ai-voice-assistant

electron based ai voice assistant utilizing gcloud text-to-speech and speech-to-text and openai chatgpt.
on machines with enough power, you can use vosk for complete offline voice recgnition.

Prerequisite

• debian based OS (tested on Ubuntu and Raspberrypi OS)
• node 16 (recommended install via nvm https://github.com/nvm-sh/nvm)

Project setup

npm install --legacy-peer-deps

Google Setup

• install gcloud cli (https://cloud.google.com/sdk/docs/install?hl=de#deb)
• you need google account with Cloud Text-to-Speech API enabled (https://cloud.google.com/text-to-speech/docs)
• you also need Cloud Speech-to-Text API enabled if not using VOSK (can be configured in assistant_config.mjs)
• after enabling both, you need to create a service account for this project and download a credentials JSON file that looks like this

{
    "type": "service_account",
    "project_id": "{{your_project_id}}",
    "private_key_id": "{{your_private_key_id}}",
    "private_key": "{{your_private_key}}",
    "client_email": "local-account@{{your_project_id}}.iam.gserviceaccount.com",
    "client_id": "{{your_client_id}}",
    "auth_uri": "https://accounts.google.com/o/oauth2/auth",
    "token_uri": "https://oauth2.googleapis.com/token",
    "auth_provider_x509_cert_url": "https://www.googleapis.com/oauth2/v1/certs",
    "client_x509_cert_url": "https://www.googleapis.com/robot/v1/metadata/x509/local-account%40{{your_project_id}}.iam.gserviceaccount.com",
    "universe_domain": "googleapis.com"
  }
  

• put this file in the root directory of the project and adjust your .env file to point to this credentials file

GOOGLE_APPLICATION_CREDENTIALS="/home/your-projects-directory/raspberry-ai-voice-assitant/google_client_secret.json"

ChatGPT Setup

• create .env file with this content replacing {{your-api-key}}
• can be obtained on openai homepage (https://platform.openai.com/account/api-keys)

OPEN_AI_APIKEY="{{your-api-key}}"

Language Model Setup

• voice assistant can be used with offline voice recognition
• language models will be loaded into RAM, so this requires a laptop or later version of raspberry pi
• to enable vosk, adjust assistant_config.mjs
• download models here: https://alphacephei.com/vosk/models
• create folder language_models in project root
• extract downloaded language model in there
• adjust modelPaths in src/voice_assistant_vosk.mjs

Run the voice assitant

npm run electron:serve

Customization

• check assistant_config.mjs

const assistantName = 'Felix';

export default {
    'useLocalSpeechToText': false, // if true uses vosk (free), if false uses google speech to text
    'lowMemoryVariant': false, // if useLocalSpeechToText is true, this determines, if big language model is used
    'language': 'de', // language used for speech to text and answers from chatGPT
    'answerWordLimit': 30, // limits the requested words of an answer from chatGPT
    'assistantName': assistantName, // this is how you activate the voice assistant using "Hey, ${name}",
    "memeTrigger": 'listiges Bild', // if you say this, assistant will fetch a meme from reddit and show it,
    "gptSystemMessage": `Du bist einer virtueller Sprachassistent. Dein Name ist ${assistantName}. Du gibst kurze genaue Antworten. Das Aktuelle Datum ist: ${new Date().toISOString()}\n\n`
};