🚀 UA-EN Voice Assistant

🔐 No API-Keys | 🖥️ Local | 🧠 LLaMA 3.1 (12K Tokens Prompt-Menu ~20 Pages in Single Request)

🎥 Demo Versions:

---

💡 Idea

This project is a proof-of-concept for an MVP, inspired by OMNI's capabilities (ChatGPT).
✅ Main advantage — fully local deployment with no restrictions.
🛠️ Use it for:

• 🌍 Language translation
• 📚 Learning enhancement (reading, writing, audio skills)
• 🎛️ Customization for specific needs

---

⚙️ Features & Tech Stack

🗣️ Language Classification: Auto-detects UA/EN using Lang-id-voxlingua107-ecapa (supports 100+ languages).
🎙️ Google Legacy Recognizer: Fast and free speech recognition.
🔊 Wav2Vec2-Bert: Best Ukrainian speech-to-text model (so far).
🗣️ Edge-TTS: High-quality natural voices for free.
🤖 Ollama-python: Download and run LLMs offline.
🖥️ Streamlit: GUI support.
📜 Dialogue saved in JSON: HISTORY.json (for main.py), short-term memory for app.py.
📝 Config.py: Customize prompt for the best user experience.

---

🚀 Getting Started

✅ Tested on:

🖥️ WSL 22.04.3
🎮 GeForce GTX 1050Ti (4GB, Mobile)
🧠 RAM 32GB

📌 Requirements:

🐍 Python 3.9+
💾 Conda 3.9+ (virtual environment)
⚡ CUDA (optional)

🔧 Installation

# 1️⃣ Clone the repository
git clone https://github.com/Katashynskyi/Voice_assistant_UA_EN.git
cd Voice_assistant_UA_EN

# 2️⃣ Create and activate a virtual environment
conda activate ./.conda

# 3️⃣ Install dependencies
pip install uv
uv pip install -r requirements.txt

---

▶️ Usage

🔹 Console Mode:

python3 main.py

🔹 Graphical Interface (GUI):

streamlit run app.py

🚀 Done! Enjoy your powerful voice assistant! (suggesting to change LLM to phi4 at least) 🔥