ChatGPT - BeagleY-AI

简介

该项目使用 BeagleY-AI 集成了语音输入、大模型响应和语音输出功能。它采用 ReSpeaker Lite 作为音频输入和输出设备，实现与 ChatGPT 和语音转文本转换服务的无缝交互。

所需硬件

入门指南

请查看入门文档来设置您的 BeagleY-AI，并将其连接到网络。

note

确保您的 Python 版本高于 Python3.7.1。
检查版本：

python3 --version

安装库

sudo apt update
sudo apt install python3-pip python3-dev
sudo apt install portaudio19-dev
sudo apt install ffmpeg -y
sudo apt-get install flac
pip3 install pyaudio
pip3 install speechrecognition
pip3 install openai
pip3 install playsound

运行以下命令配置 ReSpeaker Lite：

pw-metadata -n settings 0 clock.force-rate 16000

tip

调整 ReSpeaker Lite 音量的命令：

alsamixer

代码

以下 Python 代码实现了一个简单的语音助手，它会监听唤醒词，识别用户语音命令，将其转换为文本，使用 GPT-4 生成响应，然后将响应转换为语音并播放。

设备首先等待唤醒词，然后监听用户的命令。一旦接收到命令，程序会使用 GPT-4 生成响应并以语音形式播放。如果连续三次未能识别命令，程序会返回到监听唤醒词状态，您需要再次说出唤醒词以启动新的语音交互会话。

步骤1：配置 API 密钥

export OPENAI_API_KEY= 'your-api-key-here'

步骤2：创建一个新的 Python 文件并输入以下代码：

import speech_recognition as sr
from openai import OpenAI
from pathlib import Path
from pydub import AudioSegment
import os


client = OpenAI()

def text_to_speech(text):
    speech_file_path = Path(__file__).parent / "speech.mp3"
    response = client.audio.speech.create(
    model="tts-1",
    voice="alloy",
    input=text
    )

    response.stream_to_file(speech_file_path)
    audio = AudioSegment.from_mp3("speech.mp3")
    audio.export("speech.wav", format="wav")
    cmdline = 'aplay ' + " speech.wav" 
    os.system(cmdline)



# 初始化识别器
recognizer = sr.Recognizer()
microphone = sr.Microphone()

# 定义唤醒词
WAKE_WORD = "hi"

def listen_for_wake_word():
    with microphone as source:
        recognizer.adjust_for_ambient_noise(source, duration=0.5)
        print("Listening for wake word...")
        
        while True:
            audio = recognizer.listen(source)
            # audio = recognizer.listen(source, timeout=5, phrase_time_limit=5)
            try:
                text = recognizer.recognize_google(audio).lower()
                if WAKE_WORD in text:
                    print(f"Wake word '{WAKE_WORD}' detected.")
                    text_to_speech("hi,what can i do for you?")
                    return True
            except sr.UnknownValueError:
                continue
            except sr.RequestError as e:
                print(f"Could not request results; {e}")
                continue

i=0
def listen_for_command():
    global i
    with microphone as source:
        print("Listening for command...")
        # audio = recognizer.listen(source)
        audio = recognizer.listen(source, timeout=5, phrase_time_limit=5)
        try:
            command = recognizer.recognize_google(audio)
            print(f"You said: {command}")
            i=0
            return command
        except sr.UnknownValueError:
            print("Could not understand the audio")
            i = i+1
        except sr.RequestError as e:
            print(f"Could not request results; {e}")
            i = i+1


def get_gpt_response(prompt):
    completion = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "Your name is speaker, you can answer all kinds of questions for me"},
        {"role": "user", "content": prompt}
    ]
    )

    content_string = completion.choices[0].message.content
    paragraphs = content_string.split('\n\n')
    combined_content = ' '.join(paragraphs)
    return combined_content




def main():
    global i
    while 1:
        flag = listen_for_wake_word()
        while flag == True:
            user_input = listen_for_command()
            if i==3:
                flag = False
                i = 0
            if user_input:
                gpt_response = get_gpt_response(user_input)
                print(f"GPT says: {gpt_response}")
                text_to_speech(gpt_response)
                

if __name__ == "__main__":
    main()

步骤3：运行 Python 文件。

python LLM_beagle.py

简介​

所需硬件​

入门指南​

安装库​

代码​

简介

所需硬件

入门指南

安装库

代码