openai_api_chatbot.py

print('### IMPORTING DEPENDANCIES ###')
import os
import whisper
import pygame

from Assistant import get_audio as myaudio
from Assistant.VirtualAssistant import VirtualAssistant

os.environ['OPENAI_API_KEY']  = 'your-openai-api-key'
os.environ['IBM_API_KEY']     = 'your-ibm-cloud-watson-api-key'
os.environ['IBM_TTS_SERVICE'] = 'your-ibm-cloud-watson-tts-url'

print('DONE\n')


### MAIN
if __name__=="__main__":
    print("### SETTING UP ENVIROMENT ###")
    
    SOUND_DIR = os.path.join('sounds')
    
    print('loading whisper model...')
    whisper_model = whisper.load_model("medium") # pick the one that works best for you, but remember: only medium and large are multi language

    print('opening pygame:')
    pygame.mixer.init()
    
    # INITIATE JARVIS
    print('initiating JARVIS voice...')
    jarvis = VirtualAssistant(
        openai_api = os.getenv('OPENAI_API_KEY'),
        ibm_api    = os.getenv('IBM_API_KEY'),
        ibm_url    = os.getenv('IBM_TTS_SERVICE'),
        whisper_model= whisper_model,
        awake_with_keywords=["elephant"],
        model= "gpt-3.5-turbo",
        embed_model= "text-embedding-ada-002",
        RESPONSE_TIME = 3,
        SLEEP_DELAY = 30,
        )

    while True:
        if not(jarvis.is_awake):
            print('\n awaiting for triggering words...')
            while not(jarvis.is_awake):
                jarvis.listen_passively()
        
        jarvis.record_to_file('output.wav')
        
        if jarvis.is_awake:
            question, detected_language = myaudio.whisper_wav_to_text('output.wav',whisper_model)

            # THIS STUFF WILL NEED TO BE DONE BY SOME SORT OF "PROMPT MANAGER" (next on the list)
            # check exit command
            if "THANKS" in question.upper():
                jarvis.go_to_sleep()
                continue
            
            if "HEY" in question.upper() and "JARVIS" in question.upper() and detected_language=='en':
                question = question.upper().replace('HEY JARVIS', '')
                question = question.lower()
                VoiceIdx = 'jarvis'
            else:
                VoiceIdx = detected_language
            
            response = jarvis.get_answer(question)

            jarvis.say(response, VoiceIdx=VoiceIdx)

            print('\n')