GUI IA

Welcome! The GUI is a virtual assistant designed to answer your questions in a friendly, fun, and interactive way. Created to be your digital companion, it provides engaging responses to make your experience more enjoyable.

• Model Used: By default, we use the meta-llama/Llama-3.2-3B-Instruct model, but you can easily change it to any other model of your choice.

Recommended

• Python 3.11.0 or higher
• A Hugging Face account to obtain your API token

Installation

1. Clone the repository:

   git clone https://github.com/wendellast/Gui.git
   cd Gui

2. Install the dependencies:

   pip install -r requirements.txt

3. Configure the .env file:

   • Rename the .env-example file to .env:

     mv .env-example .env

   • Edit the .env file and add your Hugging Face token:

     HUGGINGFACEHUB_API_TOKEN=your_token_here
     

   Note: You can get your Hugging Face API token at Hugging Face Tokens.

   Note: If you're using a graphical interface with Python gui.py you don't need the Hugging Face token; it works without it.

How to Run

option 1: Running via Server

To start the server, run:

python server.py

Open your browser and go to http://localhost:7860 to interact with the AI chatbot.

Option 2: Running via Graphical Interface

To start the server and open the graphical interface, simply run the following command:

python gui.py

This will launch the application with the virtual assistant interface, where you can interact using voice or buttons.

---

Speech Configuration

The virtual assistant uses speech synthesis to respond to the user. jWe recommend using the Letícia voice, a high-quality Brazilian voice, for the best experience.

1. Using the Letícia Voice

We recommend using the Letícia voice. jTo set it up, follow these steps:

• Visit the Louderpages - Letícia website.
• Github Rhvoices
• Follow the instructions to configure the Letícia voice.

2. Other Alternatives

If you prefer, you can also use other speech synthesis options:

• Espeak: jAn open-source alternative.
• SAPI5 (Windows): jThe native speech synthesis API for Windows.

API

Example Usage via API:

from gradio_client import Client

# ==========TEST API==========

def response_gui(input_text):
    client = Client("wendellast/GUI")
    result = client.predict(
        message=input_text,
        max_tokens=512,
        temperature=0.7,
        top_p=0.95,
        api_name="/chat",
    )
    return result

# Example call:
input_text = "Hello, how are you?"
response = response_gui(input_text)
print("AI Response:", response)

Usage via LangChain Model

You can also use the model directly via LangChain:

• Define the model you want to use, such as meta-llama/Llama-3.2-3B-Instruct.
• Configure your access token in the .env file.
• Instantiate the wrapper for the model using the GuiChat class.

Supported Parameters:

• temperature: Controls the randomness of the response.
• top_p: Controls the diversity of the responses.
• repetition_penalty: Penalizes repetitions for more varied answers.
• max_new_tokens: Maximum number of tokens generated in the response.

Example usage:

from util.token_access import load_token
from your_package import GuiChat

token = load_token()
chatbot = GuiChat(auth_token=token)

while True:
    question = input("Ask here: ")
    answer = chatbot._call(question)
    print(f"Response: {answer}")