Conversational AI

A conversational AI system built with the following key components:

• Speech-to-Text (STT)
• Text-to-Speech (TTS) using ElevenLabs
• LLM Backend: meta-llama/Llama-3.2-1B-Instruct from Hugging Face
• Session & Memory Management using User ID
• Emotion Detection using bhadresh-savani/distilbert-base-uncased-emotionusing Hugging Face
• Gradio Interface for a user-friendly web app experience

This AI assistant listens, understands, analyzes emotion, remembers past conversations, and responds naturally with human-like speech.

---

🛠 Requirements

Before running this project, ensure you have:

1. A Hugging Face API Token (to access LLaMA and Emotion models)
2. An ElevenLabs API Key (for TTS)

---

Installation steps

1. Clone the Repository

git clone https://github.com/yourusername/Conversational_AI.git

cd Conversational_AI

2. Create a Virtual Environment (optional)

python -m venv venv

source venv/bin/activate # or venv\Scripts\activate on Windows

3. Install Dependencies

pip install -r requirements.txt

4. Setup API keys in app.ipynb

5. Usage

Jupyter notebook : app.ipynb

Follow the notebook cells to start the web app. This will launch a Gradio interface in your browser where you can:

1. Start the conversation only after entering a User ID eg. 1
2. Use microphone input or type your query
3. Hear voice responses

Future Enhancements

• Long-term vector memory storage
• RAG model
• Multilingual support
• Android Application
• Emotion-based response modulation
• Human-like TTS

---