None:
bitrate = int(bitrate)
# 使用config.py的get_voice_type进行音色解析
#print("volcengine_tts 被调用")
#print(f"调用 synthesize, text={text}, voice_type={voice_type}")
voice_type = get_voice_type(voice_type)
#print('映射后voice_type:', voice_type)
if not uid or uid is None:
uid = str(uuid.uuid4())
audio_bytes = tts_client.synthesize(
text,
voice_type=voice_type,
encoding=encoding,
speed_ratio=speed_ratio,
uid=uid,
cluster=cluster,
emotion=emotion,
enable_emotion=enable_emotion,
emotion_scale=emotion_scale,
rate=rate,
bitrate=bitrate,
loudness_ratio=loudness_ratio,
explicit_language=explicit_language,
context_language=context_language,
silence_duration=silence_duration,
with_timestamp=with_timestamp,
extra_param=extra_param
)
# 优先用参数，其次用环境变量，最后用当前目录
if not output_dir:
output_dir = os.getenv("OUTPUT_DIR", os.getcwd())
os.makedirs(output_dir, exist_ok=True)
import time
if output_filename:
filename = output_filename
if not filename.endswith(f".{encoding}"):
filename += f".{encoding}"
else:
filename = f"volc_tts_{int(time.time())}_{uuid.uuid4().hex[:8]}.{encoding}"
out_path = os.path.join(output_dir, filename)
with open(out_path, "wb") as f:
f.write(audio_bytes)
return TextContent(
type="text",
text=f"合成成功，音频文件路径：{out_path}"
)
@mcp.tool(
description="""
获取所有可用的火山引擎TTS音色列表。
返回：
List[dict]，每个元素包含音色的scene、desc、voice_type、lang、emotions、keywords等信息。
"""
)
def list_voice() -> list:
"""返回所有可用音色信息列表"""
return VOICE_LIST
def main():
mcp.run()
#print("准备启动 MCP 服务")
if __name__ == "__main__":
#mcp.run(transport="streamable-http")
main()
#print("MCP 服务已启动")
