ChatGPTでレシピちゃんを作ってみた

import openai
import requests

# Set OpenAI Chat Completion API Endpoint
ENDPOINT = 'https://api.openai.com/v1/chat/completions'
def get_chat_response(prompt):
　　　　# Set up API request headers
　　　　headers = {
　　　　　　　　'Content-Type': 'application/json; charset=utf-8',
　　　　　　　　'Authorization': 'Bearer ' + constants.openai_api_key
　　　　　　　　}

　　　　# Set API request data
　　　　data = {
　　　　　　　　'model': 'gpt-3.5-turbo',
　　　　　　　　'messages': [{'role': 'user', 'content': prompt }]
　　　　　　　　}

　　　　# Send API request
　　　　response = requests.post(ENDPOINT, headers=headers, json=data, timeout=60)

　　　　# Analyze API response
　　　　response_json = response.json()
　　　　message = response_json['choices'][0]['message']['content'].strip()

　　　　return message

# #### install necessary libraries
# update and upgrade recent packages
apt-get update
apt-get upgrade
# install robothat
cd /home/pi/
git clone https://github.com/sunfounder/robot-hat.git
cd robot-hat
sudo python3 setup.py install

# install pisloth
cd /home/pi/
git clone -b v2.0 https://github.com/sunfounder/pisloth.git
cd pisloth
sudo python3 setup.py install

# install i2samp
cd /home/pi/pisloth
sudo bash i2samp.sh

# install related python libraries
pip3 install openai
pip3 install pyopenjtalk
pip3 install scipy
pip3 install SpeechRecognition
sudo apt-get install flac

Ref: http://keizan3105diy.xrea.jp/rdiy200503.htm

USBが接続できるかを確認し、USBとRobotHatの順番を確認します。
$ lsusb
$　cat /proc/asound/modules
0 snd_soc_rpi_simple_soundcard
1 snd_usb_audio

ここで、soundcard(speaker)が0で、usb micが1ですので、合わせて設定します。
$ sudo nano /home/.asoundrc

pcm.!default {
type asym
capture.pcm "mic"
playback.pcm "speaker"
}

pcm.mic {
type plug
slave {
pcm "hw:0,0"
}
}

pcm.speaker {
type plug
slave {
pcm "hw:1,0"
}
}

終わったら、再起動します。
$ sudo reboot

def process_text_to_speech(text, wavfilename = "output/test.wav", stop_sound_flag=False):
    x, sr = pyopenjtalk.tts(text)
    wavfile.write(wavfilename, sr, x.astype(np.int16))
    if stop_sound_flag==True:
        mixer.music.stop()
    audio = pygame.mixer.Sound(wavfilename)
    audio.play()
return

def process_speech_to_text(speech_time, wavfilename = "output/test.wav"):
    # record voice
    cmd = "arecord -D plughw:1,0 -d {} -f cd {}".format(speech_time, wavfilename)
    print(cmd)
    os.system(cmd)

    # convert speech to text
    r = sr.Recognizer()
    audio_file = os.path.join(os.path.dirname(os.path.realpath(__file__)), wavfilename)
    print(audio_file)

    with sr.AudioFile(audio_file) as source:
        audio = r.record(source)

    try:
        text = r.recognize_google(audio, language='ja-JP')
        print(text)
    except sr.UnknownValueError:
        print("Google Speech Recognition could not understand audio")
    except sr.RequestError as e:
        print("Could not request results from Google Speech Recognition service; {0}".format(e))
    return text

import pygame
from pygame.locals import *
from pygame import mixer
from robot_hat import Music

# set up initial
pygame.init()
mixer.init()
music = Music()

# set up volume
music.music_set_volume(20) 

# load music file
mixer.music.load('output/christmas.mp3')

# start playing music
mixer.music.play() 

# stop playing music
mixer.music.stop()</span>

from pisloth import Sloth
from robot_hat import Ultrasonic, Pin

def dont_touch_me(wavfilename):
    sloth = Sloth([1,2,3,4])
    sloth.set_offset([0,0,0,0])
    sonar = Ultrasonic(Pin("D2") ,Pin("D3"))

    alert_distance = 50

    distance = sonar.read()
    if distance <= alert_distance: 
        process_text_to_speech("あー、僕が邪魔ですか。移動しますね", wavfilename) 
        sloth.do_action('turn right', 4, 90)
        sloth.do_action('forward', 2, 90) 
    else: 
        sloth.do_action('stand', 1, 90) 
        time.sleep(1)

import os
import os
import re
import time
from scipy.io import wavfile
import numpy as np
import requests
import pyaudio
import pygame
from pygame.locals import *
from pygame import mixer
import pyopenjtalk
import speech_recognition as sr
from pisloth import Sloth
from robot_hat import TTS, Music
from robot_hat import Ultrasonic, Pin
import openai
import constants


# Set OpenAI Chat Completion API Endpoint
ENDPOINT = 'https://api.openai.com/v1/chat/completions'

def get_chat_response(prompt):
    # Set up API request headers
    headers = {
        'Content-Type': 'application/json; charset=utf-8',
        'Authorization': 'Bearer ' + constants.openai_api_key
        }

    # Set API request data
    data = {
        'model': 'gpt-3.5-turbo',
        'messages': [{'role': 'user', 'content': prompt }]
        }

    # Send API request
    response = requests.post(ENDPOINT, headers=headers, json=data, timeout=60)

    # Analyze API response
    response_json = response.json()
    message = response_json['choices'][0]['message']['content'].strip()

    return message


def move_forward():
    sloth = Sloth([1,2,3,4])
    sloth.set_offset([0,0,0,0])
    sloth.do_action('forward', 2, 90)
    return


def process_speech_to_text(speech_time, wavfilename = "output/test.wav"):
    cmd = "arecord -D plughw:1,0 -d {} -f cd {}".format(speech_time, wavfilename)
    print(cmd)
    os.system(cmd)

    r = sr.Recognizer()
    audio_file = os.path.join(os.path.dirname(os.path.realpath(__file__)), wavfilename)
    print(audio_file)

    with sr.AudioFile(audio_file) as source:
        audio = r.record(source)

    try:
        text = r.recognize_google(audio, language='ja-JP')
        print(text)
    except sr.UnknownValueError:
        print("Google Speech Recognition could not understand audio")
    except sr.RequestError as e:
        print("Could not request results from Google Speech Recognition service; {0}".format(e))
    return text


def process_text_to_speech(text, wavfilename = "output/test.wav", stop_sound_flag=False):
    x, sr = pyopenjtalk.tts(text)
    wavfile.write(wavfilename, sr, x.astype(np.int16))
    
    if stop_sound_flag==True:
        mixer.music.stop()
        audio = pygame.mixer.Sound(wavfilename)
        audio.play()
    return


def simple_dance():
    sloth = Sloth([1,2,3,4])
    sloth.set_offset([0,0,0,0])


    sloth.do_action('swing',2,bpm=129)
    sloth.do_action('moon walk left',2,bpm=129)
    sloth.do_action('moon walk right',2,bpm=129)


def dont_touch_me(wavfilename):
    sloth = Sloth([1,2,3,4])
    sloth.set_offset([0,0,0,0])
    sonar = Ultrasonic(Pin("D2") ,Pin("D3"))

    alert_distance = 50
    distance = sonar.read()
    if distance <= alert_distance: 
        process_text_to_speech("あー、僕が邪魔ですか。移動しますね", wavfilename) 
        # sloth.do_action('backward', 2, 90) 
        sloth.do_action('turn right', 4, 90) 
        sloth.do_action('forward', 2, 90) 
    else: sloth.do_action('stand', 1, 90) 
         time.sleep(1) 

def main(): 
    pygame.init() 
    mixer.init() 
    music = Music() 
    music.music_set_volume(20) 

    # set up prompt filename 
    wavfilename = "output/test.wav" 

    # Greeting prompt0 = "こんにちは 御用はなんですか？" 
    process_text_to_speech(prompt0, wavfilename) 
    move_forward() 

    # Get order process_text_to_speech("10秒で指示をください〜", wavfilename) 
    prompt = process_speech_to_text(10, wavfilename) 
    mixer.music.load('output/christmas.mp3') 
    mixer.music.play() 
    process_text_to_speech("承知しました。調べますので、音楽を聴きながら、少々お待ちくださいね〜", wavfilename) 

    # Get Chatbot response and process TTS 
    chat_response = get_chat_response(prompt) 
    # print('Chatbot: ' + chat_response) 
    process_text_to_speech("お待たせしました。答えは　", wavfilename) 
    process_text_to_speech(chat_response, stop_sound_flag=True) 
    simple_dance() 
    timeout = time.time() + 60*5 # 5 minutes from now 

    while True: 
        if time.time() > timeout:
            break
        dont_touch_me(wavfilename)

if __name__ == "__main__":
main()