AI Chatbot - Text&Voice [Ollama+Vosk]

This is a console-based chatbot application that leverages machine learning models for natural language processing. It supports voice commands, text-to-speech, and syntax highlighting for programming languages.

Screenshots

Features

• Voice Assistant: Record audio and transcribe it to text for conversation.
• Model Interaction: Generate responses from a machine learning model based on conversation history.
• Syntax Highlighting: Display programming code with syntax highlighting for better readability.
• Text-to-Speech: Convert text responses into speech for an interactive experience.
• Command History: Keep track of user input across sessions.

Requirements

To run this application, you'll need to install Ollama and the following Python packages into your environment:

• sounddevice
• numpy
• requests
• pyttsx3
• vosk
• pygments
• colorama

You can install the required packages using pip:

pip install sounddevice numpy requests pyttsx3 vosk pygments colorama

Or using the requirements.txt file

Configuration

1. Config file Before running the application, you'll need to modify or create a config.json file to store your configuration settings. Here's a sample structure:

{
    "APP": {
        "name": "AI Chatbot",
        "version": 1.0
    },
    "OLLAMA": {
        "model_name": "llama3.2:latest",
        "temperature": 0.5,
        "seed": 42,
        "typing_delay": 0.007,
        "address": "http://localhost:11434/api/chat",
        "context-initialization": [
            {
                "role": "user",
                "content": "Always answer very short, but act like a professional. Start over."
            },
            {
                "role": "assistant",
                "content": "Alright"
            }
        ]
    },
    "VOSK": {
        "path": "models/vosk/",
        "model_name": "vosk-model-small-en-us-0.15"
    },
    "AUDIO_RECORDER": {
        "threshold": 0.05,
        "silence_duration": 1.5,
        "samplerate": 44100
    },
    "VOICE": {
        "id": "English",
        "rate": 125,
        "volume": 1.0
    }
}

OLLAMA

• model_name: Specify the model you want to use. Ensure the model is available. (e.g., "model_name": "llama3.2:latest")
• temperature: Adjust this value to control the randomness of responses. (0.0 = deterministic, 1.0 = highly random)
• seed: Change this to set the random seed for reproducibility.
• typing_delay: Modify this value to adjust the delay in typing simulation.
• address: Update the API address if your server runs on a different port or IP. (e.g., "address": "http://127.0.0.1:8000/api/chat")
• context-initialization: Customize the initial messages to set the assistant’s behavior and tone.

VOSK

• path: Ensure this path points to your VOSK model directory. (e.g., "path": "models/vosk/")
• model_name: Use the name of the VOSK model you want to load. (keep in mind that it should be the folder name inside vosk folder)

AUDIO_RECORDER

• threshold: Adjust this value to change the sensitivity of audio detection.
• silence_duration: Set the duration (in seconds) for silence detection.
• samplerate: Change the sample rate for audio recording as required.

VOICE

• id: Specify the voice language/identifier. (e.g., "id": "Spanish")
• rate: Adjust the speech rate (words per minute).
• volume: Change the volume level (0.0 to 1.0).

Models

Before running the application, you will also need to download the models for Ollama and Vosk. For Ollama you can just push into the console ollama pull llama3.2 and for Vosk you need to go their website. Remember to set the path for Vosk in the config.json file.

Usage

1. Run the Application: Execute the Python script to start the chatbot console.

   python main.py

2. Commands:

   • /voice: Start the voice assistant to record audio input.
   • /voice list: List all available voices.
   • /voice set "value": Set the voice by ID.
   • /help: Display available commands.
   • /clear: Clear the console.
   • /exit: End the conversation.

3. Conversation: Type your messages into the console to engage in conversation with the chatbot. The chatbot will generate responses based on your input.

Project Structure

It should look like this to make it work correctly.

Chatbot
├── models
│   └── vosk
│       └── vosk-model-small-en-us-0.15
├── main.py
├── config.json
├── audio_recorder.py
├── spinner.py
├── syntax_highlighter.py
├── text_to_speech.py
├── transcriber.py
└── requirements.txt

• main.py: The main script that runs the chatbot.
• config.json: Configuration file for application settings.
• audio_recorder.py: Handles audio recording and transcribing.
• transcriber.py: Manages audio transcription using Vosk.
• text_to_speech.py: Converts text to speech using pyttsx3.
• spinner.py: Displays a loading spinner during processing.
• syntax_highlighter.py: Provides syntax highlighting for programming languages.

Dependencies

The application uses several external libraries as mention above:

• Audio Recorder: Utilizes sounddevice and numpy for audio capture and processing.
• Spinner: Implements a loading spinner with itertools and threading.
• Syntax Highlighter: Uses pygments for syntax highlighting in various programming languages.
• Text to Speech: Integrates pyttsx3 for text-to-speech functionality.
• Transcriber: Leverages vosk for real-time speech recognition.

License

This project is licensed under the MIT License - see the LICENSE file for details.

External Licenses:

• Ollama
• Vosk