A conversational AI assistant powered by Zep knowledge graphs, custom LLM integration, and Anam AI avatar—creating natural conversations with memory and context-aware responses.

We use:

Zep for conversation memory and knowledge graph management
Anam AI for realistic avatar and voice interactions
OpenRouter with Minimax M2 (LLM)
FastAPI for streaming backend
Streamlit to wrap the logic in an interactive UI

Architecture

Set Up

Run these commands in project root

Install Dependencies

uv sync

Configure Environment

Create a .env file in the project root, similar to .env.example, and add your API keys:

Get your Zep API keys here

Ingest Data

To populate the knowledge graph with your data:

python scripts/ingest_to_graph.py

Run the Application

Start the backend server:

uvicorn backend:app --port 8000 --reload

In a separate terminal, start the frontend:

streamlit run app.py

Usage

Enter your session unique name in the sidebar
Click "Initialize New Session" to create a Zep session
Click "Start Conversation" to interact with the avatar
The assistant uses knowledge graph context and conversation history for personalized responses

Note: To understand Zep better, we recommend going through the Jupyter notebook (zep_demo.ipynb) provided in the project root.

Get a FREE Data Science eBook 📖 with 150+ essential lessons in Data Science when you subscribe to our newsletter! Stay in the loop with the latest tutorials, insights, and exclusive resources. Subscribe now!

Contribution

Contributions are welcome! Feel free to fork this repository and submit pull requests with your improvements.

README.md

AI Avatar Demo powered by Zep