Spaces:
Sleeping
Sleeping
Upload 3 files
Browse files
examples/education/mcp.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"mcpServers": {
|
3 |
+
"text_to_audio_local_sse": {
|
4 |
+
"url": "http://0.0.0.0:8888/sse"
|
5 |
+
}
|
6 |
+
}
|
7 |
+
}
|
examples/education/mcp_server_text2audio.py
ADDED
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import numpy as np
|
2 |
+
|
3 |
+
from mcp.server import FastMCP
|
4 |
+
from pydantic import Field
|
5 |
+
|
6 |
+
from aworld.utils import import_package
|
7 |
+
from aworld.logs.util import logger
|
8 |
+
|
9 |
+
# Import required packages
|
10 |
+
import_package('gtts', install_name='gTTS')
|
11 |
+
import_package('pyttsx3', install_name='pyttsx3')
|
12 |
+
import_package('librosa', install_name='librosa')
|
13 |
+
import_package('soundfile', install_name='soundfile')
|
14 |
+
import pyttsx3
|
15 |
+
from gtts import gTTS
|
16 |
+
import librosa
|
17 |
+
import soundfile as sf
|
18 |
+
|
19 |
+
mcp = FastMCP("text_to_audio")
|
20 |
+
|
21 |
+
@mcp.tool()
|
22 |
+
def convert_text_to_audio(
|
23 |
+
text: str = Field(description="Text to convert to audio"),
|
24 |
+
output_file: str = Field(description="Path to the generated audio file")
|
25 |
+
) -> str:
|
26 |
+
"""Convert input text to audio with child-friendly settings.
|
27 |
+
|
28 |
+
Args:
|
29 |
+
text: Input text to convert
|
30 |
+
output_file: Path to the generated audio file
|
31 |
+
Returns:
|
32 |
+
str: Path to the generated audio file
|
33 |
+
"""
|
34 |
+
engine = pyttsx3.init()
|
35 |
+
# Set default properties for child-friendly speech
|
36 |
+
engine.setProperty('rate', 150) # Slower speaking rate
|
37 |
+
engine.setProperty('volume', 0.9)
|
38 |
+
try:
|
39 |
+
# Use default params if none provided
|
40 |
+
params = {
|
41 |
+
"speed": 0.9,
|
42 |
+
"pitch": 1.1,
|
43 |
+
"language": "en-US",
|
44 |
+
"output_file": output_file,
|
45 |
+
"use_gtts": True
|
46 |
+
}
|
47 |
+
|
48 |
+
# Preprocess text for child-friendly output
|
49 |
+
text = _preprocess_text(text)
|
50 |
+
if params.get("use_gtts", False):
|
51 |
+
# Use gTTS for more natural sound
|
52 |
+
tts = gTTS(text=text, lang=params["language"], slow=True)
|
53 |
+
tts.save(params["output_file"])
|
54 |
+
|
55 |
+
# Post-process audio if needed (adjust volume, remove noise, etc.)
|
56 |
+
_post_process_audio(params["output_file"])
|
57 |
+
return params["output_file"]
|
58 |
+
|
59 |
+
except Exception as e:
|
60 |
+
logger.error("Error in text-to-audio conversion: %s", str(e))
|
61 |
+
raise
|
62 |
+
|
63 |
+
def _preprocess_text(text: str) -> str:
|
64 |
+
"""Preprocess text for child-friendly output.
|
65 |
+
|
66 |
+
- Add pauses between sentences
|
67 |
+
- Emphasize important words
|
68 |
+
- Handle special characters
|
69 |
+
"""
|
70 |
+
# Add slight pauses between sentences
|
71 |
+
text = text.replace('. ', '... ')
|
72 |
+
# Add emphasis on important words (can be customized)
|
73 |
+
text = text.replace('!', '! ... ')
|
74 |
+
return text
|
75 |
+
|
76 |
+
def _post_process_audio(audio_file: str) -> None:
|
77 |
+
"""Optimized post-processing for audio files."""
|
78 |
+
try:
|
79 |
+
# Load with a lower sample rate and mono channel
|
80 |
+
y, sr = librosa.load(audio_file, sr=16000, mono=True)
|
81 |
+
# Use faster normalization method
|
82 |
+
y_norm = y / np.max(np.abs(y))
|
83 |
+
# Write with optimized settings
|
84 |
+
sf.write(
|
85 |
+
audio_file,
|
86 |
+
y_norm,
|
87 |
+
sr,
|
88 |
+
format='mp4',
|
89 |
+
subtype='MP4'
|
90 |
+
)
|
91 |
+
except (IOError, ValueError, RuntimeError) as e:
|
92 |
+
logger.warning("Audio post-processing failed: %s", e)
|
93 |
+
|
94 |
+
# Main function
|
95 |
+
if __name__ == "__main__":
|
96 |
+
mcp.settings.port = 8888
|
97 |
+
mcp.run(transport='sse')
|
98 |
+
|
99 |
+
# text = "Hello, this is a test of the text-to-audio conversion."
|
100 |
+
# output_file = "output1.mp4"
|
101 |
+
# print(f"Converting text to audio: {text}")
|
102 |
+
# audio_file = convert_text_to_audio(text, output_file)
|
103 |
+
# print(f"Audio file saved to: {audio_file}")
|
examples/education/run.py
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding: utf-8
|
2 |
+
# Copyright (c) 2025 inclusionAI.
|
3 |
+
|
4 |
+
from aworld.config.conf import AgentConfig
|
5 |
+
from aworld.core.agent.base import Agent
|
6 |
+
from aworld.runner import Runners
|
7 |
+
|
8 |
+
if __name__ == '__main__':
|
9 |
+
agent_config = AgentConfig(
|
10 |
+
llm_provider="openai",
|
11 |
+
llm_model_name="gpt-4o",
|
12 |
+
llm_api_key="YOUR_API_KEY",
|
13 |
+
llm_base_url="http://localhost:5080"
|
14 |
+
)
|
15 |
+
|
16 |
+
edu_sys_prompt = "You are a helpful agent to convert text to audio for children education."
|
17 |
+
edu = Agent(
|
18 |
+
conf=agent_config,
|
19 |
+
name="edu_agent",
|
20 |
+
system_prompt=edu_sys_prompt,
|
21 |
+
mcp_servers=["text_to_audio_local_sse"] # MCP server name for agent to use
|
22 |
+
)
|
23 |
+
|
24 |
+
# run
|
25 |
+
Runners.sync_run(
|
26 |
+
input="use text_to_audio_local_sse to convert text to audio: Hello, world!",
|
27 |
+
agent=edu
|
28 |
+
)
|