Virgil AI (Assistent Interface)

Key Features • How To Use • Credits • Installing • To be considered! • Road map

---

Created principle with python3.11 e various library like SpeechRecognition and TTS library

Virgilo or Virgil is a virtual assistant like Alexa or Google Home, but integrated with an AI (GPT-3.5 turbo).Designed to give the user the ability to use it and do what they want with it from putting it on a rasperry and using it as alexa to integrating it with their device whether it's linux or windows. With the possibility to set your own settings according to the need from where and when you want ㅤㅤ

Some information 📋

Virgil-AI 🤖

💻 How VirgilAI works

🔑 Key features

You can ask lots of questions at Virgilio, like us:

• The time ⏲️
• The weather 🌧️
• The lates news 🗞️
• Change the volume 🔉
• The temperature 🌡️
• Days of the week 📆
• Interact with the domotic (Merros device only) 💡
• Timer 🔂
• Ask a Virgil to remember your commitments 🗓️
• Mediaplayer 🎵
• and ask whatever you want like: Virgilio explain quantum math 🤖

Is fast to use:

• In fact all you have to do is insert the key in the app and you're done ✅

Portable:

• You can put it on any linux/windows device including mac 🌐

📋 Installing

1. The first part of the installation is to download all the files from the repository

   • command linee git clone https://github.com/Retr0100/VirgilApp.git
   • or download the zip

2. Now we need the api (for now i am not rich and i do not pay for everything) so we are need of 3 api keys (the keys marked with * are mandatory for operation)

   • API for OpenAI and GPT, i recommend this video tutorial *
   • API for Merros (domotic socket), just create a Merros account and insert the credential
   • API for ElevenLabs This API is not required, but if you want a BEST EXPERIENCE i raccomand you to get

3. When you have all the keys/accounts, save them on any file

Notes ❗

THIS IS NOT OBLIGATORY.
Before the user on Linux (preferably, but also on Windows it would not be) create an enviroment with venv python -m venv name_enviroment after writing source name_enviroment/bin/activate. Now you can install all the dependencies without putting your PC at risk. To close the environment, just run deactivate. If you use the enviroment try Virgil only whith the text interface

📖 How to use

When you have installed and downloaded the API you can start using Virgilio 🥡

1. Open a terminal to the directory ProjectVirgil/setupAndLaunch and run the file with setup.bat for windows or run the file bash bash setup.sh on linux (first make it execuble) instead for the fast start launch python launch.py or python3 launch.py
2. At a certain point you will be asked to install the dedicated app VirgilApp, this app will allow you to configure Virgilio and change the associated settings BUT BEWARE IT IS NOT ESSENTIAL FOR THE USE OF VIRGILIO, VIRGILIO CAN ALSO BE USED WITHOUT THE USE OF THE APP if you do not want to use the app click here
3. In the app, you will have to search for the configuration page (it will not be difficult to find it) and you will have to enter the key that will be shown to you on the screen when Virgilio is launched (you only have to do this step the first time, but you can repeat it if you want to change the key)
4. Now you can change the settings of Virgil on the corresponding page (Each time you change a setting in order to have the settings updated, you will have to restart Virgil)
5. At the end of this initial configuration, which you only need to do once, you can press any key and choose to use a text interface (recommended for debugging) or a voice interface (recommended for personal use).

📚 Guide to setting

// THE VALUES ON THE JSON ARE THE DEFAULT
{
    "language": "it", //The launguage for now is useless sorry
    "wordActivation": "Virgilio", //The word on Virgil can Activate
    "volume": "100.0", // Set the start volume of Virgil
    "city": "Salerno", // City default for the Meteo
    "operation_timeout": "3", // Listening time if you don't tal 
    "dynamic_energy_threshold": "true", // Automatic and dynamic microphone sensitivity
    "energy_threshold": "3500", //Sensitivy of microphone                                     
    "elevenlabs": "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa", //Key for elevenlabs                  
    "openAI": "sk-aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa", //Key for openAI
    "merrosEmail": "email", //Credential for merros                                                                
    "merrosPassword": "password", //Credential for merros                                                               
    "temperature": "0.9", //Randomness of GPT responses
    "max_tokens": "30" //Max lenght phrase of GPT                                                                       
}                                                                                                                      

🎙️ Guide to Microphone

Problem

The recognizer tries to recognize speech even when I’m not speaking, or after I’m done speaking.

Solution

Try increasing the recognizer_instance.energy_threshold property. This is basically how sensitive the recognizer is to when recognition should start. Higher values mean that it will be less sensitive, which is useful if you are in a loud room.

I created this tool for you (the tool is included in the repository) 💓

import math
import speech_recognition as sr
import time
listener = sr.Recognizer()
def main(languageChoose:str):
    print("SAY A WORD OR PHRASE IN YOUR LANGAGE")
    resultDict = {}
    for i in range(5):
        try:
            with sr.Microphone() as source:
                print(f"{i}. SPEAK")
                start_time = time.time()
                voice = listener.listen(source,3,15)
                command = listener.recognize_google(voice,language=languageChoose)
                end_time = time.time()
                resultDict[i] = [listener.energy_threshold,command,end_time - start_time]
        except:
            pass
    return resultDict
if __name__ == "__main__":
    listener.operation_timeout = 2
    listener.dynamic_energy_threshold = True
    languageChoose = str(input("Insert your language nation and dialet if is not dialet simple repeate the nation example it-it: "))
    results  = main(languageChoose)
    sorted_keys = sorted(results.keys(), key=lambda key: results[key][2])
    sorted_dict = {key: results[key] for key in sorted_keys}
    print(f"Recommended value:  {math.ceil(list(sorted_dict.values())[0][0])}")

The tool is not 100% accurate but it is a great way to get a starting idea if you see that virgil is not accurate enough in recognizing when you speak and when you don't try adjustment

Why the key of OpenAI,ElevenLabs and Merros❓

• OpenAI: This is in fact the only mandatory key, as GPT covers 50% of the application, and this is the real difference to Alexa and Virgil.
• ElevenLabs: This key is not mandatory but it makes the experience more pleasant because ElevenLabs implements a more natural Speech To Text (TTS) and also allows you to choose your own voice. The key for the API is free and only requires registration (TIPS: If you run out of tokens and want to continue using elevenlabs free, create another account with the same email address, but put a dot anywhere before the "@" and the confirmation email will still arrive, but it will be a different email address for the site... SHHH DON'T TELL ANYONE) If you can't register, Virgil will still work, but with Google's TTS. and it's not the best choice 😅.
• Meross: This credential an required ONLY if you can use a domotic Meross but if you dont have a domotic Meross don't waste time ⏲️

How to change the voice for TTS ElevenLabs:

1. Go in this file lib/sound.py
2. Go on the site of ElevenLabs create an account (You should already have it)
3. Explore the default Voice and choice one
4. Now on this part of file sound.py

   sound = generate(
                       api_key = self.API_KEY,
                       text=text,
                       voice="Antoni",
                       model='eleven_multilingual_v1'
                   )
   

   And replace the voice whith the one you want (if after the TTS dont'work try another voice on whatch a video on YT on how to use default entries )
5. Restart Virgil

📃 Configuration without app

not updated to revision

1. In the main directory there is a file call settings.json in this file is present all the setting of Virgil open it
2. Now you can modify all the setting like API key, location etc

🔁 Change the key

Go in the directory call setup and search the key.txt (remeber the key is a hexadecimal string of 32 characters),delete it and relaunch Virgil

⚠️ TO BE CONSIDERED

1. The entire project is only at the albor and is not 100% complete
2. Virgilio without the API insert and without the connection will not function fully
3. FOR NOW many command works only in italian except for the question at GPT (so any questions not covered by the default commands)
4. The api for GPT is almost free but is necessary inser the credit/debit card (sorry i know is annoying) 💸

💎 RoadMap for the ProjectVirgil

• Make a beta of VirgilAI
• Make a first Version of App
• Increase the possibility of Virgil
• Create a website for the Project
• Improve the command recognition
• Improve the code design and make the code more efficent
• Add more language

💸 Credits and technologies used

Only me 💻 and

• openAI API
• OpenMeteo API
• Merros API
• MongoDB for Database of the setting
• Flask for create the API
• Flutter for create the mobile app (only android for now)
• Firebase for the manage of account
• EvenLabs API

ps: follow me on instagram and twitter contact me for help or support