Building a Local AI Assistant: Progress on VantaAI

As the world moves towards advanced artificial intelligence, I decided to embark on an exciting journey of creating my very own local AI assistant named VantaAI. Inspired by the capabilities of ChatGPT, my goal is to design an AI that operates entirely offline while incorporating unique features such as emotional memory and dynamic mood shifts.

What’s Unique About VantaAI?

VantaAI is built with a range of innovative functionalities that set it apart from conventional AI assistants:

1. Evolving Long-Term Memory: VantaAI retains information that adapts based on the context of our conversations, allowing for a more personalized experience.
2. Mood Tracking: A thoughtfully designed mood graph illustrates the AI’s emotional fluctuations over time, giving depth to interactions.
3. Narrative-Driven Memory Clustering: The assistant perceives itself as the protagonist of its own narrative, which influences how it remembers and responds to past conversations.
4. User-Friendly Interface: Built with a PySide6 GUI, VantaAI features separate tabs for memory, training, emotional states, and plugin management, making it intuitive and accessible.

Currently, VantaAI employs a custom Vulkan backend for accelerated model inference and training, enabling smooth operation and responsiveness. It offers personality-driven responses and supports live plugin hot-reloading, enhancing its versatility and adaptability.

Join the Conversation

I’m sharing this project not for commercial purposes, but out of pure curiosity. I would love to know if anyone else is working on similar AI projects or has suggestions for additional features I could explore.

Feel free to reach out with questions or insights; I’m eager to exchange ideas and learn from others in this fascinating field!